[PATCH] D130151: [X86][FP16] Do not split FP64->FP16 to FP64->FP32->FP16

[Phoebe Wang via Phabricator via llvm-commits]

pengfei added inline comments.


================
Comment at: llvm/test/CodeGen/X86/half-constrained.ll:208
+; X32-F16C-NEXT:    vmovq %xmm0, (%esp)
+; X32-F16C-NEXT:    calll ___truncdfhf2
+; X32-F16C-NEXT:    vpextrw $0, %xmm0, _a
----------------
LuoYuanke wrote:
> Just be curious. Why there are 4 underscore? Is it the right function name?
This exists before the FP16 patches. Notice the tests are for darwin platform. Not sure if it has special mangling.


================
Comment at: llvm/test/CodeGen/X86/vector-half-conversions.ll:727
 ; AVX2-NEXT:    vcvtpd2ps %ymm0, %xmm0
 ; AVX2-NEXT:    vcvtps2ph $0, %xmm0, %xmm0
 ; AVX2-NEXT:    vmovlhps {{.*#+}} xmm0 = xmm0[0],xmm1[0]
----------------
LuoYuanke wrote:
> It seems transform from double to float and then float to half. The same for AVX1.
Good catch! Will investigate.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D130151/new/

https://reviews.llvm.org/D130151