[llvm] [AMDGPU] Update VOP instructions for GFX12 (PR #74853)
Mariusz Sikora via llvm-commits
llvm-commits at lists.llvm.org
Mon Dec 11 07:12:01 PST 2023
================
@@ -35,16 +39,23 @@ body: |
; SCHED-NEXT: $vgpr3 = V_SUB_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
; SCHED-NEXT: $vgpr6 = V_MUL_F32_e32 killed $vgpr0, $vgpr0, implicit $mode, implicit $exec
; SCHED-NEXT: $vgpr4 = V_SUB_F32_e32 killed $vgpr1, $vgpr1, implicit $mode, implicit $exec
- ; PAIR-LABEL: name: vopd_schedule
- ; PAIR: $vgpr0 = IMPLICIT_DEF
- ; PAIR-NEXT: $vgpr1 = IMPLICIT_DEF
- ; PAIR-NEXT: $vgpr2 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
- ; PAIR-NEXT: $vgpr3, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32 $vgpr1, $vgpr1, killed $vgpr0, $vgpr0, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $mode, implicit $exec
- ; PAIR-NEXT: $vgpr4 = V_SUB_F32_e32 killed $vgpr1, $vgpr1, implicit $mode, implicit $exec
+ ;
+ ; PAIR-GFX11-LABEL: name: vopd_schedule
+ ; PAIR-GFX11: $vgpr0 = IMPLICIT_DEF
+ ; PAIR-GFX11-NEXT: $vgpr1 = IMPLICIT_DEF
+ ; PAIR-GFX11-NEXT: $vgpr2 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
+ ; PAIR-GFX11-NEXT: $vgpr3, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx11 $vgpr1, $vgpr1, killed $vgpr0, $vgpr0, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $mode, implicit $exec
+ ; PAIR-GFX11-NEXT: $vgpr4 = V_SUB_F32_e32 killed $vgpr1, $vgpr1, implicit $mode, implicit $exec
+ ;
+ ; PAIR-GFX12-LABEL: name: vopd_schedule
+ ; PAIR-GFX12: $vgpr0 = IMPLICIT_DEF
+ ; PAIR-GFX12-NEXT: $vgpr1 = IMPLICIT_DEF
+ ; PAIR-GFX12-NEXT: $vgpr2 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
+ ; PAIR-GFX12-NEXT: $vgpr3, $vgpr6 = V_DUAL_SUB_F32_e32_X_MUL_F32_e32_gfx12 $vgpr1, $vgpr1, killed $vgpr0, $vgpr0, implicit $mode, implicit $exec, implicit $mode, implicit $exec, implicit $mode, implicit $exec
+ ; PAIR-GFX12-NEXT: $vgpr4 = V_SUB_F32_e32 killed $vgpr1, $vgpr1, implicit $mode, implicit $exec
$vgpr0 = IMPLICIT_DEF
$vgpr1 = IMPLICIT_DEF
$vgpr2 = V_ADD_F32_e32 $vgpr1, $vgpr1, implicit $mode, implicit $exec
- ; can fuse vgpr3 and vgpr6 writing insts only due to reg constraints
----------------
mariusz-sikora-at-amd wrote:
Updated, thanks !
https://github.com/llvm/llvm-project/pull/74853
More information about the llvm-commits
mailing list