It may also have to do with compiler optimizations... Try rebuilding from
source to see what happens (upgrade to 4.1.0 while you're at it. I know
there were Rage 128 and Radeon fixes/enhancements in there, I suspect there
were Mach64/Rage fixes too). I don't know PowerPC arch specific options, but
on x86, I use -O3 -march=i686 -mcpu=i686 -fomit-frame-pointer -funroll-loops
-malign-functions=4 -fschedule-insns2... A quick explanation follows:
-O3: Enables a ton of little pissy optimizations (function inlining, byte
alignment, a few other things) that seldom or never have deleterious effect
on speed (may make the binaries bigger, though, and may use a bit more RAM).
-O2 is safer, but -O3 works for almost everything in my experience.
-march=i686: 80x86-specific, generates code for P6 CPUs. I use -march=athlon
with GCC 3.0, since I have an Athlon. There's probably an analogous option
for G3/G4 PPC chips, but I'm not sure what that would be. Get the latest
binutils, if you don't already have it.
-mcpu=i686: x86-specific. Optimizes instruction scheduling for a P6... There
must be a PPC analog.
-fomit-frame-pointer: Tells the compiler not to create a frame pointer for
function calls that don't absolutely require it. Makes the resulting binary
undebuggable, but saves quite a few instructions per function call. This can
really add up in frequently-called functions (like drivers). This, AFAIK, is
an XP option (should work on PPC).
-malign-functions=4: aligns functions on word boundaries. This speeds up
function calls a little. This specific alignment is almost certainly
x86-specific, but the -malign-functions=foo option isn't...
-funroll-loops: unrolls some loops, according to a heuristic. This is
distinct from -funroll-all-loops, which, like the name implies, unrolls ALL
loops. Unrolling some loops speeds things up, unrolling them all is usually
not helpful. This is XP, AFAIK.
-fschedule-insns2: does some instruction-scheduler frobbing. May speed
things up. -O3 probably already does this, but I explicitly list it anyway.
I'll bet this is XP, but RTM anyway (I haven't a Mac to test it on)
> -----Original Message-----
> From: TD - Sales Int'l Holland B.V. [mailto:[EMAIL PROTECTED]]
> Sent: Wednesday, September 05, 2001 12:24 PM
> To: [EMAIL PROTECTED]
> Subject: Re: [Newbie]XFree86 4.0.2 slower than 3.3.6 for Mach64?
>
>
> Just guessing here but maybe you have more animations turned
> on in your
> 4.0.2 than you had in 3.3.6
>
> regards
>
> ----- Original Message -----
> From: "Paul J. Lucas" <[EMAIL PROTECTED]>
> To: <[EMAIL PROTECTED]>
> Sent: Tuesday, September 04, 2001 5:28 PM
> Subject: [Newbie]XFree86 4.0.2 slower than 3.3.6 for Mach64?
>
>
> > OK, so I don't think it's my imaginiation any more, but XFree86
> > 4.0.2 seems *slower* on my Powerbook (which uses an ATI Mach64
> > LT/Pro graphics chip).
> >
> > For example, vertical scrolls in an aterm window doing
> > transparency seem a tad sluggish; dragging a window around, I
> > can *watch* the other aterm windws behind it repaint text; the
> > "ifs" xscreensaver module runs at a crawl and soaks up all
> > the CPU whereas it ran just fine under 3.3.6.
> >
> > In looking at the XFree86.log file, it seems to find and use
> > the "ati" driver for Mac64 located at 0:17:0 on the PCI bus.
> > It says it loads XXA.
> >
> > So what's the deal?
> >
> > - Paul
> >
> >
> > _______________________________________________
> > Newbie mailing list
> > [EMAIL PROTECTED]
> > *** To unsubscribe , or change message options, see:
> > http://XFree86.Org/mailman/listinfo/newbie
> >
>
> _______________________________________________
> Newbie mailing list
> [EMAIL PROTECTED]
> *** To unsubscribe , or change message options, see:
> http://XFree86.Org/mailman/listinfo/newbie
>
_______________________________________________
Newbie mailing list
[EMAIL PROTECTED]
*** To unsubscribe , or change message options, see:
http://XFree86.Org/mailman/listinfo/newbie
