[SLP]Fix crash for scalarized vectors.

Need to remove insertion of the nodes to the InVector in case of
scalarized vectors too to avoid compiler crashes.

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 1fceae5..fd4fee0 100644 (file)
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -7028,10 +7028,8 @@ public:
     auto *VecTy = FixedVectorType::get(VL.front()->getType(), VL.size());
     // If the resulting type is scalarized, do not adjust the cost.
     unsigned VecNumParts = TTI.getNumberOfParts(VecTy);
-    if (VecNumParts == VecTy->getNumElements()) {
-      InVectors.assign(1, E);
+    if (VecNumParts == VecTy->getNumElements())
       return nullptr;
-    }
     DenseMap<Value *, int> ExtractVectorsTys;
     for (auto [I, V] : enumerate(VL)) {
       // Ignore non-extractelement scalars.

diff --git a/llvm/test/Transforms/SLPVectorizer/NVPTX/buildvector-scalarized.ll b/llvm/test/Transforms/SLPVectorizer/NVPTX/buildvector-scalarized.ll
new file mode 100644 (file)
index 0000000..e1c3d9a
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/NVPTX/buildvector-scalarized.ll
@@ -0,0 +1,37 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
+; RUN: opt -passes=slp-vectorizer -S -mtriple=nvptx64-unknown-unknown < %s | FileCheck %s
+
+define <2 x float> @baz() {
+; CHECK-LABEL: define <2 x float> @baz() {
+; CHECK-NEXT:  bb:
+; CHECK-NEXT:    [[EXTRACTELEMENT:%.*]] = extractelement <2 x float> zeroinitializer, i64 0
+; CHECK-NEXT:    [[FCMP:%.*]] = fcmp uno float [[EXTRACTELEMENT]], 0.000000e+00
+; CHECK-NEXT:    [[FCMP1:%.*]] = fcmp uno float 0.000000e+00, 0.000000e+00
+; CHECK-NEXT:    [[OR:%.*]] = or i1 [[FCMP]], [[FCMP1]]
+; CHECK-NEXT:    [[FCMP2:%.*]] = fcmp oeq float 0.000000e+00, 0.000000e+00
+; CHECK-NEXT:    [[OR3:%.*]] = or i1 [[FCMP2]], [[OR]]
+; CHECK-NEXT:    [[FCMP4:%.*]] = fcmp oeq float 0.000000e+00, 0.000000e+00
+; CHECK-NEXT:    [[OR5:%.*]] = or i1 [[FCMP4]], [[OR3]]
+; CHECK-NEXT:    br i1 [[OR5]], label [[BB6:%.*]], label [[BB7:%.*]]
+; CHECK:       bb6:
+; CHECK-NEXT:    ret <2 x float> zeroinitializer
+; CHECK:       bb7:
+; CHECK-NEXT:    ret <2 x float> zeroinitializer
+;
+bb:
+  %extractelement = extractelement <2 x float> zeroinitializer, i64 0
+  %fcmp = fcmp uno float %extractelement, 0.000000e+00
+  %fcmp1 = fcmp uno float 0.000000e+00, 0.000000e+00
+  %or = or i1 %fcmp, %fcmp1
+  %fcmp2 = fcmp oeq float 0.000000e+00, 0.000000e+00
+  %or3 = or i1 %fcmp2, %or
+  %fcmp4 = fcmp oeq float 0.000000e+00, 0.000000e+00
+  %or5 = or i1 %fcmp4, %or3
+  br i1 %or5, label %bb6, label %bb7
+
+bb6:                                              ; preds = %bb
+  ret <2 x float> zeroinitializer
+
+bb7:                                              ; preds = %bb
+  ret <2 x float> zeroinitializer
+}