[llvm] 088a816 - [DAGCombiner] Use `getAnyExtOrTrunc` instead of TRUNCATE in ExtractVectorElt combine

[Pierre van Houtryve via llvm-commits]

Author: Pierre van Houtryve
Date: 2022-10-28T06:33:29Z
New Revision: 088a81682444839e58ed5deaad50175d956db777

URL: https://github.com/llvm/llvm-project/commit/088a81682444839e58ed5deaad50175d956db777
DIFF: https://github.com/llvm/llvm-project/commit/088a81682444839e58ed5deaad50175d956db777.diff

LOG: [DAGCombiner] Use `getAnyExtOrTrunc` instead of TRUNCATE in ExtractVectorElt combine

ScalarVT isn't guaranteed to be smaller than the BCSrc.

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D136849

Added: 
    llvm/test/CodeGen/AMDGPU/dagcombine-v1i8-extractvecelt-crash.ll

Modified: 
    llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
index b35fe4c224add..2e09f48f52386 100644
--- a/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
@@ -20258,7 +20258,7 @@ SDValue DAGCombiner::visitEXTRACT_VECTOR_ELT(SDNode *N) {
     unsigned BCTruncElt = IsLE ? 0 : NumElts - 1;
     SDValue BCSrc = VecOp.getOperand(0);
     if (ExtractIndex == BCTruncElt && BCSrc.getValueType().isScalarInteger())
-      return DAG.getNode(ISD::TRUNCATE, DL, ScalarVT, BCSrc);
+      return DAG.getAnyExtOrTrunc(BCSrc, DL, ScalarVT);
 
     if (LegalTypes && BCSrc.getValueType().isInteger() &&
         BCSrc.getOpcode() == ISD::SCALAR_TO_VECTOR) {

diff  --git a/llvm/test/CodeGen/AMDGPU/dagcombine-v1i8-extractvecelt-crash.ll b/llvm/test/CodeGen/AMDGPU/dagcombine-v1i8-extractvecelt-crash.ll
new file mode 100644
index 0000000000000..fee9e03c360f7
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/dagcombine-v1i8-extractvecelt-crash.ll
@@ -0,0 +1,33 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -mcpu=gfx908 < %s | FileCheck %s
+
+define void @wombat(i1 %cond, <1 x i8> addrspace(5)* %addr) {
+; CHECK-LABEL: wombat:
+; CHECK:       ; %bb.0: ; %entry
+; CHECK-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; CHECK-NEXT:    buffer_load_ubyte v2, v1, s[0:3], 0 offen
+; CHECK-NEXT:    v_and_b32_e32 v0, 1, v0
+; CHECK-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v0
+; CHECK-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; CHECK-NEXT:    s_cbranch_execz .LBB0_2
+; CHECK-NEXT:  ; %bb.1: ; %then
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    v_mov_b32_e32 v2, 0
+; CHECK-NEXT:  .LBB0_2: ; %end
+; CHECK-NEXT:    s_or_b64 exec, exec, s[4:5]
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    buffer_store_byte v2, v1, s[0:3], 0 offen
+; CHECK-NEXT:    s_waitcnt vmcnt(0)
+; CHECK-NEXT:    s_setpc_b64 s[30:31]
+entry:
+  %load = load <1 x i8>, <1 x i8> addrspace(5)* %addr, align 1
+  br i1 %cond, label %then, label %end
+
+then:
+  br label %end
+
+end:
+  %phi_value = phi <1 x i8> [%load, %entry], [zeroinitializer, %then]
+  store <1 x i8> %phi_value, <1 x i8> addrspace(5)* %addr, align 1
+  ret void
+}