AMDGPU: Use getMergedLocation in SILoadStoreOptimizer #156396
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-amdgpu Author: Matt Arsenault (arsenm) ChangesThis is merging loads and stores so use the combined DebugLoc. Not sure if computeBase should be using the merged location from Full diff: https://github.com/llvm/llvm-project/pull/156396.diff 1 Files Affected:
diff --git a/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp b/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
index 263f7127fbf10..5d2143904c873 100644
--- a/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
+++ b/llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cpp
@@ -232,10 +232,11 @@ class SILoadStoreOptimizer {
void copyToDestRegs(CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore,
- AMDGPU::OpName OpName, Register DestReg) const;
+ const DebugLoc &DL, AMDGPU::OpName OpName,
+ Register DestReg) const;
Register copyFromSrcRegs(CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore,
- AMDGPU::OpName OpName) const;
+ const DebugLoc &DL, AMDGPU::OpName OpName) const;
unsigned read2Opcode(unsigned EltSize) const;
unsigned read2ST64Opcode(unsigned EltSize) const;
@@ -1320,10 +1321,9 @@ SILoadStoreOptimizer::checkAndPrepareMerge(CombineInfo &CI,
// Paired.
void SILoadStoreOptimizer::copyToDestRegs(
CombineInfo &CI, CombineInfo &Paired,
- MachineBasicBlock::iterator InsertBefore, AMDGPU::OpName OpName,
- Register DestReg) const {
+ MachineBasicBlock::iterator InsertBefore, const DebugLoc &DL,
+ AMDGPU::OpName OpName, Register DestReg) const {
MachineBasicBlock *MBB = CI.I->getParent();
- DebugLoc DL = CI.I->getDebugLoc();
auto [SubRegIdx0, SubRegIdx1] = getSubRegIdxs(CI, Paired);
@@ -1351,9 +1351,9 @@ void SILoadStoreOptimizer::copyToDestRegs(
Register
SILoadStoreOptimizer::copyFromSrcRegs(CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore,
+ const DebugLoc &DL,
AMDGPU::OpName OpName) const {
MachineBasicBlock *MBB = CI.I->getParent();
- DebugLoc DL = CI.I->getDebugLoc();
auto [SubRegIdx0, SubRegIdx1] = getSubRegIdxs(CI, Paired);
@@ -1409,7 +1409,8 @@ SILoadStoreOptimizer::mergeRead2Pair(CombineInfo &CI, CombineInfo &Paired,
const TargetRegisterClass *SuperRC = getTargetRegisterClass(CI, Paired);
Register DestReg = MRI->createVirtualRegister(SuperRC);
- DebugLoc DL = CI.I->getDebugLoc();
+ DebugLoc DL = DebugLoc::getMergedLocation(CI.I->getDebugLoc(),
+ Paired.I->getDebugLoc());
Register BaseReg = AddrReg->getReg();
unsigned BaseSubReg = AddrReg->getSubReg();
@@ -1437,7 +1438,7 @@ SILoadStoreOptimizer::mergeRead2Pair(CombineInfo &CI, CombineInfo &Paired,
.addImm(0) // gds
.cloneMergedMemRefs({&*CI.I, &*Paired.I});
- copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdst, DestReg);
+ copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdst, DestReg);
CI.I->eraseFromParent();
Paired.I->eraseFromParent();
@@ -1491,7 +1492,8 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeWrite2Pair(
(NewOffset0 != NewOffset1) && "Computed offset doesn't fit");
const MCInstrDesc &Write2Desc = TII->get(Opc);
- DebugLoc DL = CI.I->getDebugLoc();
+ DebugLoc DL = DebugLoc::getMergedLocation(CI.I->getDebugLoc(),
+ Paired.I->getDebugLoc());
Register BaseReg = AddrReg->getReg();
unsigned BaseSubReg = AddrReg->getSubReg();
@@ -1532,7 +1534,9 @@ MachineBasicBlock::iterator
SILoadStoreOptimizer::mergeImagePair(CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore) {
MachineBasicBlock *MBB = CI.I->getParent();
- DebugLoc DL = CI.I->getDebugLoc();
+ DebugLoc DL = DebugLoc::getMergedLocation(CI.I->getDebugLoc(),
+ Paired.I->getDebugLoc());
+
const unsigned Opcode = getNewOpcode(CI, Paired);
const TargetRegisterClass *SuperRC = getTargetRegisterClass(CI, Paired);
@@ -1557,7 +1561,7 @@ SILoadStoreOptimizer::mergeImagePair(CombineInfo &CI, CombineInfo &Paired,
MachineInstr *New = MIB.addMemOperand(combineKnownAdjacentMMOs(CI, Paired));
- copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata, DestReg);
+ copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata, DestReg);
CI.I->eraseFromParent();
Paired.I->eraseFromParent();
@@ -1568,7 +1572,9 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeSMemLoadImmPair(
CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore) {
MachineBasicBlock *MBB = CI.I->getParent();
- DebugLoc DL = CI.I->getDebugLoc();
+ DebugLoc DL = DebugLoc::getMergedLocation(CI.I->getDebugLoc(),
+ Paired.I->getDebugLoc());
+
const unsigned Opcode = getNewOpcode(CI, Paired);
const TargetRegisterClass *SuperRC = getTargetRegisterClass(CI, Paired);
@@ -1589,7 +1595,7 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeSMemLoadImmPair(
New.addImm(MergedOffset);
New.addImm(CI.CPol).addMemOperand(combineKnownAdjacentMMOs(CI, Paired));
- copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::sdst, DestReg);
+ copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::sdst, DestReg);
CI.I->eraseFromParent();
Paired.I->eraseFromParent();
@@ -1600,7 +1606,9 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeBufferLoadPair(
CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore) {
MachineBasicBlock *MBB = CI.I->getParent();
- DebugLoc DL = CI.I->getDebugLoc();
+
+ DebugLoc DL = DebugLoc::getMergedLocation(CI.I->getDebugLoc(),
+ Paired.I->getDebugLoc());
const unsigned Opcode = getNewOpcode(CI, Paired);
@@ -1630,7 +1638,7 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeBufferLoadPair(
.addImm(0) // swz
.addMemOperand(combineKnownAdjacentMMOs(CI, Paired));
- copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata, DestReg);
+ copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata, DestReg);
CI.I->eraseFromParent();
Paired.I->eraseFromParent();
@@ -1641,7 +1649,9 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeTBufferLoadPair(
CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore) {
MachineBasicBlock *MBB = CI.I->getParent();
- DebugLoc DL = CI.I->getDebugLoc();
+
+ DebugLoc DL = DebugLoc::getMergedLocation(CI.I->getDebugLoc(),
+ Paired.I->getDebugLoc());
const unsigned Opcode = getNewOpcode(CI, Paired);
@@ -1681,7 +1691,7 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeTBufferLoadPair(
.addImm(0) // swz
.addMemOperand(combineKnownAdjacentMMOs(CI, Paired));
- copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata, DestReg);
+ copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata, DestReg);
CI.I->eraseFromParent();
Paired.I->eraseFromParent();
@@ -1692,12 +1702,13 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeTBufferStorePair(
CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore) {
MachineBasicBlock *MBB = CI.I->getParent();
- DebugLoc DL = CI.I->getDebugLoc();
+ DebugLoc DL = DebugLoc::getMergedLocation(CI.I->getDebugLoc(),
+ Paired.I->getDebugLoc());
const unsigned Opcode = getNewOpcode(CI, Paired);
Register SrcReg =
- copyFromSrcRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata);
+ copyFromSrcRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata);
auto MIB = BuildMI(*MBB, InsertBefore, DL, TII->get(Opcode))
.addReg(SrcReg, RegState::Kill);
@@ -1739,7 +1750,9 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeFlatLoadPair(
CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore) {
MachineBasicBlock *MBB = CI.I->getParent();
- DebugLoc DL = CI.I->getDebugLoc();
+
+ DebugLoc DL = DebugLoc::getMergedLocation(CI.I->getDebugLoc(),
+ Paired.I->getDebugLoc());
const unsigned Opcode = getNewOpcode(CI, Paired);
@@ -1757,7 +1770,7 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeFlatLoadPair(
.addImm(CI.CPol)
.addMemOperand(combineKnownAdjacentMMOs(CI, Paired));
- copyToDestRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdst, DestReg);
+ copyToDestRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdst, DestReg);
CI.I->eraseFromParent();
Paired.I->eraseFromParent();
@@ -1768,12 +1781,14 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeFlatStorePair(
CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore) {
MachineBasicBlock *MBB = CI.I->getParent();
- DebugLoc DL = CI.I->getDebugLoc();
+
+ DebugLoc DL = DebugLoc::getMergedLocation(CI.I->getDebugLoc(),
+ Paired.I->getDebugLoc());
const unsigned Opcode = getNewOpcode(CI, Paired);
Register SrcReg =
- copyFromSrcRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata);
+ copyFromSrcRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata);
auto MIB = BuildMI(*MBB, InsertBefore, DL, TII->get(Opcode))
.add(*TII->getNamedOperand(*CI.I, AMDGPU::OpName::vaddr))
@@ -2042,12 +2057,13 @@ MachineBasicBlock::iterator SILoadStoreOptimizer::mergeBufferStorePair(
CombineInfo &CI, CombineInfo &Paired,
MachineBasicBlock::iterator InsertBefore) {
MachineBasicBlock *MBB = CI.I->getParent();
- DebugLoc DL = CI.I->getDebugLoc();
+ DebugLoc DL = DebugLoc::getMergedLocation(CI.I->getDebugLoc(),
+ Paired.I->getDebugLoc());
const unsigned Opcode = getNewOpcode(CI, Paired);
Register SrcReg =
- copyFromSrcRegs(CI, Paired, InsertBefore, AMDGPU::OpName::vdata);
+ copyFromSrcRegs(CI, Paired, InsertBefore, DL, AMDGPU::OpName::vdata);
auto MIB = BuildMI(*MBB, InsertBefore, DL, TII->get(Opcode))
.addReg(SrcReg, RegState::Kill);
|
|
✅ With the latest revision this PR passed the C/C++ code formatter. |
arsenm
force-pushed
the
users/arsenm/amdgpu/use-getMergedLocation-si-load-store-opt
branch
from
5fb68b6 to
c2fbea3
Compare
Pierre-vh
reviewed
Sep 2, 2025
Contributor
Pierre-vh
left a comment
Pierre-vh
left a comment
There was a problem hiding this comment.
Would a MIR test work to test this ? https://llvm.org/docs/MIRLangRef.html#source-locations
Contributor
|
It seems like the computeBase instruction(s) should be the same as at least one of the merged stores? If not, I would imagine we don't want to include it anyway I think the existing logic in |
arsenm
force-pushed
the
users/arsenm/amdgpu/use-getMergedLocation-si-load-store-opt
branch
from
c2fbea3 to
0c5c1ce
Compare
Contributor
Author
|
Added a test, which certainly doesn't cover all the cases. The result seems to be dropping the line info |
You can test this locally with the following command:git diff -U0 --pickaxe-regex -S '([^a-zA-Z0-9#_-]undef([^a-zA-Z0-9_-]|$)|UndefValue::get)' 'HEAD~1' HEAD llvm/test/CodeGen/AMDGPU/ds-read2-write2-debug-info.ll llvm/lib/Target/AMDGPU/SILoadStoreOptimizer.cppThe following files introduce new uses of undef:
Undef is now deprecated and should only be used in the rare cases where no replacement is possible. For example, a load of uninitialized memory yields In tests, avoid using For example, this is considered a bad practice: define void @fn() {
...
br i1 undef, ...
}Please use the following instead: define void @fn(i1 %cond) {
...
br i1 %cond, ...
}Please refer to the Undefined Behavior Manual for more information. |
This is merging loads and stores so use the combined DebugLoc. Not sure if computeBase should be using the merged location from all the involved instructions. I'm also not sure how to test this sort of thing.
arsenm
force-pushed
the
users/arsenm/amdgpu/use-getMergedLocation-si-load-store-opt
branch
from
0c5c1ce to
fd8b4c2
Compare
Contributor
Author
|
ping |
1 similar comment
Contributor
Author
|
ping |
slinder1
approved these changes
Nov 10, 2025
Contributor
slinder1
left a comment
slinder1
left a comment
There was a problem hiding this comment.
LGTM, dropping is at least conservatively correct, and far preferable to incorrect info.
arsenm
deleted the
users/arsenm/amdgpu/use-getMergedLocation-si-load-store-opt
branch
Collaborator
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/138/builds/21597 Here is the relevant piece of the build log for the reference |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This is merging loads and stores so use the combined DebugLoc.
Not sure if computeBase should be using the merged location from
all the involved instructions. I'm also not sure how to test this
sort of thing.