[llvm] AMDGPU: Fold frame indexes into s_or_b32 and s_and_b32 (PR #102345)
Matt Arsenault via llvm-commits
llvm-commits at lists.llvm.org
Fri Jun 20 04:49:57 PDT 2025
https://github.com/arsenm updated https://github.com/llvm/llvm-project/pull/102345
>From f811338a6a26c45e2c761d0c72f62c9cd17d8631 Mon Sep 17 00:00:00 2001
From: Matt Arsenault <Matthew.Arsenault at amd.com>
Date: Fri, 2 Aug 2024 13:17:01 +0400
Subject: [PATCH] AMDGPU: Fold frame indexes into s_or_b32
Some pointer adds get turned into ors, and sometimes and is
performed on pointers for masking.
---
llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp | 3 +-
.../eliminate-frame-index-scalar-bit-ops.mir | 64 +++++++++----------
2 files changed, 32 insertions(+), 35 deletions(-)
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
index e41189adfb46f..03630f35e96d8 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -2803,7 +2803,8 @@ bool SIRegisterInfo::eliminateFrameIndex(MachineBasicBlock::iterator MI,
}
case AMDGPU::S_ADD_I32:
case AMDGPU::S_ADD_U32: {
- // TODO: Handle s_or_b32, s_and_b32.
+ case AMDGPU::S_OR_B32: {
+ // TODO: Handle s_and_b32 for ptrmask
unsigned OtherOpIdx = FIOperandNum == 1 ? 2 : 1;
MachineOperand &OtherOp = MI->getOperand(OtherOpIdx);
diff --git a/llvm/test/CodeGen/AMDGPU/eliminate-frame-index-scalar-bit-ops.mir b/llvm/test/CodeGen/AMDGPU/eliminate-frame-index-scalar-bit-ops.mir
index aecff1b13171d..7804d067dd790 100644
--- a/llvm/test/CodeGen/AMDGPU/eliminate-frame-index-scalar-bit-ops.mir
+++ b/llvm/test/CodeGen/AMDGPU/eliminate-frame-index-scalar-bit-ops.mir
@@ -21,13 +21,13 @@ machineFunctionInfo:
body: |
bb.0:
; MUBUFW64-LABEL: name: s_or_b32__inline_imm__fi_offset0
- ; MUBUFW64: $sgpr4 = S_LSHR_B32 $sgpr32, 6, implicit-def dead $scc
- ; MUBUFW64-NEXT: renamable $sgpr7 = S_OR_B32 12, killed $sgpr4, implicit-def $scc
+ ; MUBUFW64: renamable $sgpr4 = S_LSHR_B32 $sgpr32, 6, implicit-def dead $scc
+ ; MUBUFW64-NEXT: renamable $sgpr7 = S_OR_B32 12, $sgpr4, implicit-def $scc
; MUBUFW64-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; MUBUFW32-LABEL: name: s_or_b32__inline_imm__fi_offset0
- ; MUBUFW32: $sgpr4 = S_LSHR_B32 $sgpr32, 5, implicit-def dead $scc
- ; MUBUFW32-NEXT: renamable $sgpr7 = S_OR_B32 12, killed $sgpr4, implicit-def $scc
+ ; MUBUFW32: renamable $sgpr4 = S_LSHR_B32 $sgpr32, 5, implicit-def dead $scc
+ ; MUBUFW32-NEXT: renamable $sgpr7 = S_OR_B32 12, $sgpr4, implicit-def $scc
; MUBUFW32-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; FLATSCRW64-LABEL: name: s_or_b32__inline_imm__fi_offset0
@@ -55,25 +55,21 @@ machineFunctionInfo:
body: |
bb.0:
; MUBUFW64-LABEL: name: s_or_b32__literal__fi_offset96
- ; MUBUFW64: $sgpr4 = S_LSHR_B32 $sgpr32, 6, implicit-def $scc
- ; MUBUFW64-NEXT: $sgpr4 = S_ADD_I32 killed $sgpr4, 96, implicit-def $scc
- ; MUBUFW64-NEXT: renamable $sgpr7 = S_OR_B32 68, killed $sgpr4, implicit-def $scc
+ ; MUBUFW64: renamable $sgpr4 = S_LSHR_B32 $sgpr32, 6, implicit-def dead $scc
+ ; MUBUFW64-NEXT: renamable $sgpr7 = S_OR_B32 164, $sgpr4, implicit-def $scc
; MUBUFW64-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; MUBUFW32-LABEL: name: s_or_b32__literal__fi_offset96
- ; MUBUFW32: $sgpr4 = S_LSHR_B32 $sgpr32, 5, implicit-def $scc
- ; MUBUFW32-NEXT: $sgpr4 = S_ADD_I32 killed $sgpr4, 96, implicit-def $scc
- ; MUBUFW32-NEXT: renamable $sgpr7 = S_OR_B32 68, killed $sgpr4, implicit-def $scc
+ ; MUBUFW32: renamable $sgpr4 = S_LSHR_B32 $sgpr32, 5, implicit-def dead $scc
+ ; MUBUFW32-NEXT: renamable $sgpr7 = S_OR_B32 164, $sgpr4, implicit-def $scc
; MUBUFW32-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; FLATSCRW64-LABEL: name: s_or_b32__literal__fi_offset96
- ; FLATSCRW64: $sgpr4 = S_ADD_I32 $sgpr32, 96, implicit-def $scc
- ; FLATSCRW64-NEXT: renamable $sgpr7 = S_OR_B32 68, killed $sgpr4, implicit-def $scc
+ ; FLATSCRW64: renamable $sgpr7 = S_OR_B32 164, $sgpr32, implicit-def $scc
; FLATSCRW64-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; FLATSCRW32-LABEL: name: s_or_b32__literal__fi_offset96
- ; FLATSCRW32: $sgpr4 = S_ADD_I32 $sgpr32, 96, implicit-def $scc
- ; FLATSCRW32-NEXT: renamable $sgpr7 = S_OR_B32 68, killed $sgpr4, implicit-def $scc
+ ; FLATSCRW32: renamable $sgpr7 = S_OR_B32 164, $sgpr32, implicit-def $scc
; FLATSCRW32-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
renamable $sgpr7 = S_OR_B32 68, %stack.1, implicit-def $scc
SI_RETURN implicit $sgpr7, implicit $scc
@@ -96,31 +92,31 @@ body: |
; MUBUFW64-LABEL: name: s_or_b32__sgpr__fi_literal_offset
; MUBUFW64: liveins: $sgpr8
; MUBUFW64-NEXT: {{ $}}
- ; MUBUFW64-NEXT: $sgpr4 = S_LSHR_B32 $sgpr32, 6, implicit-def $scc
- ; MUBUFW64-NEXT: $sgpr4 = S_ADD_I32 killed $sgpr4, 80, implicit-def $scc
- ; MUBUFW64-NEXT: renamable $sgpr7 = S_OR_B32 $sgpr8, killed $sgpr4, implicit-def $scc
+ ; MUBUFW64-NEXT: renamable $sgpr4 = S_LSHR_B32 $sgpr32, 6, implicit-def dead $scc
+ ; MUBUFW64-NEXT: renamable $sgpr7 = S_ADD_I32 killed $sgpr4, $sgpr8, implicit-def $scc
+ ; MUBUFW64-NEXT: renamable $sgpr7 = S_OR_B32 killed renamable $sgpr7, 80, implicit-def $scc
; MUBUFW64-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; MUBUFW32-LABEL: name: s_or_b32__sgpr__fi_literal_offset
; MUBUFW32: liveins: $sgpr8
; MUBUFW32-NEXT: {{ $}}
- ; MUBUFW32-NEXT: $sgpr4 = S_LSHR_B32 $sgpr32, 5, implicit-def $scc
- ; MUBUFW32-NEXT: $sgpr4 = S_ADD_I32 killed $sgpr4, 80, implicit-def $scc
- ; MUBUFW32-NEXT: renamable $sgpr7 = S_OR_B32 $sgpr8, killed $sgpr4, implicit-def $scc
+ ; MUBUFW32-NEXT: renamable $sgpr4 = S_LSHR_B32 $sgpr32, 5, implicit-def dead $scc
+ ; MUBUFW32-NEXT: renamable $sgpr7 = S_ADD_I32 killed $sgpr4, $sgpr8, implicit-def $scc
+ ; MUBUFW32-NEXT: renamable $sgpr7 = S_OR_B32 killed renamable $sgpr7, 80, implicit-def $scc
; MUBUFW32-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; FLATSCRW64-LABEL: name: s_or_b32__sgpr__fi_literal_offset
; FLATSCRW64: liveins: $sgpr8
; FLATSCRW64-NEXT: {{ $}}
- ; FLATSCRW64-NEXT: $sgpr4 = S_ADD_I32 $sgpr32, 80, implicit-def $scc
- ; FLATSCRW64-NEXT: renamable $sgpr7 = S_OR_B32 $sgpr8, killed $sgpr4, implicit-def $scc
+ ; FLATSCRW64-NEXT: renamable $sgpr4 = S_ADD_I32 killed $sgpr32, $sgpr8, implicit-def $scc
+ ; FLATSCRW64-NEXT: renamable $sgpr7 = S_OR_B32 killed renamable $sgpr4, 80, implicit-def $scc
; FLATSCRW64-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; FLATSCRW32-LABEL: name: s_or_b32__sgpr__fi_literal_offset
; FLATSCRW32: liveins: $sgpr8
; FLATSCRW32-NEXT: {{ $}}
- ; FLATSCRW32-NEXT: $sgpr4 = S_ADD_I32 $sgpr32, 80, implicit-def $scc
- ; FLATSCRW32-NEXT: renamable $sgpr7 = S_OR_B32 $sgpr8, killed $sgpr4, implicit-def $scc
+ ; FLATSCRW32-NEXT: renamable $sgpr4 = S_ADD_I32 killed $sgpr32, $sgpr8, implicit-def $scc
+ ; FLATSCRW32-NEXT: renamable $sgpr7 = S_OR_B32 killed renamable $sgpr4, 80, implicit-def $scc
; FLATSCRW32-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
renamable $sgpr7 = S_OR_B32 $sgpr8, %stack.1, implicit-def $scc
SI_RETURN implicit $sgpr7, implicit $scc
@@ -143,31 +139,31 @@ body: |
; MUBUFW64-LABEL: name: s_or_b32__sgpr__fi_inlineimm_offset
; MUBUFW64: liveins: $sgpr8
; MUBUFW64-NEXT: {{ $}}
- ; MUBUFW64-NEXT: $sgpr4 = S_LSHR_B32 $sgpr32, 6, implicit-def $scc
- ; MUBUFW64-NEXT: $sgpr4 = S_ADD_I32 killed $sgpr4, 32, implicit-def $scc
- ; MUBUFW64-NEXT: renamable $sgpr7 = S_OR_B32 $sgpr8, killed $sgpr4, implicit-def $scc
+ ; MUBUFW64-NEXT: renamable $sgpr4 = S_LSHR_B32 $sgpr32, 6, implicit-def dead $scc
+ ; MUBUFW64-NEXT: renamable $sgpr7 = S_ADD_I32 killed $sgpr4, $sgpr8, implicit-def $scc
+ ; MUBUFW64-NEXT: renamable $sgpr7 = S_OR_B32 killed renamable $sgpr7, 32, implicit-def $scc
; MUBUFW64-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; MUBUFW32-LABEL: name: s_or_b32__sgpr__fi_inlineimm_offset
; MUBUFW32: liveins: $sgpr8
; MUBUFW32-NEXT: {{ $}}
- ; MUBUFW32-NEXT: $sgpr4 = S_LSHR_B32 $sgpr32, 5, implicit-def $scc
- ; MUBUFW32-NEXT: $sgpr4 = S_ADD_I32 killed $sgpr4, 32, implicit-def $scc
- ; MUBUFW32-NEXT: renamable $sgpr7 = S_OR_B32 $sgpr8, killed $sgpr4, implicit-def $scc
+ ; MUBUFW32-NEXT: renamable $sgpr4 = S_LSHR_B32 $sgpr32, 5, implicit-def dead $scc
+ ; MUBUFW32-NEXT: renamable $sgpr7 = S_ADD_I32 killed $sgpr4, $sgpr8, implicit-def $scc
+ ; MUBUFW32-NEXT: renamable $sgpr7 = S_OR_B32 killed renamable $sgpr7, 32, implicit-def $scc
; MUBUFW32-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; FLATSCRW64-LABEL: name: s_or_b32__sgpr__fi_inlineimm_offset
; FLATSCRW64: liveins: $sgpr8
; FLATSCRW64-NEXT: {{ $}}
- ; FLATSCRW64-NEXT: $sgpr4 = S_ADD_I32 $sgpr32, 32, implicit-def $scc
- ; FLATSCRW64-NEXT: renamable $sgpr7 = S_OR_B32 $sgpr8, killed $sgpr4, implicit-def $scc
+ ; FLATSCRW64-NEXT: renamable $sgpr4 = S_ADD_I32 killed $sgpr32, $sgpr8, implicit-def $scc
+ ; FLATSCRW64-NEXT: renamable $sgpr7 = S_OR_B32 killed renamable $sgpr4, 32, implicit-def $scc
; FLATSCRW64-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
;
; FLATSCRW32-LABEL: name: s_or_b32__sgpr__fi_inlineimm_offset
; FLATSCRW32: liveins: $sgpr8
; FLATSCRW32-NEXT: {{ $}}
- ; FLATSCRW32-NEXT: $sgpr4 = S_ADD_I32 $sgpr32, 32, implicit-def $scc
- ; FLATSCRW32-NEXT: renamable $sgpr7 = S_OR_B32 $sgpr8, killed $sgpr4, implicit-def $scc
+ ; FLATSCRW32-NEXT: renamable $sgpr4 = S_ADD_I32 killed $sgpr32, $sgpr8, implicit-def $scc
+ ; FLATSCRW32-NEXT: renamable $sgpr7 = S_OR_B32 killed renamable $sgpr4, 32, implicit-def $scc
; FLATSCRW32-NEXT: SI_RETURN implicit $sgpr7, implicit $scc
renamable $sgpr7 = S_OR_B32 $sgpr8, %stack.1, implicit-def $scc
SI_RETURN implicit $sgpr7, implicit $scc
More information about the llvm-commits
mailing list