1,drop inserting double SWITCH_BUFFERS scheme, which impacts performance,
because double SWITCH_BUFFER actaully doesn't switch buffer, so the CE's
ping-pong buffer is not used at all. Now only insert one SWITCH_BUFFERS at the
bottom of each GFX dmaframe. And only inserting one SWITCH_BUFFER is
-Original Message-
From: Monk Liu [mailto:monk@amd.com]
Sent: Wednesday, August 24, 2016 12:00 PM
To: amd-gfx@lists.freedesktop.org
Cc: Liu, Monk
Subject: [PATCH] drm/amdgpu:fix DMAframe for GFX8
1,drop inserting double SWITCH_BUFFERS scheme, which impacts
1,drop inserting double SWITCH_BUFFERS scheme, which impacts
performance, because double SWITCH_BUFFER actaully doesn't
switch buffer, so the CE's ping-pong buffer is not used at
all. Now only insert one SWITCH_BUFFERS at the bottom of
each GFX dmaframe. And only inserting one SWITCH_BUFFER is
a
Looks very good to me, binding time and byte is good idea according free
vram.
Did you test more games for your algorithm?
Regards,
David Zhou
On 2016年08月24日 06:19, Marek Olšák wrote:
From: Marek Olšák
The old mechanism used a per-submission limit that didn't take
Or, we also can use PM_EVENT_ messages to separate freeze, thaw, suspend,
resume, poweroff and restore.
Thanks
JimQu
发件人: amd-gfx 代表 Michel Dänzer
发送时间: 2016年8月24日 9:03:20
收件人: Alex Deucher
On 24/08/16 02:45 AM, Alex Deucher wrote:
> Some blocks require a reset to properly resume if there
> is no power down of the asic like during various
> hibernation steps.
>
> Reset the asic on freeze and force asic_init on thaw.
>
> Port of:
> 274ad65c9d02bdcbee9bae045517864c3521d530
>
From: Marek Olšák
The old mechanism used a per-submission limit that didn't take previous
submissions within the same time frame into account. It also filled VRAM
slowly when VRAM usage dropped due to a big eviction or buffer deallocation.
This new method establishes a
Replace packet2's with packet0 writes to UVD_NO_OP. The
value written to UVD_NO_OP does not matter.
Reviewed-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/radeon/cik.c| 2 +-
drivers/gpu/drm/radeon/cikd.h
Replace packet2's with packet0 writes to UVD_NO_OP. The
value written to UVD_NO_OP does not matter.
Reviewed-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 6 --
Writes to this register are the preferred way to do NOPs.
Bump the driver version as well.
Reviewed-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/radeon/r600d.h | 1 +
drivers/gpu/drm/radeon/radeon_drv.c | 3 ++-
Writes to this register are the preferred way to do NOPs.
Bump the driver version as well.
Reviewed-by: Christian König
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 3 ++-
On Tue, Aug 23, 2016 at 8:01 PM, Marek Olšák wrote:
> On Thu, Aug 18, 2016 at 10:52 AM, Christian König
> wrote:
>> In that case my patch should clearly help with that.
>>
>> Going to release what I have so far today, but looks like I need more time
>>
On Thu, Aug 18, 2016 at 10:52 AM, Christian König
wrote:
> In that case my patch should clearly help with that.
>
> Going to release what I have so far today, but looks like I need more time
> actually fixing this.
>
> In the meantime you could try the "drm/amdgpu: fix
Add support for asic_reset and asic_init to the device
suspend and resume functions.
No change in functionality.
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h| 6 --
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 14 +-
Some blocks require a reset to properly resume if there
is no power down of the asic like during various
hibernation steps.
Reset the asic on freeze and force asic_init on thaw.
Port of:
274ad65c9d02bdcbee9bae045517864c3521d530
(drm/radeon: hard reset r600 and newer GPU when hibernating.) from
The old names were dragged over from radeon. The new ones
better match the naming conventions used in the driver.
No functional change.
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h| 4 ++--
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c |
According to the hw team, it should be 16, not 8.
Cc: Peter Fang
Signed-off-by: Alex Deucher
Cc: sta...@vger.kernel.org
---
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Rex Zhu
> Sent: Monday, August 22, 2016 8:50 AM
> To: amd-gfx@lists.freedesktop.org
> Cc: Zhu, Rex
> Subject: [PATCH 4/4] drm/amd/powerplay: delete useless code in
> iceland_hwmgr.c.
>
>
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Tom St Denis
> Sent: Monday, August 22, 2016 11:01 AM
> To: amd-gfx@lists.freedesktop.org
> Cc: StDenis, Tom
> Subject: [PATCH] drm/amd/amdgpu: Print ring name in
> amdgpu_ib_schedule()
>
>
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Christian König
> Sent: Tuesday, August 23, 2016 5:20 AM
> To: amd-gfx@lists.freedesktop.org
> Subject: [PATCH 1/2] drm/amdgpu: use memcpy_to/fromio for UVD fw
> upload
>
> From: Christian
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Alexandre Demers
> Sent: Monday, August 22, 2016 11:29 PM
> To: amd-gfx@lists.freedesktop.org
> Cc: Alexander Deucher; Christian König
> Subject: amdgpu's dce_vX_0_afmt_setmode(): why so
According to the hw team, it should be 16, not 8.
Cc: Peter Fang
Signed-off-by: Alex Deucher
Cc: sta...@vger.kernel.org
---
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git
Hi Alex:
Are you working on S4 issue ?
Thanks
JimQu
发件人: amd-gfx 代表 Alex Deucher
发送时间: 2016年8月23日 0:05:24
收件人: amd-gfx@lists.freedesktop.org
抄送: Deucher, Alexander
主题: [PATCH 1/2]
On 08/22/2016 08:27 PM, Michel Dänzer wrote:
On 23/08/16 10:18 AM, Reid Hekman wrote:
I was encouraged by a commit I saw today to add Sea Islands PCI ids to
xf86-video-amdgpu. However I did not see MULLINS included. Are there
other showstoppers preventing inclusion?
I assume the commit
From: Christian König
Try to be clean here, even when it's a noop on x86.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git
OK.
Thanks
JimQu
发件人: Christian König
发送时间: 2016年8月23日 16:53:50
收件人: Qu, Jim; amd-gfx@lists.freedesktop.org
主题: Re: 答复: [PATCH] drm/amdgpu: fix the wrong value of UVD bo size
> I will update the patch.
Let me do this. I already
Hi Christian:
Yes, I have misunderstand the AMDGPU_UVD_FIRMWARE_OFFSET in
amdgpu_uvd_mc_resume().
I will update the patch.
Thanks
JimQu
发件人: Christian König
发送时间: 2016年8月23日 16:02:11
收件人: Qu, Jim; amd-gfx@lists.freedesktop.org
Am 22.08.2016 um 20:30 schrieb Alex Deucher:
Normally on shutdown or reboot we don't care about necessarily
making sure the hw is in a good state because the system is about
to be powered down or reset. However, after a shutdown or reboot
in a VM, it's best to tear down the hw properly
Am 22.08.2016 um 18:05 schrieb Alex Deucher:
Some blocks require a reset to properly resume if there
is no power down of the asic like during various
hibernation steps.
Port of:
274ad65c9d02bdcbee9bae045517864c3521d530
(drm/radeon: hard reset r600 and newer GPU when hibernating.)
from radeon.
Am 23.08.2016 um 08:20 schrieb jimqu:
Driver allocate bo for restore UVD fw ,cache, stack , etc.
the firmware size is not only ucode size but the ucode and header.
Nope that isn't correct. Only the firmware is copied to the BO, not the
header. See amdgpu_uvd_resume().
Saying this we should
Driver allocate bo for restore UVD fw ,cache, stack , etc.
the firmware size is not only ucode size but the ucode and header.
Change-Id: I886c099fa123c4814de9b7db3559c30e7b41fd1b
Signed-off-by: JimQu
---
drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c | 2 +-
1 file changed, 1
31 matches
Mail list logo