I've noticed checking through the archives that people often ask how long
conversion should take. Well, on my system it used to crawl along at about 2.4
FPS. But, now I've recompiled Transcode and the Xvid core, it speeds along at
7.5 FPS!
To do this, all you need to do is run the Configure application of Transcode
with the processor specified. Like this:
env CFLAGS="-O3 -march=athlon-xp" ./configure --enable-mmx
or you might have a P4, in which case use: "-O3 -march=pentium4" instead
(Note that the "O" is the letter, as in "Optimise", not the number!). The
"--enable-mmx" flag probably isn't needed, because I think it's the default.
Then "make" and "make install". When you've done this, transcode will now give
the message "tc_memcpy: using mmxext for memcpy" rather than the unoptimised
"tc_memcpy: using libc for memcpy" when run. To optimise the Xvid core, do:
env CFLAGS="-O3 -march=athlon-xp" ./configure
and finally "make" and "make install". Now rapid transcoding will be yours to
behold!
|