pillow-simd

You can use Hyperfine [1] instead of time for a nicer CLI benchmark. I'd also be curious to know is pillow-simd [2] gets the Python performance closer to Go/Rust, and if using Rayon [3] and changing your .iter()'s in your Rust code to .par_iter()'s will yield an improvement there. [1] https://github.com/sharkdp/hyperfine [2] https://github.com/uploadcare/pillow-simd [3] https://github.com/rayon-rs/rayon

Sort of. We run on Heroku, so we indirectly run on AWS. We're heavy users of Python and we're using the excellent Pillow-SIMD [1] library to do most of the heavy lifting. We made our own builds to link it to MozJPEG instead of libjpeg and include libwebp. [1] https://github.com/uploadcare/pillow-simd

It looks like it should, there's an installer here: https://www.lfd.uci.edu/~gohlke/pythonlibs/ But there's also a bug here: https://github.com/uploadcare/pillow-simd/issues/9 If you're using open source software like Python I'd really recommend either leaving Windows, using the bash subsystem or running a Linux VM though. Most of the open source ecosystem lives on *nix and you'll rarely find people who port their projects to Windows.

Very interesting that they have an image processor that uses SIMD for big performance gains. Using either SSE4 or AVX2 with https://github.com/uploadcare/pillow-simd

re: https://github.com/uploadcare/pillow-simd#why-do-not-contrib... They can create three versions of each affected function (fallback, SSE4 and AVX2), place them in separate files (one file for each set of compiler flags), compile each version with its own compiler flags, then link them all together, and in the main module (which is compiled for generic cpu) run cpuid and set global function pointers to the right function implementation. Then always use the global function pointer to call the right implementation of the function, and only expose calling the global function pointer if the function is exported from a shared library. They do need to make sure that function pre and post conditions are preserved in all versions and that memory alignment/layout required by optimized functions is created by the generic code. I think x264 does t