Allow using assembly optimizations
This MR allows the Meson wrap to use (or not) assembly optimizations.
This MR also fixes the issue with MSVC coughing on missing AVX2 functions, which was the symptom of a broader issue with missing symbols being hidden by the dead code optimization pass. The fix was actually done in FFmpeg 5.1, and it's been backported here.
Closes !62 (closed)
TODO
-
arm and aarch64 need the inline checks for instruction sets and incorporating NEON/VFP sources. (May be useful for Apple M1 and Android.)
Edited by amyspark