[AArch64] Copy implicit def operands when creating LDP. #164253
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Member
|
@llvm/pr-subscribers-backend-aarch64 Author: David Green (davemgreen) ChangesOtherwise we might end up with undefined register uses. Copying implicit uses can cause problems where a register is both defined and used in the same LDP, so I have not tried to add them here. Fixes #164230 Full diff: https://github.com/llvm/llvm-project/pull/164253.diff 2 Files Affected:
diff --git a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
index e69fa32967a79..96998593514a8 100644
--- a/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AArch64/AArch64LoadStoreOptimizer.cpp
@@ -1386,6 +1386,17 @@ AArch64LoadStoreOpt::mergePairedInsns(MachineBasicBlock::iterator I,
if (MOP.isReg() && MOP.isKill())
DefinedInBB.addReg(MOP.getReg());
+ // Copy over any implicit-def operands. This is like MI.copyImplicitOps, but
+ // only copies implicit defs.
+ auto CopyImplicitOps = [&](MachineBasicBlock::iterator MI) {
+ for (const MachineOperand &MO :
+ llvm::drop_begin(MI->operands(), MI->getDesc().getNumOperands()))
+ if (MO.isReg() && MO.isImplicit() && MO.isDef())
+ MIB.add(MO);
+ };
+ CopyImplicitOps(I);
+ CopyImplicitOps(Paired);
+
// Erase the old instructions.
I->eraseFromParent();
Paired->eraseFromParent();
diff --git a/llvm/test/CodeGen/AArch64/ldst-implicitop.mir b/llvm/test/CodeGen/AArch64/ldst-implicitop.mir
new file mode 100644
index 0000000000000..e29b0db0eb9de
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/ldst-implicitop.mir
@@ -0,0 +1,58 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 6
+# RUN: llc -mtriple=aarch64-- -run-pass=aarch64-ldst-opt -verify-machineinstrs -o - %s | FileCheck %s
+# Check that we copy implicit operands.
+---
+name: impdef_op1
+tracksRegLiveness: true
+body: |
+ bb.0:
+ liveins: $lr
+ ; CHECK-LABEL: name: impdef_op1
+ ; CHECK: liveins: $lr
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: renamable $q5, renamable $q20 = LDPQi renamable $lr, 3, implicit-def $q4_q5 :: (load (s128))
+ ; CHECK-NEXT: $q0 = ORRv16i8 $q4, killed $q4
+ ; CHECK-NEXT: $q1 = ORRv16i8 $q5, killed $q5
+ ; CHECK-NEXT: RET_ReallyLR
+ renamable $q5 = LDRQui renamable $lr, 3, implicit-def $q4_q5 :: (load (s128))
+ renamable $q20 = LDRQui renamable $lr, 4 :: (load (s128))
+ $q0 = ORRv16i8 $q4, killed $q4
+ $q1 = ORRv16i8 $q5, killed $q5
+ RET_ReallyLR
+...
+---
+name: impdef_op2
+body: |
+ bb.0:
+ liveins: $lr
+ ; CHECK-LABEL: name: impdef_op2
+ ; CHECK: liveins: $lr
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: renamable $q20, renamable $q5 = LDPQi renamable $lr, 3, implicit-def $q4_q5 :: (load (s128))
+ ; CHECK-NEXT: $q0 = ORRv16i8 $q4, killed $q4
+ ; CHECK-NEXT: $q1 = ORRv16i8 $q5, killed $q5
+ ; CHECK-NEXT: RET_ReallyLR
+ renamable $q20 = LDRQui renamable $lr, 3 :: (load (s128))
+ renamable $q5 = LDRQui renamable $lr, 4, implicit-def $q4_q5 :: (load (s128))
+ $q0 = ORRv16i8 $q4, killed $q4
+ $q1 = ORRv16i8 $q5, killed $q5
+ RET_ReallyLR
+...
+---
+name: impdef_both
+body: |
+ bb.0:
+ liveins: $lr
+ ; CHECK-LABEL: name: impdef_both
+ ; CHECK: liveins: $lr
+ ; CHECK-NEXT: {{ $}}
+ ; CHECK-NEXT: renamable $q5, renamable $q20 = LDPQi renamable $lr, 3, implicit-def $q4_q5, implicit-def $q20_q21 :: (load (s128))
+ ; CHECK-NEXT: $q0 = ORRv16i8 $q4, killed $q4
+ ; CHECK-NEXT: $q1 = ORRv16i8 $q5, killed $q5
+ ; CHECK-NEXT: RET_ReallyLR
+ renamable $q5 = LDRQui renamable $lr, 3, implicit-def $q4_q5 :: (load (s128))
+ renamable $q20 = LDRQui renamable $lr, 4, implicit-def $q20_q21 :: (load (s128))
+ $q0 = ORRv16i8 $q4, killed $q4
+ $q1 = ORRv16i8 $q5, killed $q5
+ RET_ReallyLR
+...
|
Otherwise we might end up with undefined register uses. Copying implicit uses can cause problems where a register is both defined and used in the same LDP, so I have not tried to add them here. Fixes llvm#164230
…implicit-defs are unique.
davemgreen
force-pushed
the
gh-a64-ldstimplicitops
branch
from
2c9752b to
152f49d
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Otherwise we might end up with undefined register uses. Copying implicit uses can cause problems where a register is both defined and used in the same LDP, so I have not tried to add them here.
Fixes #164230