On Mon, Nov 9, 2020 at 11:31 AM Hongtao Liu <crazy...@gmail.com> wrote: > > > > > + /* Support unified builtin. */ > > + || (mask2 == OPTION_MASK_ISA2_AVXVNNI) > > > > I don't think we gain anything with unified builtins. Better, just > > introduce separate builtins, e.g for > > > > Unified builtins are used for unified intrinsics, intrinsics users may prefer > same interface and let compiler decide encoding version. Separate > buitins may cause > some defination ambiguous when target attribute is used, see avx-vnni-2.c. > We also provide separate intrinsics interface for compatibility with > different compilers(llvm/msvc/icc).
Hm, the new intrinsics file introduces: +#ifdef __AVXVNNI__ +#define _mm256_dpbusd_avx_epi32(A, B, C) \ + _mm256_dpbusd_epi32((A), (B), (C)) ... +#endif /* __AVXVNNI__ */ + +#define _mm256_dpbusd_epi32(A, B, C) \ + ((__m256i) __builtin_ia32_vpdpbusd_v8si ((__v8si) (A), \ + (__v8si) (B), \ + (__v8si) (C))) + And there are two versions of intrinsics: _mm256_dpbusd_avx_epi32 _mm256_dpbusd_epi32 So, is _mm256_dpusb_epi32 active for either OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL or OPTION_MASK_ISA2_AVXVNNI ? Is _mm265_dpbusb_avx_epi32 the "compatibility intrinsics"? In case the above is correct, please expand the comment + /* Support unified builtin. */ + || (mask2 == OPTION_MASK_ISA2_AVXVNNI) with the above information, what kind of unified builtin is this. Please also note that #defines won't be tested in e.g. sse-13.c, where: --q-- Defining away "extern" and "__inline" results in all of them being compiled as proper functions. */ #define extern #define __inline --/q-- so these defines should be reimplemented as extern inline functions. Uros.
