Mismatched bfloat16 autogenerated files
Closed this issue · 3 comments
dzaima commented
Running make bf16-all
I get it generating incorrect reinterpret instructions like __riscv_reinterpret_v_bf16m8_i16m8
.
Note the lack of v
prefix there, which matches no v
in ["reinterpret"]
in bfloat16_inst.py
.
kito-cheng commented
@4vtomat could you take a look for this issue?
dzaima commented
Somewhat-relatedly, all _vf_
widening float multiply-add intrinsics, e.g. __riscv_vfwmacc_vf_f32m1
(not just, but including, bfloat16 ones) name their scalar arguments vs1
where I think it should be rs1
. This appears to have been the case for a while.