#x86-64 #generated #format #avx2 #rsum #fast-lanes

nightly fastlanes_rs

Wrapper for the FastLanes Compression Format C++ library

1 unstable release

0.1.0 Mar 3, 2024

#764 in Compression

MIT license

8KB
118 lines

FastLanes Rust Crate

This crate wraps the supplementary material for the FastLanes Compression Format Paper (https://github.com/cwida/FastLanes).

Valid Exported Functions

  • generated_rsum_fallback_scalar_rsum{,1,2,3}
  • generated_rsum_fallback_unit64_rsum{,1,2,3}
  • generated_rsum_x86_64_avx2_rsum{,1,2,3}
  • generated_rsum_x86_64_sse_rsum{,1,2,3}
  • generated_rsum_x86_64_avx512bw_rsum{,1,2,3}
  • generated_untranspose_fallback_scalar_untranspose_i{,1,2,3}
  • generated_untranspose_fallback_scalar_untranspose_o{,1,2,3}

Known Problems

  • Requires clang++ compiler available on the system
  • Requires nightly build for the avx512bw target feature flag
  • Most exported signatures are missing a definition; only use the functions listed in here
  • avx2 is enabled by default, avx512bw has to be explicitly enabled for your rustc compiler, e.g. using the RUSTFLAGS environment variable

No runtime deps

~0–2.2MB
~44K SLoC