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.
14199 lines
509 KiB
LLVM
14199 lines
509 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_v3i64_v4i64__u_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v3i64_v4i64__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 i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> poison
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__0_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__0_u_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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__0_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__0_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__1_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__1_u_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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__1_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__1_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 poison, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__2_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__2_u_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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__2_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__2_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 poison, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__3_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__3_u_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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__3_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__3_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 poison, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__4_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v3i64_v4i64__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 i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 poison, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__5_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__5_u_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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__5_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__5_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__6_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__6_u_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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__6_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__6_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 poison, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_u_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_u_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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_0_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_0_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_0_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_0_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_1_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_1_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_1_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_1_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_2_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_2_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_2_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_2_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_3_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_3_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_3_u:
|
|
; 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[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_3_u:
|
|
; 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[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_4_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_4_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_4_u:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_4_u:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_5_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_5_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_5_u:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_5_u:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_6_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_6_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_6_u:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_6_u:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_7_u(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_7_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_7_u:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_7_u:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 poison>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_7_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_7_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_7_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_7_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_7_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_7_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_7_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_7_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_7_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_7_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_7_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_7_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_7_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_7_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_7_3:
|
|
; 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[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_7_3:
|
|
; 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[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_7_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_7_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_7_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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_7_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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_7_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_7_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_7_5:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_7_5:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_7_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_7_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_7_6:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_7_6:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_7_7:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_7_7:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__u_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__u_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v0, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__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[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v0, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__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[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v0, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v0, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 poison, i32 0, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__0_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__0_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__0_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__0_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v1
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> zeroinitializer
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__1_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__1_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__1_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__1_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 0, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__2_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__2_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__2_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__2_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 0, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__3_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__3_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__3_0_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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__3_0_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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 0, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__4_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__4_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v0, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v0, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__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[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v0, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v0, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__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[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v0, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v0, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 0, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__5_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__5_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__5_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__5_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__6_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__6_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__6_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__6_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 0, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_0_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_u_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_u_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_u_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_u_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_1_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_1_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_1_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_1_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_2_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_2_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_2_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_2_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_3_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_3_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_3_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_3_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_4_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_4_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_4_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_4_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_5_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_5_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_5_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_5_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_6_0(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_6_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_6_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_6_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 0>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__u_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__u_1_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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__u_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__u_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 poison, i32 1, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__0_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__0_1_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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__0_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__0_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 0, i32 1, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__1_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__1_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__1_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__1_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 1, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__2_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__2_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__2_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__2_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 1, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__3_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__3_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__3_1_1:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__3_1_1:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 1, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__4_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__4_1_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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__4_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__4_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 1, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__5_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__5_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__5_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__5_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__6_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__6_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__6_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__6_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 1, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_1_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_u_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_u_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_u_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_u_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_0_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_0_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[10:11], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_0_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[10:11], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_0_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[10:11], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_2_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_2_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_2_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_2_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_3_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_3_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_3_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_3_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_4_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_4_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_4_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_4_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_5_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_5_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_5_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_5_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_6_1(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_6_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_6_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_6_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 1>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__u_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__u_2_2:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__u_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__u_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 poison, i32 2, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__0_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__0_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__0_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__0_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 0, i32 2, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__1_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__1_2_2:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__1_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__1_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 2, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__2_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__2_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__2_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__2_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 2, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__3_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__3_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__3_2_2:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__3_2_2:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 2, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__4_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__4_2_2:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__4_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__4_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 2, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__5_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__5_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[8:11], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__5_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[8:11], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__5_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v11, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[8:11], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__6_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__6_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__6_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__6_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 2, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_2_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_u_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_u_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_u_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_u_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_0_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_0_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[12:13], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_0_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[12:13], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_0_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[12:13], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_1_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_1_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[10:11], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_1_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[10:11], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_1_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[10:11], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_3_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_3_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_3_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_3_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_4_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_4_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_4_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_4_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_5_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_5_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_5_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_5_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_6_2(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_6_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_6_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_6_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 2>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__u_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__u_3_3:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__u_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__u_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 poison, i32 3, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__0_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__0_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__0_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__0_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 0, i32 3, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__1_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__1_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; 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, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__1_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__1_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 3, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__2_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__2_3_3:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__2_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__2_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 3, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__3_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__3_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__3_3_3:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__3_3_3:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 3, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__4_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__4_3_3:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__4_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__4_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 3, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__5_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__5_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__5_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[10:13], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__5_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v13, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[10:13], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__6_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__6_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__6_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[12:15], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__6_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v15, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[12:15], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 3, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_3_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_u_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_u_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_u_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_u_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_0_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_0_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[14:15], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_0_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[14:15], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_0_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[14:15], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_1_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_1_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[12:13], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_1_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[12:13], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_1_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[12:13], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_2_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_2_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[10:11], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_2_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[10:11], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_2_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[10:11], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_4_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_4_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_4_3:
|
|
; 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[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_4_3:
|
|
; 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[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_5_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_5_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_5_3:
|
|
; 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[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_5_3:
|
|
; 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[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_6_3(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_6_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_6_3:
|
|
; 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[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_6_3:
|
|
; 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[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 3>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__u_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v3i64_v4i64__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 i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 poison, i32 4, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__0_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__0_4_4:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__0_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__0_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 0, i32 4, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__1_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__1_4_4:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__1_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__1_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 4, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__2_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__2_4_4:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__2_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__2_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 4, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__3_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__3_4_4:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__3_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__3_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 4, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__4_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX9-LABEL: v_shuffle_v3i64_v4i64__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 i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 4, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__5_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__5_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__5_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__5_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v1
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__6_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__6_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__6_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__6_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v1
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 4, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_4_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_4_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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_4_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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_u_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_u_4:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_u_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_u_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_0_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_0_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_0_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_0_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_1_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_1_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_1_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_1_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_2_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_2_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_2_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_2_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_3_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_3_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_3_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_3_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_5_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_5_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_5_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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_5_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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_6_4(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_6_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_6_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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[0:1], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_6_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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[0:1], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 4>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__u_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__u_5_5:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__u_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__u_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 poison, i32 5, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__0_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__0_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__0_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__0_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 0, i32 5, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__1_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__1_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__1_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__1_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 1, i32 5, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__2_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__2_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__2_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__2_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 2, i32 5, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__3_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__3_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[10:11], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__3_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[10:11], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__3_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[10:11], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 3, i32 5, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__4_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__4_5_5:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__4_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__4_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 4, i32 5, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__5_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__5_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__5_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__5_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v3
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__6_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__6_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__6_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__6_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v3
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 5, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_5_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_5_5:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_5_5:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_u_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_u_5:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_u_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_u_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_0_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_0_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_0_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_0_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_1_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_1_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_1_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_1_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_2_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_2_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_2_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_2_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_3_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_3_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_3_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_3_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_4_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_4_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_4_5:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_4_5:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_6_5(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_6_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_6_5:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[2:3], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_6_5:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[2:3], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 5>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__u_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__u_6_6:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__u_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__u_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 poison, i32 6, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__0_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__0_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__0_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__0_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 0, i32 6, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__1_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__1_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__1_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__1_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 1, i32 6, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__2_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__2_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[10:11], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__2_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[10:11], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__2_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[10:11], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 2, i32 6, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__3_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__3_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[12:13], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__3_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[12:13], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__3_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[12:13], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 3, i32 6, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__4_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__4_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__4_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__4_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v5
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 4, i32 6, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__5_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__5_6_6:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__5_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__5_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 6, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__6_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__6_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__6_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__6_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v5
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 6, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_6_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_6_6:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_6_6:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_u_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_u_6:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_u_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_u_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_0_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_0_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_0_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_0_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_1_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_1_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_1_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_1_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_2_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_2_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_2_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_2_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_3_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_3_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_3_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_3_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_4_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_4_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_4_6:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_4_6:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_5_6(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_5_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_5_6:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[4:5], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_5_6:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[4:5], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 6>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__u_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__u_7_7:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__u_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__u_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 poison, i32 7, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__0_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__0_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__0_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[8:9], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__0_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[8:9], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v8
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v9
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 0, i32 7, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__1_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__1_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX900-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[10:11], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__1_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[10:11], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__1_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[10:11], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v10
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v11
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 1, i32 7, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__2_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__2_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX900-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[12:13], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__2_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX90A-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[12:13], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__2_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[12:13], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v6, v12
|
|
; GFX942-NEXT: v_mov_b32_e32 v7, v13
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 2, i32 7, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__3_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__3_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[14:15], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__3_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[14:15], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__3_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v14
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[14:15], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 3, i32 7, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__4_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__4_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX900-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX900-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__4_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[0:3], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__4_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v2, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v3, v7
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[0:3], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 4, i32 7, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__5_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__5_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; 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, v7
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__5_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX90A-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[2:5], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__5_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: v_mov_b32_e32 v4, v6
|
|
; GFX942-NEXT: v_mov_b32_e32 v5, v7
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[2:5], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 7, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__6_7_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__6_7_7:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__6_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[4:7], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__6_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[4:7], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 7, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_u_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_u_7:
|
|
; 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[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_u_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v8, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v8, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_u_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v8, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v8, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_0_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_0_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v16, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_0_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v16, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v16, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_0_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: v_mov_b32_e32 v16, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v16, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v16, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_1_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_1_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[6:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v14, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_1_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[6:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v14, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v14, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_1_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[6:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v14, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v14, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v14, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_2_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_2_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v12, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_2_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v12, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v12, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_2_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v12, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v12, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v12, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_3_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_3_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[2:9]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_3_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[2:9]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def v[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_3_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[2:9]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def v[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_4_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_4_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_4_7:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_4_7:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v0
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v1
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_5_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_5_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_5_7:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_5_7:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v2
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v3
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @v_shuffle_v3i64_v4i64__7_6_7(ptr addrspace(1) inreg %ptr) {
|
|
; GFX900-LABEL: v_shuffle_v3i64_v4i64__7_6_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: v_mov_b32_e32 v10, 0
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def v[0:7]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX900-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX900-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX900-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: v_shuffle_v3i64_v4i64__7_6_7:
|
|
; 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[0:7]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX90A-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX90A-NEXT: global_store_dwordx2 v10, v[6:7], s[16:17] offset:16
|
|
; GFX90A-NEXT: global_store_dwordx4 v10, v[6:9], s[16:17]
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: v_shuffle_v3i64_v4i64__7_6_7:
|
|
; 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[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: global_store_dwordx2 v10, v[6:7], s[0:1] offset:16
|
|
; GFX942-NEXT: v_mov_b32_e32 v8, v4
|
|
; GFX942-NEXT: v_mov_b32_e32 v9, v5
|
|
; GFX942-NEXT: global_store_dwordx4 v10, v[6:9], s[0:1]
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0)
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=v"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=v"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 7>
|
|
store <3 x i64> %shuf, ptr addrspace(1) %ptr, align 32
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__u_u_u() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__u_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> poison
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__0_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__0_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__0_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__0_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 0, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__1_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__1_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__1_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__1_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__2_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__2_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__2_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__2_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__3_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__3_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__3_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__3_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__4_u_u() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__4_u_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__5_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__5_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__5_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__5_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__6_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__6_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__6_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__6_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_u_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_u_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_u_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_u_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_0_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_0_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_0_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_0_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_1_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_1_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_1_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_1_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_2_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_2_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_2_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_2_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_3_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_3_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_3_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_3_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_4_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_4_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_4_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_4_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_5_u() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_5_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_6_u() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_6_u:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_6_u:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_6_u:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_7_u() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_7_u:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 poison>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_7_0() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_7_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_7_1() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_7_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_7_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_7_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_7_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_7_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_7_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_7_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_7_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_7_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_7_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_7_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s14
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s13, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_7_4() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_7_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s12, s4
|
|
; GFX9-NEXT: s_mov_b32 s13, s5
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_7_5() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_7_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s12, s6
|
|
; GFX9-NEXT: s_mov_b32 s13, s7
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_7_6() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_7_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: s_mov_b32 s10, s14
|
|
; GFX9-NEXT: s_mov_b32 s11, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_7_7() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_7_7:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__u_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__u_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 poison, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__0_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__0_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s8
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: s_mov_b32 s12, s8
|
|
; GFX9-NEXT: s_mov_b32 s13, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> zeroinitializer
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__1_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__1_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__2_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__2_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s4
|
|
; GFX9-NEXT: s_mov_b32 s11, s5
|
|
; GFX9-NEXT: s_mov_b32 s12, s4
|
|
; GFX9-NEXT: s_mov_b32 s13, s5
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__3_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__3_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__4_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__4_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__5_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__5_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__5_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__5_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__6_0_0() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__6_0_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_0_0() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_0_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_0_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_0_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_u_0() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_u_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_u_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_u_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_1_0() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_1_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s14
|
|
; GFX900-NEXT: s_mov_b32 s11, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_1_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s14
|
|
; GFX90A-NEXT: s_mov_b32 s11, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_1_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s14
|
|
; GFX942-NEXT: s_mov_b32 s11, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_2_0() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_2_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_2_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_2_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s16
|
|
; GFX942-NEXT: s_mov_b32 s11, s17
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_3_0() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_3_0:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: s_mov_b32 s12, s4
|
|
; GFX9-NEXT: s_mov_b32 s13, s5
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_4_0() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_4_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_4_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_4_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_5_0() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_5_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s4
|
|
; GFX900-NEXT: s_mov_b32 s13, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_5_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s4
|
|
; GFX90A-NEXT: s_mov_b32 s13, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_5_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s12, s0
|
|
; GFX942-NEXT: s_mov_b32 s13, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_6_0() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_6_0:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_6_0:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_6_0:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 0>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__u_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__u_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 poison, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__0_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__0_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 0, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__1_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__1_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__2_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__2_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s6
|
|
; GFX9-NEXT: s_mov_b32 s11, s7
|
|
; GFX9-NEXT: s_mov_b32 s12, s6
|
|
; GFX9-NEXT: s_mov_b32 s13, s7
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__3_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__3_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__4_1_1() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__4_1_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__5_1_1() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__5_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__5_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__5_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s12, s10
|
|
; GFX942-NEXT: s_mov_b32 s13, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__6_1_1() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__6_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s14
|
|
; GFX900-NEXT: s_mov_b32 s11, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__6_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s14
|
|
; GFX90A-NEXT: s_mov_b32 s11, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__6_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: s_mov_b32 s11, s3
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_1_1() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_1_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_1_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_1_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s10
|
|
; GFX942-NEXT: s_mov_b32 s13, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_u_1() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_u_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s6
|
|
; GFX900-NEXT: s_mov_b32 s13, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_u_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s6
|
|
; GFX90A-NEXT: s_mov_b32 s13, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_u_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_0_1() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_0_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: s_mov_b32 s12, s6
|
|
; GFX900-NEXT: s_mov_b32 s13, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_0_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: s_mov_b32 s12, s6
|
|
; GFX90A-NEXT: s_mov_b32 s13, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_0_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_2_1() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_2_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_2_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_2_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_3_1() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_3_1:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: s_mov_b32 s12, s6
|
|
; GFX9-NEXT: s_mov_b32 s13, s7
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_4_1() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_4_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: s_mov_b32 s12, s6
|
|
; GFX900-NEXT: s_mov_b32 s13, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_4_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: s_mov_b32 s12, s6
|
|
; GFX90A-NEXT: s_mov_b32 s13, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_4_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_5_1() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_5_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s6
|
|
; GFX900-NEXT: s_mov_b32 s13, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_5_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s6
|
|
; GFX90A-NEXT: s_mov_b32 s13, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_5_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_6_1() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_6_1:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: s_mov_b32 s12, s6
|
|
; GFX900-NEXT: s_mov_b32 s13, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_6_1:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: s_mov_b32 s12, s6
|
|
; GFX90A-NEXT: s_mov_b32 s13, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_6_1:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 1>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__u_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__u_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 poison, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__0_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__0_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 0, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__1_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__1_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__2_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__2_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s8
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: s_mov_b32 s12, s8
|
|
; GFX9-NEXT: s_mov_b32 s13, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__3_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__3_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__4_2_2() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__4_2_2:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__5_2_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__5_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__5_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__5_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__6_2_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__6_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__6_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__6_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_2_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_2_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_2_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_2_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_u_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_u_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_u_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_u_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_0_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_0_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: s_mov_b32 s12, s16
|
|
; GFX900-NEXT: s_mov_b32 s13, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_0_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: s_mov_b32 s12, s16
|
|
; GFX90A-NEXT: s_mov_b32 s13, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_0_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: s_mov_b32 s12, s4
|
|
; GFX942-NEXT: s_mov_b32 s13, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_1_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_1_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_1_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_1_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_3_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_3_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s14
|
|
; GFX900-NEXT: s_mov_b32 s11, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_3_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s14
|
|
; GFX90A-NEXT: s_mov_b32 s11, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_3_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s14
|
|
; GFX942-NEXT: s_mov_b32 s11, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_4_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_4_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_4_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_4_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_5_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_5_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s16
|
|
; GFX900-NEXT: s_mov_b32 s13, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_5_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s16
|
|
; GFX90A-NEXT: s_mov_b32 s13, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_5_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s12, s4
|
|
; GFX942-NEXT: s_mov_b32 s13, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_6_2() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_6_2:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_6_2:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_6_2:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 2>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__u_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__u_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 poison, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__0_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__0_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s14
|
|
; GFX9-NEXT: s_mov_b32 s11, s15
|
|
; GFX9-NEXT: s_mov_b32 s12, s14
|
|
; GFX9-NEXT: s_mov_b32 s13, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 0, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__1_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__1_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s6
|
|
; GFX9-NEXT: s_mov_b32 s9, s7
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__2_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__2_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__3_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__3_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__4_3_3() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__4_3_3:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__5_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__5_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__5_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__5_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s13, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__6_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__6_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s14
|
|
; GFX900-NEXT: s_mov_b32 s11, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__6_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s14
|
|
; GFX90A-NEXT: s_mov_b32 s11, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__6_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s14
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s14
|
|
; GFX942-NEXT: s_mov_b32 s11, s15
|
|
; GFX942-NEXT: s_mov_b32 s13, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_3_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_3_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_3_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_3_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s12, s10
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s13, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_u_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_u_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_u_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_u_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_0_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_0_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: s_mov_b32 s12, s18
|
|
; GFX900-NEXT: s_mov_b32 s13, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_0_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: s_mov_b32 s12, s18
|
|
; GFX90A-NEXT: s_mov_b32 s13, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_0_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_1_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_1_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_1_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_1_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s14
|
|
; GFX942-NEXT: s_mov_b32 s13, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_2_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_2_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_2_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_2_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_4_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_4_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_4_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_4_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_5_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_5_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s18
|
|
; GFX900-NEXT: s_mov_b32 s13, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_5_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s18
|
|
; GFX90A-NEXT: s_mov_b32 s13, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_5_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_6_3() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_6_3:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_6_3:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_6_3:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 3>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__u_4_4() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__u_4_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 poison, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__0_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__0_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__0_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__0_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 0, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__1_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__1_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__1_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__1_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 1, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__2_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__2_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__2_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__2_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 2, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__3_4_4() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__3_4_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__3_4_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__3_4_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 3, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__4_4_4() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__4_4_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> poison, <3 x i32> <i32 4, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__5_4_4() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__5_4_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__6_4_4() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__6_4_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s4
|
|
; GFX9-NEXT: s_mov_b32 s11, s5
|
|
; GFX9-NEXT: s_mov_b32 s12, s4
|
|
; GFX9-NEXT: s_mov_b32 s13, s5
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_4_4() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_4_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_u_4() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_u_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_0_4() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_0_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_0_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_0_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_1_4() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_1_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_1_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_1_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_2_4() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_2_4:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s20
|
|
; GFX900-NEXT: s_mov_b32 s11, s21
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_2_4:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s20
|
|
; GFX90A-NEXT: s_mov_b32 s11, s21
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_2_4:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_3_4() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_3_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_5_4() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_5_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: s_mov_b32 s10, s14
|
|
; GFX9-NEXT: s_mov_b32 s11, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_6_4() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_6_4:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[12:19]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s18
|
|
; GFX9-NEXT: s_mov_b32 s9, s19
|
|
; GFX9-NEXT: s_mov_b32 s10, s16
|
|
; GFX9-NEXT: s_mov_b32 s11, s17
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 4>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__u_5_5() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__u_5_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 poison, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__0_5_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__0_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s14
|
|
; GFX900-NEXT: s_mov_b32 s11, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__0_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s14
|
|
; GFX90A-NEXT: s_mov_b32 s11, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__0_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: s_mov_b32 s11, s3
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 0, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__1_5_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__1_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__1_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__1_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s12, s10
|
|
; GFX942-NEXT: s_mov_b32 s13, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 1, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__2_5_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__2_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s14
|
|
; GFX900-NEXT: s_mov_b32 s11, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__2_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s14
|
|
; GFX90A-NEXT: s_mov_b32 s11, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__2_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s2
|
|
; GFX942-NEXT: s_mov_b32 s11, s3
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 2, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__3_5_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__3_5_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__3_5_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__3_5_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s10
|
|
; GFX942-NEXT: s_mov_b32 s13, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 3, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__4_5_5() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__4_5_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 4, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__5_5_5() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__5_5_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__6_5_5() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__6_5_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s6
|
|
; GFX9-NEXT: s_mov_b32 s11, s7
|
|
; GFX9-NEXT: s_mov_b32 s12, s6
|
|
; GFX9-NEXT: s_mov_b32 s13, s7
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_5_5() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_5_5:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_u_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_u_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s12, s6
|
|
; GFX900-NEXT: s_mov_b32 s13, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_u_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s12, s6
|
|
; GFX90A-NEXT: s_mov_b32 s13, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_u_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_0_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_0_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_0_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_0_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s10
|
|
; GFX942-NEXT: s_mov_b32 s9, s11
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_1_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_1_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_1_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_1_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_2_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_2_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: s_mov_b32 s12, s6
|
|
; GFX900-NEXT: s_mov_b32 s13, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_2_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: s_mov_b32 s12, s6
|
|
; GFX90A-NEXT: s_mov_b32 s13, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_2_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[12:19]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s18
|
|
; GFX942-NEXT: s_mov_b32 s9, s19
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: s_mov_b32 s12, s14
|
|
; GFX942-NEXT: s_mov_b32 s13, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_3_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_3_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_3_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_3_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_4_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_4_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: s_mov_b32 s12, s6
|
|
; GFX900-NEXT: s_mov_b32 s13, s7
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_4_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: s_mov_b32 s12, s6
|
|
; GFX90A-NEXT: s_mov_b32 s13, s7
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_4_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_6_5() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_6_5:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_6_5:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_6_5:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: s_mov_b32 s12, s2
|
|
; GFX942-NEXT: s_mov_b32 s13, s3
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 5>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__u_6_6() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__u_6_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 poison, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__0_6_6() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__0_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: s_mov_b32 s12, s16
|
|
; GFX900-NEXT: s_mov_b32 s13, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__0_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: s_mov_b32 s12, s16
|
|
; GFX90A-NEXT: s_mov_b32 s13, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__0_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: s_mov_b32 s12, s4
|
|
; GFX942-NEXT: s_mov_b32 s13, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 0, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__1_6_6() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__1_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s6
|
|
; GFX900-NEXT: s_mov_b32 s9, s7
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__1_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s6
|
|
; GFX90A-NEXT: s_mov_b32 s9, s7
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__1_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 1, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__2_6_6() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__2_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: s_mov_b32 s12, s16
|
|
; GFX900-NEXT: s_mov_b32 s13, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__2_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: s_mov_b32 s12, s16
|
|
; GFX90A-NEXT: s_mov_b32 s13, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__2_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: s_mov_b32 s12, s4
|
|
; GFX942-NEXT: s_mov_b32 s13, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 2, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__3_6_6() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__3_6_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__3_6_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__3_6_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s12
|
|
; GFX942-NEXT: s_mov_b32 s11, s13
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 3, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__4_6_6() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__4_6_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 4, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__5_6_6() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__5_6_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s10
|
|
; GFX9-NEXT: s_mov_b32 s9, s11
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__6_6_6() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__6_6_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s8
|
|
; GFX9-NEXT: s_mov_b32 s11, s9
|
|
; GFX9-NEXT: s_mov_b32 s12, s8
|
|
; GFX9-NEXT: s_mov_b32 s13, s9
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_6_6() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_6_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: s_mov_b32 s10, s12
|
|
; GFX9-NEXT: s_mov_b32 s11, s13
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_u_6() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_u_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_0_6() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_0_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_0_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_0_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_1_6() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_1_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s16
|
|
; GFX900-NEXT: s_mov_b32 s13, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_1_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s16
|
|
; GFX90A-NEXT: s_mov_b32 s13, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_1_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s4
|
|
; GFX942-NEXT: s_mov_b32 s13, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_2_6() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_2_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s16
|
|
; GFX900-NEXT: s_mov_b32 s11, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_2_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s16
|
|
; GFX90A-NEXT: s_mov_b32 s11, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_2_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_3_6() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_3_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s16
|
|
; GFX900-NEXT: s_mov_b32 s13, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_3_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s16
|
|
; GFX90A-NEXT: s_mov_b32 s13, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_3_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s4
|
|
; GFX942-NEXT: s_mov_b32 s13, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_4_6() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_4_6:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: s_mov_b32 s12, s16
|
|
; GFX900-NEXT: s_mov_b32 s13, s17
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_4_6:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: s_mov_b32 s12, s16
|
|
; GFX90A-NEXT: s_mov_b32 s13, s17
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_4_6:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: s_mov_b32 s12, s4
|
|
; GFX942-NEXT: s_mov_b32 s13, s5
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_5_6() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_5_6:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 6>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__u_7_7() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__u_7_7:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 poison, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__0_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__0_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s18
|
|
; GFX900-NEXT: s_mov_b32 s11, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s18
|
|
; GFX900-NEXT: s_mov_b32 s13, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__0_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s18
|
|
; GFX90A-NEXT: s_mov_b32 s11, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s18
|
|
; GFX90A-NEXT: s_mov_b32 s13, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__0_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s6
|
|
; GFX942-NEXT: s_mov_b32 s11, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 0, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__1_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__1_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__1_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__1_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s2
|
|
; GFX942-NEXT: s_mov_b32 s9, s3
|
|
; GFX942-NEXT: s_mov_b32 s12, s10
|
|
; GFX942-NEXT: s_mov_b32 s13, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 1, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__2_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__2_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s10, s18
|
|
; GFX900-NEXT: s_mov_b32 s11, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s18
|
|
; GFX900-NEXT: s_mov_b32 s13, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__2_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s10, s18
|
|
; GFX90A-NEXT: s_mov_b32 s11, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s18
|
|
; GFX90A-NEXT: s_mov_b32 s13, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__2_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s10, s6
|
|
; GFX942-NEXT: s_mov_b32 s11, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 2, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__3_7_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__3_7_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__3_7_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__3_7_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s12, s10
|
|
; GFX942-NEXT: s_mov_b32 s13, s11
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 3, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__4_7_7() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__4_7_7:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s10, s14
|
|
; GFX9-NEXT: s_mov_b32 s11, s15
|
|
; GFX9-NEXT: s_mov_b32 s12, s14
|
|
; GFX9-NEXT: s_mov_b32 s13, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 4, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__5_7_7() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__5_7_7:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s6
|
|
; GFX9-NEXT: s_mov_b32 s9, s7
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 5, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__6_7_7() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__6_7_7:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[4:11]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s12, s10
|
|
; GFX9-NEXT: s_mov_b32 s13, s11
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 6, i32 7, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_u_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_u_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s10
|
|
; GFX900-NEXT: s_mov_b32 s9, s11
|
|
; GFX900-NEXT: s_mov_b32 s12, s10
|
|
; GFX900-NEXT: s_mov_b32 s13, s11
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_u_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s10
|
|
; GFX90A-NEXT: s_mov_b32 s9, s11
|
|
; GFX90A-NEXT: s_mov_b32 s12, s10
|
|
; GFX90A-NEXT: s_mov_b32 s13, s11
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_u_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 poison, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_0_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_0_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s4
|
|
; GFX900-NEXT: s_mov_b32 s11, s5
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_0_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s4
|
|
; GFX90A-NEXT: s_mov_b32 s11, s5
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_0_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: s_mov_b32 s12, s14
|
|
; GFX942-NEXT: s_mov_b32 s13, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 0, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_1_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_1_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s18
|
|
; GFX900-NEXT: s_mov_b32 s13, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_1_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s18
|
|
; GFX90A-NEXT: s_mov_b32 s13, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_1_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 1, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_2_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_2_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[16:23]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s22
|
|
; GFX900-NEXT: s_mov_b32 s9, s23
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: s_mov_b32 s12, s22
|
|
; GFX900-NEXT: s_mov_b32 s13, s23
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_2_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[16:23]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s22
|
|
; GFX90A-NEXT: s_mov_b32 s9, s23
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: s_mov_b32 s12, s22
|
|
; GFX90A-NEXT: s_mov_b32 s13, s23
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_2_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[8:15]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s14
|
|
; GFX942-NEXT: s_mov_b32 s9, s15
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: s_mov_b32 s12, s14
|
|
; GFX942-NEXT: s_mov_b32 s13, s15
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 2, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_3_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_3_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[4:11]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s12, s18
|
|
; GFX900-NEXT: s_mov_b32 s13, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_3_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[4:11]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s12, s18
|
|
; GFX90A-NEXT: s_mov_b32 s13, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_3_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[4:11]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_nop 0
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 3, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_4_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_4_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[12:19]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s18
|
|
; GFX900-NEXT: s_mov_b32 s9, s19
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: s_mov_b32 s12, s18
|
|
; GFX900-NEXT: s_mov_b32 s13, s19
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_4_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[12:19]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s18
|
|
; GFX90A-NEXT: s_mov_b32 s9, s19
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: s_mov_b32 s12, s18
|
|
; GFX90A-NEXT: s_mov_b32 s13, s19
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_4_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s0
|
|
; GFX942-NEXT: s_mov_b32 s11, s1
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 4, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_5_7() {
|
|
; GFX9-LABEL: s_shuffle_v3i64_v4i64__7_5_7:
|
|
; GFX9: ; %bb.0:
|
|
; GFX9-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; def s[8:15]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_mov_b32 s8, s14
|
|
; GFX9-NEXT: s_mov_b32 s9, s15
|
|
; GFX9-NEXT: s_mov_b32 s12, s14
|
|
; GFX9-NEXT: s_mov_b32 s13, s15
|
|
; GFX9-NEXT: ;;#ASMSTART
|
|
; GFX9-NEXT: ; use s[8:13]
|
|
; GFX9-NEXT: ;;#ASMEND
|
|
; GFX9-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 5, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
|
|
define void @s_shuffle_v3i64_v4i64__7_6_7() {
|
|
; GFX900-LABEL: s_shuffle_v3i64_v4i64__7_6_7:
|
|
; GFX900: ; %bb.0:
|
|
; GFX900-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; def s[8:15]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_mov_b32 s8, s14
|
|
; GFX900-NEXT: s_mov_b32 s9, s15
|
|
; GFX900-NEXT: s_mov_b32 s10, s12
|
|
; GFX900-NEXT: s_mov_b32 s11, s13
|
|
; GFX900-NEXT: s_mov_b32 s12, s14
|
|
; GFX900-NEXT: s_mov_b32 s13, s15
|
|
; GFX900-NEXT: ;;#ASMSTART
|
|
; GFX900-NEXT: ; use s[8:13]
|
|
; GFX900-NEXT: ;;#ASMEND
|
|
; GFX900-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX90A-LABEL: s_shuffle_v3i64_v4i64__7_6_7:
|
|
; GFX90A: ; %bb.0:
|
|
; GFX90A-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; def s[8:15]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_mov_b32 s8, s14
|
|
; GFX90A-NEXT: s_mov_b32 s9, s15
|
|
; GFX90A-NEXT: s_mov_b32 s10, s12
|
|
; GFX90A-NEXT: s_mov_b32 s11, s13
|
|
; GFX90A-NEXT: s_mov_b32 s12, s14
|
|
; GFX90A-NEXT: s_mov_b32 s13, s15
|
|
; GFX90A-NEXT: ;;#ASMSTART
|
|
; GFX90A-NEXT: ; use s[8:13]
|
|
; GFX90A-NEXT: ;;#ASMEND
|
|
; GFX90A-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; GFX942-LABEL: s_shuffle_v3i64_v4i64__7_6_7:
|
|
; GFX942: ; %bb.0:
|
|
; GFX942-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; def s[0:7]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_mov_b32 s8, s6
|
|
; GFX942-NEXT: s_mov_b32 s9, s7
|
|
; GFX942-NEXT: s_mov_b32 s10, s4
|
|
; GFX942-NEXT: s_mov_b32 s11, s5
|
|
; GFX942-NEXT: s_mov_b32 s12, s6
|
|
; GFX942-NEXT: s_mov_b32 s13, s7
|
|
; GFX942-NEXT: ;;#ASMSTART
|
|
; GFX942-NEXT: ; use s[8:13]
|
|
; GFX942-NEXT: ;;#ASMEND
|
|
; GFX942-NEXT: s_setpc_b64 s[30:31]
|
|
%vec0 = call <4 x i64> asm "; def $0", "=s"()
|
|
%vec1 = call <4 x i64> asm "; def $0", "=s"()
|
|
%shuf = shufflevector <4 x i64> %vec0, <4 x i64> %vec1, <3 x i32> <i32 7, i32 6, i32 7>
|
|
call void asm sideeffect "; use $0", "{s[8:13]}"(<3 x i64> %shuf)
|
|
ret void
|
|
}
|
|
;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
|
|
; GFX90APLUS: {{.*}}
|