[PATCH] D85553: [X86] Don't produce bad x86andp nodes for i1 vectors

[Keno Fischer via Phabricator via llvm-commits]

This revision was landed with ongoing or failed builds.
This revision was automatically updated to reflect the committed changes.
Closed by commit rGc58674df147a: [X86] Don't produce bad x86andp nodes for i1 vectors (authored by loladiro).

Changed prior to commit:
  https://reviews.llvm.org/D85553?vs=284028&id=284100#toc

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D85553/new/

https://reviews.llvm.org/D85553

Files:
  llvm/lib/Target/X86/X86ISelLowering.cpp
  llvm/test/CodeGen/X86/avx512-select.ll


Index: llvm/test/CodeGen/X86/avx512-select.ll
===================================================================
--- llvm/test/CodeGen/X86/avx512-select.ll
+++ llvm/test/CodeGen/X86/avx512-select.ll
@@ -705,3 +705,64 @@
   store <1 x i1> %c, <1 x i1>* %x
   ret void
 }
+
+; Regression test from https://github.com/JuliaLang/julia/issues/36955
+define i8 @julia_issue36955(<8 x i1> %mask, <8 x double> %a) {
+; X86-AVX512F-LABEL: julia_issue36955:
+; X86-AVX512F:       # %bb.0:
+; X86-AVX512F-NEXT:    vpmovsxwq %xmm0, %zmm0
+; X86-AVX512F-NEXT:    vpsllq $63, %zmm0, %zmm0
+; X86-AVX512F-NEXT:    vxorpd %xmm2, %xmm2, %xmm2
+; X86-AVX512F-NEXT:    vcmplepd %zmm2, %zmm1, %k1
+; X86-AVX512F-NEXT:    vptestmq %zmm0, %zmm0, %k0 {%k1}
+; X86-AVX512F-NEXT:    korw %k0, %k1, %k0
+; X86-AVX512F-NEXT:    kmovw %k0, %eax
+; X86-AVX512F-NEXT:    # kill: def $al killed $al killed $eax
+; X86-AVX512F-NEXT:    vzeroupper
+; X86-AVX512F-NEXT:    retl
+;
+; X64-AVX512F-LABEL: julia_issue36955:
+; X64-AVX512F:       # %bb.0:
+; X64-AVX512F-NEXT:    vpmovsxwq %xmm0, %zmm0
+; X64-AVX512F-NEXT:    vpsllq $63, %zmm0, %zmm0
+; X64-AVX512F-NEXT:    vxorpd %xmm2, %xmm2, %xmm2
+; X64-AVX512F-NEXT:    vcmplepd %zmm2, %zmm1, %k1
+; X64-AVX512F-NEXT:    vptestmq %zmm0, %zmm0, %k0 {%k1}
+; X64-AVX512F-NEXT:    korw %k0, %k1, %k0
+; X64-AVX512F-NEXT:    kmovw %k0, %eax
+; X64-AVX512F-NEXT:    # kill: def $al killed $al killed $eax
+; X64-AVX512F-NEXT:    vzeroupper
+; X64-AVX512F-NEXT:    retq
+;
+; X86-AVX512BW-LABEL: julia_issue36955:
+; X86-AVX512BW:       # %bb.0:
+; X86-AVX512BW-NEXT:    vpsllw $15, %xmm0, %xmm0
+; X86-AVX512BW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; X86-AVX512BW-NEXT:    vxorpd %xmm3, %xmm3, %xmm3
+; X86-AVX512BW-NEXT:    vcmplepd %zmm3, %zmm1, %k1
+; X86-AVX512BW-NEXT:    vpcmpgtw %zmm0, %zmm2, %k0 {%k1}
+; X86-AVX512BW-NEXT:    korw %k0, %k1, %k0
+; X86-AVX512BW-NEXT:    kmovd %k0, %eax
+; X86-AVX512BW-NEXT:    # kill: def $al killed $al killed $eax
+; X86-AVX512BW-NEXT:    vzeroupper
+; X86-AVX512BW-NEXT:    retl
+;
+; X64-AVX512BW-LABEL: julia_issue36955:
+; X64-AVX512BW:       # %bb.0:
+; X64-AVX512BW-NEXT:    vpsllw $15, %xmm0, %xmm0
+; X64-AVX512BW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; X64-AVX512BW-NEXT:    vxorpd %xmm3, %xmm3, %xmm3
+; X64-AVX512BW-NEXT:    vcmplepd %zmm3, %zmm1, %k1
+; X64-AVX512BW-NEXT:    vpcmpgtw %zmm0, %zmm2, %k0 {%k1}
+; X64-AVX512BW-NEXT:    korw %k0, %k1, %k0
+; X64-AVX512BW-NEXT:    kmovd %k0, %eax
+; X64-AVX512BW-NEXT:    # kill: def $al killed $al killed $eax
+; X64-AVX512BW-NEXT:    vzeroupper
+; X64-AVX512BW-NEXT:    retq
+  %fcmp = fcmp ugt <8 x double> %a, zeroinitializer
+  %xor = xor <8 x i1> %fcmp, <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>
+  %select1 = select <8 x i1> %fcmp, <8 x i1> zeroinitializer, <8 x i1> %mask
+  %select2 = select <8 x i1> %xor, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i1> %select1
+  %ret = bitcast <8 x i1> %select2 to i8
+  ret i8 %ret
+}
Index: llvm/lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- llvm/lib/Target/X86/X86ISelLowering.cpp
+++ llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -39633,10 +39633,14 @@
 
   // vselect Cond, 000..., X -> andn Cond, X
   if (TValIsAllZeros) {
-    MVT AndNVT = MVT::getVectorVT(MVT::i64, CondVT.getSizeInBits() / 64);
-    SDValue CastCond = DAG.getBitcast(AndNVT, Cond);
-    SDValue CastRHS = DAG.getBitcast(AndNVT, RHS);
-    SDValue AndN = DAG.getNode(X86ISD::ANDNP, DL, AndNVT, CastCond, CastRHS);
+    SDValue CastRHS = DAG.getBitcast(CondVT, RHS);
+    SDValue AndN;
+    // The canonical form differs for i1 vectors - x86andnp is not used
+    if (CondVT.getScalarType() == MVT::i1)
+      AndN = DAG.getNode(ISD::AND, DL, CondVT, DAG.getNOT(DL, Cond, CondVT),
+                         CastRHS);
+    else
+      AndN = DAG.getNode(X86ISD::ANDNP, DL, CondVT, Cond, CastRHS);
     return DAG.getBitcast(VT, AndN);
   }
 


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D85553.284100.patch
Type: text/x-patch
Size: 4011 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20200808/5fbc61bb/attachment.bin>