ValueTracking: Allow getUnderlyingObject to look at vectors #114311
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
This was referenced Oct 30, 2024
Contributor
Author
arsenm
added
the
llvm:analysis
Graphite App
Member
|
@llvm/pr-subscribers-backend-amdgpu @llvm/pr-subscribers-llvm-analysis Author: Matt Arsenault (arsenm) ChangesWe can identify some easy vector of pointer cases, such as Full diff: https://github.com/llvm/llvm-project/pull/114311.diff 2 Files Affected:
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index aa5142f33624099..dd7cac9f4fc3259 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -6686,11 +6686,11 @@ static bool isSameUnderlyingObjectInLoop(const PHINode *PN,
}
const Value *llvm::getUnderlyingObject(const Value *V, unsigned MaxLookup) {
- if (!V->getType()->isPointerTy())
- return V;
for (unsigned Count = 0; MaxLookup == 0 || Count < MaxLookup; ++Count) {
if (auto *GEP = dyn_cast<GEPOperator>(V)) {
V = GEP->getPointerOperand();
+ if (!V->getType()->isPointerTy()) // Only handle scalar pointer base.
+ return nullptr;
} else if (Operator::getOpcode(V) == Instruction::BitCast ||
Operator::getOpcode(V) == Instruction::AddrSpaceCast) {
Value *NewV = cast<Operator>(V)->getOperand(0);
diff --git a/llvm/test/CodeGen/AMDGPU/promote-alloca-vector-gep.ll b/llvm/test/CodeGen/AMDGPU/promote-alloca-vector-gep.ll
index 355a2b8796b24dc..76e1868b3c4b9e8 100644
--- a/llvm/test/CodeGen/AMDGPU/promote-alloca-vector-gep.ll
+++ b/llvm/test/CodeGen/AMDGPU/promote-alloca-vector-gep.ll
@@ -35,17 +35,30 @@ bb:
ret void
}
-; TODO: Should be able to promote this
define amdgpu_kernel void @scalar_alloca_ptr_with_vector_gep_offset_select(i1 %cond) {
; CHECK-LABEL: define amdgpu_kernel void @scalar_alloca_ptr_with_vector_gep_offset_select(
; CHECK-SAME: i1 [[COND:%.*]]) {
; CHECK-NEXT: [[BB:.*:]]
-; CHECK-NEXT: [[ALLOCA:%.*]] = alloca [4 x i32], align 4, addrspace(5)
-; CHECK-NEXT: [[GETELEMENTPTR0:%.*]] = getelementptr inbounds i8, ptr addrspace(5) [[ALLOCA]], <4 x i64> <i64 0, i64 1, i64 2, i64 3>
-; CHECK-NEXT: [[GETELEMENTPTR1:%.*]] = getelementptr inbounds i8, ptr addrspace(5) [[ALLOCA]], <4 x i64> <i64 3, i64 2, i64 1, i64 0>
-; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], <4 x ptr addrspace(5)> [[GETELEMENTPTR0]], <4 x ptr addrspace(5)> [[GETELEMENTPTR1]]
-; CHECK-NEXT: [[EXTRACTELEMENT:%.*]] = extractelement <4 x ptr addrspace(5)> [[SELECT]], i64 1
-; CHECK-NEXT: store i32 0, ptr addrspace(5) [[EXTRACTELEMENT]], align 4
+; CHECK-NEXT: [[TMP0:%.*]] = call noalias nonnull dereferenceable(64) ptr addrspace(4) @llvm.amdgcn.dispatch.ptr()
+; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr addrspace(4) [[TMP0]], i64 1
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr addrspace(4) [[TMP1]], align 4, !invariant.load [[META0]]
+; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr addrspace(4) [[TMP0]], i64 2
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr addrspace(4) [[TMP3]], align 4, !range [[RNG1]], !invariant.load [[META0]]
+; CHECK-NEXT: [[TMP5:%.*]] = lshr i32 [[TMP2]], 16
+; CHECK-NEXT: [[TMP6:%.*]] = call range(i32 0, 1024) i32 @llvm.amdgcn.workitem.id.x()
+; CHECK-NEXT: [[TMP7:%.*]] = call range(i32 0, 1024) i32 @llvm.amdgcn.workitem.id.y()
+; CHECK-NEXT: [[TMP8:%.*]] = call range(i32 0, 1024) i32 @llvm.amdgcn.workitem.id.z()
+; CHECK-NEXT: [[TMP9:%.*]] = mul nuw nsw i32 [[TMP5]], [[TMP4]]
+; CHECK-NEXT: [[TMP10:%.*]] = mul i32 [[TMP9]], [[TMP6]]
+; CHECK-NEXT: [[TMP11:%.*]] = mul nuw nsw i32 [[TMP7]], [[TMP4]]
+; CHECK-NEXT: [[TMP12:%.*]] = add i32 [[TMP10]], [[TMP11]]
+; CHECK-NEXT: [[TMP13:%.*]] = add i32 [[TMP12]], [[TMP8]]
+; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds [1024 x [4 x i32]], ptr addrspace(3) @scalar_alloca_ptr_with_vector_gep_offset_select.alloca, i32 0, i32 [[TMP13]]
+; CHECK-NEXT: [[GETELEMENTPTR0:%.*]] = getelementptr inbounds i8, ptr addrspace(3) [[TMP14]], <4 x i64> <i64 0, i64 1, i64 2, i64 3>
+; CHECK-NEXT: [[GETELEMENTPTR1:%.*]] = getelementptr inbounds i8, ptr addrspace(3) [[TMP14]], <4 x i64> <i64 3, i64 2, i64 1, i64 0>
+; CHECK-NEXT: [[SELECT:%.*]] = select i1 [[COND]], <4 x ptr addrspace(3)> [[GETELEMENTPTR0]], <4 x ptr addrspace(3)> [[GETELEMENTPTR1]]
+; CHECK-NEXT: [[EXTRACTELEMENT:%.*]] = extractelement <4 x ptr addrspace(3)> [[SELECT]], i64 1
+; CHECK-NEXT: store i32 0, ptr addrspace(3) [[EXTRACTELEMENT]], align 4
; CHECK-NEXT: ret void
;
bb:
|
nikic
approved these changes
Nov 1, 2024
Contributor
nikic
left a comment
nikic
left a comment
There was a problem hiding this comment.
A tentative LGTM. I think this particular change is fine, but it's a dangerous area because all of AA basically does not support pointers of vectors at all and treats them as escapes. Wouldn't surprise me if this causes a miscompile.
arsenm
force-pushed
the
users/arsenm/amdgpu-promote-alloca-handle-vector-of-pointers
branch
from
b5022a1 to
5c33264
Compare
arsenm
force-pushed
the
users/arsenm/getunderlyingobject-look-through-vector-gep
branch
from
6e8c27a to
e518d75
Compare
Base automatically changed from
users/arsenm/amdgpu-promote-alloca-handle-vector-of-pointers
to
main
November 6, 2024 16:47
We can identify some easy vector of pointer cases, such as a getelementptr with a scalar base.
arsenm
force-pushed
the
users/arsenm/getunderlyingobject-look-through-vector-gep
branch
from
e518d75 to
5a2566f
Compare
Contributor
Author
arsenm
deleted the
users/arsenm/getunderlyingobject-look-through-vector-gep
branch
Member
|
This patch causes opt to crash: Reported by dtcxzyw/llvm-opt-benchmark#1644 |
Contributor
Author
|
hvdijk
added a commit
to hvdijk/llvm-project
that referenced
this pull request
Nov 8, 2024
Fixes a crash seen after llvm#114311
hvdijk
added a commit
to hvdijk/llvm-project
that referenced
this pull request
Nov 9, 2024
Fixes a crash seen after llvm#114311
Groverkss
pushed a commit
to iree-org/llvm-project
that referenced
this pull request
Nov 15, 2024
…lvm#115569) Fixes a crash seen after llvm#114311
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.
We can identify some easy vector of pointer cases, such as
a getelementptr with a scalar base.