[SLP]Fix a crash when trying to find reduced ops for the reduced value.

Need to use original reduced value, not the one the compiler gets after
reduction, it may be replaced by the extractelement instruction already.

Author: alexey-bataev

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

@@ -13043,7 +13043,7 @@ class HorizontalReduction {
           Value *OrigV = TrackedToOrig.find(RdxVal)->second;
           unsigned NumOps =
               VectorizedVals.lookup(RdxVal) + SameValuesCounter[OrigV];
-          if (NumOps != ReducedValsToOps.find(RdxVal)->second.size())
+          if (NumOps != ReducedValsToOps.find(OrigV)->second.size())
             LocalExternallyUsedValues[RdxVal];
         }
         // Do not need the list of reused scalars in regular mode anymore.

llvm/test/Transforms/SLPVectorizer/X86/reduction-extracted-value.ll

@@ -0,0 +1,58 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu -mattr=+avx < %s | FileCheck %s
+
+define i32 @foo() {
+; CHECK-LABEL: @foo(
+; CHECK-NEXT:  bb:
+; CHECK-NEXT:    [[TMP0:%.*]] = extractelement <2 x i32> zeroinitializer, i32 0
+; CHECK-NEXT:    [[TMP1:%.*]] = or <4 x i32> zeroinitializer, zeroinitializer
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i32> [[TMP1]], i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = mul <4 x i32> [[TMP1]], zeroinitializer
+; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i32> [[TMP3]], zeroinitializer
+; CHECK-NEXT:    [[TMP5:%.*]] = call i32 @llvm.vector.reduce.mul.v4i32(<4 x i32> [[TMP4]])
+; CHECK-NEXT:    [[OP_RDX:%.*]] = mul i32 0, [[TMP5]]
+; CHECK-NEXT:    [[OP_RDX1:%.*]] = mul i32 [[OP_RDX]], 0
+; CHECK-NEXT:    [[OP_RDX2:%.*]] = mul i32 [[TMP0]], [[TMP0]]
+; CHECK-NEXT:    [[OP_RDX3:%.*]] = mul i32 [[TMP0]], [[TMP0]]
+; CHECK-NEXT:    [[OP_RDX4:%.*]] = mul i32 [[OP_RDX1]], [[OP_RDX2]]
+; CHECK-NEXT:    [[OP_RDX5:%.*]] = mul i32 [[OP_RDX3]], [[TMP2]]
+; CHECK-NEXT:    [[OP_RDX6:%.*]] = mul i32 [[OP_RDX4]], [[OP_RDX5]]
+; CHECK-NEXT:    ret i32 [[OP_RDX6]]
+;
+bb:
+  %inst5 = add i32 0, 0
+  %0 = extractelement <2 x i32> zeroinitializer, i32 0
+  %inst7 = mul i32 %0, %inst5
+  %1 = extractelement <2 x i32> zeroinitializer, i32 0
+  %inst13 = mul i32 %1, %inst7
+  %inst14 = mul i32 %inst13, 0
+  %2 = extractelement <2 x i32> zeroinitializer, i32 0
+  %inst19 = mul i32 %2, %inst14
+  %inst20 = mul i32 %inst19, 0
+  %3 = extractelement <2 x i32> zeroinitializer, i32 0
+  %inst26 = mul i32 %3, %inst20
+  %inst27 = mul i32 %inst26, 0
+  %4 = or <4 x i32> zeroinitializer, zeroinitializer
+  %5 = extractelement <4 x i32> %4, i32 0
+  %inst31 = mul i32 %5, 0
+  %inst32 = add i32 %inst31, 0
+  %inst33 = mul i32 %5, %inst27
+  %inst34 = mul i32 %inst33, %inst32
+  %6 = extractelement <4 x i32> %4, i32 1
+  %inst39 = mul i32 %6, 0
+  %inst40 = add i32 %inst39, 0
+  %inst41 = mul i32 0, %inst34
+  %inst42 = mul i32 %inst41, %inst40
+  %7 = extractelement <4 x i32> %4, i32 2
+  %inst47 = mul i32 %7, 0
+  %inst48 = add i32 %inst47, 0
+  %inst49 = mul i32 0, %inst42
+  %inst50 = mul i32 %inst49, %inst48
+  %8 = extractelement <4 x i32> %4, i32 3
+  %inst55 = mul i32 %8, 0
+  %inst56 = add i32 %inst55, 0
+  %inst57 = mul i32 0, %inst50
+  %inst58 = mul i32 %inst57, %inst56
+  ret i32 %inst58
+}
+