[Nasm-devel] AVX512-FP16 instructions
Agner Fog
agner at agner.org
Thu Apr 28 02:04:02 PDT 2022
Hi nasm developers
There are 110 new instructions for half-precision floating point
calculations in the instruction set extension AVX512-FP16:
https://www.intel.com/content/www/us/en/develop/download/intel-avx512-fp16-architecture-specification.html
These instructions are using a hitherto unused m bit in the EVEX prefix
to indicate half precision.
I hope that you will add support for these instructions in NASM.
These instructions are supported by Gcc, Clang, and my disassembler (
https://www.agner.org/optimize/#objconv )
- Agner
More information about the Nasm-devel
mailing list