[XviD-devel] [RFC] memcpy thingy

[Christoph Lampert]

On Fri, 31 Oct 2003, Edouard Gomez wrote:
> I've seen on  doom9 a thread a so called "AMD  optimized" build that was
> supposed  to  give  very  large  improvement over  Koepi's  builds  just
> replacing the libc memcpy with the AMD copyrighted one.
> 
> I know that we use memcpy in very few places (mostly setedges) but i was
> curious to confirm or refute the  results given by users so i ported the
> well known xine memcpy functions to nasm.
> 
> The  results are  much more  close  to what  i expected:  ie very  small
> improvement (<1%),  far from the  10% advertised by the  "AMD optimized"
> build users.
> 
> Do you think it's worth commiting this code ?

1% on what plattform?  Windows, I guess?
I find 1% just by memcpy rather much, considering that the OS should have
MMX-optmized memcpy already (Linux does, Windows I don't know).

If it's really close to 1%, we could of course add a function pointer 
for memcpy just as for DCT etc. It's not much of a deal to change the few
calls. 
On the other hand, if memcpy really needs a nontrivial part of time, it
might be better to write e.g. a special NASMed setedges() routine.

Not for 1.0, of course... 

gruel