New blog series: @folkertdev shows how we use SIMD in the zlib-rs project.
SIMD is crucial to good performance, but learning how to use it can be daunting. In this series we'll show concrete examples of using SIMD in a real world project.
Part 1 explains how the compiler already uses SIMD for us, how to evaluate whether it's doing a good job, and how to use a more optimal version when the current CPU supports it.
https://tweedegolf.nl/en/blog/153/simd-in-zlib-rs-part-1-autovectorization-and-target-features
tweedegolf.nlSIMD in zlib-rs (part 1): Autovectorization and target features - Blog - Tweede golfI'm fascinated by the creative use of SIMD instructions. When you first learn about SIMD, it is clear that doing more multiplications in a single instruction is useful for speeding up matrix multi ...
Meanwhile, Raph's 
![static inline __m256 avx2_fft_kerneld4(__m256 vec) { __m256 top, bot, sum, diff; const __m256 NEGATE_MASK = _mm256_set_ps(-0.0F, 0.0F, 0.0F, 0.0F, 0.0F, 0.0F, 0.0F, 0.0F); top = vec; bot = _mm256_permute2f128_ps(vec, vec, 1); sum = _mm256_add_ps(top, bot); diff = _mm256_sub_ps(top, bot); constexpr auto PMASK1 = _MM_SHUFFLE(2, 3, 1, 0); constexpr auto PMASK2 = (2 << 4) | 0; _.m256 z = _mm256_permute2f128_ps(sum ] _mm256_permute_ps(diff, PMASK1), PMASK2); z = _mm256_xor_ps(z, NEGATE_MASK); constexpr int PMASK3 = 0 | (2 << 2) | (1 << 4) | (3 << 6); z = _mm256_castpd_ps(_mm256_permutedx64_pd(_mm256_castps_pd(z), PMASK3)); top = z; bot = _mm256_permute2f128_ps(z, z, 1); sum = _mm256_add_ps(top, bot); diff = _mm256_sub_ps(top, bot); _.m256 y = _mm256_permute2f128_ps(sum, diff, PMASK2); return y; I](https://media.hachyderm.io/media_attachments/files/114/186/000/897/101/998/original/a661a690f7f428c4.png "static inline __m256 avx2_fft_kerneld4(__m256 vec) { __m256 top, bot, sum, diff; const __m256 NEGATE_MASK = _mm256_set_ps(-0.0F, 0.0F, 0.0F, 0.0F, 0.0F, 0.0F, 0.0F, 0.0F); top = vec; bot = _mm256_permute2f128_ps(vec, vec, 1); sum = _mm256_add_ps(top, bot); diff = _mm256_sub_ps(top, bot); constexpr auto PMASK1 = _MM_SHUFFLE(2, 3, 1, 0); constexpr auto PMASK2 = (2 << 4) | 0; _.m256 z = _mm256_permute2f128_ps(sum ] _mm256_permute_ps(diff, PMASK1), PMASK2); z = _mm256_xor_ps(z, NEGATE_MASK); constexpr int PMASK3 = 0 | (2 << 2) | (1 << 4) | (3 << 6); z = _mm256_castpd_ps(_mm256_permutedx64_pd(_mm256_castps_pd(z), PMASK3)); top = z; bot = _mm256_permute2f128_ps(z, z, 1); sum = _mm256_add_ps(top, bot); diff = _mm256_sub_ps(top, bot); _.m256 y = _mm256_permute2f128_ps(sum, diff, PMASK2); return y; I")
, only to discover it trips over its own feet 
. Our hero's memory seems as patchy as their 
