[FFmpeg-devel] [RFC] an aes "optimization"

Reimar Döffinger Reimar.Doeffinger
Fri Jul 2 21:10:20 CEST 2010


Hello,
I have a patch I'd like some people to test and report numbers on
(and of course Michael to untimately OK/NOK it).
On AMD64 (AMD Phenom II X2 550) with gcc 4.4.4 the situation is:
1) code size decreases from 4100 to 3480 bytes
2) decicycles as measured by aes-test increases from 12803 to 12882

The reason for doing this was to increase the readability of the code,
but gcc is very unstable and this is what I'd consider the best result
so far, but it might behave horrible on other architectures.
Removing the av_noinline results in a size around 4080 bytes and
12972 decicycles btw...
On PPC64 (using gcc 4.3.5) the timer seems seems to be broken, because
it claims a ridiculously low 213 vs. 216 decicycles (the difference
is below the precision of measurement)...
And code size only decreases from 4364 to 4064 bytes.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: aesopt.diff
Type: text/x-diff
Size: 1825 bytes
Desc: not available
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100702/68ef71c1/attachment.diff>



More information about the ffmpeg-devel mailing list