Skip to contentSkip to conversation

AGP : ATI128,powerpc

DISCUSSION Vincent Katz 4 participants 8 messages Nov 14, 2001 – Nov 15, 2001

lore
Share & tools
lore
#0 lore raw

Hi,

Linux version : 2.4.14
File with a supposed probleme : linux/drivers/char/agp/agpgart_be.c

I have a powerpc with an ati128 AGP.

When i compile the kernel i have this error : error "Please define flush_cache."

this error come from the function flush_cache in agpgart_be.c where definition for powerpc are not setting . Why ?

Quoted 13 lines

static inline void flush_cache(void)
{

#if defined(__i386__) || defined(__x86_64__)
       asm volatile ("wbinvd":::"memory");

#elif defined(__alpha__) || defined(__ia64__) || defined(__sparc__)
/* ??? I wonder if we'll really need to flush caches, or if the
core logic can manage to keep the system coherent. The ARM speaks only of using `cflush' to get things in memory in preparation for power failure.

If we do need to call `cflush', we'll need a target page, as we can only flush one page at a time.

Quoted 6 lines

Ditto for IA-64. --davidm 00/08/07 */
mb();
#else
#error "Please define flush_cache."
#endif
}

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

#1 lore raw

Hi,

Try Ben Herrenschmidt's tree. There is support for powerpc in that tree.

Kevin

Quoted 5 lines · Vincent Katz

Hi,

Linux version : 2.4.14
File with a supposed probleme : linux/drivers/char/agp/agpgart_be.c

Quoted 28 lines

I have a powerpc with an ati128 AGP.

When i compile the kernel i have this error : error "Please define flush_cache."

this error come from the function flush_cache in agpgart_be.c where definition for powerpc are not setting . Why ?

Quoted 19 lines

static inline void flush_cache(void)
{

#if defined(__i386__) || defined(__x86_64__)
       asm volatile ("wbinvd":::"memory");

#elif defined(__alpha__) || defined(__ia64__) || defined(__sparc__)
/* ??? I wonder if we'll really need to flush caches, or if the
core logic can manage to keep the system coherent. The ARM speaks only of using `cflush' to get things in memory in preparation for power failure.

If we do need to call `cflush', we'll need a target page, as we can only flush one page at a time.

Ditto for IA-64. --davidm 00/08/07 */
mb();
#else
#error "Please define flush_cache."
#endif
}

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

#2 lore raw
Quoted 7 lines · Kevin B. Hendricks

Hi,

Try Ben Herrenschmidt's tree. There is support for powerpc in that tree.

Kevin

i've been using ben's tree for a powerbook for quite some time, and i used to see the same error, so i disabled agp and dri.

has this been fixed lately ?
i'm desperate for faster video. 2d performance under x is pathetic, and scrolling is a major pita.

does anyone know how to speed up a pismo powerbook x server ?

regards,

jörn

--
Jörn Nettingsmeier
home://Kurfürstenstr.49.45138.Essen.Germany
phone://+49.201.491621
http://spunk.dnsalias.org
http://www.linuxdj.com/audio/lad/

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

#3 lore raw
Quoted 5 lines

i've been using ben's tree for a powerbook for quite some time, and i used to see the same error, so i disabled agp and dri.

has this been fixed lately ?

Yes.

Quoted 4 lines

i'm desperate for faster video. 2d performance under x is pathetic, and scrolling is a major pita.

does anyone know how to speed up a pismo powerbook x server ?

XFree 4.1 should work fine. AGP won't help 2D accel, make sure you use the "ati" driver and not "fbdev".

Ben.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

#4 lore raw
In reply to Vincent Katz
Quoted 11 lines

Linux version : 2.4.14
File with a supposed probleme : linux/drivers/char/agp/agpgart_be.c

I have a powerpc with an ati128 AGP.

When i compile the kernel i have this error : error "Please define flush_cache."

this error come from the function flush_cache in agpgart_be.c where definition for powerpc are not setting . Why ?

What PPC machine is this ? I added support for Apple's UniNorth chipset in my kernel (patch submited to dri-devel a few days ago) along with the necessary r128 DRM changes to work around HW bugs.

(My rsync tree is at rsync.penguinppc.org::linux-2.4-benh).

If you are using a different chipset, then my code may not help

Ben.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

#5 lore raw

Hi,
thanks for your reply.

PPC machine : TiBook (Powerbook Titanium)
ATY : aty128fb: Rage Mobility M3 (AGP) [chip rev 0x0] 8M 128-bit SDR SGRAM (1:1)

Is it possible to add in the flush_cache function a "defined(__powerpc__)" like alpha, sparc ..or i386 .. ?

When the function flush_cache in the file "agpgart_be.c" ll be updated in the offical kernel for ppc cpu ?

thanks.

Quoted 15 lines · Benjamin Herrenschmidt
Quoted 10 lines

Linux version : 2.4.14
File with a supposed probleme : linux/drivers/char/agp/agpgart_be.c

I have a powerpc with an ati128 AGP.

When i compile the kernel i have this error : error "Please define flush_cache."

this error come from the function flush_cache in agpgart_be.c where definition for powerpc are not setting . Why ?

What PPC machine is this ? I added support for Apple's UniNorth chipset in my kernel (patch submited to dri-devel a few days ago) along with the necessary r128 DRM changes to work around HW bugs.

Quoted 7 lines

(My rsync tree is at rsync.penguinppc.org::linux-2.4-benh).

If you are using a different chipset, then my code may not help

Ben.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

#6 lore raw
Quoted 10 lines

PPC machine : TiBook (Powerbook Titanium)
ATY : aty128fb: Rage Mobility M3 (AGP) [chip rev 0x0] 8M 128-bit SDR SGRAM (1:1)

Is it possible to add in the flush_cache function a "defined(__powerpc__)" like alpha, sparc ..or i386 .. ?

When the function flush_cache in the file "agpgart_be.c" ll be updated in the offical kernel for ppc cpu ?

It's more than flush cache that is needed ;) flush_cache alone will eventually allow AGP to compile, not to work.

Get my current rsync tree from rsync.penguinppc.org::linux-2.4-benh, it has complete support for UniNorth AGP on r128. (the M3 you have in your Pismo is an r128). Beware however that accelerated GL via DRI is known to crash with this chip for reasons not yet determined.

I've sent my current patches to the DRI mailing list, I can't say yet when all this will appear upstream as some merging has to be done with other archs using the "special" agp_ioremap. I will submit the AGP part of the patch directly to linus/marcello once I'm in sync with DRI people, there's no need to send it until something can use it ;)

Ben.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/

#7 lore raw

Ok thanks, i understand the problem ;-)
this, go against my frame of mind but i yield (i m not narrow-minded ;-) ), meanwhile the update of the official, i'll use your rsync.
thx for all, good luck.

thx also for my gpm problem :-)

vince.

Quoted 18 lines · Benjamin Herrenschmidt
Quoted 9 lines

PPC machine : TiBook (Powerbook Titanium)
ATY : aty128fb: Rage Mobility M3 (AGP) [chip rev 0x0] 8M 128-bit SDR SGRAM (1:1)

Is it possible to add in the flush_cache function a "defined(__powerpc__)" like alpha, sparc ..or i386 .. ?

When the function flush_cache in the file "agpgart_be.c" ll be updated in the offical kernel for ppc cpu ?

It's more than flush cache that is needed ;) flush_cache alone will eventually allow AGP to compile, not to work.

Get my current rsync tree from rsync.penguinppc.org::linux-2.4-benh, it has complete support for UniNorth AGP on r128. (the M3 you have in your Pismo is an r128). Beware however that accelerated GL via DRI is known to crash with this chip for reasons not yet determined.

Quoted 9 lines

I've sent my current patches to the DRI mailing list, I can't say yet when all this will appear upstream as some merging has to be done with other archs using the "special" agp_ioremap. I will submit the AGP part of the patch directly to linus/marcello once I'm in sync with DRI people, there's no need to send it until something can use it ;)

Ben.

** Sent via the linuxppc-dev mail list. See http://lists.linuxppc.org/