Skip to content

hiscoa-compress: faster msb calculation#48

Open
loskutov wants to merge 1 commit intomounaiban:masterfrom
loskutov:patch-1
Open

hiscoa-compress: faster msb calculation#48
loskutov wants to merge 1 commit intomounaiban:masterfrom
loskutov:patch-1

Conversation

@loskutov
Copy link

No description provided.

cryptoluks pushed a commit to cryptoluks/captdriver that referenced this pull request Feb 28, 2026
Performance:
- try_match: precompute match limit (buffer end, max length, line
  boundary) instead of checking 3 conditions per byte in inner loop
- push_bits: defer XOR obfuscation to a single final pass instead of
  2 XOR operations per byte during bit writing
- Remove xorval field from compressor state (no longer needed)

Bug fix:
- Fix 32-bit alignment padding: when already aligned, the expression
  32 - (bitpos % 32) evaluated to 32, adding 4 unnecessary bytes per
  band. Use (32 - (bitpos % 32)) % 32 to correctly produce 0 when
  already aligned. This also fixes the pre-existing roundtrip test
  failures ("band not fully decompressed: 4 bytes left").

Covers the same __builtin_clz optimization as upstream PR mounaiban#48, but
with a portable fallback for non-GCC/Clang compilers.

https://claude.ai/code/session_01HbeKJzNreDZgAHFz32FMfE
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant