X11 and mtrr's ..anyone any clues on this problem ?
Hello, I have mttr support configured in my kernel (2.4.4 and 2.4.10). This is supposed to dramatically improve graphics speed. However I get the following when I start X up.. Oct 14 01:30:27 tanya last message repeated 12 times Oct 14 01:30:31 tanya kernel: mtrr: no MTRR for f6000000,800000 found Oct 14 01:30:31 tanya kernel: mtrr: no MTRR for f6800000,800000 found Oct 14 01:30:31 tanya kernel: mtrr: no MTRR for f7000000,800000 found Oct 14 01:30:32 tanya kernel: mtrr: no MTRR for f6000000,800000 found Oct 14 01:30:32 tanya kernel: mtrr: no MTRR for f6800000,800000 found Oct 14 01:30:32 tanya kernel: mtrr: no MTRR for f7000000,800000 found Oct 14 01:30:32 tanya kernel: mtrr: no MTRR for f7800000,800000 found Oct 14 01:30:34 tanya kernel: mtrr: base(0xf6000000) is not aligned on a size(0x1800000) boundary There is a document on this in the kernel documentation directory, but following it's suggestions makes no difference. Anyone know anything about this ? I am using a Matrox G450 with Asus m/board .. Intel 815 chipset. Thanks. -- Regards Cliff
On Sun, 14 Oct 2001, Cliff Sarginson wrote:
Oct 14 01:30:27 tanya last message repeated 12 times Oct 14 01:30:31 tanya kernel: mtrr: no MTRR for f6000000,800000 found Oct 14 01:30:31 tanya kernel: mtrr: no MTRR for f6800000,800000 found Oct 14 01:30:31 tanya kernel: mtrr: no MTRR for f7000000,800000 found Oct 14 01:30:32 tanya kernel: mtrr: no MTRR for f7800000,800000 found
Its due to the way the g450 (and other dual head matrox cards) split memory when its in dual head mode. X sees the cards as two seperate devices, and complains at the lack of mtrrs for each memrange. The kernel should already have set up a range covering the entire on-card memory. regards, Dave. -- | Dave Jones. http://www.codemonkey.org.uk | SuSE Labs
On Sun, Oct 14, 2001 at 01:39:06AM +0200, Dave Jones wrote:
On Sun, 14 Oct 2001, Cliff Sarginson wrote:
Oct 14 01:30:27 tanya last message repeated 12 times Oct 14 01:30:31 tanya kernel: mtrr: no MTRR for f6000000,800000 found Oct 14 01:30:31 tanya kernel: mtrr: no MTRR for f6800000,800000 found Oct 14 01:30:31 tanya kernel: mtrr: no MTRR for f7000000,800000 found Oct 14 01:30:32 tanya kernel: mtrr: no MTRR for f7800000,800000 found
Its due to the way the g450 (and other dual head matrox cards) split memory when its in dual head mode. X sees the cards as two seperate devices, and complains at the lack of mtrrs for each memrange.
The kernel should already have set up a range covering the entire on-card memory.
regards,
Dave.
I see. Is the implication here that I should not worry about it ? If so then I will be happy :). Am I getting max performance anyway (I don't use the dual head feature yet). -- Regards Cliff
On Sun, 14 Oct 2001, Cliff Sarginson wrote:
I see. Is the implication here that I should not worry about it ? If so then I will be happy :). Am I getting max performance anyway (I don't use the dual head feature yet).
As long as you've an entry in /proc/mtrr covering the whole memory range (lspci -v will you the range) then you're fine. regards, Dave. -- | Dave Jones. http://www.codemonkey.org.uk | SuSE Labs
On Sun, Oct 14, 2001 at 03:35:11PM +0200, Dave Jones wrote:
On Sun, 14 Oct 2001, Cliff Sarginson wrote:
I see. Is the implication here that I should not worry about it ? If so then I will be happy :). Am I getting max performance anyway (I don't use the dual head feature yet).
As long as you've an entry in /proc/mtrr covering the whole memory range (lspci -v will you the range) then you're fine.
regards,
Dave.
This is what lspci -v tells me..(why does it say G400 when it is a G450?) 01:00.0 VGA compatible controller: Matrox Graphics, Inc. MGA G400 AGP (rev 82) ( Subsystem: Matrox Graphics, Inc.: Unknown device 0641 Flags: bus master, medium devsel, latency 64, IRQ 11 Memory at f6000000 (32-bit, prefetchable) [size=32M] Memory at f5000000 (32-bit, non-prefetchable) [size=16K] Memory at f4800000 (32-bit, non-prefetchable) [size=8M] Expansion ROM at f5fe0000 [disabled] [size=128K] Capabilities: [dc] Power Management version 2 Capabilities: [f0] AGP version 2.0 Thanks. -- Regards Cliff
On Sun, 14 Oct 2001, Cliff Sarginson wrote:
This is what lspci -v tells me..(why does it say G400 when it is a G450?)
They share the same PCI ID. The difference between them is the revision. Rev < 80 = G400 Rev >=80 = G450
Memory at f6000000 (32-bit, prefetchable) [size=32M] Memory at f5000000 (32-bit, non-prefetchable) [size=16K] Memory at f4800000 (32-bit, non-prefetchable) [size=8M]
The important one here is the first one (as its prefetchable) If this entry is in /proc/mtrr (and the size matches) you're home and dry. regards, Dave. -- | Dave Jones. http://www.codemonkey.org.uk | SuSE Labs
On Sun, Oct 14, 2001 at 06:16:07PM +0200, Dave Jones wrote:
On Sun, 14 Oct 2001, Cliff Sarginson wrote:
This is what lspci -v tells me..(why does it say G400 when it is a G450?)
They share the same PCI ID. The difference between them is the revision. Rev < 80 = G400 Rev >=80 = G450
Memory at f6000000 (32-bit, prefetchable) [size=32M] Memory at f5000000 (32-bit, non-prefetchable) [size=16K] Memory at f4800000 (32-bit, non-prefetchable) [size=8M]
The important one here is the first one (as its prefetchable) If this entry is in /proc/mtrr (and the size matches) you're home and dry.
regards,
Dave.
[root@tanya]/usr/local/bin# cat /proc/mtrr reg00: base=0x00000000 ( 0MB), size= 512MB: write-back, count=1 reg01: base=0xf8000000 (3968MB), size= 64MB: write-combining, count=2 reg02: base=0xf6000000 (3936MB), size= 32MB: write-combining, count=1 After starting X the count field on reg02: bumps upto 12. So, if I read this correct, it is all tickety boo ? Thanks -- Regards Cliff
On Sun, 14 Oct 2001, Cliff Sarginson wrote:
[root@tanya]/usr/local/bin# cat /proc/mtrr reg02: base=0xf6000000 (3936MB), size= 32MB: write-combining, count=1 After starting X the count field on reg02: bumps upto 12. So, if I read this correct, it is all tickety boo ?
Yup :-) regards, Dave. -- | Dave Jones. http://www.codemonkey.org.uk | SuSE Labs
On Sun, Oct 14, 2001 at 06:35:23PM +0200, Dave Jones wrote:
On Sun, 14 Oct 2001, Cliff Sarginson wrote:
[root@tanya]/usr/local/bin# cat /proc/mtrr reg02: base=0xf6000000 (3936MB), size= 32MB: write-combining, count=1 After starting X the count field on reg02: bumps upto 12. So, if I read this correct, it is all tickety boo ?
Yup :-)
regards,
Dave.
Thanks ! I can worry about more important things now ... :) -- Regards Cliff
participants (2)
-
Cliff Sarginson -
Dave Jones