Andi Kleen <a...@linux.intel.com> 于2024年8月5日周一 06:31写道： > > > BTW, I noticed that in LLVM there is FP8 support for ARM currently > > undergoing. I will have a look on it to see if everything is mature. > > There's even FP8 work for ARM work under way for gcc, see > https://gcc.gnu.org/pipermail/gcc-patches/2024-August/659248.html >
The FP8 for ARM was defined as an opaque type used for intrinsic only. It behaves the same as int8 on reg move/load/store, but no conversion or scalar operation permitted. As we only have convert instruction for FP8 currently, I think there is no need to add such scalar type. We didn't add ABI and scalar type for BF16 until the arithmetic instruction was introduced. > > -andi
