4 ms·

Does anyone have any tips for similar wizardry-level SIMD optimization on ARM?

by lumb63 2y ago

anonymoushn 2y ago
If you learn AVX2 programming via highload, my impression is that NEON is quite similar. The main difference is the lack of movemask. You can read these[0] articles[1] about what to do instead. For SVE, prior to very recent versions of SVE, there was no tblq (pshufb equivalent) so I didn't have much hope for using it for general-purpose programming, though of course it would be fine for stuff like TFA. [0]: https://community.arm.com/arm-community-blogs/b/infrastructure-solutions-blog/posts/porting-x86-vector-bitmask-optimizations-to-arm-neon https://community.arm.com/arm-community-blogs/b/infrastructu... [1]: https://www.corsix.org/content/whirlwind-tour-aarch64-vector-instructions https://www.corsix.org/content/whirlwind-tour-aarch64-vector...
TinkersW 2y ago
It isn't as wide on ARM so the gains will be smaller(most ARM is only 128 bit wide neon)