96 lines
4.0 KiB
LLVM
96 lines
4.0 KiB
LLVM
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
|
|
; RUN: llc -global-isel=0 -mtriple=amdgcn--amdhsa -mcpu=fiji < %s | FileCheck -check-prefixes=CHECK,CHECK-SDAG %s
|
|
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn--amdhsa -mcpu=fiji < %s | FileCheck -check-prefixes=CHECK,CHECK-GISEL %s
|
|
|
|
declare i32 @llvm.amdgcn.ds.permute(i32, i32) #0
|
|
|
|
define amdgpu_kernel void @ds_permute(ptr addrspace(1) %out, i32 %index, i32 %src) nounwind {
|
|
; CHECK-LABEL: ds_permute:
|
|
; CHECK: ; %bb.0:
|
|
; CHECK-NEXT: s_load_dwordx4 s[0:3], s[8:9], 0x0
|
|
; CHECK-NEXT: s_add_i32 s12, s12, s17
|
|
; CHECK-NEXT: s_mov_b32 flat_scratch_lo, s13
|
|
; CHECK-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
|
|
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s2
|
|
; CHECK-NEXT: v_mov_b32_e32 v1, s3
|
|
; CHECK-NEXT: ds_permute_b32 v2, v0, v1
|
|
; CHECK-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-NEXT: v_mov_b32_e32 v1, s1
|
|
; CHECK-NEXT: s_waitcnt lgkmcnt(0)
|
|
; CHECK-NEXT: flat_store_dword v[0:1], v2
|
|
; CHECK-NEXT: s_endpgm
|
|
%bpermute = call i32 @llvm.amdgcn.ds.permute(i32 %index, i32 %src) #0
|
|
store i32 %bpermute, ptr addrspace(1) %out, align 4
|
|
ret void
|
|
}
|
|
|
|
define amdgpu_kernel void @ds_permute_imm_offset(ptr addrspace(1) %out, i32 %base_index, i32 %src) nounwind {
|
|
; CHECK-SDAG-LABEL: ds_permute_imm_offset:
|
|
; CHECK-SDAG: ; %bb.0:
|
|
; CHECK-SDAG-NEXT: s_load_dwordx4 s[0:3], s[8:9], 0x0
|
|
; CHECK-SDAG-NEXT: s_add_i32 s12, s12, s17
|
|
; CHECK-SDAG-NEXT: s_mov_b32 flat_scratch_lo, s13
|
|
; CHECK-SDAG-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
|
|
; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s2
|
|
; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s3
|
|
; CHECK-SDAG-NEXT: ds_permute_b32 v2, v0, v1 offset:4
|
|
; CHECK-SDAG-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-SDAG-NEXT: v_mov_b32_e32 v1, s1
|
|
; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; CHECK-SDAG-NEXT: flat_store_dword v[0:1], v2
|
|
; CHECK-SDAG-NEXT: s_endpgm
|
|
;
|
|
; CHECK-GISEL-LABEL: ds_permute_imm_offset:
|
|
; CHECK-GISEL: ; %bb.0:
|
|
; CHECK-GISEL-NEXT: s_load_dwordx4 s[0:3], s[8:9], 0x0
|
|
; CHECK-GISEL-NEXT: s_add_i32 s12, s12, s17
|
|
; CHECK-GISEL-NEXT: s_mov_b32 flat_scratch_lo, s13
|
|
; CHECK-GISEL-NEXT: s_lshr_b32 flat_scratch_hi, s12, 8
|
|
; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; CHECK-GISEL-NEXT: s_add_i32 s2, s2, 4
|
|
; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s2
|
|
; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s3
|
|
; CHECK-GISEL-NEXT: ds_permute_b32 v2, v0, v1
|
|
; CHECK-GISEL-NEXT: v_mov_b32_e32 v0, s0
|
|
; CHECK-GISEL-NEXT: v_mov_b32_e32 v1, s1
|
|
; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; CHECK-GISEL-NEXT: flat_store_dword v[0:1], v2
|
|
; CHECK-GISEL-NEXT: s_endpgm
|
|
%index = add i32 %base_index, 4
|
|
%bpermute = call i32 @llvm.amdgcn.ds.permute(i32 %index, i32 %src) #0
|
|
store i32 %bpermute, ptr addrspace(1) %out, align 4
|
|
ret void
|
|
}
|
|
|
|
define void @ds_permute_vv(ptr addrspace(1) %out, i32 %base_index, i32 %src) nounwind {
|
|
; CHECK-SDAG-LABEL: ds_permute_vv:
|
|
; CHECK-SDAG: ; %bb.0:
|
|
; CHECK-SDAG-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; CHECK-SDAG-NEXT: v_lshlrev_b32_e32 v2, 2, v2
|
|
; CHECK-SDAG-NEXT: ds_permute_b32 v2, v2, v3 offset:4
|
|
; CHECK-SDAG-NEXT: s_waitcnt lgkmcnt(0)
|
|
; CHECK-SDAG-NEXT: flat_store_dword v[0:1], v2
|
|
; CHECK-SDAG-NEXT: s_waitcnt vmcnt(0)
|
|
; CHECK-SDAG-NEXT: s_setpc_b64 s[30:31]
|
|
;
|
|
; CHECK-GISEL-LABEL: ds_permute_vv:
|
|
; CHECK-GISEL: ; %bb.0:
|
|
; CHECK-GISEL-NEXT: s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
|
|
; CHECK-GISEL-NEXT: v_lshlrev_b32_e32 v2, 2, v2
|
|
; CHECK-GISEL-NEXT: v_add_u32_e32 v2, vcc, 4, v2
|
|
; CHECK-GISEL-NEXT: ds_permute_b32 v2, v2, v3
|
|
; CHECK-GISEL-NEXT: s_waitcnt lgkmcnt(0)
|
|
; CHECK-GISEL-NEXT: flat_store_dword v[0:1], v2
|
|
; CHECK-GISEL-NEXT: s_waitcnt vmcnt(0)
|
|
; CHECK-GISEL-NEXT: s_setpc_b64 s[30:31]
|
|
%index = add i32 %base_index, 1
|
|
%byte_index = shl i32 %index, 2
|
|
%permute = call i32 @llvm.amdgcn.ds.permute(i32 %byte_index, i32 %src) #0
|
|
store i32 %permute, ptr addrspace(1) %out, align 4
|
|
ret void
|
|
}
|
|
|
|
attributes #0 = { nounwind readnone convergent }
|