[Ffmpeg-devel] [PATCH] dead code elimination

Måns Rullgård mans
Mon Feb 26 22:37:25 CET 2007

Luca Barbato <lu_zero at gentoo.org> writes:

> M?ns Rullg?rd wrote:
>>> sparc64 is unlikely to be faster than sparc32. I can benchmark it if
>>> you tell me how.
>> Why?
> Because on saner arches 64bit variant means just 64bit registers, not
> more registers and less cruft as for the x86_64
> You end up paying a bit more address load and other tasks and most of
> the apps don't use 64bit in an efficient way...

IIRC, Sparc64 is often used with 32-bit pointers and 64-bit longs,
including 64-bit loads and stores.  The conditional code in
simple_idct would certainly benefit from that.

You are correct that most programs don't need a 64-bit address space.

M?ns Rullg?rd
mans at mansr.com

More information about the ffmpeg-devel mailing list