radeon and EXA

[Michael Olbrich]

On Tue, Sep 20, 2005 at 02:35:04PM -0400, Michel Dänzer wrote:
> On Tue, 2005-09-20 at 10:12 +0200, Michael Olbrich wrote:
> > 
> > Thinkpad with Radeon Mobility 7500
> > Kernel 2.6.13-mm1
> > debian experimental xorg packages (6.8.99.900.dfsg.1-0pre1) + EXA and
> > radeon_drv from cvs
> 
> How much video RAM, what resolution and depth? It might be useful to see
> your config and log files.

(--) RADEON(0): VideoRAM: 32768 kByte (64 bit DDR SDRAM)
...
(--) RADEON(0): Virtual size is 1400x1050 (pitch 1408)
...
(--) Depth 24 pixmap format is 32 bpp

The relevant part of my config:
-------------------------------------------------------
Section "Extensions"
        Option "Composite" "enable"
EndSection


Section "Device"
        Option     "AGPMode"            "4"     # <i>
        Option     "AGPFastWrite"       "on"    # [<bool>]
        Option     "EnablePageFlip" "on"
        Identifier  "Radeon0"
        Driver      "radeon"
        VendorName  "ATI Technologies Inc"
        BoardName   "Radeon Mobility M7 LW [Radeon Mobility 7500]"
        BusID       "PCI:1:0:0"
        Option      "AccelMethod"       "EXA"
        Option      "DynamicClocks"     "on"
        Option      "BackingStore"      "off"
        Option      "BIOSHotkeys"       "on"
EndSection
------------------------------------------------------

oh, and I found something interesting in my log. Not sure if that's a
problem (whole log appended):
------------------------------------------------------
(II) LoadModule: "radeon"
(II) Reloading /usr/X11R6/lib/modules/drivers/radeon_drv.so
(WW) ****INVALID MEM ALLOCATION**** b: 0xe0000000 e: 0xe8000000 correcting^G
(II) window:
        [0] -1  0       0xe0000000 - 0xe7ffffff (0x8000000) MX[B]
(II) resSize:
        [0] -1  0       0x00000000 - 0xffffffff (0x0) MX[B]
(II) window fixed:
        [0] -1  0       0xe0000000 - 0xe7ffffff (0x8000000) MX[B]
(WW) ****INVALID IO ALLOCATION**** b: 0x3000 e: 0x3100 correcting^G
(II) window:
        [0] -1  0       0x00003000 - 0x000030ff (0x100) IX[B]
        [1] -1  0       0x00003400 - 0x000034ff (0x100) IX[B]
        [2] -1  0       0x00003800 - 0x000038ff (0x100) IX[B]
        [3] -1  0       0x00003c00 - 0x00003cff (0x100) IX[B]
(II) resSize:
        [0] -1  0       0x00000000 - 0xffffffff (0x0) IX[B]
(II) window fixed:
        [0] -1  0       0x00003000 - 0x000030ff (0x100) IX[B]
        [1] -1  0       0x00003400 - 0x000034ff (0x100) IX[B]
        [2] -1  0       0x00003800 - 0x000038ff (0x100) IX[B]
        [3] -1  0       0x00003c00 - 0x00003cff (0x100) IX[B]
Requesting insufficient memory window!: start: 0x3000 end: 0x30ff size 0xc0120100
Requesting insufficient memory window!: start: 0x3400 end: 0x34ff size 0xc0120100
Requesting insufficient memory window!: start: 0x3800 end: 0x38ff size 0xc0120100
Requesting insufficient memory window!: start: 0x3c00 end: 0x3cff size 0xc0120100
(EE) Cannot find a replacement memory range
------------------------------------------------------


> > The basics are pretty much what was reported so far.
> > Without xcompmgr scrolling is rather unusable.
> > With xcompmgr scrolling is better (but not as good as XAA) but switching
> > between desktops (and windows to a lesser extend) is very slow.
> 
> Do you have Eric Anholt's latest exaGlyphs improvements? They made a big
> difference here, at least for scrolling.

I have it now.
Another point about scrolling: a tiled background seems to make it a lot
worse.
 
> > What worries me most is the influence of CPU load. I've been running a
> > compiler in background and the system was more or less unusable.
> > Switching between windows takes seconds. The clock on my panel (which is
> > usually the last thing that still works when the load gets to high)
> > sometimes skips 2-3 second. Even the cursor gets jerky. While that can
> > be improved with a higher priority for the xserver it is far from the
> > responsiveness I get with XAA.
> 
> Note that EXA vs. XAA doesn't have any direct impact on how cursor
> movement is handled, unless you're using SW cursor maybe.

I'm not, although with EXA it sometimes feels like I am.
(II) RADEON(0): Using hardware cursor (offset 25640960)

> > (This is all with xcompmgr which is what we want EXA for anyway, right?)
> 
> If you don't run the compile with nice, it might help to renice the X
> server and/or xcompmgr to -1 or something.

Hmmm, seems to help a bit.

A few more observations:
Most actions have a warmup phase (for the lack of a better word):
Draging a window or scrollbar starts with on or two jumps before running
smoothly.
After some time the smooth part doesn't come. Everything is very slow
until I switch desktops or restart xcompmgr. I have no idea what
triggers the problem nor am I sure if that's really what stops it.
Switching desktops takes forever. The clock on my pannel usually skips 4
seconds.

And very annoying: pressing a key during the mentioned "warmup phase"
usually results in getting it twice. High cpu load makes it worse. High
priority for xcompmgr makes it better.

michael