On 30 Jan, Yosi wrote:
> As the maintainer of the Lame spec files for Linux RPM packages, I'd
> like to ask for your help. I would like to use the best gcc
> (version 3.x) flags for differnet platforms. By "best" I mean the ones
> that produce the fastest code. Here are the values that I currently use
> for Pentium III or higher CPUs:
>
> -O3 -fomit-frame-pointer -mcpu=athlon -march=athlon \
> -malign-functions=4 -funroll-loops -fexpensive-optimizations \
> -fschedule-insns2 -mfancy-math-387 -pipe
gcc 3.0.3 produces a broken lame executable if you use -O3 with
-funroll-loops.
> And these are the values I use when Lame is compiled on an Athlon:
>
> -O3 -fomit-frame-pointer -fno-strength-reduce -mcpu=athlon \
> -march=athlon -malign-functions=4 -funroll-loops \
> -fexpensive-optimizations -fschedule-insns2 -mfancy-math-387 -pipe
- Is the use of -m{cpu,arch}=athlon in both cases intendet?
- How large is the speed improvement with "-fno-strength-reduce"?
> Can you recommend *out of your experience* other optimizations flags
> for these platforms? Note, that I'm talking about gcc v3.x
I already collect them in configure.in (at least the CVS version of it
contains some gcc 3.x fixes/improvements). So _not_ specifying CFLAGS
and using --enable-expopt should DTRT (at least with 3.92, for 3.91 you
may want to adopt some of the CVS changes).
Bye,
Alexander.
--
Weird enough for government work.
http://www.Leidinger.net Alexander @ Leidinger.net
GPG fingerprint = C518 BC70 E67F 143F BE91 3365 79E2 9C60 B006 3FE7
_______________________________________________
mp3encoder mailing list
[EMAIL PROTECTED]
http://minnie.tuhs.org/mailman/listinfo/mp3encoder
