Improving on std:count_if()'s auto-vectorization
1–10 of 51 posts
Re: Improving on std:count_if()'s auto-vectorization
#2It’s a good example to illustrate how to get more simd from the compiler
But the overly specific constraint means this is not a general count_if algorithm.
For this to be useful I have to: - know there are only 255 true values - but have a large dataset so it’s worth optimizing - not want to stop early when some threshold is met
This is so specialized it’s not even worth having a generic predicate argument for.
Re: Improving on std:count_if()'s auto-vectorization
#3Meta-question: Given how common :: is in programming languages, and how rare of a typo it is, is it really worth it for HN's title filtering logic to include a rule for automatically replacing it with a single :?
Re: Improving on std:count_if()'s auto-vectorization
#4another solution is to just cast the result to an uint8_t; with this, clang 19.1.0 gives the same assembly:
Re: Improving on std:count_if()'s auto-vectorization
#5Soon I am wondering if rather than rely on finicky auto-vectorization we’ll just have LLMs help “hand-optimize” more routines. Just like how memcmp and memcpy are optimized by hand today maybe like 20% of the program could just be LLM-assisted assembly. @ffmpeg on X thinks maybe they are starting to get it [1] and I had some success having an LLM generate working WebAssembly [2]
https://x.com/ffmpeg/status/1898408922769223994?s=46
Re: Improving on std:count_if()'s auto-vectorization
#6another solution is to just cast the result to an uint8_t; with this, clang 19.1.0 gives the same assembly: https://gcc.godbolt.org/z/E5oTW5eKe
Which is discussed in the post and doesn’t work in GCC.
Re: Improving on std:count_if()'s auto-vectorization
#7Soon I am wondering if rather than rely on finicky auto-vectorization we’ll just have LLMs help “hand-optimize” more routines. Just like how memcmp and memcpy are optimized by hand today maybe like 20% of the program could just be LLM-assisted assembly. @ffmpeg on X thinks maybe they are starting to get it [1] and I had some success having an LLM generate working WebAssembly [2] https://x.com/ffmpeg/status/1898408922…
Instead of replacing one finicky and temperamental approach (auto-vectorizers) with another (LLM codegen) I'd much rather see more exploration of explicit SIMD abstractions like Intel's ISPC language. Shader languages for GPUs had this figured out forever ago, there is a sensible middle-ground to be had between brittle compiler magic and no compiler at all.
Re: Improving on std:count_if()'s auto-vectorization
#8another solution is to just cast the result to an uint8_t; with this, clang 19.1.0 gives the same assembly: https://gcc.godbolt.org/z/E5oTW5eKe
Which is discussed in the post and doesn’t work in GCC.
Oh right, I didn't see it in a couple of passes (and searching for cast); for anyone else looking it's in the 3rd footnote. Thanks.
Re: Improving on std:count_if()'s auto-vectorization
#9[flagged]
Re: Improving on std:count_if()'s auto-vectorization
#10Seems like you can do this sort of speed up even without the 256 constraint. Just run this sped up version, but after each set of 256 iterations, dump all the 8-bit counters to the 64-bit final result.