[llvm] a3508e0 - [SLP]Small buidlvector only graph should contains scalars from same block

[Alexey Bataev via llvm-commits]

Author: Alexey Bataev
Date: 2024-10-23T10:46:38-07:00
New Revision: a3508e0246741851f06cb9ba86203be8d89febed

URL: https://github.com/llvm/llvm-project/commit/a3508e0246741851f06cb9ba86203be8d89febed
DIFF: https://github.com/llvm/llvm-project/commit/a3508e0246741851f06cb9ba86203be8d89febed.diff

LOG: [SLP]Small buidlvector only graph should contains scalars from same block

If the graph is small and has single buildvector node, all scalars
instructions must be from the same basic block to prevent compiler
crash.

Fixes #113451

Added: 
    llvm/test/Transforms/SLPVectorizer/X86/small-graph-diff-block-instructions.ll

Modified: 
    llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 756b25ac985612..d824c40c401df7 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -11588,7 +11588,8 @@ bool BoUpSLP::isTreeTinyAndNotFullyVectorizable(bool ForReduction) const {
 
   if (VectorizableTree.back()->isGather() &&
       VectorizableTree.back()->isAltShuffle() &&
-      VectorizableTree.back()->getVectorFactor() > 2)
+      VectorizableTree.back()->getVectorFactor() > 2 &&
+      allSameBlock(VectorizableTree.back()->Scalars))
     return false;
 
   assert(VectorizableTree.empty()

diff  --git a/llvm/test/Transforms/SLPVectorizer/X86/small-graph-
diff -block-instructions.ll b/llvm/test/Transforms/SLPVectorizer/X86/small-graph-
diff -block-instructions.ll
new file mode 100644
index 00000000000000..82c8b1d707cf4f
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/X86/small-graph-
diff -block-instructions.ll
@@ -0,0 +1,33 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -mattr=+avx512dq -slp-threshold=-99999 < %s | FileCheck %s
+
+define i32 @test(i32 %arg, i32 %arg1) {
+; CHECK-LABEL: define i32 @test(
+; CHECK-SAME: i32 [[ARG:%.*]], i32 [[ARG1:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT:  [[BB:.*:]]
+; CHECK-NEXT:    [[ZEXT:%.*]] = zext i32 [[ARG1]] to i64
+; CHECK-NEXT:    [[ZEXT2:%.*]] = zext i32 [[ARG]] to i64
+; CHECK-NEXT:    [[SEXT:%.*]] = sext i32 [[ARG]] to i64
+; CHECK-NEXT:    br label %[[BB3:.*]]
+; CHECK:       [[BB3]]:
+; CHECK-NEXT:    [[GETELEMENTPTR:%.*]] = getelementptr i64, ptr addrspace(1) null, i64 [[ZEXT]]
+; CHECK-NEXT:    [[GETELEMENTPTR4:%.*]] = getelementptr i64, ptr addrspace(1) null, i64 [[ZEXT2]]
+; CHECK-NEXT:    [[GETELEMENTPTR5:%.*]] = getelementptr i64, ptr addrspace(1) null, i64 [[SEXT]]
+; CHECK-NEXT:    [[ZEXT6:%.*]] = zext i32 0 to i64
+; CHECK-NEXT:    [[GETELEMENTPTR7:%.*]] = getelementptr i64, ptr addrspace(1) null, i64 [[ZEXT6]]
+; CHECK-NEXT:    ret i32 0
+;
+bb:
+  %zext = zext i32 %arg1 to i64
+  %zext2 = zext i32 %arg to i64
+  %sext = sext i32 %arg to i64
+  br label %bb3
+
+bb3:
+  %getelementptr = getelementptr i64, ptr addrspace(1) null, i64 %zext
+  %getelementptr4 = getelementptr i64, ptr addrspace(1) null, i64 %zext2
+  %getelementptr5 = getelementptr i64, ptr addrspace(1) null, i64 %sext
+  %zext6 = zext i32 0 to i64
+  %getelementptr7 = getelementptr i64, ptr addrspace(1) null, i64 %zext6
+  ret i32 0
+}