Code

[mono] simdhash tweaks#101082

Code

Merged

kg merged 9 commits intodotnet:maindotnet/runtime:mainfrom kg:simdhash-tweakskg/runtime:simdhash-tweaksCopy head branch name to clipboardApr 19, 2024

+134-116

Commits

Commits on Apr 19, 2024

• Inlining this is very profitable on MSVC, and clang already does it anyway

  kg

  committed
  041810a

  Copy full SHA for 041810a

• Use rand_s on msvc for benchmark data so that simdhash isn't at an automatic disadvantage

  kg

  committed
  7cca7ef

  Copy full SHA for 7cca7ef

• MSVC-compatible random number generation

  kg

  committed
  9ef2473

  Copy full SHA for 9ef2473

• Remove unnecessary assert

  kg

  committed
  1f25dfa

  Copy full SHA for 1f25dfa

• Optimize out unnecessary unrolled iterations from scalar bucket scan

  Show description for 4ad58cc
  kg

  committed
  4ad58cc

  Copy full SHA for 4ad58cc

• Use xoshiro256 for consistent random value set in benchmark across all targets

  Show description for e643967
  kg

  committed
  e643967

  Copy full SHA for e643967

• In unrolled scalar search, do equivalent of chained cmovs instead of conditional returns. This makes find_random much faster and find_missing only slightly slower

  kg

  committed
  bb54b1f

  Copy full SHA for bb54b1f

• ght compatible simdhash build fixes

  Show description for ebb0949
  kg

  committed
  ebb0949

  Copy full SHA for ebb0949

• MSVC build fix

  kg

  committed
  300acb1

  Copy full SHA for 300acb1