bc323b609b
Use the default, which freely coalesces anything it can. This mostly shows improvements, with a handful of regressions. The main concern would be if introducing wider registers is more likely to push the register usage up to the next occupancy tier.
14317 lines
513 KiB
LLVM
14317 lines
513 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 < %s | FileCheck -check-prefixes=GFX9,GFX900 %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a < %s | FileCheck -check-prefixes=GFX9,GFX90APLUS,GFX90A %s
|
|
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx942 < %s | FileCheck -check-prefixes=GFX9,GFX90APLUS,GFX942 %s
|
|
|
|
|
|
define void @v_shuffle_v3p3_v4p3__u_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v3p3_v4p3__u_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> poison
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__0_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__0_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__0_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__0_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__1_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__1_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__1_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__1_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 poison, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__2_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__2_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__2_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__2_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 poison, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__3_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__3_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__3_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__3_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 poison, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__4_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v3p3_v4p3__4_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 poison, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__5_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__5_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__5_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__5_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 poison, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__6_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__6_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__6_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__6_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 poison, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_0_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_0_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_0_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_0_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_1_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_1_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_1_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_1_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_2_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_2_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_2_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_2_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_3_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_3_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_3_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_3_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_4_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_4_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_4_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_4_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_5_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_5_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_5_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_5_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_6_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_6_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_6_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_6_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_7_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_7_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_7_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_7_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 poison>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_7_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_7_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[5:8]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v9, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_7_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_7_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_7_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_7_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_7_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_7_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_7_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_7_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_7_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_7_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_7_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_7_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_7_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_7_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_7_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_7_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_7_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_7_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_7_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_7_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_7_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_7_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_7_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_7_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_7_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_7_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__u_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__u_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v0, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__u_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v0, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__u_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v0, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 poison, i32 0, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__0_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__0_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__0_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__0_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> zeroinitializer
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__1_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__1_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__1_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__1_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 0, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__2_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__2_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__2_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__2_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 0, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__3_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__3_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__3_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__3_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 0, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__4_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__4_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v0, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__4_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v0, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__4_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v0, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 0, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__5_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__5_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__5_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__5_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 0, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__6_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__6_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__6_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__6_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 0, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_u_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_u_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[5:8]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_u_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: global_store_dwordx3 v1, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_u_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v1, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_1_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_1_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[5:8]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v9, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_1_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_1_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_2_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_2_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[5:8]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v9, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_2_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v9
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX90A-NEXT: global_store_dwordx3 v10, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_2_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v9
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX942-NEXT: global_store_dwordx3 v10, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_3_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_3_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[5:8]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v9, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_3_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_3_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_4_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_4_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[5:8]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v9, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_4_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_4_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_5_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_5_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[5:8]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v9, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_5_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_5_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_6_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_6_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[5:8]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v9, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_6_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v6
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_6_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v6
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 0>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__u_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__u_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__u_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__u_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 poison, i32 1, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__0_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__0_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__0_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__0_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 0, i32 1, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__1_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__1_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__1_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__1_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 1, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__2_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__2_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__2_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__2_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 1, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__3_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__3_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__3_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__3_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 1, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__4_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__4_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__4_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__4_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 1, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__5_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__5_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__5_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__5_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 1, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__6_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__6_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__6_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__6_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 1, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_u_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_u_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_u_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_u_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_0_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_0_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_0_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_0_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_2_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_2_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_2_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[4:6], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_2_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[4:6], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_3_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_3_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_3_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_3_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_4_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_4_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_4_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[8:10], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_4_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[8:10], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_5_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_5_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_5_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_5_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_6_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_6_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_6_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_6_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 1>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__u_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__u_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__u_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__u_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 poison, i32 2, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__0_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__0_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__0_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__0_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 0, i32 2, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__1_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__1_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__1_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__1_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 2, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__2_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__2_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__2_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__2_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 2, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__3_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__3_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__3_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__3_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 2, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__4_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__4_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__4_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__4_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 2, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__5_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__5_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[4:6], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__5_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__5_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 2, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__6_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__6_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__6_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__6_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 2, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_u_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_u_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_u_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_u_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_0_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_0_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_0_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[4:6], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_0_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[4:6], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_1_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_1_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_1_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_1_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_3_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_3_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_3_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_3_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_4_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_4_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_4_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_4_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_5_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_5_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_5_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[4:6], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_5_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[4:6], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_6_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_6_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_6_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v6
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_6_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v6
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 2>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__u_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__u_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__u_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__u_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 poison, i32 3, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__0_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__0_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__0_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__0_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 0, i32 3, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__1_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__1_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__1_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__1_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 3, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__2_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__2_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__2_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__2_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 3, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__3_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__3_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__3_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__3_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 3, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__4_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__4_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__4_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__4_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 3, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__5_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__5_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[5:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__5_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__5_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 3, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__6_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__6_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__6_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__6_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 3, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_u_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_u_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_u_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_u_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_0_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_0_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_0_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[4:6], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_0_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[4:6], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_1_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_1_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_1_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_1_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_2_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_2_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_2_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_2_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_4_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_4_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_4_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_4_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_5_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_5_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_5_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[4:6], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_5_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[4:6], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_6_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_6_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_6_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_6_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v6
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 3>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__u_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v3p3_v4p3__u_4_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 poison, i32 4, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__0_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__0_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__0_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__0_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 0, i32 4, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__1_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__1_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__1_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__1_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 4, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__2_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__2_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__2_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__2_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 4, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__3_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__3_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__3_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__3_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 4, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__4_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v3p3_v4p3__4_4_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 4, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__5_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__5_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__5_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__5_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 4, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__6_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__6_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__6_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__6_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 4, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_u_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_u_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_u_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v1, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_u_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v1, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_0_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_0_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_0_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v6
|
|
; GFX90A-NEXT: global_store_dwordx3 v10, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_0_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v6
|
|
; GFX942-NEXT: global_store_dwordx3 v10, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_1_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_1_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_1_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_1_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_2_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_2_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_2_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v6
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_2_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_3_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_3_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_3_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_3_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_5_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_5_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_5_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_5_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_6_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_6_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v0
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_6_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_6_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 4>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__u_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__u_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__u_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__u_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 poison, i32 5, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__0_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__0_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__0_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__0_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 0, i32 5, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__1_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__1_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__1_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__1_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 1, i32 5, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__2_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__2_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__2_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__2_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 2, i32 5, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__3_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__3_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__3_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[4:6], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__3_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[4:6], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 3, i32 5, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__4_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__4_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__4_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__4_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 4, i32 5, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__5_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__5_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__5_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__5_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 5, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__6_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__6_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__6_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__6_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 5, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_u_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_u_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_u_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_u_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_0_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_0_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_0_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[8:10], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_0_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[8:10], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_1_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_1_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_1_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_1_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_2_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_2_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_2_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_2_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_3_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_3_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_3_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_3_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_4_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_4_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_4_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[6:8], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_4_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[6:8], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_6_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_6_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_6_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v1
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[6:8], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_6_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v1
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[6:8], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 5>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__u_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__u_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__u_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__u_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 poison, i32 6, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__0_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__0_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__0_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__0_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 0, i32 6, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__1_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__1_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__1_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__1_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v4
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 1, i32 6, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__2_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__2_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__2_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__2_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 2, i32 6, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__3_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__3_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__3_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v6
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[4:6], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__3_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[4:6], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 3, i32 6, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__4_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__4_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__4_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__4_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 4, i32 6, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__5_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__5_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__5_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__5_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 6, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__6_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__6_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__6_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__6_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 6, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_u_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_u_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_u_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_u_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_0_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_0_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_0_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_0_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_1_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_1_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_1_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_1_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_2_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_2_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_2_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[4:6], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_2_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[4:6], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_3_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_3_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_3_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_3_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_4_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_4_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_4_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[6:8], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_4_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[6:8], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_5_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_5_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v2
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_5_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_5_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 6>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__u_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__u_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__u_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__u_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 poison, i32 7, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__0_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__0_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__0_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__0_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 0, i32 7, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__1_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__1_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__1_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__1_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 1, i32 7, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__2_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__2_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__2_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__2_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 2, i32 7, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__3_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__3_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__3_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__3_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 3, i32 7, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__4_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__4_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__4_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__4_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 4, i32 7, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__5_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__5_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[1:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__5_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__5_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v1
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 7, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__6_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__6_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__6_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v5, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__6_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v5, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 7, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_u_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_u_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v4, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_u_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_u_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_0_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_0_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v8, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_0_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_0_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_1_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_1_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[3:6]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v7, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_1_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:5]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX90A-NEXT: global_store_dwordx3 v6, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_1_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:5]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v5
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v5
|
|
; GFX942-NEXT: global_store_dwordx3 v6, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_2_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_2_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:5]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_2_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_2_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_3_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_3_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[1:4]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_3_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX90A-NEXT: global_store_dwordx3 v8, v[2:4], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_3_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v7
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v7
|
|
; GFX942-NEXT: global_store_dwordx3 v8, v[2:4], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_4_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_4_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_4_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[6:8], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_4_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[6:8], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_5_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_5_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v1
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_5_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_5_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3p3_v4p3__7_6_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3p3_v4p3__7_6_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:3]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx3 v6, v[3:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3p3_v4p3__7_6_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:3]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX90A-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX90A-NEXT: global_store_dwordx3 v4, v[0:2], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3p3_v4p3__7_6_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v0, v3
|
|
; GFX942-NEXT: v_mov_b32_e32 v1, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v3
|
|
; GFX942-NEXT: global_store_dwordx3 v4, v[0:2], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 7>
|
|
store <3 x ptr addrspace(3)> %shuf, ptr addrspace(1) %ptr, align 16
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__u_u_u() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__u_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> poison
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__0_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__0_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__0_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__0_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__1_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__1_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__1_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__1_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__2_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__2_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__2_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__2_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__3_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__3_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__3_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__3_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__4_u_u() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__4_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__5_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__5_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__5_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__5_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__6_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__6_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__6_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__6_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_0_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_0_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_0_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_0_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_1_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_1_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_1_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_1_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_2_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_2_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_2_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_2_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_3_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_3_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_3_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_3_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_4_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_4_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_4_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_4_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_5_u() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__7_5_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_6_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_6_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_6_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_6_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_7_u() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_7_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_7_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_7_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_7_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_7_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_7_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_7_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_7_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_7_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_7_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_7_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_7_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_7_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_7_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_7_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_7_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_7_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_7_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_7_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_7_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_7_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_7_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_7_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_7_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_7_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_7_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_7_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_7_6() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__7_7_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__u_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__u_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__u_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__u_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 poison, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__0_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__0_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s8
|
|
; GFX9-NEXT: s_mov_b32 s10, s8
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> zeroinitializer
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__1_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__1_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__1_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__1_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__2_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__2_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__2_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__2_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__3_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__3_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__3_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__3_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__4_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__4_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__4_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__4_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__5_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__5_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s9
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__5_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s9
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__5_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s5
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__6_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__6_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__6_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__6_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_u_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_u_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_u_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_u_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_1_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_1_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s5
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_1_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s5
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_1_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s1
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_2_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_2_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_2_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_2_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_3_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_3_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_3_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_3_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_4_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_4_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s15
|
|
; GFX900-NEXT: s_mov_b32 s9, s12
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_4_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s15
|
|
; GFX90A-NEXT: s_mov_b32 s9, s12
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_4_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s4
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_5_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_5_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_5_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_5_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_6_0() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_6_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_6_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_6_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s6
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__u_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__u_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 poison, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__0_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__0_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 0, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__1_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__1_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__2_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__2_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__3_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__3_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__4_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__4_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__5_1_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__5_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s10, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__5_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s10, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__5_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: s_mov_b32 s10, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__6_1_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__6_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__6_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__6_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_1_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_u_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_u_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_u_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_u_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_0_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_0_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_0_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_0_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_2_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_2_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_2_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_2_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_3_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_3_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_3_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_3_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_4_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_4_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s15
|
|
; GFX900-NEXT: s_mov_b32 s9, s12
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_4_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s15
|
|
; GFX90A-NEXT: s_mov_b32 s9, s12
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_4_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s4
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_5_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_5_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_5_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_5_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_6_1() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_6_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_6_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_6_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s6
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__u_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__u_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 poison, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__0_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__0_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 0, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__1_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__1_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__2_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__2_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__3_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__3_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__4_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__4_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__5_2_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__5_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__5_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__5_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: s_mov_b32 s9, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__6_2_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__6_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__6_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__6_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_2_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_u_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_u_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_u_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_u_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_0_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_0_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_0_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_0_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_1_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_1_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_1_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_1_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_3_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_3_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_3_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_3_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_4_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_4_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_4_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_4_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_5_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_5_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_5_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_5_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_6_2() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_6_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_6_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_6_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__u_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__u_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__u_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__u_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 poison, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__0_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__0_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 0, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__1_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__1_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__1_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__1_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__2_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__2_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__2_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__2_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__3_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__3_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__3_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__3_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__4_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__4_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__4_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__4_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__5_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__5_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s9
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__5_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s9
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__5_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s5
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__6_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__6_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__6_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__6_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_u_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_u_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_u_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_u_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_0_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_0_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_0_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_0_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_1_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_1_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_1_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_1_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_2_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_2_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_2_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_2_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_4_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_4_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s15
|
|
; GFX900-NEXT: s_mov_b32 s9, s12
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_4_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s15
|
|
; GFX90A-NEXT: s_mov_b32 s9, s12
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_4_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s4
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_5_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_5_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_5_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_5_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_6_3() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_6_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_6_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_6_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s6
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__u_4_4() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__u_4_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 poison, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__0_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__0_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__0_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__0_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 0, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__1_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__1_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__1_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__1_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 1, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__2_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__2_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__2_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__2_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 2, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__3_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__3_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__3_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__3_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 3, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__4_4_4() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__4_4_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> poison, <3 x i32> <i32 4, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__5_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__5_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__5_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__5_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__6_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__6_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__6_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__6_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_u_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_u_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_u_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_u_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_0_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_0_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s15
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_0_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s15
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_0_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_1_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_1_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_1_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_1_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_2_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_2_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s15
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_2_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s15
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_2_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_3_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_3_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s15
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_3_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s15
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_3_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_5_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_5_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s5
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_5_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s5
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_5_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s1
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_6_4() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_6_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_6_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_6_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__u_5_5() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__u_5_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 poison, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__0_5_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__0_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s9, s5
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__0_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s9, s5
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__0_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s1
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 0, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__1_5_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__1_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s10, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__1_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s10, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__1_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: s_mov_b32 s10, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 1, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__2_5_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__2_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__2_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__2_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 2, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__3_5_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__3_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s9
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__3_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s9
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__3_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s9
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 3, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__4_5_5() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__4_5_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 4, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__5_5_5() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__5_5_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__6_5_5() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__6_5_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_5_5() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__7_5_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_u_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_u_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_u_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_u_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_0_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_0_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s15
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_0_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s15
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_0_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_1_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_1_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_1_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_1_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_2_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_2_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s15
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_2_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s15
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_2_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_3_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_3_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s15
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_3_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s15
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_3_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_4_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_4_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_4_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_4_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_6_5() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_6_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_6_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_6_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__u_6_6() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__u_6_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 poison, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__0_6_6() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__0_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__0_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__0_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 0, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__1_6_6() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__1_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__1_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__1_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: s_mov_b32 s9, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 1, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__2_6_6() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__2_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__2_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__2_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 2, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__3_6_6() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__3_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s10
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__3_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s10
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__3_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 3, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__4_6_6() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__4_6_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 4, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__5_6_6() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__5_6_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s9
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__6_6_6() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__6_6_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_6_6() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__7_6_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s9, s10
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_u_6() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__7_u_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_0_6() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_0_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_0_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_0_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s11
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_1_6() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_1_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_1_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_1_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_2_6() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_2_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_2_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_2_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s11
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_3_6() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_3_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_3_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_3_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s11
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_4_6() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_4_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s6
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_4_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s6
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_4_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_5_6() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__7_5_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__u_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__u_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__u_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__u_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 poison, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__0_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__0_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__0_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__0_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 0, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__1_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__1_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__1_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__1_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 1, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__2_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__2_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__2_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__2_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 2, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__3_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__3_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__3_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__3_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 3, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__4_7_7() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__4_7_7:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 4, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__5_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__5_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s5
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__5_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s5
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__5_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s1
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 5, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__6_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__6_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__6_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__6_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 6, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_u_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_u_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_u_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_u_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 poison, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_0_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_0_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_0_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_0_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 0, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_1_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_1_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_1_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_1_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 1, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_2_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_2_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_2_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_2_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 2, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_3_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_3_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s11
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_3_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s11
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_3_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s7
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 3, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_4_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_4_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s4
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_4_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s4
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_4_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s0
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 4, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_5_7() {
|
|
; GFX9-LABEL: s_shuffle_v3p3_v4p3__7_5_7:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:10]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 5, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3p3_v4p3__7_6_7() {
|
|
; GFX900-LABEL: s_shuffle_v3p3_v4p3__7_6_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s7
|
|
; GFX900-NEXT: s_mov_b32 s9, s6
|
|
; GFX900-NEXT: s_mov_b32 s10, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:10]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3p3_v4p3__7_6_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s7
|
|
; GFX90A-NEXT: s_mov_b32 s9, s6
|
|
; GFX90A-NEXT: s_mov_b32 s10, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:10]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3p3_v4p3__7_6_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:3]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s3
|
|
; GFX942-NEXT: s_mov_b32 s9, s2
|
|
; GFX942-NEXT: s_mov_b32 s10, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:10]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x ptr addrspace(3)> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x ptr addrspace(3)> %vec0, <4 x ptr addrspace(3)> %vec1, <3 x i32> <i32 7, i32 6, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:10]}"(<3 x ptr addrspace(3)> %shuf)
|
|
ret void
|
|
}
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; GFX90APLUS: {{.*}}
|