[SLP]Check the whole def-use chain in the tree to find proper dominance, if the last instruction is the same

If the insertion point (last instruction) of the user nodes is the same,
need to check the whole def-use chain in the tree to find proper
dominance to prevent a compiler crash.

Fixes #131818

Author: alexey-bataev

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

@@ -14086,11 +14086,18 @@ BoUpSLP::isGatherShuffledSingleRegisterEntry(
   auto CheckParentNodes = [&](const TreeEntry *User1, const TreeEntry *User2,
                               unsigned EdgeIdx) {
     const TreeEntry *Ptr1 = User1;
+    const TreeEntry *Ptr2 = User2;
+    SmallDenseMap<const TreeEntry *, unsigned> PtrToIdx;
+    while (Ptr2) {
+      PtrToIdx.try_emplace(Ptr2, EdgeIdx);
+      EdgeIdx = Ptr2->UserTreeIndex.EdgeIdx;
+      Ptr2 = Ptr2->UserTreeIndex.UserTE;
+    }
     while (Ptr1) {
       unsigned Idx = Ptr1->UserTreeIndex.EdgeIdx;
       Ptr1 = Ptr1->UserTreeIndex.UserTE;
-      if (Ptr1 == User2)
-        return Idx < EdgeIdx;
+      if (auto It = PtrToIdx.find(Ptr1); It != PtrToIdx.end())
+        return Idx < It->second;
     }
     return false;
   };

llvm/test/Transforms/SLPVectorizer/X86/same-last-instruction-different-parents.ll

@@ -0,0 +1,64 @@
+; 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 < %s | FileCheck %s
+
+define i32 @test(i32 %0, i1 %1) {
+; CHECK-LABEL: define i32 @test(
+; CHECK-SAME: i32 [[TMP0:%.*]], i1 [[TMP1:%.*]]) {
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i32> poison, i32 [[TMP0]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <2 x i32> [[TMP3]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP5:%.*]] = sitofp <2 x i32> [[TMP4]] to <2 x double>
+; CHECK-NEXT:    [[TMP6:%.*]] = sitofp <2 x i32> [[TMP4]] to <2 x double>
+; CHECK-NEXT:    br i1 [[TMP1]], label %[[BB7:.*]], label %[[BB9:.*]]
+; CHECK:       [[BB7]]:
+; CHECK-NEXT:    [[TMP8:%.*]] = call <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[TMP6]], <2 x double> zeroinitializer, <2 x double> zeroinitializer)
+; CHECK-NEXT:    br label %[[BB16:.*]]
+; CHECK:       [[BB9]]:
+; CHECK-NEXT:    br i1 false, label %[[BB14:.*]], label %[[BB10:.*]]
+; CHECK:       [[BB10]]:
+; CHECK-NEXT:    [[TMP11:%.*]] = call <2 x double> @llvm.copysign.v2f64(<2 x double> zeroinitializer, <2 x double> [[TMP5]])
+; CHECK-NEXT:    [[TMP12:%.*]] = shufflevector <2 x double> [[TMP6]], <2 x double> <double 0.000000e+00, double poison>, <2 x i32> <i32 2, i32 1>
+; CHECK-NEXT:    [[TMP13:%.*]] = call <2 x double> @llvm.fmuladd.v2f64(<2 x double> [[TMP11]], <2 x double> [[TMP12]], <2 x double> zeroinitializer)
+; CHECK-NEXT:    br label %[[BB14]]
+; CHECK:       [[BB14]]:
+; CHECK-NEXT:    [[TMP15:%.*]] = phi <2 x double> [ [[TMP13]], %[[BB10]] ], [ zeroinitializer, %[[BB9]] ]
+; CHECK-NEXT:    br label %[[BB16]]
+; CHECK:       [[BB16]]:
+; CHECK-NEXT:    [[TMP17:%.*]] = phi <2 x double> [ [[TMP15]], %[[BB14]] ], [ [[TMP8]], %[[BB7]] ]
+; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <2 x double> [[TMP17]], i32 0
+; CHECK-NEXT:    [[TMP19:%.*]] = extractelement <2 x double> [[TMP17]], i32 1
+; CHECK-NEXT:    [[TMP20:%.*]] = fmul double [[TMP19]], [[TMP18]]
+; CHECK-NEXT:    [[TMP21:%.*]] = fptosi double [[TMP20]] to i32
+; CHECK-NEXT:    ret i32 [[TMP21]]
+;
+  %3 = sitofp i32 %0 to double
+  %4 = sitofp i32 %0 to double
+  %5 = sitofp i32 %0 to double
+  br i1 %1, label %6, label %9
+
+6:
+  %7 = tail call double @llvm.fmuladd.f64(double %4, double 0.000000e+00, double 0.000000e+00)
+  %8 = tail call double @llvm.fmuladd.f64(double %5, double 0.000000e+00, double 0.000000e+00)
+  br label %18
+
+9:
+  br i1 false, label %15, label %10
+
+10:
+  %11 = call double @llvm.copysign.f64(double 0.000000e+00, double %3)
+  %12 = tail call double @llvm.fmuladd.f64(double %11, double %4, double 0.000000e+00)
+  %13 = call double @llvm.copysign.f64(double 0.000000e+00, double %5)
+  %14 = tail call double @llvm.fmuladd.f64(double %13, double 0.000000e+00, double 0.000000e+00)
+  br label %15
+
+15:
+  %16 = phi double [ %12, %10 ], [ 0.000000e+00, %9 ]
+  %17 = phi double [ %14, %10 ], [ 0.000000e+00, %9 ]
+  br label %18
+
+18:
+  %19 = phi double [ %17, %15 ], [ %8, %6 ]
+  %20 = phi double [ %16, %15 ], [ %7, %6 ]
+  %21 = fmul double %20, %19
+  %22 = fptosi double %21 to i32
+  ret i32 %22
+}