[PATCH] D134897: [AMDGPU] Fix V_CMP_CLASS_F16_t16_e64 src1 type.

Dmitry Preobrazhensky via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu Sep 29 13:34:35 PDT 2022


dp added a comment.

In D134897#3824888 <https://reviews.llvm.org/D134897#3824888>, @arsenm wrote:

> 5f5f566b265d <https://reviews.llvm.org/rG5f5f566b265db00f577ead268400d99f34ba9cdd> is really a hardware bug, so it's possible it's been fixed on newer hardware (don't think it was ever properly communicated as one to fix though)

Last time I read SP3 documentation for GFX11, this issue was presented as a feature, not a bug. There was also a special flag to label opcodes which correctly handled fp inline constants for 16 bit integer operands.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D134897/new/

https://reviews.llvm.org/D134897



More information about the llvm-commits mailing list