[PATCH] D71592: [X86] Enable STRICT_FP_TO_SINT/UINT on X86 backend.

LiuChen via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Dec 18 16:52:55 PST 2019


LiuChen3 marked an inline comment as done and an inline comment as not done.
LiuChen3 added inline comments.


================
Comment at: llvm/test/CodeGen/X86/vec-strict-fptoint-128.ll:397
+; AVX512DQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512DQ-NEXT:    vcvttpd2uqq %zmm0, %zmm0
+; AVX512DQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
----------------
craig.topper wrote:
> LiuChen3 wrote:
> > pengfei wrote:
> > > Just noticed: is this an unsafe behavior for strict FP, because the upper elements may cause exceptions?
> > Maybe we can set these strict-fptoint legal only when target supports avx512vl  and avx512dq?
> Can we use 0.0 to pad instead of undef?
I think it's reasonable. I'll work on it.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71592/new/

https://reviews.llvm.org/D71592





More information about the llvm-commits mailing list