Fix HEXL benchmark performance degradation #416
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR fixes some performance degradation on the HEXL-enabled benchmarks, observed on ICX with clang-12. See below for one example, where this PR yields ~3.8x speedup on the N=8192 low-level NTT benchmarks.
To the best of my knowledge, using
seal::UniformRandomGenerator
turns out to be slow enough to cause some AVX512 transitions between benchmark iterations. AVX512 transitions also likely explains some of the strange runtime behavior described in #386. This PR switches touniform_int_distribution
for the benchmarks, which is fast enough to avoid the AVX512 transitions.