[FFmpeg-devel] Fw: [foms] Paper submissions to LCA

[Frank Barchard]

On Fri, Jul 17, 2009 at 1:26 AM, Carl Eugen Hoyos <cehoyos at ag.or.at> wrote:

> M?ns Rullg?rd <mans <at> mansr.com> writes:
>
> > On that topic, does anyone know why so many FATE tests are failing
> > with icc?
> Perhaps you could have a look what is so special about this (short)
> function?
> I could upload the assembly of O0 and O1 if you want.


A work around, or technique you can use to track down the bug, is use a
pragma to disable optimization around a function, or a few lines of code.
 #pragma optimize("", off)

I considered doing this
port myself, and started with x264.  It was relatively easy.  Visual C
was also not too hard, and both the Visual C, and especially the Intel
C versions were substantially faster than GCC (4.3).  Until I turned
on the (32
bit) assembly, which made them all the same.  So I passed that to gcc
people.  One downside of icc
builds, compared to gcc, is code size is alot bigger.