Greetings

I tried to compile lame 3.91 with gcc-3.0.1 and I tried to allow as much optimizations 
as possible while having as small result in 'make test' as possible

I have found that it doesn;t matter if I use --enable-expopt=full or not, the result 
is varying only because of two gcc options

'make test' result 18 is the best I am able to get, even with all optimizations off 
(is it possible to get smaller number?)

with --enable-expopts=full I got result 413

with full expopt, but with -fno-caller-saves added to OPTIMIZE variable in configure 
script I got result 21

after deleting -ffast-math from the same place I got the result 18 - same as without 
any optimizations

-----

I understand that -ffast-math may change something, but this -fno-caller-saves, may it 
be the compiler bug?

anyway, all I wanted is, that maybe you could allow all optimizations by default and 
only --ffast-math and -fcaller-saves to expopt=full (because it looks they 
are only two which can decrease quality of output)

-fcaller-saves is enabled by default with -O2 and higher



--------

and one more thing, I thought that when I add -mcpu=athlon and -march=athlon to 
gcc-3.0.1 parameters, compiler will use 3dnow and 3dnowext when possible 
(vector operations or something, in fact I don't know anything about mp3 compression 
but I thought it could help). I have Duron 700

Result was very bad, with 'pentiumpro' I have playtime/cputime 2.5 (or near by), with 
'athlon' I have playtime/cputime 0.6 (or near by). Can be compiler bug or 
it simply doesn't make sense to use 3dnow for mp3 encoding?

-------
I don't send all this stuff to gcc creators because I don't know if this is really 
their problem and becuase I don't have last released version of gcc (3.0.3 
is out, I think). So if you would take a look at it, it might help somehow

all for now bye



_______________________________________________
mp3encoder mailing list
[EMAIL PROTECTED]
http://minnie.tuhs.org/mailman/listinfo/mp3encoder

Reply via email to