New 8088 LZSA1 jumptable version 1% to 3% faster

emmanuel-marty / lzsa

Byte-aligned, efficient lossless packer that is optimized for fast decompression on 8-bit micros

Other

236 stars 30 forks source link

New 8088 LZSA1 jumptable version 1% to 3% faster #44

Closed MobyGamer closed 4 years ago

MobyGamer commented 4 years ago

Pavel Zagrebin was able to optimize the LZSA1 speed-optimized jumptable version by an additional 1-3% by adding more jump cases as well as a few optimizations I missed. I've tested his code on my test harness, as well as an RLE2 optimization, and everything checks out.

emmanuel-marty commented 4 years ago

Merged Pavel's version manually in commit 668204d95389d24d9122db785709086d0014f57f (due to conflicts with the merge for some reason). Thanks! An extra 1-3% speed is impressive at this stage!