Instead of decoding LZW data into an internal buffer, and copying that buffer into another buffer, and then into a frame, it decodes the LZW data directly into the frame where possible.
RGBA data is processed in one place, in a regular function, without a need for a big macro.
Weezl Decoder is recycled if min_code_size doesn't change (which is more efficient according to weezl's docs).
kornelski
commented
7 months ago
Fixes #87
Closes #150 and #151.
Instead of decoding LZW data into an internal buffer, and copying that buffer into another buffer, and then into a frame, it decodes the LZW data directly into the frame where possible.
RGBA data is processed in one place, in a regular function, without a need for a big macro.
Weezl Decoder is recycled if min_code_size doesn't change (which is more efficient according to weezl's docs).
Overall 2%-5% improvement in benchmarks.