[llvm] 2b1e037 - [SLP]Fix createInsertVector mask emission

[Alexey Bataev via llvm-commits]

Author: Alexey Bataev
Date: 2025-01-18T11:48:53-08:00
New Revision: 2b1e037adb274c515b6ebe7808cc7da6a5b9c3b3

URL: https://github.com/llvm/llvm-project/commit/2b1e037adb274c515b6ebe7808cc7da6a5b9c3b3
DIFF: https://github.com/llvm/llvm-project/commit/2b1e037adb274c515b6ebe7808cc7da6a5b9c3b3.diff

LOG: [SLP]Fix createInsertVector mask emission

Added: 
    

Modified: 
    llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
    llvm/test/Transforms/SLPVectorizer/X86/insert-subvector.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 34d9abb4dc7a35..fc3afed391a06d 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -4972,7 +4972,7 @@ static Value *createInsertVector(
     // the subvector length.
     const unsigned VecVF = getNumElements(Vec->getType());
     SmallVector<int> Mask(VecVF, PoisonMaskElem);
-    std::iota(Mask.begin(), std::next(Mask.begin(), Index), 0);
+    std::iota(Mask.begin(), Mask.end(), 0);
     for (unsigned I : seq<unsigned>(SubVecVF))
       Mask[I + Index] = I + VecVF;
     if (Generator) {
@@ -13976,11 +13976,12 @@ Value *BoUpSLP::gather(
     Instruction *InsElt;
     if (auto *VecTy = dyn_cast<FixedVectorType>(Scalar->getType())) {
       assert(SLPReVec && "FixedVectorType is not expected.");
-      Vec = InsElt = cast<Instruction>(createInsertVector(
-          Builder, Vec, Scalar, Pos * getNumElements(VecTy)));
-      auto *II = dyn_cast<IntrinsicInst>(InsElt);
+      Vec =
+          createInsertVector(Builder, Vec, Scalar, Pos * getNumElements(VecTy));
+      auto *II = dyn_cast<IntrinsicInst>(Vec);
       if (!II || II->getIntrinsicID() != Intrinsic::vector_insert)
         return Vec;
+      InsElt = II;
     } else {
       Vec = Builder.CreateInsertElement(Vec, Scalar, Builder.getInt32(Pos));
       InsElt = dyn_cast<InsertElementInst>(Vec);

diff  --git a/llvm/test/Transforms/SLPVectorizer/X86/insert-subvector.ll b/llvm/test/Transforms/SLPVectorizer/X86/insert-subvector.ll
index 456af1f0b3c70f..80b62c3cfffaca 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/insert-subvector.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/insert-subvector.ll
@@ -17,7 +17,7 @@ define <16 x double> @test(ptr %x, double %v, double %a) {
 ; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <2 x double> [[TMP8]], <2 x double> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP10:%.*]] = call <16 x double> @llvm.vector.insert.v16f64.v6f64(<16 x double> poison, <6 x double> [[TMP1]], i64 0)
 ; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <4 x double> [[TMP3]], <4 x double> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP12:%.*]] = shufflevector <16 x double> [[TMP10]], <16 x double> [[TMP11]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 16, i32 17, i32 18, i32 19, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP12:%.*]] = shufflevector <16 x double> [[TMP10]], <16 x double> [[TMP11]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 16, i32 17, i32 18, i32 19, i32 14, i32 15>
 ; CHECK-NEXT:    [[TMP13:%.*]] = call <16 x double> @llvm.vector.insert.v16f64.v2f64(<16 x double> [[TMP12]], <2 x double> [[TMP6]], i64 6)
 ; CHECK-NEXT:    [[TMP14:%.*]] = call <16 x double> @llvm.vector.insert.v16f64.v2f64(<16 x double> [[TMP13]], <2 x double> [[TMP7]], i64 8)
 ; CHECK-NEXT:    [[TMP15:%.*]] = call <16 x double> @llvm.vector.insert.v16f64.v2f64(<16 x double> [[TMP14]], <2 x double> [[TMP9]], i64 10)