[PATCHES] radeon kms pm patches

[PATCHES] radeon kms pm patches

[Alex Deucher]

This weekend I finally got some time to sit down with kms pm, and I
created a new patch set (against drm-radeon-testing).  For those that
want to play with the i2c stuff for thermal chips, you'll probably
grab my recent i2c patches as well.  You can grab my latest pm patches
here:
http://people.freedesktop.org/~agd5f/pm2/

So far I haven't seen any corruption when changing power modes.

What the patches do:
- implement gui idle irq support
- only change clocks when the engine is idle
- add support for turning down the number of active simds in lower
power modes (r6xx+)
- add a pm_fini function
- move set/get power state logic into asic specific callbacks.
Different strategies for handling
different power tables formats.

Things left to do:
- reset clocks to default on module unload (in pm_fini function)
- add request module support for hwmon i2c thermal chip drivers
- add hwmon support for internal thermal/fan support used on some
r6xx/r7xx boards
- add more robust power state selection
- tie power state selection into external events (manual power mode
selection, AC/DC state, etc.)
- hook up memory reclocking
- hook up pcie lane setting
- hook up voltage setting

Alex

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

2010/3/15 Alex Deucher <alexdeucher@gmail.com>:
> This weekend I finally got some time to sit down with kms pm, and I
> created a new patch set (against drm-radeon-testing).  For those that
> want to play with the i2c stuff for thermal chips, you'll probably
> grab my recent i2c patches as well.  You can grab my latest pm patches
> here:
> http://people.freedesktop.org/~agd5f/pm2/
>
> So far I haven't seen any corruption when changing power modes.

Why don't you post them inline? It would be possible to comment any
part of code.

1) What is GUI?
2) What is SIMD?

-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

2010/3/15 Alex Deucher <alexdeucher@gmail.com>:
> This weekend I finally got some time to sit down with kms pm, and I
> created a new patch set (against drm-radeon-testing).  For those that
> want to play with the i2c stuff for thermal chips, you'll probably
> grab my recent i2c patches as well.  You can grab my latest pm patches
> here:
> http://people.freedesktop.org/~agd5f/pm2/

http://people.freedesktop.org/~agd5f/pm2/0003-drm-radeon-kms-add-some-comments-about-radeon_asic.patch

Code you removed was needed for PM debugfs, to do not display memory
clock on IGP


-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

2010/3/15 Alex Deucher <alexdeucher@gmail.com>:
> This weekend I finally got some time to sit down with kms pm, and I
> created a new patch set (against drm-radeon-testing).  For those that
> want to play with the i2c stuff for thermal chips, you'll probably
> grab my recent i2c patches as well.  You can grab my latest pm patches
> here:
> http://people.freedesktop.org/~agd5f/pm2/

http://people.freedesktop.org/~agd5f/pm2/0006-drm-radeon-kms-wait-for-gpu-idle-before-changing-po.patch

You made same mistake I did earlier with "0" argument in condition.

You should use same trick as I did in:
http://people.freedesktop.org/~agd5f/pm2/0013-drm-radeon-kms-switch-to-condition-waiting-for-recl.patch

-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

2010/3/15 Alex Deucher <alexdeucher@gmail.com>:
> This weekend I finally got some time to sit down with kms pm, and I
> created a new patch set (against drm-radeon-testing).  For those that
> want to play with the i2c stuff for thermal chips, you'll probably
> grab my recent i2c patches as well.  You can grab my latest pm patches
> here:
> http://people.freedesktop.org/~agd5f/pm2/

You can consider porting
[PATCH 2/2] drm/radeon/kms: prepare for more reclocking operations

You will need to sync with VBLANK few times I believe.

Not sure about GUI syncing, do not know what is this.

-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

2010/3/15 Alex Deucher <alexdeucher@gmail.com>:
> This weekend I finally got some time to sit down with kms pm, and I
> created a new patch set (against drm-radeon-testing).  For those that
> want to play with the i2c stuff for thermal chips, you'll probably
> grab my recent i2c patches as well.  You can grab my latest pm patches
> here:
> http://people.freedesktop.org/~agd5f/pm2/

http://people.freedesktop.org/~agd5f/pm2/0009-drm-radeon-kms-r6xx-enable-simd-pm.patch

Didn't you miss there something actually *setting* SIMD(s?)?

-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

2010/3/15 Rafał Miłecki <zajec5@gmail.com>:
> 2010/3/15 Alex Deucher <alexdeucher@gmail.com>:
>> This weekend I finally got some time to sit down with kms pm, and I
>> created a new patch set (against drm-radeon-testing).  For those that
>> want to play with the i2c stuff for thermal chips, you'll probably
>> grab my recent i2c patches as well.  You can grab my latest pm patches
>> here:
>> http://people.freedesktop.org/~agd5f/pm2/
>>
>> So far I haven't seen any corruption when changing power modes.
>
> Why don't you post them inline? It would be possible to comment any
> part of code.

It was a lot of patches.  I can post them with the message next time.

>
> 1) What is GUI?

Graphical User Interface.  In this case it refers to the drawing
engine and related blocks (2D, 3D, CP, etc.).

> 2) What is SIMD?

SIMD (Single Instruction Multiple Data) blocks are
the computational blocks of the shader engine.

Alex

>
> --
> Rafał
>

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

2010/3/15 Rafał Miłecki <zajec5@gmail.com>:
> 2010/3/15 Alex Deucher <alexdeucher@gmail.com>:
>> This weekend I finally got some time to sit down with kms pm, and I
>> created a new patch set (against drm-radeon-testing).  For those that
>> want to play with the i2c stuff for thermal chips, you'll probably
>> grab my recent i2c patches as well.  You can grab my latest pm patches
>> here:
>> http://people.freedesktop.org/~agd5f/pm2/
>
> http://people.freedesktop.org/~agd5f/pm2/0003-drm-radeon-kms-add-some-comments-about-radeon_asic.patch
>
> Code you removed was needed for PM debugfs, to do not display memory
> clock on IGP

Oh, right... debugfs support is also currently broken with multiple
cards.  I'll just drop this patch for now.

Alex

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

2010/3/15 Rafał Miłecki <zajec5@gmail.com>:
> 2010/3/15 Alex Deucher <alexdeucher@gmail.com>:
>> This weekend I finally got some time to sit down with kms pm, and I
>> created a new patch set (against drm-radeon-testing).  For those that
>> want to play with the i2c stuff for thermal chips, you'll probably
>> grab my recent i2c patches as well.  You can grab my latest pm patches
>> here:
>> http://people.freedesktop.org/~agd5f/pm2/
>
> http://people.freedesktop.org/~agd5f/pm2/0009-drm-radeon-kms-r6xx-enable-simd-pm.patch
>
> Didn't you miss there something actually *setting* SIMD(s?)?

heh... whoops.  fixed.

Alex

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

2010/3/15 Rafał Miłecki <zajec5@gmail.com>:
> 2010/3/15 Alex Deucher <alexdeucher@gmail.com>:
>> This weekend I finally got some time to sit down with kms pm, and I
>> created a new patch set (against drm-radeon-testing).  For those that
>> want to play with the i2c stuff for thermal chips, you'll probably
>> grab my recent i2c patches as well.  You can grab my latest pm patches
>> here:
>> http://people.freedesktop.org/~agd5f/pm2/
>
> You can consider porting
> [PATCH 2/2] drm/radeon/kms: prepare for more reclocking operations
>
> You will need to sync with VBLANK few times I believe.
>
> Not sure about GUI syncing, do not know what is this.

I've gone ahead and rebased my patches against Dave's current
drm-radeon-testing branch, so both patches should be in now.

Alex

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

2010/3/15 Rafał Miłecki <zajec5@gmail.com>:
> 2010/3/15 Alex Deucher <alexdeucher@gmail.com>:
>> This weekend I finally got some time to sit down with kms pm, and I
>> created a new patch set (against drm-radeon-testing).  For those that
>> want to play with the i2c stuff for thermal chips, you'll probably
>> grab my recent i2c patches as well.  You can grab my latest pm patches
>> here:
>> http://people.freedesktop.org/~agd5f/pm2/
>
> http://people.freedesktop.org/~agd5f/pm2/0006-drm-radeon-kms-wait-for-gpu-idle-before-changing-po.patch
>
> You made same mistake I did earlier with "0" argument in condition.
>
> You should use same trick as I did in:
> http://people.freedesktop.org/~agd5f/pm2/0013-drm-radeon-kms-switch-to-condition-waiting-for-recl.patch

Good catch!  fixed.

Alex

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

[-- Attachment #1: Type: text/plain, Size: 1533 bytes --]

On Sun, Mar 14, 2010 at 7:01 PM, Alex Deucher <alexdeucher@gmail.com> wrote:
> This weekend I finally got some time to sit down with kms pm, and I
> created a new patch set (against drm-radeon-testing).  For those that
> want to play with the i2c stuff for thermal chips, you'll probably
> grab my recent i2c patches as well.  You can grab my latest pm patches
> here:
> http://people.freedesktop.org/~agd5f/pm2/

Updated patches rebased against airlied's new drm-radeon-testing along
with the fixes noted in Rafal's comments:
http://people.freedesktop.org/~agd5f/pm2/

Alex


>
> So far I haven't seen any corruption when changing power modes.
>
> What the patches do:
> - implement gui idle irq support
> - only change clocks when the engine is idle
> - add support for turning down the number of active simds in lower
> power modes (r6xx+)
> - add a pm_fini function
> - move set/get power state logic into asic specific callbacks.
> Different strategies for handling
> different power tables formats.
>
> Things left to do:
> - reset clocks to default on module unload (in pm_fini function)
> - add request module support for hwmon i2c thermal chip drivers
> - add hwmon support for internal thermal/fan support used on some
> r6xx/r7xx boards
> - add more robust power state selection
> - tie power state selection into external events (manual power mode
> selection, AC/DC state, etc.)
> - hook up memory reclocking
> - hook up pcie lane setting
> - hook up voltage setting
>
> Alex
>

[-- Attachment #2: 0001-drm-radeon-kms-pm-fix-typo-in-power-table-parsing.patch --]
[-- Type: application/mbox, Size: 1267 bytes --]

[-- Attachment #3: 0002-drm-radeon-kms-gfx-init-fixes-for-r6xx-r7xx.patch --]
[-- Type: application/mbox, Size: 2782 bytes --]

[-- Attachment #4: 0003-i2c-algo-bit-Add-pre-and-post-xfer-hooks.patch --]
[-- Type: application/mbox, Size: 1864 bytes --]

[-- Attachment #5: 0004-drm-radeon-kms-use-new-pre-post_xfer-i2c-bit-algo-h.patch --]
[-- Type: application/mbox, Size: 9563 bytes --]

[-- Attachment #6: 0005-drm-radeon-kms-add-gui_idle-callback.patch --]
[-- Type: application/mbox, Size: 7073 bytes --]

[-- Attachment #7: 0006-drm-radeon-kms-add-support-for-gui-idle-interrupts.patch --]
[-- Type: application/mbox, Size: 6387 bytes --]

[-- Attachment #8: 0007-drm-radeon-kms-wait-for-gpu-idle-before-changing-po.patch --]
[-- Type: application/mbox, Size: 3631 bytes --]

[-- Attachment #9: 0008-drm-radeon-kms-pm-rework-power-mode-parsing.patch --]
[-- Type: application/mbox, Size: 3130 bytes --]

[-- Attachment #10: 0009-drm-radeon-kms-pm-add-asic-specific-callbacks-for-s.patch --]
[-- Type: application/mbox, Size: 13306 bytes --]

[-- Attachment #11: 0010-drm-radeon-kms-pm-add-asic-specific-callbacks-for-g.patch --]
[-- Type: application/mbox, Size: 32120 bytes --]

[-- Attachment #12: 0011-drm-radeon-kms-r6xx-add-support-for-SIMD-pm.patch --]
[-- Type: application/mbox, Size: 5844 bytes --]

[-- Attachment #13: Type: text/plain, Size: 345 bytes --]

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev

[-- Attachment #14: Type: text/plain, Size: 161 bytes --]

--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Sedat Dilek]

Hi,

I tried the pm2 patches from [1] on top of Linus-tree -
2.6.34-rc1-git6 to be precise.

I have prepared a single patch of drm-radeon-testing GIT branch (as of
15-Mar-2010) [2].

The patch-series file against 2.6.34-rc1 looks like:

[series]
# upstream updates
upstream/patch-2.6.34-rc1-git6

# patch from <http://files.iniza.org/drm-radeon-kms/drm-radeon-testing/>
drm-radeon-testing/drm-radeon-testing.patch

# patches from <http://people.freedesktop.org/~agd5f/pm2/>
agd5f/pm2/0001-drm-radeon-kms-pm-fix-typo-in-power-table-parsing.patch
agd5f/pm2/0002-drm-radeon-kms-gfx-init-fixes-for-r6xx-r7xx.patch
#agd5f/pm2/0003-i2c-algo-bit-Add-pre-and-post-xfer-hooks.patch
agd5f/pm2/0004-drm-radeon-kms-use-new-pre-post_xfer-i2c-bit-algo-h.patch
agd5f/pm2/0005-drm-radeon-kms-add-gui_idle-callback.patch
agd5f/pm2/0006-drm-radeon-kms-add-support-for-gui-idle-interrupts.patch
agd5f/pm2/0007-drm-radeon-kms-wait-for-gpu-idle-before-changing-po.patch
agd5f/pm2/0008-drm-radeon-kms-pm-rework-power-mode-parsing.patch
agd5f/pm2/0009-drm-radeon-kms-pm-add-asic-specific-callbacks-for-s.patch
agd5f/pm2/0010-drm-radeon-kms-pm-add-asic-specific-callbacks-for-g.patch
agd5f/pm2/0011-drm-radeon-kms-r6xx-add-support-for-SIMD-pm.patch
[/series]


Happy testing...
- Sedat -

References:
------------------
[1] http://people.freedesktop.org/~agd5f/pm2/
[2] http://files.iniza.org/drm-radeon-kms/drm-radeon-testing/

NOTE:
---------
These two patches are now in latest drm-radeon-testing GIT branch:
* 0001-drm-radeon-kms-pm-fix-typo-in-power-table-parsing.patch
* 0002-drm-radeon-kms-gfx-init-fixes-for-r6xx-r7xx.patch

You can drop this patch it is included in upstream:
* 0003-i2c-algo-bit-Add-pre-and-post-xfer-hooks.patch

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

On Mon, Mar 15, 2010 at 3:30 AM, Alex Deucher <alexdeucher@gmail.com> wrote:
> On Sun, Mar 14, 2010 at 7:01 PM, Alex Deucher <alexdeucher@gmail.com> wrote:
>> This weekend I finally got some time to sit down with kms pm, and I
>> created a new patch set (against drm-radeon-testing).  For those that
>> want to play with the i2c stuff for thermal chips, you'll probably
>> grab my recent i2c patches as well.  You can grab my latest pm patches
>> here:
>> http://people.freedesktop.org/~agd5f/pm2/
>
> Updated patches rebased against airlied's new drm-radeon-testing along
> with the fixes noted in Rafal's comments:
> http://people.freedesktop.org/~agd5f/pm2/

Another set of updated patches against drm-radeon-testing:
http://people.freedesktop.org/~agd5f/pm2/

These implement much the remaining pm functionality.  So far they are
working well here.
these patches add:
- memory reclocking
- pcie lane changes
- update display watermarks as bandwidth changes
- allow power management with multi-head
- reset power mode on exit

Alex


>
> Alex
>
>
>>
>> So far I haven't seen any corruption when changing power modes.
>>
>> What the patches do:
>> - implement gui idle irq support
>> - only change clocks when the engine is idle
>> - add support for turning down the number of active simds in lower
>> power modes (r6xx+)
>> - add a pm_fini function
>> - move set/get power state logic into asic specific callbacks.
>> Different strategies for handling
>> different power tables formats.
>>
>> Things left to do:
>> - reset clocks to default on module unload (in pm_fini function)
>> - add request module support for hwmon i2c thermal chip drivers
>> - add hwmon support for internal thermal/fan support used on some
>> r6xx/r7xx boards
>> - add more robust power state selection
>> - tie power state selection into external events (manual power mode
>> selection, AC/DC state, etc.)
>> - hook up memory reclocking
>> - hook up pcie lane setting
>> - hook up voltage setting
>>
>> Alex
>>
>

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

2010/3/17 Alex Deucher <alexdeucher@gmail.com>:
> On Mon, Mar 15, 2010 at 3:30 AM, Alex Deucher <alexdeucher@gmail.com> wrote:
>> On Sun, Mar 14, 2010 at 7:01 PM, Alex Deucher <alexdeucher@gmail.com> wrote:
>>> This weekend I finally got some time to sit down with kms pm, and I
>>> created a new patch set (against drm-radeon-testing).  For those that
>>> want to play with the i2c stuff for thermal chips, you'll probably
>>> grab my recent i2c patches as well.  You can grab my latest pm patches
>>> here:
>>> http://people.freedesktop.org/~agd5f/pm2/
>>
>> Updated patches rebased against airlied's new drm-radeon-testing along
>> with the fixes noted in Rafal's comments:
>> http://people.freedesktop.org/~agd5f/pm2/
>
> Another set of updated patches against drm-radeon-testing:
> http://people.freedesktop.org/~agd5f/pm2/
>
> These implement much the remaining pm functionality.  So far they are
> working well here.
> these patches add:
> - memory reclocking
> - pcie lane changes
> - update display watermarks as bandwidth changes
> - allow power management with multi-head
> - reset power mode on exit

Nice, looks alright, thanks. Will test later if I manage before
leaving for skiing.

Do you know if we could ever recolck memory on multiple CRTCs? If not,
what's the reason?

-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

2010/3/17 Rafał Miłecki <zajec5@gmail.com>:
> 2010/3/17 Alex Deucher <alexdeucher@gmail.com>:
>> On Mon, Mar 15, 2010 at 3:30 AM, Alex Deucher <alexdeucher@gmail.com> wrote:
>>> On Sun, Mar 14, 2010 at 7:01 PM, Alex Deucher <alexdeucher@gmail.com> wrote:
>>>> This weekend I finally got some time to sit down with kms pm, and I
>>>> created a new patch set (against drm-radeon-testing).  For those that
>>>> want to play with the i2c stuff for thermal chips, you'll probably
>>>> grab my recent i2c patches as well.  You can grab my latest pm patches
>>>> here:
>>>> http://people.freedesktop.org/~agd5f/pm2/
>>>
>>> Updated patches rebased against airlied's new drm-radeon-testing along
>>> with the fixes noted in Rafal's comments:
>>> http://people.freedesktop.org/~agd5f/pm2/
>>
>> Another set of updated patches against drm-radeon-testing:
>> http://people.freedesktop.org/~agd5f/pm2/
>>
>> These implement much the remaining pm functionality.  So far they are
>> working well here.
>> these patches add:
>> - memory reclocking
>> - pcie lane changes
>> - update display watermarks as bandwidth changes
>> - allow power management with multi-head
>> - reset power mode on exit
>
> Nice, looks alright, thanks. Will test later if I manage before
> leaving for skiing.

Thanks.

>
> Do you know if we could ever recolck memory on multiple CRTCs? If not,
> what's the reason?

I think so.  It works now more or less, there's just an occasional
jitter on the head not synced.  I'm trying to find out how we do it
internally.

Alex

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

2010/3/17 Alex Deucher <alexdeucher@gmail.com>:
> Another set of updated patches against drm-radeon-testing:
> http://people.freedesktop.org/~agd5f/pm2/

Not serious thing, but still it's warning...

# git am pm2/0001*
Applying: drm/radeon/kms/atom: make sure tables are valid
/home/drm-2.6/.git/rebase-apply/patch:475: space before tab in indent.
        if (atom_parse_data_header(mode_info->atom_context, index,
NULL, &frev, &crev, &data_offset)) {
/home/drm-2.6/.git/rebase-apply/patch:476: space before tab in indent.
                power_info = (union power_info
*)(mode_info->atom_context->bios + data_offset);
warning: 2 lines add whitespace errors.


-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

2010/3/17 Alex Deucher <alexdeucher@gmail.com>
> Another set of updated patches against drm-radeon-testing:
> http://people.freedesktop.org/~agd5f/pm2/
>
> These implement much the remaining pm functionality.  So far they are
> working well here.
> these patches add:
> - memory reclocking
> - pcie lane changes
> - update display watermarks as bandwidth changes
> - allow power management with multi-head
> - reset power mode on exit

It seems memory reclocking and SIMDs setting is disabled for r6xx with
currently available patches...?

Otherwise it seems to work fine.

--
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

2010/3/17 Rafał Miłecki <zajec5@gmail.com>:
> 2010/3/17 Alex Deucher <alexdeucher@gmail.com>
>> Another set of updated patches against drm-radeon-testing:
>> http://people.freedesktop.org/~agd5f/pm2/
>>
>> These implement much the remaining pm functionality.  So far they are
>> working well here.
>> these patches add:
>> - memory reclocking
>> - pcie lane changes
>> - update display watermarks as bandwidth changes
>> - allow power management with multi-head
>> - reset power mode on exit
>
> It seems memory reclocking and SIMDs setting is disabled for r6xx with
> currently available patches...?
>

It re-clocks memory (for single head anyway), r600_set_power_state():

                /* set memory clock */
                if (rdev->asic->set_memory_clock && (mclk !=
rdev->pm.current_mclk)) {
                        radeon_sync_with_vblank(rdev);
                        radeon_pm_debug_check_in_vbl(rdev, false);
                        radeon_set_memory_clock(rdev, mclk);
                        radeon_pm_debug_check_in_vbl(rdev, true);
			rdev->pm.current_mclk = mclk;
                        DRM_INFO("Setting: m: %d\n", mclk);
                }

The simd stuff is still disabled at the moment.

> Otherwise it seems to work fine.

Thanks for testing.

Alex

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

2010/3/17 Rafał Miłecki <zajec5@gmail.com>:
> 2010/3/17 Alex Deucher <alexdeucher@gmail.com>:
>> Another set of updated patches against drm-radeon-testing:
>> http://people.freedesktop.org/~agd5f/pm2/
>
> Not serious thing, but still it's warning...
>
> # git am pm2/0001*
> Applying: drm/radeon/kms/atom: make sure tables are valid
> /home/drm-2.6/.git/rebase-apply/patch:475: space before tab in indent.
>        if (atom_parse_data_header(mode_info->atom_context, index,
> NULL, &frev, &crev, &data_offset)) {
> /home/drm-2.6/.git/rebase-apply/patch:476: space before tab in indent.
>                power_info = (union power_info
> *)(mode_info->atom_context->bios + data_offset);
> warning: 2 lines add whitespace errors.

Looks like I was a little too quirk with that patch :)  I'll get that fixed up.

Alex

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

W dniu 18 marca 2010 04:27 użytkownik Alex Deucher
<alexdeucher@gmail.com> napisał:
> 2010/3/17 Rafał Miłecki <zajec5@gmail.com>:
>> 2010/3/17 Alex Deucher <alexdeucher@gmail.com>
>>> Another set of updated patches against drm-radeon-testing:
>>> http://people.freedesktop.org/~agd5f/pm2/
>>>
>>> These implement much the remaining pm functionality.  So far they are
>>> working well here.
>>> these patches add:
>>> - memory reclocking
>>> - pcie lane changes
>>> - update display watermarks as bandwidth changes
>>> - allow power management with multi-head
>>> - reset power mode on exit
>>
>> It seems memory reclocking and SIMDs setting is disabled for r6xx with
>> currently available patches...?
>>
>
> It re-clocks memory (for single head anyway), r600_set_power_state():
>
>                /* set memory clock */
>                if (rdev->asic->set_memory_clock && (mclk !=
> rdev->pm.current_mclk)) {
>                        radeon_sync_with_vblank(rdev);
>                        radeon_pm_debug_check_in_vbl(rdev, false);
>                        radeon_set_memory_clock(rdev, mclk);
>                        radeon_pm_debug_check_in_vbl(rdev, true);
>                        rdev->pm.current_mclk = mclk;
>                        DRM_INFO("Setting: m: %d\n", mclk);
>                }
>
> The simd stuff is still disabled at the moment.

Whoops, I missed one patch. My mistake.

-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Sedat Dilek]

Hi,

the last days I am testing Alex's pm2-patches with Linus-tree - this
report based on patchset as of 17-Mar-2010 22:42 [0].

Unfortunately, I cannot load radeon kernel-module with dynpm=1 (0 =
disabled works fine).
All testing was done in init-3 in a virtual-console.
Looking into the log-files show (a) regression(s).

I have prepared patches against Linux-2.6.34-rc1 [1] and also attached
logs in [2].
The agdf5-pm2-with-linustree.tar.bz2 tarball contains all data (see [3]).

Have fun.


Kind Regards,
- Sedat -

References:
------------------

[0] http://people.freedesktop.org/~agd5f/pm2/
[1] http://files.iniza.org/agdf5-pm2-with-linustree/patches/
[2] http://files.iniza.org/agdf5-pm2-with-linustree/logs/
[3] http://files.iniza.org/agdf5-pm2-with-linustree/archive/

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCHES] radeon kms pm patches

[Sedat Dilek]

[1] shows:
...
Mar 18 10:07:35 seduxbox kernel: [  127.005632]
[drm:r500_hw_i2c_xfer], i2c write error 0x00000002
...

- Sedat -

[1] http://files.iniza.org/agdf5-pm2-with-linustree/logs/debug.txt

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

W dniu 18 marca 2010 10:05 użytkownik Rafał Miłecki <zajec5@gmail.com> napisał:
> Whoops, I missed one patch. My mistake.

I've just applied all patches dated as "17-Mar-2010 22:42" and tested.

Now memory reclocking is really enabled, but it causes corruptions for
me. Maybe that's the same thing you experienced with dual head?

Video is *far* from perfect, but maybe will give you some idea of what
do I experience:
http://estudent.put.poznan.pl/rafal.milecki/kernel/MOV00225.MP4

The same happened for me before all your patches, when I tried to
enable memory reclocking on my own.

-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Sedat Dilek]

Hi,

hereby my new experiences with my RV515.

# lspci -nnvv | grep "VGA compatible controller"
01:00.0 VGA compatible controller [0300]: ATI Technologies Inc M52
[Mobility Radeon X1300] [1002:714a] (prog-if 00 [VGA controller])

I applied "[PATCH] drm/radeon/kms: add hw_i2c module option" from [1]
in addition to pm2-patches (see my last email) and rebuild radeon
kernel-module:

$ cd $build-dir
$ make M=drivers/gpu/drm/radeon clean
$ apply_patch
$ make M=drivers/gpu/drm/radeon

We have now a new module-option "hw_i2c" (0 = disabled, 1 = enabled),
This requires to re-generate modules.dep and map files:

# depmod

Then I rebooted to check new radeon kernel-module.

[init-3]
# modprobe -r -v radeon
# modprobe -v radeon modeset=1 dynpm=1 hw_i2c=1

That is working - no blank screen.

But [2] shows also same error when there is a failure (does this error
really matter?):
...
Mar 18 11:56:40 seduxbox kernel: [   62.350145]
[drm:r500_hw_i2c_xfer], i2c write error 0x00000002
...

Reclocking seems to work (see below) - when I start glxgears and let
in run a while you see increasing "e: XXXXX m: YYYYY" - closing
glxgears results in downclocking.

I will check if starting and closing OpenArena (Anholt benchmark) will
result in a blank screen (but this was yesterday).

Kind Regards,
- Sedat -

References:
------------------
[1] http://marc.info/?l=dri-devel&m=126880645519474&w=2
[2] http://files.iniza.org/agdf5-pm2-with-linustree/logs/debug_ok.txt

----- INVESTIGATIONS -----

[ 2213.179060] [drm] Requested: e: 12900 m: 13500 p: 16
[ 2213.602308] [drm] Requested: e: 21000 m: 13500 p: 16
[ 2213.621661] [drm] Setting: e: 21000
[ 2214.021304] [drm] Requested: e: 32400 m: 13500 p: 16
[ 2214.046572] [drm] Setting: e: 32400
[ 2214.446302] [drm] Requested: e: 39200 m: 30000 p: 16
[ 2214.465277] [drm] Setting: e: 39200
[ 2214.465848] [drm] not in vbl for pm change 00020002 00000000 at entry
[ 2214.468161] [drm] not in vbl for pm change 00020002 00000000 at exit
[ 2214.468166] [drm] Setting: m: 30000
[ 2214.868305] [drm] Requested: e: 39200 m: 30000 p: 16
[ 2235.680072] [drm] Requested: e: 32400 m: 13500 p: 16
[ 2235.680281] [drm] Setting: e: 32400
[ 2235.690389] [drm] not in vbl for pm change 00020002 00000000 at entry
[ 2235.692412] [drm] not in vbl for pm change 00020002 00000000 at exit
[ 2235.692417] [drm] Setting: m: 13500
[ 2236.092069] [drm] Requested: e: 21000 m: 13500 p: 16
[ 2236.092105] [drm] GUI not idle!!!
[ 2236.492071] [drm] Requested: e: 21000 m: 13500 p: 16
[ 2236.492249] [drm] Setting: e: 21000
[ 2237.292060] [drm] Requested: e: 12900 m: 13500 p: 16
[ 2237.292250] [drm] Setting: e: 12900
[ 2237.692070] [drm] Requested: e: 12900 m: 13500 p: 16
[ 2238.092068] [drm] Requested: e: 12900 m: 13500 p: 16
- EOT -

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Sedat Dilek]

Experiences (II) - OpenArena

Mostly, I am using radeon Gallium3D driver r300g dri/statetracker.
Here I can start openarena (OA) but entering fight-mode closes OA immediately.
There seems to be a problem in mesa master GIT branch.

commit 9d48a621d2a0e55a76a2cfd0aea3b773e907ed50
"llvmpipe: Fix crashes when there is no depth buffer bound."

OA is playable with "classic" mesa (KMS/DRI2).

After Anholt's OA benchmark is finished - I get a blank screen.

Log-file see [1].

- Sedat -

[1] http://files.iniza.org/agdf5-pm2-with-linustree/logs/dmesg_openarena-r300classic.txt

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

2010/3/18 Rafał Miłecki <zajec5@gmail.com>:
> W dniu 18 marca 2010 10:05 użytkownik Rafał Miłecki <zajec5@gmail.com> napisał:
>> Whoops, I missed one patch. My mistake.
>
> I've just applied all patches dated as "17-Mar-2010 22:42" and tested.
>
> Now memory reclocking is really enabled, but it causes corruptions for
> me. Maybe that's the same thing you experienced with dual head?
>
> Video is *far* from perfect, but maybe will give you some idea of what
> do I experience:
> http://estudent.put.poznan.pl/rafal.milecki/kernel/MOV00225.MP4
>
> The same happened for me before all your patches, when I tried to
> enable memory reclocking on my own.

yeah, the similar to what I saw with dualhead.  I guess we are either
missing the vblank window or the reclock is taking too much time.  I
wonder if it would make sense to wait for a vline rather than vblank,
say a vline 1/2 or 2/3 of the way down the screen, to give us some
buffer time between calling the reclock function and whatever setup it
does and the beginning of the actual vblank period.

Alex

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

W dniu 18 marca 2010 17:40 użytkownik Alex Deucher
<alexdeucher@gmail.com> napisał:
> 2010/3/18 Rafał Miłecki <zajec5@gmail.com>:
>> W dniu 18 marca 2010 10:05 użytkownik Rafał Miłecki <zajec5@gmail.com> napisał:
>>> Whoops, I missed one patch. My mistake.
>>
>> I've just applied all patches dated as "17-Mar-2010 22:42" and tested.
>>
>> Now memory reclocking is really enabled, but it causes corruptions for
>> me. Maybe that's the same thing you experienced with dual head?
>>
>> Video is *far* from perfect, but maybe will give you some idea of what
>> do I experience:
>> http://estudent.put.poznan.pl/rafal.milecki/kernel/MOV00225.MP4
>>
>> The same happened for me before all your patches, when I tried to
>> enable memory reclocking on my own.
>
> yeah, the similar to what I saw with dualhead.  I guess we are either
> missing the vblank window or the reclock is taking too much time.  I
> wonder if it would make sense to wait for a vline rather than vblank,
> say a vline 1/2 or 2/3 of the way down the screen, to give us some
> buffer time between calling the reclock function and whatever setup it
> does and the beginning of the actual vblank period.

If you suspect too slow resuming reclokcing context you can try tricks from:
http://marc.info/?l=linux-kernel&m=126730151530139&w=2

It's about setting higher priority for reclocking context or switching
to busy waiting.

-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Rafał Miłecki]

W dniu 18 marca 2010 17:40 użytkownik Alex Deucher
<alexdeucher@gmail.com> napisał:
> 2010/3/18 Rafał Miłecki <zajec5@gmail.com>:
>> W dniu 18 marca 2010 10:05 użytkownik Rafał Miłecki <zajec5@gmail.com> napisał:
>>> Whoops, I missed one patch. My mistake.
>>
>> I've just applied all patches dated as "17-Mar-2010 22:42" and tested.
>>
>> Now memory reclocking is really enabled, but it causes corruptions for
>> me. Maybe that's the same thing you experienced with dual head?
>>
>> Video is *far* from perfect, but maybe will give you some idea of what
>> do I experience:
>> http://estudent.put.poznan.pl/rafal.milecki/kernel/MOV00225.MP4
>>
>> The same happened for me before all your patches, when I tried to
>> enable memory reclocking on my own.
>
> yeah, the similar to what I saw with dualhead.  I guess we are either
> missing the vblank window or the reclock is taking too much time.  I
> wonder if it would make sense to wait for a vline rather than vblank,
> say a vline 1/2 or 2/3 of the way down the screen, to give us some
> buffer time between calling the reclock function and whatever setup it
> does and the beginning of the actual vblank period.

Did you play with this some more?

Is this possible reclocking memory in wrong moment (so missing VBLANK
probably) cause lock up? I experienced few already, when I forced
myself to use this code ignoring corruptions. Any other idea what can
cause lock up? I'm 100% sure it's memory reclocking code related.

-- 
Rafał

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Alex Deucher]

2010/3/26 Rafał Miłecki <zajec5@gmail.com>:
> W dniu 18 marca 2010 17:40 użytkownik Alex Deucher
> <alexdeucher@gmail.com> napisał:
>> 2010/3/18 Rafał Miłecki <zajec5@gmail.com>:
>>> W dniu 18 marca 2010 10:05 użytkownik Rafał Miłecki <zajec5@gmail.com> napisał:
>>>> Whoops, I missed one patch. My mistake.
>>>
>>> I've just applied all patches dated as "17-Mar-2010 22:42" and tested.
>>>
>>> Now memory reclocking is really enabled, but it causes corruptions for
>>> me. Maybe that's the same thing you experienced with dual head?
>>>
>>> Video is *far* from perfect, but maybe will give you some idea of what
>>> do I experience:
>>> http://estudent.put.poznan.pl/rafal.milecki/kernel/MOV00225.MP4
>>>
>>> The same happened for me before all your patches, when I tried to
>>> enable memory reclocking on my own.
>>
>> yeah, the similar to what I saw with dualhead.  I guess we are either
>> missing the vblank window or the reclock is taking too much time.  I
>> wonder if it would make sense to wait for a vline rather than vblank,
>> say a vline 1/2 or 2/3 of the way down the screen, to give us some
>> buffer time between calling the reclock function and whatever setup it
>> does and the beginning of the actual vblank period.
>
> Did you play with this some more?
>
> Is this possible reclocking memory in wrong moment (so missing VBLANK
> probably) cause lock up? I experienced few already, when I forced
> myself to use this code ignoring corruptions. Any other idea what can
> cause lock up? I'm 100% sure it's memory reclocking code related.

I get occasional lockups as well with the memory reclock.  I suspect
we may have to disable crtc mem requests when changing the mclk, but I
haven't really had a change to play more with this.  Matthew's been
looking it some more this week.

Alex

------------------------------------------------------------------------------
Download Intel&#174; Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--

Re: [PATCHES] radeon kms pm patches

[Matthew Garrett]

On Fri, Mar 26, 2010 at 08:49:21PM +0100, Rafał Miłecki wrote:

> Is this possible reclocking memory in wrong moment (so missing VBLANK
> probably) cause lock up? I experienced few already, when I forced
> myself to use this code ignoring corruptions. Any other idea what can
> cause lock up? I'm 100% sure it's memory reclocking code related.

The first thing is that memory reclocking is definitely missing vblank 
occasionally - but sometimes that's because the interrupt fires before 
we get there, and sometimes it's because we're scheduled later. I have 
some fixes for that. Other corruption is likely to be due to the engine 
reclocking, which seems to also need to be done in vblank.

As far as the lockups go, I think this is due to memory access during 
reclock. Nothing should be coming from the cp, but we may be getting 
accesses from the crtc if we miss vblank (there's a bit in the crtc 
control that can deal with this), but more importantly we're doing 
nothing to prevent unaccelerated fallbacks touching vram during reclock. 
I can provoke lockups more easily using x11perf which probably triggers 
several of those, so I lean towards that being something we need to fix 
before looking for other potential issues.

-- 
Matthew Garrett | mjg59@srcf.ucam.org

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Re: [PATCHES] radeon kms pm patches

[Michel Dänzer]

On Fri, 2010-03-26 at 19:56 +0000, Matthew Garrett wrote: 
> 
> As far as the lockups go, I think this is due to memory access during 
> reclock. Nothing should be coming from the cp, but we may be getting 
> accesses from the crtc if we miss vblank (there's a bit in the crtc 
> control that can deal with this), but more importantly we're doing 
> nothing to prevent unaccelerated fallbacks touching vram during reclock. 
> I can provoke lockups more easily using x11perf which probably triggers 
> several of those, so I lean towards that being something we need to fix 
> before looking for other potential issues.

Have you tried making RADEONPrepareAccess_CS() in the X driver always
return FALSE (you may also need to prevent RADEONUploadToScreenCS() from
returning FALSE if the BO isn't busy) to confirm this? That should
prevent any direct CPU access to VRAM, at least as long as you don't hit
any software fallbacks in the 3D driver.


-- 
Earthling Michel Dänzer           |                http://www.vmware.com
Libre software enthusiast         |          Debian, X and DRI developer

------------------------------------------------------------------------------
Download Intel® Parallel Studio Eval
Try the new software tools for yourself. Speed compiling, find bugs
proactively, and fine-tune applications for parallel performance.
See why Intel Parallel Studio got high marks during beta.
http://p.sf.net/sfu/intel-sw-dev
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel