faster and fully conformant

farindk released this 03 Jun 17:58

· 1257 commits to master since this release

all conformance streams decode correctly
faster, parallel SAO
image memory can be allocated externally
frame-dropping API
selection which temporal sub-stream to decode
fake speedup options to turn off deblocking and SAO
new parallelization architecture
sherlock265 can show slice boundaries and tiles
core rewritten in C++ for better maintainability (API still C only)

Assets 3