Joerg Schilling wrote:
> Ian Collins <ian at ianshome.com> wrote:
>
>> That's an odd comment, considering most reviews of FF3 I've seen
>> commends it for being both faster and leaking less memory tan its
>> predecessor.
>>
>> It's certainly been stable long term on my desktop since 3.0.3.
>>
>
> What is the last time you used it on Sparc?
>
> Netscape always was slower on sparc systems then on even slower x86 machines.
> I would guess that there is strange byte order dependent code inside the
> graphics engine.
I don't know of any "strange byte order dependent code" in the Cairo graphics
engine used in current Firefox, but I do know of several simpler reasons it's
faster on x86 than SPARC:
- it makes heavy use of the Render extension if the X server supports it.
Xorg supports Render on almost all cards, with several providing hardware
acceleration for it, which helps x86. On SPARC, only a few drivers for
Xsun were updated to add Render support, and it's off by default. For Xorg,
unless you've added Martin's drivers, only XVR-2500 has a Xorg driver
provided so far - and I don't know what level of Render acceleration any
of those provide.
- The pixman backend used for various graphics rendering has code for MMX & SSE
acceleration, but none for VIS. Given the upstream projects' target
audience are approximately 99% x86 and 0% SPARC, they've concentrated their
optimization efforts there.
- Virtually any graphics card on a PC of today is several generations newer
than any SPARC graphics card you've used.
--
-Alan Coopersmith- alan.coopersmith at sun.com
Sun Microsystems, Inc. - X Window System Engineering
