[all-commits] [llvm/llvm-project] bede46: [Fuzzer] Optimize UpdateFeatureFrequency (#65288)

[Arseny Kapoulkine via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: bede46f57eec7e9b4665a2b0f1df59a2fb1cc74d
      https://github.com/llvm/llvm-project/commit/bede46f57eec7e9b4665a2b0f1df59a2fb1cc74d
  Author: Arseny Kapoulkine <arseny.kapoulkine at gmail.com>
  Date:   2023-09-07 (Thu, 07 Sep 2023)

  Changed paths:
    M compiler-rt/lib/fuzzer/FuzzerCorpus.h

  Log Message:
  -----------
  [Fuzzer] Optimize UpdateFeatureFrequency (#65288)

Instead of a linear scan, use a bitset to track rarity of features. This
improves fuzzer throughput rather dramatically (close to 2x) in early
exploratory phases; in steady state this seems to improve fuzzing
throughput by ~15% according to perf.

The benchmarks are done on an executable with ~100k features, so the
results may change based on the executable that's being fuzzed.

kFeatureSetSize is 2M so the bitset is adding 256 KB to
sizeof(InputCorpus), but this should be fine since there's already three
arrays indexed by feature index for a total of 200 MB.