Skip to content

[GVNSink] Fix incorrect codegen with respect to GEPs #85333 #92047

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
May 14, 2024

Conversation

hiraditya
Copy link
Collaborator

As mentioned in #68882 and https://discourse.llvm.org/t/rfc-replacing-getelementptr-with-ptradd/68699

Gep arithmetic isn't consistent with different types. GVNSink didn't realize this and sank all geps as long as their operands can be wired via PHIs
in a post-dominator.

Fixes: #85333
Reapply: #88440 after fixing the non-determinism issues in #90995

As mentioned in llvm#68882 and https://discourse.llvm.org/t/rfc-replacing-getelementptr-with-ptradd/68699

Gep arithmetic isn't consistent with different types. GVNSink didn't realize this and sank all geps
as long as their operands can be wired via PHIs
in a post-dominator.

Fixes: llvm#85333
@hiraditya hiraditya requested a review from efriedma-quic May 13, 2024 23:20
@hiraditya hiraditya requested a review from nikic May 13, 2024 23:20
@llvmbot
Copy link
Member

llvmbot commented May 13, 2024

@llvm/pr-subscribers-llvm-transforms

Author: AdityaK (hiraditya)

Changes

As mentioned in #68882 and https://discourse.llvm.org/t/rfc-replacing-getelementptr-with-ptradd/68699

Gep arithmetic isn't consistent with different types. GVNSink didn't realize this and sank all geps as long as their operands can be wired via PHIs
in a post-dominator.

Fixes: #85333
Reapply: #88440 after fixing the non-determinism issues in #90995


Full diff: https://github.com/llvm/llvm-project/pull/92047.diff

2 Files Affected:

  • (modified) llvm/lib/Transforms/Scalar/GVNSink.cpp (+4-5)
  • (added) llvm/test/Transforms/GVNSink/different-gep-types.ll (+101)
diff --git a/llvm/lib/Transforms/Scalar/GVNSink.cpp b/llvm/lib/Transforms/Scalar/GVNSink.cpp
index 95a4c644a91ad..7a183e4d3aa8f 100644
--- a/llvm/lib/Transforms/Scalar/GVNSink.cpp
+++ b/llvm/lib/Transforms/Scalar/GVNSink.cpp
@@ -763,12 +763,11 @@ GVNSink::analyzeInstructionForSinking(LockstepReverseIterator &LRI,
   // try and continue making progress.
   Instruction *I0 = NewInsts[0];
 
-  // If all instructions that are going to participate don't have the same
-  // number of operands, we can't do any useful PHI analysis for all operands.
-  auto hasDifferentNumOperands = [&I0](Instruction *I) {
-    return I->getNumOperands() != I0->getNumOperands();
+  auto isNotSameOperation = [&I0](Instruction *I) {
+    return !I0->isSameOperationAs(I);
   };
-  if (any_of(NewInsts, hasDifferentNumOperands))
+
+  if (any_of(NewInsts, isNotSameOperation))
     return std::nullopt;
 
   for (unsigned OpNum = 0, E = I0->getNumOperands(); OpNum != E; ++OpNum) {
diff --git a/llvm/test/Transforms/GVNSink/different-gep-types.ll b/llvm/test/Transforms/GVNSink/different-gep-types.ll
new file mode 100644
index 0000000000000..659be51cc56ee
--- /dev/null
+++ b/llvm/test/Transforms/GVNSink/different-gep-types.ll
@@ -0,0 +1,101 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -passes=gvn-sink -S %s | FileCheck %s
+
+target datalayout = "e-m:e-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"
+
+%"struct.std::pair" = type <{ i32, %struct.substruct, [2 x i8] }>
+%struct.substruct = type { i8, i8 }
+%"struct.std::random_access_iterator_tag" = type { i8 }
+
+; Check that gep is not sunk as they are of different types.
+define void @bar(ptr noundef nonnull align 4 dereferenceable(4) %__i, i32 noundef %__n) {
+; CHECK-LABEL: define void @bar(
+; CHECK-SAME: ptr noundef nonnull align 4 dereferenceable(4) [[__I:%.*]], i32 noundef [[__N:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[__N]], 1
+; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[__I]], align 4
+; CHECK-NEXT:    [[INCDEC_PTR4:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 -8
+; CHECK-NEXT:    br label [[IF_END6:%.*]]
+; CHECK:       if.else:
+; CHECK-NEXT:    [[TMP1:%.*]] = load ptr, ptr [[__I]], align 4
+; CHECK-NEXT:    [[ADD_PTR:%.*]] = getelementptr inbounds %"struct.std::pair", ptr [[TMP1]], i32 [[__N]]
+; CHECK-NEXT:    br label [[IF_END6]]
+; CHECK:       if.end6:
+; CHECK-NEXT:    [[INCDEC_PTR_SINK:%.*]] = phi ptr [ [[INCDEC_PTR4]], [[IF_THEN]] ], [ [[ADD_PTR]], [[IF_ELSE]] ]
+; CHECK-NEXT:    store ptr [[INCDEC_PTR_SINK]], ptr [[__I]], align 4
+; CHECK-NEXT:    ret void
+;
+entry:
+  %cmp = icmp eq i32 %__n, 1
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  %3 = load ptr, ptr %__i, align 4
+  %incdec.ptr4 = getelementptr inbounds i8, ptr %3, i32 -8
+  br label %if.end6
+
+if.else:
+  %4 = load ptr, ptr %__i, align 4
+  %add.ptr = getelementptr inbounds %"struct.std::pair", ptr %4, i32 %__n
+  br label %if.end6
+
+if.end6:
+  %incdec.ptr.sink = phi ptr [ %incdec.ptr4, %if.then ], [ %add.ptr, %if.else ]
+  store ptr %incdec.ptr.sink, ptr %__i, align 4
+  ret void
+}
+
+; Check that load,gep, and store are all sunk as they are safe to do.
+define void @foo(ptr noundef nonnull align 4 dereferenceable(4) %__i, i32 noundef %__n) {
+; CHECK-LABEL: define void @foo(
+; CHECK-SAME: ptr noundef nonnull align 4 dereferenceable(4) [[__I:%.*]], i32 noundef [[__N:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i32 [[__N]], 1
+; CHECK-NEXT:    br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK:       if.then:
+; CHECK-NEXT:    br label [[IF_END6:%.*]]
+; CHECK:       if.else:
+; CHECK-NEXT:    [[CMP2:%.*]] = icmp eq i32 [[__N]], -1
+; CHECK-NEXT:    br i1 [[CMP2]], label [[IF_THEN3:%.*]], label [[IF_ELSE5:%.*]]
+; CHECK:       if.then3:
+; CHECK-NEXT:    br label [[IF_END6]]
+; CHECK:       if.else5:
+; CHECK-NEXT:    br label [[IF_END6]]
+; CHECK:       if.end6:
+; CHECK-NEXT:    [[DOTSINK1:%.*]] = phi i32 [ -4, [[IF_ELSE5]] ], [ -8, [[IF_THEN3]] ], [ 8, [[IF_THEN]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = load ptr, ptr [[__I]], align 4
+; CHECK-NEXT:    [[INCDEC_PTR:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 [[DOTSINK1]]
+; CHECK-NEXT:    store ptr [[INCDEC_PTR]], ptr [[__I]], align 4
+; CHECK-NEXT:    ret void
+;
+entry:
+  %cmp = icmp eq i32 %__n, 1
+  br i1 %cmp, label %if.then, label %if.else
+
+if.then:
+  %1 = load ptr, ptr %__i, align 4
+  %incdec.ptr = getelementptr inbounds i8, ptr %1, i32 8
+  store ptr %incdec.ptr, ptr %__i, align 4
+  br label %if.end6
+
+if.else:
+  %cmp2 = icmp eq i32 %__n, -1
+  br i1 %cmp2, label %if.then3, label %if.else5
+
+if.then3:
+  %3 = load ptr, ptr %__i, align 4
+  %incdec.ptr4 = getelementptr inbounds i8, ptr %3, i32 -8
+  store ptr %incdec.ptr4, ptr %__i, align 4
+  br label %if.end6
+
+if.else5:
+  %4 = load ptr, ptr %__i, align 4
+  %add.ptr = getelementptr inbounds i8, ptr %4, i32 -4
+  store ptr %add.ptr, ptr %__i, align 4
+  br label %if.end6
+
+if.end6:
+  ret void
+}

Copy link
Contributor

@nikic nikic left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

@hiraditya hiraditya merged commit bf7a0f9 into llvm:main May 14, 2024
6 checks passed
@hiraditya hiraditya changed the title Fix incorrect codegen with respect to GEPs #85333 [GVNSink] Fix incorrect codegen with respect to GEPs #85333 May 14, 2024
@hiraditya hiraditya deleted the gvnsink_pr85333 branch December 4, 2024 21:55
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

Successfully merging this pull request may close these issues.

GVNSink produces incorrect codegen with respect to GEPs
3 participants