GCC: and inlining failed in call to always_inline 'FOO': target specific option mismatch

AVX512 comes with the number of variants, and a compiler must know which AVX512 version it compiles.

GCC error inlining failed in call to always_inline 'FOO': target specific option mismatch occurs when a program containing some SIMD-intrinsics, and compiler has wrong or missing target options. The target option are introduced by "-m".

Lets look at the error from real world:

/usr/lib/gcc/x86_64-linux-gnu/5/include/avx512vlbwintrin.h:790:1: error: inlining failed in call to always_inline ‘_mm_movm_epi8’: target specific option mismatch

Now, when open avx512vlbwintrin.h, we see at the beginning of file:

...
#pragma GCC push_options
#pragma GCC target("avx512vl,avx512bw")
#define __DISABLE_AVX512VLBW__
...

Thus, in order to properly compile the program, gcc have to be feed with the two options listed at the target line: -mavx512vl and -mavx512bw.

Not everything in AVX2 is 256-bit

AVX2 has added support for 256-bit arguments for many operations on packed integers, although not all. Some instructions accept the 256-bit registers, but operates on 128-bit lanes rather whole register.

There are three major groups of instructions: packing (narrowing conversion), unpacking (interleave) and permutations; below is a full list of instructions (with intrinsics):

• valignr (_mm256_alignr_epi8)
• vpslldq (_mm256_bslli_epi128)
• vpsrldq (_mm256_bsrli_epi128)
• vmpsadbw (_mm256_mpsadbw_epu8)
• vpacksswb (_mm256_packs_epi16)
• vpackssdw (_mm256_packs_epi32)
• vpackuswb (_mm256_packus_epi16)
• vpackusdw (_mm256_packus_epi32)
• vperm2i128 (_mm256_permute2x128_si256)
• vpermq (_mm256_permute4x64_epi64)
• vpermpd (_mm256_permute4x64_pd)
• vpshufd (_mm256_shuffle_epi32)
• vpshufb (_mm256_shuffle_epi8)
• vpshufhw (_mm256_shufflehi_epi16)
• vpshuflw (_mm256_shufflelo_epi16)
• vpslldq (_mm256_slli_si256)
• vpsrldq (_mm256_srli_si256)
• vpunpckhwd (_mm256_unpackhi_epi16)
• vpunpckhdq (_mm256_unpackhi_epi32)
• vpunpckhqdq (_mm256_unpackhi_epi64)
• vpunpckhbw (_mm256_unpackhi_epi8)
• vpunpcklwd (_mm256_unpacklo_epi16)
• vpunpckldq (_mm256_unpacklo_epi32)
• vpunpcklqdq (_mm256_unpacklo_epi64)
• vpunpcklbw (_mm256_unpacklo_epi8)

For me the most surprising are packing instructions (vpack*) as they require additional shuffling (after or before the instruction) if we want to keep order of values. In some cases the order is crucial.