[SLP] Fix crash on propagate IR flags for undef operands of min/max

reductions.

If both operands of the newly created SelectInst are Undefs the
resulting operation is also Undef, not SelectInst. It may cause crashes
when trying to propagate IR flags because function expects exactly
SelectInst instruction, nothing else.

llvm-svn: 314323

2 files changed, 92 insertions, 3 deletions

diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 7f767dcc7d6..d201387debd 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -4938,7 +4938,8 @@ class HorizontalReduction {
       case RK_Max:
       case RK_UMin:
       case RK_UMax:
-        propagateIRFlags(cast<SelectInst>(Op)->getCondition(), ReductionOps[0]);
+        if (auto *SI = dyn_cast<SelectInst>(Op))
+          propagateIRFlags(SI->getCondition(), ReductionOps[0]);
         propagateIRFlags(Op, ReductionOps[1]);
         return Op;
       case RK_None:
@@ -4961,8 +4962,10 @@ class HorizontalReduction {
       case RK_Max:
       case RK_UMin:
       case RK_UMax:
-        propagateIRFlags(cast<SelectInst>(Op)->getCondition(),
-                         cast<SelectInst>(I)->getCondition());
+        if (auto *SI = dyn_cast<SelectInst>(Op)) {
+          propagateIRFlags(SI->getCondition(),
+                           cast<SelectInst>(I)->getCondition());
+        }
         propagateIRFlags(Op, I);
         return Op;
       case RK_None:
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/undef_vect.ll b/llvm/test/Transforms/SLPVectorizer/X86/undef_vect.ll
new file mode 100644
index 00000000000..7ecd5805f93
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/X86/undef_vect.ll
@@ -0,0 +1,86 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -slp-vectorizer -slp-vectorize-hor -slp-vectorize-hor-store -S < %s -mtriple=x86_64-unknown-linux-gnu -mcpu=bdver2 | FileCheck %s
+
+%"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76" = type { i32, i32 }
+
+define void @_Z2azv() local_unnamed_addr {
+; CHECK-LABEL: @_Z2azv(
+; CHECK-NEXT:  for.body.lr.ph:
+; CHECK-NEXT:    [[DOTSROA_CAST_4:%.*]] = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 4, i32 0
+; CHECK-NEXT:    [[DOTSROA_RAW_IDX_4:%.*]] = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 4, i32 1
+; CHECK-NEXT:    [[DOTSROA_CAST_5:%.*]] = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 5, i32 0
+; CHECK-NEXT:    [[DOTSROA_RAW_IDX_5:%.*]] = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 5, i32 1
+; CHECK-NEXT:    [[DOTSROA_CAST_6:%.*]] = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 6, i32 0
+; CHECK-NEXT:    [[DOTSROA_RAW_IDX_6:%.*]] = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 6, i32 1
+; CHECK-NEXT:    [[DOTSROA_CAST_7:%.*]] = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 7, i32 0
+; CHECK-NEXT:    [[DOTSROA_RAW_IDX_7:%.*]] = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 7, i32 1
+; CHECK-NEXT:    [[TMP0:%.*]] = bitcast i32* [[DOTSROA_CAST_4]] to <8 x i32>*
+; CHECK-NEXT:    [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* [[TMP0]], align 4
+; CHECK-NEXT:    [[CMP_I1_4:%.*]] = icmp slt i32 undef, undef
+; CHECK-NEXT:    [[DOTSROA_SPECULATED_4:%.*]] = select i1 [[CMP_I1_4]], i32 undef, i32 undef
+; CHECK-NEXT:    [[CMP_I1_5:%.*]] = icmp slt i32 [[DOTSROA_SPECULATED_4]], undef
+; CHECK-NEXT:    [[DOTSROA_SPECULATED_5:%.*]] = select i1 [[CMP_I1_5]], i32 undef, i32 [[DOTSROA_SPECULATED_4]]
+; CHECK-NEXT:    [[CMP_I1_6:%.*]] = icmp slt i32 [[DOTSROA_SPECULATED_5]], undef
+; CHECK-NEXT:    [[DOTSROA_SPECULATED_6:%.*]] = select i1 [[CMP_I1_6]], i32 undef, i32 [[DOTSROA_SPECULATED_5]]
+; CHECK-NEXT:    [[CMP_I1_7:%.*]] = icmp slt i32 [[DOTSROA_SPECULATED_6]], undef
+; CHECK-NEXT:    [[DOTSROA_SPECULATED_7:%.*]] = select i1 [[CMP_I1_7]], i32 undef, i32 [[DOTSROA_SPECULATED_6]]
+; CHECK-NEXT:    [[CMP_I1_8:%.*]] = icmp slt i32 undef, undef
+; CHECK-NEXT:    [[RDX_SHUF:%.*]] = shufflevector <8 x i32> [[TMP1]], <8 x i32> undef, <8 x i32> <i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[RDX_MINMAX_CMP:%.*]] = icmp sgt <8 x i32> [[TMP1]], [[RDX_SHUF]]
+; CHECK-NEXT:    [[RDX_MINMAX_SELECT:%.*]] = select <8 x i1> [[RDX_MINMAX_CMP]], <8 x i32> [[TMP1]], <8 x i32> [[RDX_SHUF]]
+; CHECK-NEXT:    [[RDX_SHUF1:%.*]] = shufflevector <8 x i32> [[RDX_MINMAX_SELECT]], <8 x i32> undef, <8 x i32> <i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[RDX_MINMAX_CMP2:%.*]] = icmp sgt <8 x i32> [[RDX_MINMAX_SELECT]], [[RDX_SHUF1]]
+; CHECK-NEXT:    [[RDX_MINMAX_SELECT3:%.*]] = select <8 x i1> [[RDX_MINMAX_CMP2]], <8 x i32> [[RDX_MINMAX_SELECT]], <8 x i32> [[RDX_SHUF1]]
+; CHECK-NEXT:    [[RDX_SHUF4:%.*]] = shufflevector <8 x i32> [[RDX_MINMAX_SELECT3]], <8 x i32> undef, <8 x i32> <i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>
+; CHECK-NEXT:    [[RDX_MINMAX_CMP5:%.*]] = icmp sgt <8 x i32> [[RDX_MINMAX_SELECT3]], [[RDX_SHUF4]]
+; CHECK-NEXT:    [[RDX_MINMAX_SELECT6:%.*]] = select <8 x i1> [[RDX_MINMAX_CMP5]], <8 x i32> [[RDX_MINMAX_SELECT3]], <8 x i32> [[RDX_SHUF4]]
+; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <8 x i32> [[RDX_MINMAX_SELECT6]], i32 0
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp sgt i32 [[TMP2]], undef
+; CHECK-NEXT:    [[OP_EXTRA:%.*]] = select i1 [[TMP3]], i32 [[TMP2]], i32 undef
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp sgt i32 [[OP_EXTRA]], undef
+; CHECK-NEXT:    [[OP_EXTRA7:%.*]] = select i1 [[TMP4]], i32 [[OP_EXTRA]], i32 undef
+; CHECK-NEXT:    [[DOTSROA_SPECULATED_8:%.*]] = select i1 [[CMP_I1_8]], i32 undef, i32 undef
+; CHECK-NEXT:    [[DOTSROA_SPECULATED_9:%.*]] = select i1 undef, i32 undef, i32 [[OP_EXTRA7]]
+; CHECK-NEXT:    [[CMP_I1_10:%.*]] = icmp slt i32 [[DOTSROA_SPECULATED_9]], undef
+; CHECK-NEXT:    ret void
+;
+for.body.lr.ph:
+  %.sroa_cast.4 = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 4, i32 0
+  %retval.sroa.0.0.copyload.i5.4 = load i32, i32* %.sroa_cast.4, align 4
+  %.sroa_raw_idx.4 = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 4, i32 1
+  %retval.sroa.0.0.copyload.i7.4 = load i32, i32* %.sroa_raw_idx.4, align 4
+  %cmp.i2.4 = icmp slt i32 %retval.sroa.0.0.copyload.i5.4, %retval.sroa.0.0.copyload.i7.4
+  %0 = select i1 %cmp.i2.4, i32 %retval.sroa.0.0.copyload.i7.4, i32 %retval.sroa.0.0.copyload.i5.4
+  %cmp.i1.4 = icmp slt i32 undef, %0
+  %.sroa.speculated.4 = select i1 %cmp.i1.4, i32 %0, i32 undef
+  %.sroa_cast.5 = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 5, i32 0
+  %retval.sroa.0.0.copyload.i5.5 = load i32, i32* %.sroa_cast.5, align 4
+  %.sroa_raw_idx.5 = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 5, i32 1
+  %retval.sroa.0.0.copyload.i7.5 = load i32, i32* %.sroa_raw_idx.5, align 4
+  %cmp.i2.5 = icmp slt i32 %retval.sroa.0.0.copyload.i5.5, %retval.sroa.0.0.copyload.i7.5
+  %1 = select i1 %cmp.i2.5, i32 %retval.sroa.0.0.copyload.i7.5, i32 %retval.sroa.0.0.copyload.i5.5
+  %cmp.i1.5 = icmp slt i32 %.sroa.speculated.4, %1
+  %.sroa.speculated.5 = select i1 %cmp.i1.5, i32 %1, i32 %.sroa.speculated.4
+  %.sroa_cast.6 = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 6, i32 0
+  %retval.sroa.0.0.copyload.i5.6 = load i32, i32* %.sroa_cast.6, align 4
+  %.sroa_raw_idx.6 = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 6, i32 1
+  %retval.sroa.0.0.copyload.i7.6 = load i32, i32* %.sroa_raw_idx.6, align 4
+  %cmp.i2.6 = icmp slt i32 %retval.sroa.0.0.copyload.i5.6, %retval.sroa.0.0.copyload.i7.6
+  %2 = select i1 %cmp.i2.6, i32 %retval.sroa.0.0.copyload.i7.6, i32 %retval.sroa.0.0.copyload.i5.6
+  %cmp.i1.6 = icmp slt i32 %.sroa.speculated.5, %2
+  %.sroa.speculated.6 = select i1 %cmp.i1.6, i32 %2, i32 %.sroa.speculated.5
+  %.sroa_cast.7 = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 7, i32 0
+  %retval.sroa.0.0.copyload.i5.7 = load i32, i32* %.sroa_cast.7, align 4
+  %.sroa_raw_idx.7 = getelementptr inbounds %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76", %"struct.std::h.0.4.8.12.16.20.24.28.248.0.1.2.3.76"* undef, i64 7, i32 1
+  %retval.sroa.0.0.copyload.i7.7 = load i32, i32* %.sroa_raw_idx.7, align 4
+  %cmp.i2.7 = icmp slt i32 %retval.sroa.0.0.copyload.i5.7, %retval.sroa.0.0.copyload.i7.7
+  %3 = select i1 %cmp.i2.7, i32 %retval.sroa.0.0.copyload.i7.7, i32 %retval.sroa.0.0.copyload.i5.7
+  %cmp.i1.7 = icmp slt i32 %.sroa.speculated.6, %3
+  %.sroa.speculated.7 = select i1 %cmp.i1.7, i32 %3, i32 %.sroa.speculated.6
+  %cmp.i1.8 = icmp slt i32 %.sroa.speculated.7, undef
+  %.sroa.speculated.8 = select i1 %cmp.i1.8, i32 undef, i32 %.sroa.speculated.7
+  %.sroa.speculated.9 = select i1 undef, i32 undef, i32 %.sroa.speculated.8
+  %cmp.i1.10 = icmp slt i32 %.sroa.speculated.9, undef
+  ret void
+}
+