[PATCH] LUT: reduce number of writes to DC_LUT_RW_INDEX and properly barrier (was: Mucked up colours on HD2400 with weird behaviour)

newer
0x9490:0x1787:0x2003: HIS Radeon...

Yang Zhao

26 Mar 2009 26 Mar '09

02:56

Proper patch for incorrect LUT writing when out-of-order execution is used. Also reduces the overall number of writes to DC_LUT_RW_INDEX. Cheers, -- Yang Zhao http://yangman.ca

Attachments:

0001-LUT-reduce-number-of-writes-to-DC_LUT_RW_INDEX-and.patch (text/x-patch — 3.8 KB)

Show replies by date

Michael Cree

26 Mar 26 Mar

03:47

New subject: [PATCH] LUT: reduce number of writes to DC_LUT_RW_INDEX and properly barrier (was: Mucked up colours on HD2400 with weird behaviour)

On 26/03/2009, at 3:56 PM, Yang Zhao wrote:

...

Proper patch for incorrect LUT writing when out-of-order execution is used. Also reduces the overall number of writes to DC_LUT_RW_INDEX.

I suggest holding off applying the patch as the proper fix is to fix the RHDRegRead and RHDRegWrite routines to work for all architectures. Cheers Michael. -- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org

Yang Zhao

05:22

New subject: [PATCH] LUT: reduce number of writes to DC_LUT_RW_INDEX and properly barrier (was: Mucked up colours on HD2400 with weird behaviour)

2009/3/25 Michael Cree :

...

On 26/03/2009, at 3:56 PM, Yang Zhao wrote:

...
Proper patch for incorrect LUT writing when out-of-order execution is used. Also reduces the overall number of writes to DC_LUT_RW_INDEX.

I suggest holding off applying the patch as the proper fix is to fix the RHDRegRead and RHDRegWrite routines to work for all architectures.

Are the MMIO routines in compiler.h always barriered for Alpha? If that's the case, the cleanup still applies, just without the calls to write_mem_barrier(). -- Yang Zhao http://yangman.ca -- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org

Michael Cree

22:08

New subject: [PATCH] LUT: reduce number of writes to DC_LUT_RW_INDEX and properly barrier (was: Mucked up colours on HD2400 with weird behaviour)

On 26/03/2009, at 6:22 PM, Yang Zhao wrote:

...

2009/3/25 Michael Cree :

...
On 26/03/2009, at 3:56 PM, Yang Zhao wrote:

...
Proper patch for incorrect LUT writing when out-of-order execution is used. Also reduces the overall number of writes to DC_LUT_RW_INDEX.

I suggest holding off applying the patch as the proper fix is to fix the RHDRegRead and RHDRegWrite routines to work for all architectures.

Are the MMIO routines in compiler.h always barriered for Alpha?

Yes, they include barriers for a number of architectures, not just Alpha. They also include byte swapping for big-endian architectures. That is the point of them - they provide a portable means of accessing the IO ports. Michael. -- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org

Yang Zhao

23:02

New subject: [PATCH] LUT: reduce number of writes to DC_LUT_RW_INDEX and properly barrier (was: Mucked up colours on HD2400 with weird behaviour)

2009/3/26 Michael Cree :

...

On 26/03/2009, at 6:22 PM, Yang Zhao wrote:

...
2009/3/25 Michael Cree :

...
On 26/03/2009, at 3:56 PM, Yang Zhao wrote:

...
Proper patch for incorrect LUT writing when out-of-order execution is used. Also reduces the overall number of writes to DC_LUT_RW_INDEX.

I suggest holding off applying the patch as the proper fix is to fix the RHDRegRead and RHDRegWrite routines to work for all architectures.

Are the MMIO routines in compiler.h always barriered for Alpha?

Yes, they include barriers for a number of architectures, not just Alpha. They also include byte swapping for big-endian architectures.

Thanks. Revised patch attached. This is orthogonal to the proposed fixes to RHDRegWrite(). Cheers, -- Yang Zhao http://yangman.ca

Matthias Hopf

3 Apr 3 Apr

10:03

New subject: [PATCH] LUT: reduce number of writes to DC_LUT_RW_INDEX and properly barrier (was: Mucked up colours on HD2400 with weird behaviour)

On Mar 26, 09 16:02:36 -0700, Yang Zhao wrote:

...

...
...
...
...
Proper patch for incorrect LUT writing when out-of-order execution is used. Also reduces the overall number of writes to DC_LUT_RW_INDEX. Thanks. Revised patch attached. This is orthogonal to the proposed fixes to RHDRegWrite().

Looks good. Applied. Thanks Matthias -- Matthias Hopf __ __ __ Maxfeldstr. 5 / 90409 Nuernberg (_ | | (_ |__ mat@mshopf.de Phone +49-911-74053-715 __) |_| __) |__ R & D www.mshopf.de -- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org

5503

Age (days ago)

5511

Last active (days ago)

List overview

Download

5 comments

3 participants

participants (3)

Matthias Hopf
Michael Cree
Yang Zhao