On Saturday 26 May 2001 18:11, ralf willenbacher wrote:
> Steven Newbury wrote:
> > On Friday 25 May 2001 01:22, ralf willenbacher wrote:
> > > thanks for trying it out.
> > > i fixed a typo (again)..
> >
> > <patch snipped>
> > In your patch you changed 'totalcard = b->size' and 'totalagp=b->size'
> > to 'totalcard += b->size' and 'totalagp += b->size' in
> > xc/lib/GL/mesa/src/drv/mga/mgatexmem.c. This change stopped things
> > working quite right.
>
> i use the total size to figure out where to kick a texture out.
> without the right numbers it could end up returning 0 and that would be
> bad
> since my g200 has no space on the card at all.
>
> It starts up fine but seems to only have a couple of Mb of
>
> > texture space available - while I have 44Mb allocated giving 32Mb for
> > textures.
>
> it only uploads into card space, this is slower then into agp space.
>
> > The XF86Config option AGPSize still does not take effect,
>
> i have
> Section "Device"
> Identifier "g200"
> Driver "mga"
> Option "AGPSize" "32"
> EndSection
> and get
> (II) MGA(0): [agp] 32768 kB allocated with handle 0xd1041000
I am really quite confused about this. I have it set the same as you expect
for 44 instead of 32. I have added an extra message to report the size of
the AGPSize setting during startup - but it reports 0Mb?!?
(**) MGA(0): Option "AGPMode" "2"
(**) MGA(0): Option "AGPSize" "44"
(II) Loading sub module "mga_hal"
(II) LoadModule: "mga_hal"
(II) Loading /usr/X11R6/lib/modules/drivers/mga_hal_drv.o
(II) Module mga_hal: vendor="The XFree86 Project"
compiled for 4.0.99.900, module version = 1.0.0
ABI class: XFree86 Video Driver, version 0.4
(II) MGA(0): Matrox HAL module used
(--) MGA(0): Chipset: "mgag400"
(**) MGA(0): Allocating 0Mb for AGP memory
(**) MGA(0): Using AGP 2x mode
It all seems correct in mga_driver.c, but for some reason it doesn't seem to
be setting pMga->agpSize from AGPSize
>
> im currently running with this pice of code..
>
> static int mgaChooseTexHeap( mgaContextPtr mmesa, mgaTextureObjectPtr t
> )
> {
> int freeagp, freecard;
> int fitincard, fitinagp;
> int totalcard, totalagp;
> TMemBlock *b;
>
> totalcard = totalagp = fitincard = fitinagp = freeagp = freecard = 0;
>
> b = mmesa->texHeap[0];
> while(b)
> {
> totalcard += b->size;
> if(b->free)
> {
> if(t->totalSize <= b->size)fitincard = 1;
> }
> b = b->next;
> }
>
> b = mmesa->texHeap[1];
> while(b)
> {
> totalagp += b->size;
> if(b->free)
> {
> if(t->totalSize <= b->size)fitinagp = 1;
> }
> b = b->next;
> }
>
> if(fitincard)return 0;
> if(fitinagp)return 1;
>
> if(totalcard && totalagp)
> {
> int ages;
> int ratio = (totalcard > totalagp) ? totalcard / totalagp :
> totalagp / totalcard;
> ages = mmesa->sarea->texAge[0] + mmesa->sarea->texAge[1];
> if( (ages % ratio) == 0)
> {
> fprintf(stderr,"goes into heap %d\n", totalcard > totalagp ? 1 :
> 0);
> return totalcard > totalagp ? 1 : 0;
> }
> else
> {
> fprintf(stderr,"goes into heap %d\n", totalcard > totalagp ? 0 :
> 1);
> return totalcard > totalagp ? 0 : 1;
> }
> }
>
> fprintf(stderr, "goes into heap %d\n", totalagp ? 1 : 0);
> if(totalagp) return 1;
> return 0;
> }
>
Yes, that works.
The code from the end of that function in the patch you posted ended:
if(totalcard >= t->totalSize)return 0;
if(totalagp >= t->totalSize)return 1;
This didn't work, I guess it was always returning 0 in my case?
> using the sarea ages is really wishfull thinking but it should work to
> select the smaller memory space from time to time.
_______________________________________________
Dri-devel mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/lists/listinfo/dri-devel
