You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
You can improve the performance of the CRC32 with slice-by-16; POC here. I've tested it extensively and all of the weird quirks (e.g. incrementing 16 times per loop instead of just adding and incrementing after each iteration) are the fastest possible method.
Shipped in 3.0.0, thanks a lot for this! The slice-by-16 CRC is now used for PNG chunk checksums (about 2.4× the throughput of the old byte-wise loop). It’s credited in the source and in THIRD_PARTY_NOTICES.md, along with fflate.
Activity
slice-by-16 crc32 (#12)
slice-by-16 crc32 (#12)
matmen commentedon Oct 14, 2026
Shipped in 3.0.0, thanks a lot for this! The slice-by-16 CRC is now used for PNG chunk checksums (about 2.4× the throughput of the old byte-wise loop). It’s credited in the source and in THIRD_PARTY_NOTICES.md, along with fflate.