xla icon indicating copy to clipboard operation
xla copied to clipboard

[XLA:CPU] Support limiting LLVM codegen in Aarch64 and other new x86 instructions

Open penpornk opened this issue 1 year ago • 0 comments

PR https://github.com/openxla/xla/pull/17722 supports limiting the CPU ISA that LLVM will codegen. It currently only supports x86 ISAs from SSE4_2 up to AMX_FP16 and imposes a strict ordering (SSE4_2 < AVX < ... < AMX_FP16). We should

  1. Add Aarch64 support too.
  2. Investigate if we need to support AVX10 and it would go in the mix with other x86 instructions.

penpornk avatar Sep 30 '24 14:09 penpornk