Add AVX-512 Ice Lake decoder and retarget to .NET 10 - #38
Merged
Conversation
Port simdutf's Ice Lake base64 kernel (VPERMI2B lookup, VPCOMPRESSB whitespace compression, masked 48-byte stores) using the .NET 10 AVX-512 VBMI/VBMI2 intrinsics. Dispatch prefers AVX-512 VBMI2, then AVX2, SSSE3, or scalar. Benchmarks on a Xeon Gold 6548N reach 11.3 GB/s versus 4.7 GB/s for System.Buffers.Text.Base64.DecodeFromUtf8.
This makes same-machine kernel comparisons part of the regular run.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Summary
VPERMI2Blookup,VPCOMPRESSBto strip white space, and a masked 48-byte store.NEON → AVX-512 VBMI2 → AVX2 → SSSE3 → scalar.Benchmarks (big4, Intel Xeon Gold 6548N, 2.8 GHz, .NET 10.0.11)
Enron email corpus (
benchmark/data/email), BenchmarkDotNet, one core:DecodeFromUtf8)FromBase64StringvsConvert.FromBase64String)Avx512Vbmi2.IsSupportedandVector512.IsHardwareAcceleratedare both true on this machine.Tests
EnronChoppedUTF8failures (SSE, AVX2, AVX-512) are a pre-existing 1-bytebytesConsumeddisagreement with scalar on a chopped suffix; AVX-512 matches AVX2/SSE.