The test data (in avx2neontestdata.h) of the following intrinsic are incorrect.
MM512_CMP_PD_MASK
MM256_CMP_PD
MM256_CMP_PS
MM512_CMP_PS_MASK
I think it is because the interpretation of imm8 is inconsistent with the corresponding intrinsic of AVX. For example, in _mm512_cmp_pd_mask when when imm8=_CMP_EQ_UQ,
The test data (in
avx2neontestdata.h
) of the following intrinsic are incorrect.imm8
is inconsistent with the corresponding intrinsic of AVX. For example, in_mm512_cmp_pd_mask
whenwhen imm8=_CMP_EQ_UQ
,AvxToSve gives 81 whereas AVX gives 0, indicating that the comparing results of elements at index 0, 4, 6 are incorrect. And it is where the
NAN
sits.The experiment is performed using Intel icx compiler and Intel sde (emulate tigerlake).