[MemCpyOpt] No need to create memcpy(a <- a)
#98321
Merged
Conversation
This file contains 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
|
@llvm/pr-subscribers-llvm-transforms Author: DianQK (DianQK) ChangesWhen forwarding Full diff: https://github.com/llvm/llvm-project/pull/98321.diff 2 Files Affected:
diff --git a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index b9efd9aaa28c5..1a4b3657cedae 100644
--- a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -1161,6 +1161,14 @@ bool MemCpyOptPass::processMemCpyMemCpyDependence(MemCpyInst *M,
MSSA->getMemoryAccess(MDep), MSSA->getMemoryAccess(M)))
return false;
+ // No need to create `memcpy(a <- a)`.
+ if (BAA.isMustAlias(M->getDest(), MDep->getRawSource())) {
+ // Remove the instruction we're replacing.
+ eraseInstruction(M);
+ ++NumMemCpyInstr;
+ return true;
+ }
+
// If the dest of the second might alias the source of the first, then the
// source and dest might overlap. In addition, if the source of the first
// points to constant memory, they won't overlap by definition. Otherwise, we
diff --git a/llvm/test/Transforms/MemCpyOpt/memcpy.ll b/llvm/test/Transforms/MemCpyOpt/memcpy.ll
index 48698f25174e7..a19cfe6e75b70 100644
--- a/llvm/test/Transforms/MemCpyOpt/memcpy.ll
+++ b/llvm/test/Transforms/MemCpyOpt/memcpy.ll
@@ -139,6 +139,56 @@ define void @test6_memcpy(ptr %src, ptr %dest) nounwind {
ret void
}
+; FIXME: When forwarding to memcpy(arg+1, arg+1), we don't need to create this memcpy.
+define void @test6_memcpy_forward_back(ptr %arg) nounwind {
+; CHECK-LABEL: @test6_memcpy_forward_back(
+; CHECK-NEXT: [[DEST:%.*]] = getelementptr inbounds i8, ptr [[ARG:%.*]], i64 1
+; CHECK-NEXT: ret void
+;
+ %tmp = alloca [16 x i8], align 1
+ %src = getelementptr inbounds i8, ptr %arg, i64 1
+ %dest = getelementptr inbounds i8, ptr %arg, i64 1
+ call void @llvm.memcpy.inline.p0.p0.i32(ptr align 1 %tmp, ptr align 1 %src, i32 16, i1 false)
+ call void @llvm.memcpy.inline.p0.p0.i32(ptr align 1 %dest, ptr align 1 %tmp, i32 16, i1 false)
+ ret void
+}
+
+; We have to retain this `memcpy(arg+2, arg+1)` forwarding.
+define void @test6_memcpy_forward_not_back(ptr %arg) nounwind {
+; CHECK-LABEL: @test6_memcpy_forward_not_back(
+; CHECK-NEXT: [[TMP:%.*]] = alloca [16 x i8], align 1
+; CHECK-NEXT: [[SRC:%.*]] = getelementptr inbounds i8, ptr [[ARG:%.*]], i64 1
+; CHECK-NEXT: [[DEST:%.*]] = getelementptr inbounds i8, ptr [[ARG]], i64 2
+; CHECK-NEXT: call void @llvm.memcpy.inline.p0.p0.i32(ptr align 1 [[TMP]], ptr align 1 [[SRC]], i32 16, i1 false)
+; CHECK-NEXT: call void @llvm.memcpy.inline.p0.p0.i32(ptr align 1 [[DEST]], ptr align 1 [[TMP]], i32 16, i1 false)
+; CHECK-NEXT: ret void
+;
+ %tmp = alloca [16 x i8], align 1
+ %src = getelementptr inbounds i8, ptr %arg, i64 1
+ %dest = getelementptr inbounds i8, ptr %arg, i64 2
+ call void @llvm.memcpy.inline.p0.p0.i32(ptr align 1 %tmp, ptr align 1 %src, i32 16, i1 false)
+ call void @llvm.memcpy.inline.p0.p0.i32(ptr align 1 %dest, ptr align 1 %tmp, i32 16, i1 false)
+ ret void
+}
+
+; There is data writing between the two memcpy operations.
+define void @test6_memcpy_forward_back_failed(ptr %arg) nounwind {
+; CHECK-LABEL: @test6_memcpy_forward_back_failed(
+; CHECK-NEXT: [[TMP:%.*]] = alloca [16 x i8], align 1
+; CHECK-NEXT: [[SRC:%.*]] = getelementptr inbounds i8, ptr [[ARG:%.*]], i64 1
+; CHECK-NEXT: [[DEST:%.*]] = getelementptr inbounds i8, ptr [[ARG]], i64 1
+; CHECK-NEXT: call void @llvm.memcpy.inline.p0.p0.i32(ptr align 1 [[TMP]], ptr align 1 [[SRC]], i32 16, i1 false)
+; CHECK-NEXT: call void @llvm.memcpy.inline.p0.p0.i32(ptr align 1 [[DEST]], ptr align 1 [[TMP]], i32 16, i1 false)
+; CHECK-NEXT: ret void
+;
+ %tmp = alloca [16 x i8], align 1
+ %src = getelementptr inbounds i8, ptr %arg, i64 1
+ %dest = getelementptr inbounds i8, ptr %arg, i64 1
+ call void @llvm.memcpy.inline.p0.p0.i32(ptr align 1 %tmp, ptr align 1 %src, i32 16, i1 false)
+ store i8 1, ptr %dest, align 1
+ call void @llvm.memcpy.inline.p0.p0.i32(ptr align 1 %dest, ptr align 1 %tmp, i32 16, i1 false)
+ ret void
+}
@x = external global %0
|
dtcxzyw
reviewed
Jul 10, 2024
dtcxzyw
reviewed
Jul 10, 2024
dtcxzyw
requested changes
Jul 10, 2024
nikic
reviewed
Jul 10, 2024
|
LLVM Buildbot has detected a new failure on builder Full details are available at: https://lab.llvm.org/buildbot/#/builders/109/builds/861 Here is the relevant piece of the build log for the reference: |
aaryanshukla
pushed a commit
to aaryanshukla/llvm-project
that referenced
this pull request
Jul 14, 2024
When forwarding `memcpy`, we don't need to create `memcpy(a, a)`.
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.
When forwarding
memcpy, we don't need to creatememcpy(a, a).I fixed a related test case at 62b3e68.