thanks for this catch!
Reviewed-by: Monk Liu
发件人: Colin King
发送时间: 2017年2月4日 4:23:42
收件人: Deucher, Alexander; Koenig, Christian; David Airlie; Liu, Monk; Yu,
Xiangliang; amd-gfx@lists.freedesktop.org;
Instead of taking the first pipe and givint the rest to kfd, take the
first 2 queues of each pipe.
Effectively, amdgpu and amdkfd own the same number of queues. But
because the queues are spread over multiple pipes the hardware will be
able to better handle concurrent compute workloads.
amdgpu
The gfxv7 contains a slightly different version of cik_mqd called
bonaire_mqd. This can introduce subtle bugs if fixes are not applied in
both places.
Signed-off-by: Andres Rodriguez
---
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 135 ++
1 file
Take ownership of pipe initialization away from KFD.
Note that hpd_eop_gpu_addr was already large enough to accomodate all
pipes.
Signed-off-by: Andres Rodriguez
---
drivers/gpu/drm/radeon/cik.c| 27 ++-
drivers/gpu/drm/radeon/radeon_kfd.c |
Use the same gfx_*_mqd_commit function for kfd and amdgpu codepaths.
This removes the last duplicates of this programming sequence.
Signed-off-by: Andres Rodriguez
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 51 ++-
Pipes provide better concurrency than queues, therefore we want to make
sure that apps use queues from different pipes whenever possible.
Optimize for the trivial case where an app will consume rings in order,
therefore we don't want adjacent rings to belong to the same pipe.
Signed-off-by:
Update the KGD to KFD interface to allow sharing pipes with queue
granularity instead of pipe granularity.
This allows for more interesting pipe/queue splits.
Signed-off-by: Andres Rodriguez
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 22 -
Rename straggler instances of r(adeon)dev to a(mdgpu)dev
Signed-off-by: Andres Rodriguez
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 70 +++---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h | 14 +++---
drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c |
This information is already available in adev.
Signed-off-by: Andres Rodriguez
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c | 12 ++--
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c | 12 ++--
2 files changed, 12 insertions(+), 12 deletions(-)
Previously the queue/pipe split with kfd operated with pipe
granularity. This patch allows amdgpu to take ownership of an arbitrary
set of queues.
It also consolidates the last few magic numbers in the compute
initialization process into mec_init.
Signed-off-by: Andres Rodriguez
The current queue/pipe split policy is for amdgpu to take the first pipe of
MEC0 and leave the rest for amdkfd to use. This policy is taken as an
assumption in a few areas of the implementation.
This patch series aims to allow for flexible/tunable queue/pipe split policies
between kgd and kfd. It
The CP_MEC_DOORBELL_RANGE_* and CP_PQ_STATUS.DOORBELL_ENABLE registers
are not HQD specific.
They only need to be set once if at least 1 pipe requested doorbell
support.
Signed-off-by: Andres Rodriguez
---
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 +
From: Colin Ian King
bo_va is being kfree'd twice, once in the call to amdgpu_vm_bo_rmv
and then a short while later. Fix this double free by removing
the 2nd kfree.
Detected by CoverityScan, CID#1399524 ("Double Free")
Signed-off-by: Colin Ian King
On Thu, Feb 2, 2017 at 4:56 PM, Andrey Grodzovsky
wrote:
> This series is a folow-up on
> https://patchwork.kernel.org/patch/9501787/
>
> The first patch makes changes to atomic helpers to allow for drives with
> ASYNC flip support to use them.
> Patch 2 is to use this
On Fri, Feb 3, 2017 at 12:23 PM, Kent Russell wrote:
> This cleans up a remnant from Christian's patch (drm/ttm: revert "add
> optional LRU removal callback v2")
>
> Kent
>
FWIW, I already squashed the same fix into the patch when it went
upstream. Feel free to apply to
This cleans up a remnant from Christian's patch (drm/ttm: revert "add
optional LRU removal callback v2")
Kent
>From dfdf89a8dff0e378db6fa9578f75b02eec02ca0e Mon Sep 17 00:00:00 2001
From: Kent Russell
Date: Tue, 31 Jan 2017 06:44:10 -0500
Subject: [PATCH] drm/ttm: Fix
This cleans up a remnant from Christian's patch (drm/ttm: revert "add
optional LRU removal callback v2")
Kent
>From dfdf89a8dff0e378db6fa9578f75b02eec02ca0e Mon Sep 17 00:00:00 2001
From: Kent Russell
Date: Tue, 31 Jan 2017 06:44:10 -0500
Subject: [PATCH] drm/ttm: Fix
gcc-4.8 warns about '{0}' being used an an initializer for nested structures:
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c: In function
‘gfx_v8_0_ring_emit_ce_meta_init’:
drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c:7263:2: warning: missing braces around
initializer [-Wmissing-braces]
} ce_payload = {0};
Reviewed-by: Alex Deucher
From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf Of Zhu,
Rex
Sent: Friday, February 03, 2017 7:10 AM
To: Christian König; amd-gfx@lists.freedesktop.org
Cc: Alex Deucher
Subject: Re: [PATCH 00/12] mise patch related
On 02/02/17 06:36 PM, Christian König wrote:
> Am 02.02.2017 um 07:09 schrieb Michel Dänzer:
>> [SNIP]
>> OTOH the people running the kernel aren't always the same people
>> building it, so the downside is that this would potentially delay
>> getting X86_PAT enabled.
>
> And exactly for this
On 02.02.2017 11:25, Christian König wrote:
From: Christian König
Enable/disable the handling globally for now and
print a warning when we enable it for the first time.
v2: write to the correct register, adjust bits to that hw generation
Signed-off-by: Christian
Thanks Alex and Christian.
Patch7 is just a workaround for issue: when dpm disabled, can't initialize uvd
on Ci.
Now I have found the root cause. so no need to add timeout time.
Please review the attached patch.
Best Regards
Rex
From: amd-gfx
22 matches
Mail list logo