[SLP][REVEC] The vectorized result for ShuffleVector may not be ShuffleVectorInst. (#116940)

Author: HanKuanChen

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

@@ -15716,16 +15716,18 @@ Value *BoUpSLP::vectorizeTree(TreeEntry *E, bool PostponedPHIs) {
           LLVM_DEBUG(dbgs() << "SLP: Diamond merged for " << *VL0 << ".\n");
           return E->VectorizedValue;
         }
-        assert(isa<ShuffleVectorInst>(Src) &&
-               "Not supported shufflevector usage.");
-        auto *SVSrc = cast<ShuffleVectorInst>(Src);
-        assert(isa<PoisonValue>(SVSrc->getOperand(1)) &&
-               "Not supported shufflevector usage.");
         SmallVector<int> ThisMask(calculateShufflevectorMask(E->Scalars));
-        SmallVector<int> NewMask(ThisMask.size());
-        transform(ThisMask, NewMask.begin(),
-                  [&SVSrc](int Mask) { return SVSrc->getShuffleMask()[Mask]; });
-        V = Builder.CreateShuffleVector(SVSrc->getOperand(0), NewMask);
+        if (auto *SVSrc = dyn_cast<ShuffleVectorInst>(Src)) {
+          assert(isa<PoisonValue>(SVSrc->getOperand(1)) &&
+                 "Not supported shufflevector usage.");
+          SmallVector<int> NewMask(ThisMask.size());
+          transform(ThisMask, NewMask.begin(), [&SVSrc](int Mask) {
+            return SVSrc->getShuffleMask()[Mask];
+          });
+          V = Builder.CreateShuffleVector(SVSrc->getOperand(0), NewMask);
+        } else {
+          V = Builder.CreateShuffleVector(Src, ThisMask);
+        }
         propagateIRFlags(V, E->Scalars, VL0);
         if (auto *I = dyn_cast<Instruction>(V))
           V = propagateMetadata(I, E->Scalars);

llvm/test/Transforms/SLPVectorizer/revec-shufflevector.ll

@@ -105,3 +105,19 @@ entry:
   store <4 x i32> %2, ptr %6, align 4
   ret void
 }
+
+define void @test5(ptr %out) {
+; CHECK-LABEL: @test5(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    store <8 x i32> zeroinitializer, ptr [[OUT:%.*]], align 4
+; CHECK-NEXT:    ret void
+;
+entry:
+  %0 = shufflevector <8 x i32> zeroinitializer, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
+  %1 = shufflevector <8 x i32> zeroinitializer, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7>
+  %2 = getelementptr inbounds i32, ptr %out, i64 0
+  %3 = getelementptr inbounds i32, ptr %out, i64 4
+  store <4 x i32> %0, ptr %2, align 4
+  store <4 x i32> %1, ptr %3, align 4
+  ret void
+}