[PATCH] D60403: [CostModel][X86] Add bool anyof/allof reduction costs

[Sanjay Patel via Phabricator via llvm-commits]

spatel added inline comments.


================
Comment at: lib/Target/X86/X86TargetTransformInfo.cpp:2493
+  static const CostTblEntry AVX2BoolReduction[] = {
+    { ISD::AND,  MVT::v16i16,  2 },
+    { ISD::AND,  MVT::v32i8,   2 },
----------------
Add a code comment here with the expected lowering?
  vpmovmskb %xmm, %eax
  cmpb  $IMM, %al



================
Comment at: lib/Target/X86/X86TargetTransformInfo.cpp:2502
+    { ISD::AND,  MVT::v8i32,   2 },
+    { ISD::AND,  MVT::v16i16,  4 },
+    { ISD::AND,  MVT::v32i8,   4 },
----------------
Similarly:
  vextractf128 %ymm0, %xmm1, 1
  vpand  %xmm1, %xmm0, %xmm0 
  vpmovmskb %xmm0, %eax
  cmpb  $IMM, %al


================
Comment at: lib/Target/X86/X86TargetTransformInfo.cpp:2523
+  // Lowered to MOVMSK+CMP on pre-AVX512 targets.
+  if (ValTy->getScalarSizeInBits() == 1) {
+    if (ST->hasAVX2())
----------------
Better to make this 'isIntegerTy(1)'? We should not have a vector type here?


Repository:
  rL LLVM

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D60403/new/

https://reviews.llvm.org/D60403