[SLp]Fix a crash because of wrong deps between vectorized nodes.

Need to change the order of the nodes vectorization to avoid too early
insertion of the first node.

Author: alexey-bataev

llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp

@@ -8959,7 +8959,7 @@ BoUpSLP::isGatherShuffledEntry(const TreeEntry *TE, ArrayRef<Value *> VL,
         // vectorized nodes - make it depend on index.
         if (TE->UserTreeIndices.front().UserTE !=
                 TEPtr->UserTreeIndices.front().UserTE &&
-            TE->Idx > TEPtr->Idx)
+            TE->Idx < TEPtr->Idx)
           continue;
       }
       // Check if the user node of the TE comes after user node of EntryPtr,

llvm/test/Transforms/SLPVectorizer/X86/multi-uses-with-deps-in-first.ll

@@ -0,0 +1,40 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 3
+; RUN: opt -S --passes=slp-vectorizer -mtriple=x86_64-pc-linux-gnu < %s | FileCheck %s
+
+define void @test(double %add) {
+; CHECK-LABEL: define void @test(
+; CHECK-SAME: double [[ADD:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x double> <double poison, double 0.000000e+00>, double [[ADD]], i32 0
+; CHECK-NEXT:    [[TMP1:%.*]] = fmul <2 x double> [[TMP0]], zeroinitializer
+; CHECK-NEXT:    [[TMP2:%.*]] = fmul <2 x double> [[TMP0]], zeroinitializer
+; CHECK-NEXT:    br label [[COND_TRUE45:%.*]]
+; CHECK:       cond.true45:
+; CHECK-NEXT:    [[TMP3:%.*]] = fmul <2 x double> [[TMP1]], zeroinitializer
+; CHECK-NEXT:    [[TMP4:%.*]] = fmul <2 x double> [[TMP2]], zeroinitializer
+; CHECK-NEXT:    [[TMP5:%.*]] = fadd <2 x double> zeroinitializer, [[TMP3]]
+; CHECK-NEXT:    [[TMP6:%.*]] = fsub <2 x double> [[TMP5]], zeroinitializer
+; CHECK-NEXT:    [[TMP7:%.*]] = fsub <2 x double> [[TMP6]], [[TMP4]]
+; CHECK-NEXT:    ret void
+;
+entry:
+  %mul1 = fmul double %add, 0.000000e+00
+  %mul2 = fmul double %add, 0.000000e+00
+  %mul3 = fmul double 0.000000e+00, 0.000000e+00
+  br label %cond.true45
+
+cond.true45:
+  %mul26 = fmul double %mul1, 0.000000e+00
+  %add27 = fadd double 0.000000e+00, %mul26
+  %sub28 = fsub double %add27, 0.000000e+00
+  %mul30 = fmul double %mul2, 0.000000e+00
+  %sub31 = fsub double %sub28, %mul30
+  %mul7 = fmul double %mul3, 0.000000e+00
+  %add8 = fadd double %mul7, 0.000000e+00
+  %sub9 = fsub double %add8, 0.000000e+00
+  %mul11 = fmul double %mul3, 0.000000e+00
+  %sub12 = fsub double %sub9, %mul11
+  %0 = insertelement <2 x double> zeroinitializer, double %sub31, i32 0
+  %1 = insertelement <2 x double> %0, double %sub12, i32 1
+  ret void
+}