#19314: ffmpeg: enable mmx optimizations by default ---------------------------------+------------------------------------------ Reporter: devans@… | Owner: devans@… Type: defect | Status: assigned Priority: Normal | Milestone: Port Bugs Component: ports | Version: 1.7.1 Keywords: mmx Intel | Port: ffmpeg ---------------------------------+------------------------------------------ Comment(by devans@…): Yes, I agree. I was just trying to get people to upgrade to the latest on general principal and you caught me! ;-) In addition are these issues which I previously glossed over * technically --disable-mmx only addresses a subset of Intel-like optimizations. The full set of configure options are {{{ --disable-amd3dnow disable 3DNow! optimizations --disable-amd3dnowext disable 3DNow! extended optimizations --disable-mmx disable MMX optimizations --disable-mmx2 disable MMX2 optimizations --disable-sse disable SSE optimizations --disable-ssse3 disable SSSE3 optimizations }}} * the --disable-mmx configure option for ffmpeg is not completely effective in the current release 0.5 or on svn trunk (ffmpeg-devel.) The ifdef logic in libswscale/rgb2rgb.c keys on the definition of ARCH_X86 only and ignores and even overwrites HAVE_MMX, HAVE_MMX2, etc that result from the configure options. The resulting behavior is to compile separate versions of the accelerated functions for each possible combination of acceleration options. The correct optimization function is then selected based on a check at run time. Thus, at least in this one instance, there is no protection from the XCode 3.0 gcc compilation bug. -- Ticket URL: <http://trac.macports.org/ticket/19314#comment:6> MacPorts <http://www.macports.org/> Ports system for Mac OS