Am 09.03.2018 um 00:24 schrieb Oak Zeng:
This is preparation for sharing client ID definitions
between amdgpu and amdkfd
Change-Id: Ie0b7b14c9c2366fd896745a51b74a9ba97ee3224
Signed-off-by: Oak Zeng
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h | 44 +--
Hi Rex,
I think still initializing the PSP even when you don't need it for
firmware upload sounds like a good idea to me.
But take that with a grain of salt since I really on don't know that
part of the hardware so well.
Christian.
Am 09.03.2018 um 06:10 schrieb Zhu, Rex:
Hi Alex,
How
That would be a possibility, but I don't think reading the register once
during driver startup is a real issue.
See you need to hammer on the problematic registers to trigger the
misbehavior, e.g. not read it once or twice but thousands of times.
Additional to that as I said before, VM
On 2018年03月09日 07:24, Oak Zeng wrote:
This is preparation for sharing client ID definitions
between amdgpu and amdkfd
Change-Id: Ie0b7b14c9c2366fd896745a51b74a9ba97ee3224
Signed-off-by: Oak Zeng
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h | 44 +--
Hi Alex,
How about keep the firmware type checking in set_ip_blocks.
and remove the same check code in psp module.
also no need to change load type if psp load firmware failed in psp module.
Please review the attached patch.
Best Regards
Rex
From: amd-gfx
On Thu, Mar 8, 2018 at 11:26 PM, Alex Deucher wrote:
> On Thu, Mar 8, 2018 at 6:24 PM, Oak Zeng wrote:
>> This is preparation for sharing client ID definitions
>> between amdgpu and amdkfd
>>
>> Change-Id: Ie0b7b14c9c2366fd896745a51b74a9ba97ee3224
>>
On Thu, Mar 8, 2018 at 8:45 PM, Harry Wentland wrote:
> They were used by amdgpu_dm at some point but since it has its own
> amdgpu_dm_connector now these aren't needed anymore.
>
> Signed-off-by: Harry Wentland
Reviewed-by: Alex Deucher
On Thu, Mar 8, 2018 at 6:24 PM, Oak Zeng wrote:
> This is preparation for sharing client ID definitions
> between amdgpu and amdkfd
>
> Change-Id: Ie0b7b14c9c2366fd896745a51b74a9ba97ee3224
> Signed-off-by: Oak Zeng
> ---
>
This is preparation for sharing client ID definitions
between amdgpu and amdkfd
Change-Id: Ie0b7b14c9c2366fd896745a51b74a9ba97ee3224
Signed-off-by: Oak Zeng
---
drivers/gpu/drm/amd/amdgpu/amdgpu_ih.h | 44 +--
Can GIM read the value and pass it to the VF's via the vbios or the shared area?
Alex
From: Liu, Monk
Sent: Thursday, March 8, 2018 9:59:34 PM
To: Koenig, Christian; Deucher, Alexander; Mao, David
Cc: Jin, Jian-Rong; amd-gfx@lists.freedesktop.org
Subject: RE:
> Although KMD now use a cache way to give libdrm the CC_RB_BACKEND_DISABLE
> value, but KMD still need to read it at least one shot during loading driver
> right ?
Correct, but we do this while having exclusive access.
The problem is although for VF0 it is reading this register in exclusive
They were used by amdgpu_dm at some point but since it has its own
amdgpu_dm_connector now these aren't needed anymore.
Signed-off-by: Harry Wentland
---
drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h | 29 -
1 file changed, 29 deletions(-)
diff
We already checked and returned early in most of the IP
functions, fill in the rest as well.
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c | 9 +
1 file changed, 9 insertions(+)
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c
We already handle the firmware loading type checks in the
psp module directly, no need for an additional check.
Signed-off-by: Alex Deucher
---
drivers/gpu/drm/amd/amdgpu/soc15.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)
diff --git
Save/restore the backlight level scratch register in S3/S4 so the
backlight level comes back at the previously requested level.
Bug: https://bugzilla.kernel.org/show_bug.cgi?id=199047
Fixes: 4ec6ecf48c64d (drm/amdgpu: drop scratch regs save and restore from S3/S4
handling)
Signed-off-by: Alex
On Thu, Mar 8, 2018 at 12:31 PM, Marek Olšák wrote:
> From: Qiang Yu
>
> to support SRIOV and MCBP, need 16 IBs per submit
>
> Signed-off-by: Qiang Yu
> Reviewed-by: Junwei Zhang
Reviewed-by: Alex Deucher
Am 08.03.2018 um 18:31 schrieb Marek Olšák:
From: Qiang Yu
to support SRIOV and MCBP, need 16 IBs per submit
Signed-off-by: Qiang Yu
Reviewed-by: Junwei Zhang
Acked-by: Christian König
---
From: Michel Dänzer
amdgpu_scanout_extents_intersect could leave the scanout damage region
in an invalid state, triggering debugging checks in pixman:
*** BUG ***
In pixman_region_append_non_o: The expression r->x1 < r->x2 was false
Set a breakpoint on
From: Qiang Yu
to support SRIOV and MCBP, need 16 IBs per submit
Signed-off-by: Qiang Yu
Reviewed-by: Junwei Zhang
---
amdgpu/amdgpu.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/amdgpu/amdgpu.h
On 2018-03-07 10:33 AM, Mario Kleiner wrote:
> On 03/07/2018 09:50 AM, Michel Dänzer wrote:
>> On 2018-03-07 04:45 AM, Mario Kleiner wrote:
>>> On 03/05/2018 10:56 PM, Alex Deucher wrote:
On Mon, Mar 5, 2018 at 12:44 PM, Michel Dänzer
wrote:
> From: Michel Dänzer
+ Harry, Leo
I saw this as well with tonga. It seems to be related to the CTM/regamma
changes.
Alex
From: amd-gfx on behalf of Hubert
Łępicki
Sent: Thursday, March 8, 2018 7:03:06 AM
To:
Hey,
I was giving a shot your work in progress branch for the AMDGPU
driver, hoping maybe it brings some fixes to Ryzen 5 2400g. It does
not look like it, but I might have discovered a bug (or you may be
well aware of it since it's work in progress...). Either way, just to
let you know:
Graphics
On Thu, Mar 8, 2018 at 5:02 AM, Hawking Zhang wrote:
> The vram type for dGPU is stored in umc_info while sys mem type
> for APU is stored in integratedsysteminfo
>
> Change-Id: Iec14a0cac407f3da37245786c8f1b688968f8726
> Signed-off-by: Hawking Zhang
On Wed, Mar 7, 2018 at 11:42 PM, Zhu, Rex wrote:
>>Is this consistent with our other smu implementations? If we never use the
>>errors, I'm ok with removing them, but I don't want to have to add them again
>>later if we decide we
>>actually need them.
>
> Rex: until
Although KMD now use a cache way to give libdrm the
CC_RB_BACKEND_DISABLE value, but KMD still need to read it at least
one shot during loading driver right ?
Correct, but we do this while having exclusive access.
Not only tlb flush, other ring’s vm flush of VF2 can all be ruined by
this
To be more specific let me give you an example we face with now:
Although KMD now use a cache way to give libdrm the CC_RB_BACKEND_DISABLE
value, but KMD still need to read it at least one shot during loading driver
right ?
For SR-IOV there is still chance that VF2 is doing CPU based tlb flush
Yeah, I agree with you we’d better find all those registers
Stitch together the REQ and ACK part only avoid this issue for vm flush only,
and we may still hit the issue in other place, I knew it.
The frustrating job is how can we find all those registers ?
And more is since this
The vram type for dGPU is stored in umc_info while sys mem type
for APU is stored in integratedsysteminfo
Change-Id: Iec14a0cac407f3da37245786c8f1b688968f8726
Signed-off-by: Hawking Zhang
---
drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c | 95 +++-
Please ignore this patch since it doesn't sync up to the latest code base. I
will re-send another review later
Regards,
Hawking
-Original Message-
From: Hawking Zhang
Sent: 2018年3月8日 17:37
To: amd-gfx@lists.freedesktop.org
Cc: Zhang, Hawking
Hi Monk,
While we can avoid such vm flush failure by stitch together of the
sending REQ and reading ACK part, at least for compute ring this is
confirmed.
Well there are two misunderstanding here.
First of all this solution doesn't really work, it just hides the
problem because we don't do
The vram type info for dGPU is saved in umc_info while sys mem type info
for APU is saved in integratedsysteminfo
Change-Id: I08b84c65e4e31c65e2e5a277a1acdf032ec04d2e
Signed-off-by: Hawking Zhang
---
drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c | 95
Am 08.03.2018 um 09:35 schrieb Michel Dänzer:
On 2018-03-07 10:47 AM, Christian König wrote:
Am 07.03.2018 um 09:42 schrieb Michel Dänzer:
On 2018-03-06 07:23 PM, Christian König wrote:
E.g. the last time I tested it placing things into GTT still resulted
in quite a performance penalty for
On 2018-03-07 10:47 AM, Christian König wrote:
> Am 07.03.2018 um 09:42 schrieb Michel Dänzer:
>> On 2018-03-06 07:23 PM, Christian König wrote:
>>
>>> E.g. the last time I tested it placing things into GTT still resulted
>>> in quite a performance penalty for rendering.
>> FWIW, I think the
33 matches
Mail list logo