Dirac, GCC-4.0.0 and SIMD optimisations on x86 architecture

[Anuradha Suraparaju]

I used GCC-4.0.0 to build the Dirac project. I used the --enable-mmx
flag to enable MMX optimisations. Other versions of gcc (3.3.4 and
3.4.3) give a speedup of approximately 30% when the mmx optimisations
are enabled. However with gcc-4.0.0, the performance of the Dirac
encoder with MMX opts enabled is lots slower than with them disabled.

The configuration of my test machine is
CPU -    Intel(R) Pentium(R) 4 CPU 2.66GHz
Memory - 512 KB
OS: Suse Linux 9.2

My question is how do I report this as a bug? What information do I
need to provide in the bug report? Did anybody else face similar
problems with GCC-4.0.0 and MMX-enabled programs.

Regards,
A. Suraparaju

--