> The order of job->base.s_fence and preventing concurrent CS was guaranteed by
> the VM PD being reserved, that is now no longer the case and we need a new
> lock for protection.
Thought of this either, but since UMD/APP submits their jobs in one thread upon
one ctx (if submits jobs within
On Fri, Sep 8, 2017 at 11:41 AM, Darren Salt wrote:
> Noticed while playing “Valley”, which was causing some 8MB of leakage per
> second. kmemleak listed many entries looking like this:
>
> unreferenced object 0x8802c2951800 (size 1024):
> comm "Xorg", pid
On 2017年09月11日 18:58, Christian König wrote:
From: Christian König
There is no guarantee that the last BO_VA actually needed an update.
Good catch. One comment inline
Additional to that all command submissions must wait for moved BOs to
be cleared, not just the
On 2017年09月11日 17:04, Christian König wrote:
NAK, as Roger pointed, the performance of Vulkan Dota2 drops from
72fps to 24fps.
This patches will kill Per-vm-BO advantages, even drop to 1/3 of
previous non-per-vm-bo.
This is irrelevant and actually a foreseen consequence of per VM BOs.
If
On Mon, Sep 11, 2017 at 09:22:03AM +0800, Evan Quan wrote:
> Change-Id: I53bcacd531d8df801a49acd0409c3f27b49fb1be
> Signed-off-by: Evan Quan
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 2 ++
> drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h | 2 ++
>
Am 11.09.2017 um 11:10 schrieb zhoucm1:
On 2017年09月11日 17:04, Christian König wrote:
NAK, as Roger pointed, the performance of Vulkan Dota2 drops from
72fps to 24fps.
This patches will kill Per-vm-BO advantages, even drop to 1/3 of
previous non-per-vm-bo.
This is irrelevant and actually
Patch 1, 3 are Reviewed-by: Huang Rui
With minor comments on Patch 2 and 4, with that fixed, please feel free to
add Reviewed-by: Huang Rui
Thanks,
Rui
On Mon, Sep 11, 2017 at 09:22:02AM +0800, Evan Quan wrote:
> Change-Id:
Reviewed-by: Roger He
Thanks
Roger(Hongbo.He)
-Original Message-
From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf Of
Christian K?nig
Sent: Monday, September 11, 2017 6:53 PM
To: Zhou, David(ChunMing) ;
need to unreserve ttm bo before "cs_add_fence" and "entity_push_job"
otherwise there will be deadlock between "recover_vram_from_shadow"
and previous two routines on the ttm bo's resv lock.
Change-Id: I9c18b677e474ce5b45a9cc24da3fa255d77b3d44
Signed-off-by: Monk Liu
---
Am 11.09.2017 um 10:10 schrieb Michel Dänzer:
From: Michel Dänzer
Fixes crash when trying to unload the amdgpu module before the fbdev
framebuffer was initialized, which can happen since the DRM fbdev helper
code supports deferred setup.
Signed-off-by: Michel Dänzer
From: Michel Dänzer
Fixes crash when trying to unload the amdgpu module before the fbdev
framebuffer was initialized, which can happen since the DRM fbdev helper
code supports deferred setup.
Signed-off-by: Michel Dänzer
---
NAK, as Roger pointed, the performance of Vulkan Dota2 drops from
72fps to 24fps.
This patches will kill Per-vm-BO advantages, even drop to 1/3 of
previous non-per-vm-bo.
This is irrelevant and actually a foreseen consequence of per VM BOs.
If the performance drops is in-acceptable then the
Am 11.09.2017 um 11:02 schrieb Monk Liu:
need to unreserve ttm bo before "cs_add_fence" and "entity_push_job"
otherwise there will be deadlock between "recover_vram_from_shadow"
and previous two routines on the ttm bo's resv lock.
Change-Id: I9c18b677e474ce5b45a9cc24da3fa255d77b3d44
From: Christian König
There is no guarantee that the last BO_VA actually needed an update.
Additional to that all command submissions must wait for moved BOs to
be cleared, not just the first one.
Signed-off-by: Christian König
---
From: Michel Dänzer
Fixes crash when trying to unload the amdgpu module before the fbdev
framebuffer was initialized, which can happen since the DRM fbdev helper
code supports deferred setup.
Signed-off-by: Michel Dänzer
---
Thanks for the review, please see some comments inline.
Am 08.09.2017 um 23:12 schrieb Felix Kuehling:
I think this change could use a little more explanation. I also have
some ideas for simplifying the code and making it easier to understand.
See comments inline.
But it looks good to me
NAK, as Roger pointed, the performance of Vulkan Dota2 drops from 72fps
to 24fps.
This patches will kill Per-vm-BO advantages, even drop to 1/3 of
previous non-per-vm-bo.
all moved and relocated fence have already synced, we just need to catch
the va mapping but which is CS itself required,
On Mon, Sep 11, 2017 at 09:22:05AM +0800, Evan Quan wrote:
> Change-Id: I7a0427555b34111e4e7600aa1e8b9cae31b44af7
> Signed-off-by: Evan Quan
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 2 --
> drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c | 2 +-
> 2 files changed, 1
Ok then we have a misunderstanding here. This applies to all per VM BOs,
not just kernel BOs.
It's just that as long as the per VM BOs aren't used the handling is
still correct and we don't have any updates to UMD BOs we implicitly
wait on.
Sorry I should have waited a bit more, but since
Am 10.09.2017 um 09:30 schrieb Maarten Lankhorst:
Op 04-09-17 om 21:02 schreef Christian König:
From: Christian König
Stop requiring that the src reservation object is locked for this operation.
Signed-off-by: Christian König
---
Hi Nicolai,
I don't get this patch, 'x' starts at 0 and goes to sgpr_size but that
doesn't include the offset into the SGPR space right?
I mean I get the patch in umr_read_sgprs() but in print_waves() won't
that mean you're printing out SGPRS[0..size]?
Or are you saying having the base
Op 11-09-17 om 14:53 schreef Christian König:
> Am 10.09.2017 um 09:30 schrieb Maarten Lankhorst:
>> Op 04-09-17 om 21:02 schreef Christian König:
>>> From: Christian König
>>>
>>> Stop requiring that the src reservation object is locked for this operation.
>>>
>>>
My mean is you should add a condition in code to check if BO is kernel BO or
not. I expect V3.
Regards,
David
发自坚果 Pro
He, Roger 于 2017年9月11日 下午6:57写道:
Reviewed-by: Roger He
Thanks
Roger(Hongbo.He)
-Original Message-
From: amd-gfx
It still doesn't make sense, I don't understand why we need all existed update
fence for CS, which could implicitly contain unrelated fence next CS used. For
moved list case, we already synced. We should explicitly sync the needed fence.
For per VM BOs, we lack mapping fence, that's why VM
I don't understand why we need all existed update fence for CS
We don't need all updates for CS, we just need all updates for per VM
BOs for CS because we don't know which BOs are used in the CS.
That's the overall problem of using per VM BOs, the kernel doesn't know
what is actually used by
From: Christian König
All users of a VM must always wait for updates with always
valid BOs to be completed.
v2: remove debugging leftovers, rename struct member
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c | 10
Am 11.09.2017 um 03:42 schrieb Zhang, Jerry (Junwei):
On 09/11/2017 09:22 AM, Evan Quan wrote:
Change-Id: I9b7ebc99b7c75c03fb46d16c4c49348dd551325e
Signed-off-by: Evan Quan
The series patches is
Reviewed-by: Junwei Zhang
A bit more commit message
Am 11.09.2017 um 02:55 schrieb Michel Dänzer:
On 09/09/17 12:52 AM, Christian König wrote:
From: Christian König
Just some cleanup.
Signed-off-by: Christian König
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h| 25 +---
Am 09.09.2017 um 12:49 schrieb Nicolai Hähnle:
From: Nicolai Hähnle
This is already hooked up to the "amdgpu_gpr" debugfs file used by
the umr userspace debugging tool.
Signed-off-by: Nicolai Hähnle
Acked-by: Christian König
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Christian König
> Sent: Thursday, September 07, 2017 4:17 AM
> To: amd-gfx@lists.freedesktop.org
> Subject: [PATCH libdrm] amdgpu: make userptr unit test more interesting
>
> From: Christian
From: Christian König
There is no guarantee that the last BO_VA actually needed an update.
Additional to that all command submissions must wait for moved BOs to
be cleared, not just the first one.
v2: Don't overwrite any newer fence.
Signed-off-by: Christian König
Am 11.09.2017 um 17:13 schrieb Maarten Lankhorst:
Op 11-09-17 om 16:45 schreef Christian König:
Am 11.09.2017 um 15:56 schrieb Maarten Lankhorst:
Op 11-09-17 om 14:53 schreef Christian König:
Am 10.09.2017 um 09:30 schrieb Maarten Lankhorst:
[SNIP]
To be honest that looks rather ugly to me
call to memset to assign 0 value immediately after allocating
memory with kzalloc is unnecesaary as kzalloc allocates the memory
filled with 0 value.
Semantic patch used to resolve this issue:
@@
expression e,e2; constant c;
statement S;
@@
e = kzalloc(e2, c);
if(e == NULL) S
- memset(e, 0,
On Fri, Sep 8, 2017 at 2:40 AM, Rex Zhu wrote:
> Change-Id: If3ab520f8c65bec070e97cd8787041919d687ad3
> Signed-off-by: Rex Zhu
Series is:
Reviewed-by: Alex Deucher
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c | 79
>
Am 11.09.2017 um 15:56 schrieb Maarten Lankhorst:
Op 11-09-17 om 14:53 schreef Christian König:
Am 10.09.2017 um 09:30 schrieb Maarten Lankhorst:
Op 04-09-17 om 21:02 schreef Christian König:
From: Christian König
Stop requiring that the src reservation object is
Op 11-09-17 om 16:45 schreef Christian König:
> Am 11.09.2017 um 15:56 schrieb Maarten Lankhorst:
>> Op 11-09-17 om 14:53 schreef Christian König:
>>> Am 10.09.2017 um 09:30 schrieb Maarten Lankhorst:
Op 04-09-17 om 21:02 schreef Christian König:
> From: Christian König
From: Jan Andres
The line in question seems to assume some kind of correlation between
the numbering of the audio converters and the stream encoders, which
doesn't exist on my card: Here, the active stream encoders are indexed
2 and 3 while audios have index 0 and 1, so audio
On Mon, Sep 11, 2017 at 4:24 AM, Christian König
wrote:
> Am 11.09.2017 um 10:10 schrieb Michel Dänzer:
>>
>> From: Michel Dänzer
>>
>> Fixes crash when trying to unload the amdgpu module before the fbdev
>> framebuffer was initialized, which can
On 11.09.2017 15:31, Tom St Denis wrote:
Hi Nicolai,
I don't get this patch, 'x' starts at 0 and goes to sgpr_size but that
doesn't include the offset into the SGPR space right?
I mean I get the patch in umr_read_sgprs() but in print_waves() won't
that mean you're printing out
Am 11.09.2017 um 17:22 schrieb Christian König:
Am 11.09.2017 um 17:13 schrieb Maarten Lankhorst:
Op 11-09-17 om 16:45 schreef Christian König:
Am 11.09.2017 um 15:56 schrieb Maarten Lankhorst:
Op 11-09-17 om 14:53 schreef Christian König:
Am 10.09.2017 um 09:30 schrieb Maarten Lankhorst:
Op 11-09-17 om 17:24 schreef Christian König:
> Am 11.09.2017 um 17:22 schrieb Christian König:
>> Am 11.09.2017 um 17:13 schrieb Maarten Lankhorst:
>>> Op 11-09-17 om 16:45 schreef Christian König:
Am 11.09.2017 um 15:56 schrieb Maarten Lankhorst:
> Op 11-09-17 om 14:53 schreef Christian
> -Original Message-
> From: Kuehling, Felix
> Sent: Wednesday, September 06, 2017 5:54 PM
> To: amd-gfx@lists.freedesktop.org; Deucher, Alexander; Oded Gabbay;
> Koenig, Christian
> Subject: Re: [PATCH 0/8] Retry page fault handling for Vega10
>
> I realized that the drm-next-4.15-wip
Several users have complained that the tile table update broke Oland
support. Despite several attempts to fix it, the root cause is still
unknown at this point and no solution is available. As it is not
acceptable to leave a known regression breaking a major functionality
in the kernel for several
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Alexandre Demers
> Sent: Thursday, September 07, 2017 10:49 PM
> To: amd-gfx@lists.freedesktop.org
> Subject: [PATCH 2/6] Reorganize, rename, move and split some VCE 1
> defines.
>
> For
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Alexandre Demers
> Sent: Thursday, September 07, 2017 10:49 PM
> To: amd-gfx@lists.freedesktop.org
> Subject: [PATCH 1/6] Add initial VCE_V1_0 files.
>
> Signed-off-by: Alexandre Demers
Am 11.09.2017 um 17:29 schrieb Maarten Lankhorst:
Op 11-09-17 om 17:24 schreef Christian König:
Am 11.09.2017 um 17:22 schrieb Christian König:
Am 11.09.2017 um 17:13 schrieb Maarten Lankhorst:
Op 11-09-17 om 16:45 schreef Christian König:
Am 11.09.2017 um 15:56 schrieb Maarten Lankhorst:
From: Yongqiang Sun
In case of vstartup happens before vsync, set vertical line
start to 0.
Signed-off-by: Yongqiang Sun
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
From: Wenjing Liu
Signed-off-by: Wenjing Liu
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff
From: Hersen Wu
TODO: Current change only work for sucessive last fused pipe,
like p3, or p3,p2. It does not work for fused p1,p2.
Signed-off-by: Hersen Wu
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
From: Yue Hin Lau
Signed-off-by: Yue Hin Lau
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dwb.c | 288 +
From: Tony Cheng
bug: default is mapped to no dithering.
default to spatial dithering based on color depth
Signed-off-by: Tony Cheng
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
From: Yue Hin Lau
Signed-off-by: Yue Hin Lau
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
drivers/gpu/drm/amd/display/dc/dcn10/Makefile | 2 +-
drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.h
* Bunch of Raven fixes and other work
* Stop powering down clock sources in atomic check
* Stop using amdgpu_connector and use our own
* Fix display unplug during S3 or fbcon
Andrew Jiang (1):
drm/amd/display: Fix context alloc failed logging
Charlene Liu (1):
drm/amd/display: fix
From: Eric Bernstein
Signed-off-by: Eric Bernstein
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.c | 3 +--
From: Yongqiang Sun
Signed-off-by: Yongqiang Sun
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c | 6 --
1 file changed, 6 deletions(-)
From: Andrew Jiang
Since there was no return statement in the fail block immediately
preceding the context_alloc_fail block, any failure within the
function caused a context alloc failed error message to be printed.
Since the context_alloc_fail block is only used once, move
From: Logatharshan Thothiralingam
Signed-off-by: Logatharshan Thothiralingam
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
Powering down the clock source during unref is unsafe as we might want
to unref during atomic_check
Signed-off-by: Harry Wentland
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
From: Shirish S
For MPO to work with per surface rendering and flipping,
the previous logic of restricting plane[1] only as underlay
needs to be removed.
validate_surface_sets() now checks only the width and height
bounds in case of underlay rather than checking format.
This should be taken care of in validate now.
All of timing sync is quite broken at the moment anyways. Will submit
another patch set to address that.
Signed-off-by: Harry Wentland
Reviewed-by: Andrey Grodzovsky
Reviewed-by: Tony Cheng
Still one more in dc_validate_global
Signed-off-by: Harry Wentland
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
drivers/gpu/drm/amd/display/dc/core/dc_resource.c | 19 ++---
From: Hersen Wu
Signed-off-by: Hersen Wu
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 9 ---
drivers/gpu/drm/amd/display/dc/core/dc_link.c
From: Wenjing Liu
[Description]
We originally use TPS4 phy test pattern for test pattern 7.
On RV we switched to a new method to use CP2520.
CP2520 should produce the same result.
However in reality, it fails DP PHY automation test.
We use the original method instead.
From: Charlene Liu
Signed-off-by: Charlene Liu
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
.../drm/amd/display/dc/dce110/dce110_hw_sequencer.c | 20 +++-
1 file changed, 19
This should only happen on full update. If this ever happens
on regular pageflips it needs to be debugged.
Signed-off-by: Harry Wentland
Reviewed-by: Tony Cheng
Acked-by: Harry Wentland
---
On Mon, Sep 11, 2017 at 10:29 PM, Deucher, Alexander
wrote:
>> -Original Message-
>> From: Kuehling, Felix
>> Sent: Wednesday, September 06, 2017 5:54 PM
>> To: amd-gfx@lists.freedesktop.org; Deucher, Alexander; Oded Gabbay;
>> Koenig, Christian
>> Subject: Re:
On Mon, Sep 11, 2017 at 8:37 AM, Himanshu Jha
wrote:
> call to memset to assign 0 value immediately after allocating
> memory with kzalloc is unnecesaary as kzalloc allocates the memory
> filled with 0 value.
>
> Semantic patch used to resolve this issue:
>
> @@
>
From: Hersen Wu
RV1 support only USB-C active DP-2-HDMI dongle. HPD short pulse is
generated only for DP signal.
When processing HPD short pulse, it must be DP active dongle. No need
for I2C-Over-AUX detection.
v2: Add description
Signed-off-by: Hersen Wu
This seems to lack a commit msg
Dave.
On 12 Sep. 2017 4:15 am, "Harry Wentland" wrote:
> From: Hersen Wu
>
> Signed-off-by: Hersen Wu
> Reviewed-by: Tony Cheng
> Acked-by: Harry Wentland
69 matches
Mail list logo