[llvm] [CodeGen] Simplify updateLiveIn in MachineSink (PR #79831)
Jay Foad via llvm-commits
llvm-commits at lists.llvm.org
Mon Jan 29 06:21:45 PST 2024
https://github.com/jayfoad created https://github.com/llvm/llvm-project/pull/79831
When a whole register is added a basic block's liveins, use
LaneBitmask::getAll for the live lanes instead of trying to calculate an
accurate mask of the lanes that comprise the register.
This simplifies the code and matches other places where a whole register
is marked as livein.
This also avoids problems when regunits that are synthesized by TableGen
to represent ad hoc aliasing have a lane mask of 0.
Fixes #78942
>From f875c011bb236fd6678181a58a88ea796a299cc6 Mon Sep 17 00:00:00 2001
From: Jay Foad <jay.foad at amd.com>
Date: Mon, 29 Jan 2024 11:04:22 +0000
Subject: [PATCH] [CodeGen] Simplify updateLiveIn in MachineSink
When a whole register is added a basic block's liveins, use
LaneBitmask::getAll for the live lanes instead of trying to calculate an
accurate mask of the lanes that comprise the register.
This simplifies the code and matches other places where a whole register
is marked as livein.
This also avoids problems when regunits that are synthesized by TableGen
to represent ad hoc aliasing have a lane mask of 0.
Fixes #78942
---
llvm/lib/CodeGen/MachineSink.cpp | 9 ++-------
llvm/test/CodeGen/AMDGPU/av_spill_cross_bb_usage.mir | 2 +-
.../CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll | 2 +-
llvm/test/CodeGen/AMDGPU/postra-machine-sink.mir | 2 +-
4 files changed, 5 insertions(+), 10 deletions(-)
diff --git a/llvm/lib/CodeGen/MachineSink.cpp b/llvm/lib/CodeGen/MachineSink.cpp
index e7e8f602683480..c3a1d3759882d8 100644
--- a/llvm/lib/CodeGen/MachineSink.cpp
+++ b/llvm/lib/CodeGen/MachineSink.cpp
@@ -1949,13 +1949,8 @@ static void updateLiveIn(MachineInstr *MI, MachineBasicBlock *SuccBB,
for (unsigned DefReg : DefedRegsInCopy)
for (MCPhysReg S : TRI->subregs_inclusive(DefReg))
SuccBB->removeLiveIn(S);
- for (auto U : UsedOpsInCopy) {
- Register SrcReg = MI->getOperand(U).getReg();
- LaneBitmask Mask;
- for (MCRegUnitMaskIterator S(SrcReg, TRI); S.isValid(); ++S)
- Mask |= (*S).second;
- SuccBB->addLiveIn(SrcReg, Mask);
- }
+ for (auto U : UsedOpsInCopy)
+ SuccBB->addLiveIn(MI->getOperand(U).getReg());
SuccBB->sortUniqueLiveIns();
}
diff --git a/llvm/test/CodeGen/AMDGPU/av_spill_cross_bb_usage.mir b/llvm/test/CodeGen/AMDGPU/av_spill_cross_bb_usage.mir
index 65648bacd55679..c1da29ecc2c2f5 100644
--- a/llvm/test/CodeGen/AMDGPU/av_spill_cross_bb_usage.mir
+++ b/llvm/test/CodeGen/AMDGPU/av_spill_cross_bb_usage.mir
@@ -47,7 +47,7 @@ body: |
; GCN-NEXT: {{ $}}
; GCN-NEXT: bb.1:
; GCN-NEXT: successors: %bb.2(0x80000000)
- ; GCN-NEXT: liveins: $exec:0x000000000000000F, $sgpr30, $sgpr31, $vgpr0:0x0000000000000003, $vgpr1:0x0000000000000003, $vgpr2:0x0000000000000003, $vgpr3:0x0000000000000003, $vgpr4:0x0000000000000003, $vgpr5:0x0000000000000003, $vgpr6:0x0000000000000003, $vgpr7:0x0000000000000003, $vgpr8:0x0000000000000003, $vgpr9:0x0000000000000003, $vgpr40, $sgpr30_sgpr31, $vgpr10_vgpr11:0x000000000000000F, $vgpr14_vgpr15:0x000000000000000F, $vgpr41_vgpr42:0x000000000000000F, $vgpr43_vgpr44:0x000000000000000F, $vgpr45_vgpr46:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F
+ ; GCN-NEXT: liveins: $exec, $sgpr30, $sgpr31, $vgpr0, $vgpr1, $vgpr2, $vgpr3, $vgpr4, $vgpr5, $vgpr6, $vgpr7, $vgpr8, $vgpr9, $vgpr40, $sgpr30_sgpr31, $vgpr10_vgpr11:0x000000000000000F, $vgpr14_vgpr15:0x000000000000000F, $vgpr41_vgpr42:0x000000000000000F, $vgpr43_vgpr44:0x000000000000000F, $vgpr45_vgpr46:0x000000000000000F, $vgpr56_vgpr57:0x000000000000000F, $vgpr58_vgpr59:0x000000000000000F, $vgpr60_vgpr61:0x000000000000000F
; GCN-NEXT: {{ $}}
; GCN-NEXT: renamable $vgpr57 = COPY $vgpr9, implicit $exec
; GCN-NEXT: renamable $vgpr56 = COPY $vgpr8, implicit $exec
diff --git a/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll b/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll
index 5a128c7541d1ec..24ef8ce1beb2db 100644
--- a/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll
+++ b/llvm/test/CodeGen/AMDGPU/branch-folding-implicit-def-subreg.ll
@@ -808,7 +808,7 @@ define amdgpu_kernel void @f1(ptr addrspace(1) %arg, ptr addrspace(1) %arg1, i64
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: bb.58:
; GFX90A-NEXT: successors: %bb.7(0x80000000)
- ; GFX90A-NEXT: liveins: $exec:0x000000000000000F, $sgpr12, $sgpr13, $sgpr14, $sgpr15:0x0000000000000003, $sgpr23:0x0000000000000003, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7:0x000000000000000F, $sgpr8_sgpr9, $sgpr10_sgpr11, $sgpr24_sgpr25, $sgpr26_sgpr27, $sgpr28_sgpr29, $sgpr42_sgpr43, $sgpr16_sgpr17_sgpr18_sgpr19:0x00000000000000F0, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $vgpr2_vgpr3:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr18_vgpr19:0x000000000000000F, $vgpr20_vgpr21:0x000000000000000F, $vgpr22_vgpr23:0x000000000000000F, $vgpr24_vgpr25:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
+ ; GFX90A-NEXT: liveins: $exec, $sgpr12, $sgpr13, $sgpr14, $sgpr15, $sgpr23, $vgpr30, $vgpr31, $sgpr4_sgpr5, $sgpr6_sgpr7:0x000000000000000F, $sgpr8_sgpr9, $sgpr10_sgpr11, $sgpr24_sgpr25, $sgpr26_sgpr27, $sgpr28_sgpr29, $sgpr42_sgpr43, $sgpr16_sgpr17_sgpr18_sgpr19:0x00000000000000F0, $sgpr20_sgpr21_sgpr22_sgpr23:0x000000000000003C, $vgpr2_vgpr3:0x000000000000000F, $vgpr10_vgpr11:0x000000000000000F, $vgpr18_vgpr19:0x000000000000000F, $vgpr20_vgpr21:0x000000000000000F, $vgpr22_vgpr23:0x000000000000000F, $vgpr24_vgpr25:0x000000000000000F, $sgpr0_sgpr1_sgpr2_sgpr3
; GFX90A-NEXT: {{ $}}
; GFX90A-NEXT: renamable $vgpr15 = COPY killed renamable $sgpr23, implicit $exec
; GFX90A-NEXT: renamable $vgpr17 = COPY killed renamable $sgpr15, implicit $exec
diff --git a/llvm/test/CodeGen/AMDGPU/postra-machine-sink.mir b/llvm/test/CodeGen/AMDGPU/postra-machine-sink.mir
index ff7a40bfde9580..86863c31753642 100644
--- a/llvm/test/CodeGen/AMDGPU/postra-machine-sink.mir
+++ b/llvm/test/CodeGen/AMDGPU/postra-machine-sink.mir
@@ -5,7 +5,7 @@
# CHECK-LABEL: bb.0:
# CHECK: renamable $sgpr1 = COPY renamable $sgpr2
# CHECK-LABEL: bb.1:
-# CHECK: liveins: $sgpr0_sgpr1:0x000000000000000F
+# CHECK: liveins: $sgpr0_sgpr1
# CHECK: renamable $vgpr1_vgpr2 = COPY renamable $sgpr0_sgpr1
---
More information about the llvm-commits
mailing list