Closed PSeitz closed 1 year ago
Merging #85 (c2d7613) into main (9e77c40) will decrease coverage by
0.14%
. The diff coverage is100.00%
.
@@ Coverage Diff @@
## main #85 +/- ##
==========================================
- Coverage 90.81% 90.67% -0.14%
==========================================
Files 11 11
Lines 2144 2112 -32
==========================================
- Hits 1947 1915 -32
Misses 197 197
Impacted Files | Coverage Δ | |
---|---|---|
src/block/compress.rs | 99.82% <100.00%> (-0.01%) |
:arrow_down: |
src/block/decompress.rs | 95.46% <100.00%> (-0.03%) |
:arrow_down: |
Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.
Remove some code which is used only in a rare edge case to increase compression speed. When counting the same bytes, the 7 last bytes of the input are now ignored.
Used tool: cargo asm + llvm-mca shows increased throughput
cargo asm --release --no-default-features --features checked-decode --example compress_block compress_internal --mca-intel | grep Through
The same optimization can't be used for the safe version (removing the cold marked internal function). It produces (not obviously) worse assembly for some reason. That can be seen via the tool and in the benchmark.