The bug I hit is because in my local branch that vmid counter logic is changed
( I optimized unnecessary vm-flush after gpu recover), Christian is right on
that,
So with drm-next branch my bug won't hit, I'm fine with sched_dep there as long
as not such issue.
Thanks
/Monk
-Original Messa
> So you have a pipeline sync when you don't need one and that is really really
> bad for things shared between processes, e.g. X/Wayland and it's clients.
Oh, that may explain the thing here: My environment is a no-X-window system
(customer's cloud gaming user case), so I don't launch X at all,
On Fri, Nov 02, 2018 at 04:26:49PM +0800, Chunming Zhou wrote:
> Signed-off-by: Chunming Zhou
> ---
> tests/amdgpu/Makefile.am | 3 +-
> tests/amdgpu/amdgpu_test.c | 12 ++
> tests/amdgpu/amdgpu_test.h | 21 +++
> tests/amdgpu/meson.build | 2 +-
> tests/amdgpu/syncobj_tests.c |
> -Original Message-
> From: Daniel Vetter On Behalf Of Daniel Vetter
> Sent: Monday, November 05, 2018 5:39 PM
> To: Zhou, David(ChunMing)
> Cc: dri-de...@lists.freedesktop.org; amd-gfx@lists.freedesktop.org
> Subject: Re: [PATCH libdrm 5/5] [libdrm] add syncobj timeline tests
>
> On
> BTW: could we let the Job remember the hw fence seq that it need to sync up
> to ? e.g. in "drm_sched_entity_clear_dep" we not only wake up scheduler but
> also set the hw fence seq number to the job (and keep the big one), so in the
> end in amdgpu_ib_schedule(), we knows exactly the last seq
From: Likun Gao
Hi,
Those series of patch modified the code of RLC to improve the resuabillity of
RLC's code. The process was separate into three part:
Part1[PATCH 1/3]: Unify RLC's function into the struct amdgpu_rlc_funcs and
change the
method of calling RLC.
Part2[PATCH 2/3]: Abstract RLC's
From: Likun Gao
Put function rlc_init,rlc_fini,rlc_resume,rlc_stop,rlc_start into structure
amdgpu_rlc_funcs and change the method to call rlc function for each verssion of
GFX.
Signed-off-by: Likun Gao
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 6 ++
drivers/gpu/drm/amd/amdgpu/gfx_v6_
From: Likun Gao
separate the function and struct of RLC from the file of GFX
Signed-off-by: Likun Gao
---
drivers/gpu/drm/amd/amdgpu/Makefile | 1 +
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c | 202 +---
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.h | 70 +-
drive
From: Likun Gao
abstract the function of amdgpu_gfx_rlc_enter/exit_safe_mode,
amdgpu_gfx_rlc_fini and some part of rlc_init to improve the reusability of RLC.
Signed-off-by: Likun Gao
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gfx.c| 201
drivers/gpu/drm/amd/amdgpu/
Hi Likun,
I'm not deep enough into what and how the RLC does it's job to
completely judge but in general that looks like a really nice cleanup.
Just two notes:
1. Don't add function first to amdgpu_gfx.[ch] and in the next patch
move them over to amdgpu_rlc.[ch]. Just add them directly to the
Hi Christian
For scenario: Bad Job (hang, vmid1) -->Job A (context 10, explicit dep for Job
B, vmid2) --> Job B(context 10, vmid2) --> Job C (context 11, vmid3)
Assume "job_hang_limit" is 0, and assume "sched_hw_submission" is 4, I give a
second thought on the logic after GPU reset:
1) the ba
> That won't work correctly because the sequence wraps around from time to time
> and you can't correctly handle that in PM4 AFAIK. Well maybe with two waits,
> but not 100% sure.
Hmmm, I still want to give a try, and I still feel current logic is wrong after
TDR , see my reply in another email
> and later its VMID's "current_gpu_reset_count" is updated to
> "adev->gpu_reset_count"
The question is how much later that is done. My recollection is that we
don't reset that for resubmission, but that could be wrong.
Anyway I think the cleanest approach to always handle that correctly
would
> Anyway I think the cleanest approach to always handle that correctly would be
> to always insert a vm flush before all jobs on resubmission.
That is most likely better for VM flush handling as well.
Yeah, that’s true and more simple
/Monk
-Original Message-
From: Koenig, Christian
S
> The question is how much later that is done. My recollection is that we don't
> reset that for resubmission, but that could be wrong.
According to my code (drm-next) VMID's "current_gpu_rest_counter" is updated in
"vm_flush" routine, so it is always there for either resubmission or not ...
Se
+amdgfx, amdgpu specific patches should go here
On 2018-11-05 05:33 AM, Shaokun Zhang wrote:
> RETIMER_REDRIVER_INFO shows the buffer as a decimal value with a '0x'
> prefix, which is somewhat misleading.
>
> Fix it to print hexadecimal, as was intended.
>
> Fixes: 2f14bc89("drm/amd/display: add
On 2018-11-01 9:51 p.m., Lyude Paul wrote:
> [why]
> Removing connector reusage from DM to match the rest of the tree ended
> up revealing an issue that was surprisingly subtle. The original amdgpu
> code for DC that was submitted appears to have left a chunk in
> dm_dp_create_fake_mst_encoder() th
On 9 Oct 2018, Michel Dänzer stated:
> On 2018-10-05 7:14 p.m., Nick Alcock wrote:
>> On 5 Oct 2018, Michel Dänzer told this:
>>
>>> On 2018-10-04 9:58 p.m., Nick Alcock wrote:
So a few days ago I started getting sprays of these warnings --
sorry, but because it was a few days ago I'm n
The value is dependent on whether fbc is available.
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c
b/drivers/gpu/drm/am
Causes a black screen on Stoney laptop.
bug: https://bugs.freedesktop.org/show_bug.cgi?id=108577
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce
[Why]
drm_plane_cleanup does not free the plane.
[How]
Call drm_primary_helper_destroy which will also free the plane.
Signed-off-by: Harry Wentland
---
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/amd/disp
It's unnecessarily duplicating drm_plane_type.
Signed-off-by: Harry Wentland
---
drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h | 8 +---
.../gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c | 38 +--
2 files changed, 20 insertions(+), 26 deletions(-)
diff --git a/drivers/gpu/drm/amd/a
> -Original Message-
> From: amd-gfx On Behalf Of
> Harry Wentland
> Sent: Monday, November 5, 2018 3:45 PM
> To: amd-gfx@lists.freedesktop.org
> Cc: Wentland, Harry
> Subject: [PATCH 2/2] drm/amd/display: Stop leaking planes
>
> [Why]
> drm_plane_cleanup does not free the plane.
>
> [H
> -Original Message-
> From: amd-gfx On Behalf Of
> Harry Wentland
> Sent: Monday, November 5, 2018 3:45 PM
> To: amd-gfx@lists.freedesktop.org
> Cc: Wentland, Harry
> Subject: [PATCH 1/2] drm/amdgpu: Drop amdgpu_plane
>
> It's unnecessarily duplicating drm_plane_type.
>
> Signed-off-by
On 2018-11-04 2:20 p.m., Christian König wrote:
> Am 02.11.18 um 19:59 schrieb Kuehling, Felix:
>> On 2018-11-02 9:48 a.m., Christian König wrote:
>>> Vega10 has multiple interrupt rings,
>> I don't think I've seen your code that implements multiple interrupt
>> rings. So it's a bit hard to comment
On 2018-11-05 2:27 p.m., Alex Deucher wrote:
> The value is dependent on whether fbc is available.
>
> Signed-off-by: Alex Deucher
> ---
> drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c | 4 +++-
> 1 file changed, 3 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/amd/
On 2018-11-05 2:27 p.m., Alex Deucher wrote:
> Causes a black screen on Stoney laptop.
>
> bug: https://bugs.freedesktop.org/show_bug.cgi?id=108577
> Signed-off-by: Alex Deucher
> ---
> drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion
On 2018-11-02 9:26 a.m., David Francis wrote:
> This will clean up powerplay code, as we are no longer
> multiplying the clocks by 1000 in DM and then dividing them
> by 1000 in powerplay
>
> Signed-off-by: David Francis
Series is
Reviewed-by: Harry Wentland
Harry
> ---
> drivers/gpu/drm/amd
These are some recent patches that are easy to upstream (part 1). For
part 2 (hopefully still this month) I'll need to advance the merging
of KFD into amdgpu a little further to avoid upstreaming duplicated
data structures that no longer need to be duplicated.
Eric Huang (1):
drm/amdkfd: change
From: Yong Zhao
This will make reading code much easier. This fixes a few spots missed in a
previous commit with the same title.
Signed-off-by: Yong Zhao
Reviewed-by: Felix Kuehling
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c | 12 ++--
1 f
The adev parameter in amdgpu_sync_fence and amdgpu_sync_resv is only
needed for updating sync->last_vm_update. This breaks if different
adevs are passed to calls for the same sync object.
Always pass NULL for calls from KFD because sync objects used for
KFD don't belong to any particular device, a
From: Harish Kasiviswanathan
PD or PT might have to be moved during validation and this move has to be
completed before updating it. If page table updates are done using SDMA
then this serializing is done by SDMA command submission.
And if PD/PT updates are done by CPU, then explicit waiting for
From: Harish Kasiviswanathan
Instead of waiting for each KFD BO after validation just wait for the
last BO moving fence.
Signed-off-by: Harish Kasiviswanathan
Reviewed-by: Felix Kuehling
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 13 -
1
From: Gang Ba
Add Vega12 and Polaris12 device info and device IDs to KFD.
Signed-off-by: Gang Ba
Reviewed-by: Felix Kuehling
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c | 2 +
drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c | 3 +-
drivers/gpu/dr
From: Yong Zhao
This makes debug message get printed even when there is early return.
Signed-off-by: Yong Zhao
Reviewed-by: Felix Kuehling
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c | 14 +++---
1 file changed, 7 insertions(+), 7 deletions(-)
d
This change prepares for adding SG BOs that will be used for mapping
doorbells into GPUVM address space.
This type of BO would be mistaken for an invalid userptr BO. Improve
that check to test that it's actually a userptr BO so that SG BOs that
are still in the CPU domain can be validated and mapp
From: Yong Zhao
This is a known gfx9 HW issue, and this change can perfectly workaround
the issue.
Signed-off-by: Yong Zhao
Reviewed-by: Felix Kuehling
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c | 25 ++---
1 file changed, 22 inserti
From: Eric Huang
It is to improve system limit by:
1. replacing userptrlimit with a total memory limit that
conunts TTM memory usage and userptr usage.
2. counting acc size for all BOs.
Signed-off-by: Eric Huang
Reviewed-by: Felix Kuehling
Signed-off-by: Felix Kuehling
---
drivers/gpu/drm/am
From: Hawking Zhang
set_topology_info is used for driver to set current topology info
to xgmi ta
Signed-off-by: Hawking Zhang
Reviewed-by: Shaoyun Liu
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/psp_v11_0.c | 24 +++-
1 file changed, 23 insertions(+), 1 del
From: Hawking Zhang
Add helper functions for the psp xgmi ta.
Signed-off-by: Hawking Zhang
Acked-by: Alex Deucher
Reviewed-by: Huang Rui
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 98 +
1 file changed, 98 insertions(+)
diff --g
This adds the psp interface for xgmi discovery and configuration.
Hawking Zhang (11):
drm/amdgpu/psp: add structure for xgmi ta and its shared buffer
drm/amdgpu/psp: init/de-init xgmi ta microcode
drm/amdgpu/psp: add helper function to load/unload xgmi ta
drm/amdgpu/psp: add xgmi ta header
From: Hawking Zhang
get_topology_info function is used for driver to query topology_info
for current device from xgmi ta
Signed-off-by: Hawking Zhang
Reviewed-by: Shaoyun Liu
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/psp_v11_0.c | 39 ++
1 fil
From: Hawking Zhang
get_hive_id is used for driver to query hive_id for current device
from xgmi ta
Signed-off-by: Hawking Zhang
Acked-by: Alex Deucher
Reviewed-by: Huang Rui
Reviewed-by: Shaoyun Liu
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/psp_v11_0.c | 17 --
From: Hawking Zhang
psp_xgmi_invoke is the helper function to issue ta cmd to firmware
Signed-off-by: Hawking Zhang
Acked-by: Alex Deucher
Reviewed-by: Huang Rui
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 36 +
drivers/gpu/drm/a
From: Hawking Zhang
Add data structures for xgmi trusted application.
Signed-off-by: Hawking Zhang
Acked-by: Alex Deucher
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h | 18 +-
1 file changed, 17 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu
From: Hawking Zhang
Add ucode handling for psp xgmi ta firmware.
Signed-off-by: Hawking Zhang
Acked-by: Alex Deucher
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 2 ++
drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h | 12 ++
drivers/gpu/drm/amd/amdgpu/psp_v1
From: Hawking Zhang
topology info structure needs to match with the one defined
in xgmi ta
Signed-off-by: Hawking Zhang
Reviewed-by: Shaoyun Liu
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.h | 29 +
drivers/gpu/drm/amd/amdgpu/amdgpu_xgmi
From: Hawking Zhang
get_node_id function is used for driver to get node_id
for current device from xgmi ta
Signed-off-by: Hawking Zhang
Acked-by: Alex Deucher
Reviewed-by: Huang Rui
Reviewed-by: Shaoyun Liu
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_gmc.h | 2 +-
d
From: Hawking Zhang
Add the psp xgmi driver interface.
Signed-off-by: Hawking Zhang
Acked-by: Alex Deucher
Reviewed-by: Huang Rui
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/ta_xgmi_if.h | 130
1 file changed, 130 insertions(+)
create mode 10
From: Hawking Zhang
Setup and tear down xgmi as part of psp.
v2:
- make psp_xgmi_terminate static
- squash in:
drm/amdgpu: only issue xgmi cmd when it is enabled
drm/amdgpu/psp: terminate xgmi ta in suspend and hw_fini phase
Signed-off-by: Hawking Zhang
Acked-by: Alex Deucher
Reviewed-by: Hua
50 matches
Mail list logo
