slow performance of libggi on svgalib target due to bad libc ?

[Tijs van Bakel]

Hi,

I've noticed a performance loss in using ggiPutBox() on top of the
svgalib target on some machines, of which those using Suse 6.1.

The problem seems to be in the memcpy()'s used by ggiPutBox(), which
are not optimized on these systems.

Using this small snippet in svgalib directly speeds up the putbox
operation by about 100%.

  uint32* d;
  uint32* s;
  int count = 320*200/4;
  while (count--)
    *d++ = *s++;

I'm almost sure the _real_ problem is in the way libc was compiled or
set up, but still it's quite frustrating to see libggi perform twice
as bad as this silly routine.  Would it be wise to adapt the
ggiPutBox() implemenation for the svgalib target, or should I be
telling people that Suse delivers a lousy libc?

-- 
Tijs van Bakel, <[EMAIL PROTECTED]>