the CSA is used for world switch, and each amdgpu device should have one and
only one CSA,
and this CSA will pined, and mapped to each virtual memory /process.
CP/RLCV will use this CSA buffer when preemption occurred, and will write some
hardware status into this CSA buffer, within the
> -Original Message-
> From: Alex Deucher [mailto:alexdeuc...@gmail.com]
> Sent: Tuesday, December 20, 2016 7:11 AM
> To: Yu, Xiangliang
> Cc: amd-gfx list ;
> dl.SRDC_SW_GPUVirtualization
> ; Liu,
> -Original Message-
> From: Alex Deucher [mailto:alexdeuc...@gmail.com]
> Sent: Tuesday, December 20, 2016 7:23 AM
> To: Yu, Xiangliang
> Cc: amd-gfx list ;
> dl.SRDC_SW_GPUVirtualization
> ;
Thank monk’s comments.
From: Liu, Monk
Sent: Tuesday, December 20, 2016 12:09 PM
To: Alex Deucher ; Yu, Xiangliang
Cc: amd-gfx list ; dl.SRDC_SW_GPUVirtualization
; Min, Frank
> -Original Message-
> From: Alex Deucher [mailto:alexdeuc...@gmail.com]
> Sent: Tuesday, December 20, 2016 7:25 AM
> To: Yu, Xiangliang
> Cc: amd-gfx list ;
> dl.SRDC_SW_GPUVirtualization
> ;
> -Original Message-
> From: Alex Deucher [mailto:alexdeuc...@gmail.com]
> Sent: Tuesday, December 20, 2016 7:14 AM
> To: Yu, Xiangliang
> Cc: amd-gfx list ;
> dl.SRDC_SW_GPUVirtualization
>
> -Original Message-
> From: Alex Deucher [mailto:alexdeuc...@gmail.com]
> Sent: Tuesday, December 20, 2016 7:27 AM
> To: Yu, Xiangliang
> Cc: amd-gfx list ;
> dl.SRDC_SW_GPUVirtualization
> ;
Thank monk’s detail expanation.
And I think this patch is only support virtualization world switch, not touch
whole amdpgu preemption.
Thanks!
Xiangliang Yu
From: Liu, Monk
Sent: Tuesday, December 20, 2016 11:58 AM
To: Alex Deucher ; Yu, Xiangliang
Hi Alex
I agree with you that this patch's GOLDEN setting programming should be put in
VI.C, but I found a hardware issue :
original linux logic is that we set golden setting registers separately within
each IP's hw init routine, but for TONGA VF, it is really strange that we must
set all
Hi Christian,
the CSA VA update should be correct, we pin it after it allocated per device
during init stage, and we map it upon each process creates their vm page tables,
since this CSA is pined, I assume there is no need to validate it prior to
submission
please correct me if I missed
> -Original Message-
> From: Alex Deucher [mailto:alexdeuc...@gmail.com]
> Sent: Tuesday, December 20, 2016 7:35 AM
> To: Yu, Xiangliang
> Cc: amd-gfx list ;
> dl.SRDC_SW_GPUVirtualization
>
On Dec 19, 2016 6:48 AM, Serguei Sagalovitch
wrote:
> If compute queue is occupied only by you, the efficiency
> is equal with setting job queue to high priority I think.
The only risk is the situation when graphics will take all
needed CUs. But in any case it
Hi Emil,
On 2016-12-19 07:46 AM, Emil Velikov wrote:
Hi Harry,
On 14 December 2016 at 16:26, Harry Wentland wrote:
They are still used all over the place (e.g.
dc/dce110/dce110_resource.c:413).
We should at least do an spatch to use kzalloc/krealloc/kfree across the
On Sun, Dec 18, 2016 at 8:46 AM, Daniel Vetter wrote:
> On Sat, Dec 17, 2016 at 04:35:09PM +, 'Liviu Dudau' wrote:
>> On Fri, Dec 16, 2016 at 07:16:25PM +, Deucher, Alexander wrote:
>> > > -Original Message-
>> > > From: Liviu Dudau [mailto:li...@dudau.co.uk]
>> >
From: Junwei Zhang
v2: agd: squash in various fixes
Signed-off-by: Junwei Zhang
Reviewed-by: Alex Deucher
Reviewed-by: Christian König
Reviewed-by: Ken Wang
Signed-off-by:
From: Junwei Zhang
Signed-off-by: Junwei Zhang
Reviewed-by: Alex Deucher
Reviewed-by: Christian König
Reviewed-by: Ken Wang
Signed-off-by: Alex Deucher
Signed-off-by: Huang Rui
---
drivers/gpu/drm/amd/amdgpu/vi_dpm.h | 4
1 file changed, 4 deletions(-)
diff --git a/drivers/gpu/drm/amd/amdgpu/vi_dpm.h
b/drivers/gpu/drm/amd/amdgpu/vi_dpm.h
index fc120ba..c43e03f 100644
--- a/drivers/gpu/drm/amd/amdgpu/vi_dpm.h
+++
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Huang Rui
> Sent: Monday, December 19, 2016 2:16 AM
> To: amd-gfx@lists.freedesktop.org
> Cc: Huang, Ray
> Subject: [PATCH] drm/amd/powerplay: fix request smc_sk firmware case
>
> This patch
amdgpu and radeon development has moved to this list.
Signed-off-by: Alex Deucher
---
MAINTAINERS | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/MAINTAINERS b/MAINTAINERS
index 1a7a7731..3e21c61 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@
From: Junwei Zhang
Signed-off-by: Junwei Zhang
Reviewed-by: Alex Deucher
Reviewed-by: Christian König
Reviewed-by: Ken Wang
Signed-off-by: Alex Deucher
This patch set adds support for Polaris 12 GPUs.
Junwei Zhang (3):
drm/amd/amdgpu: add Polaris12 support (v2)
drm/amdgpu/powerplay: add Polaris12 support
drm/amd/amdgpu: add Polaris12 PCI ID
drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c| 6 +
Hey Guys,
One particular piece I'd like to discuss is how to get around the below issue:
> Known current obstacles:
>
>
> The SQ is currently programmed to disregard the HQD priorities, and instead
> it picks
> jobs at random. Settings from the shader itself are also
On Sat, Dec 17, 2016 at 11:16 AM, Xiangliang Yu wrote:
> KIQ is queue-memory based initialization method: setup KIQ queue
> firstly, then send command to KIQ to setup other queues, without
> accessing registers.
>
> For virtualization, need KIQ to access virtual function
On Sat, Dec 17, 2016 at 11:16 AM, Xiangliang Yu wrote:
> KIQ has some behavior as compute ring.
>
> Signed-off-by: Xiangliang Yu
Should be squashed into patch 5.
Alex
> ---
> drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 3 ++-
> 1 file changed, 2
On Sat, Dec 17, 2016 at 11:16 AM, Xiangliang Yu wrote:
> According to chip device id to set VF flag, and call virtual
> interface to setup all realted IP blocks.
>
> Signed-off-by: Xiangliang Yu
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 5
On Sat, Dec 17, 2016 at 11:16 AM, Xiangliang Yu wrote:
> GPU guest driver send mailbox messages to hyperverisor to request
> full access to all of registers and release it when access is done.
>
> Signed-off-by: Xiangliang Yu
> Signed-off-by:
On Sat, Dec 17, 2016 at 11:16 AM, Xiangliang Yu wrote:
> One of important role of KIQ is provide one way to access VF
> registers. This patch implement the feature and export interfaces.
>
> Signed-off-by: Monk Liu
> Signed-off-by: Xiangliang Yu
On Sat, Dec 17, 2016 at 11:16 AM, Xiangliang Yu wrote:
> GPU virtualization component need vim common ip block and the
> block was also public before. Export it again.
>
> Signed-off-by: Xiangliang Yu
> ---
> drivers/gpu/drm/amd/amdgpu/vi.c | 2 +-
>
On Sat, Dec 17, 2016 at 11:16 AM, Xiangliang Yu wrote:
> Different chips will have different virtual behaviors, so need to
> implemnt different virtual feature according to hardware design.
>
> This patch will implemnt Vi family virtualization, it will call
> CSA, mailbox
On Sat, Dec 17, 2016 at 11:16 AM, Xiangliang Yu wrote:
> GPU virtualization has different sequence from normal, change it.
>
> Signed-off-by: Frank Min
> Signed-off-by: Monk Liu
> Signed-off-by: Xiangliang Yu
>
On Mon, Dec 19, 2016 at 6:17 AM, Christian König
wrote:
> Yeah, agree totally with Monk here.
>
> If you find that you need some KIQ helpers which are independent of the
> hardware generation you can put them into amdgpu_kiq.c.
>
> Monk, David or maybe even Michel and
> -Original Message-
> From: amd-gfx [mailto:amd-gfx-boun...@lists.freedesktop.org] On Behalf
> Of Xiangliang Yu
> Sent: Saturday, December 17, 2016 11:17 AM
> To: amd-gfx@lists.freedesktop.org; dl.SRDC_SW_GPUVirtualization
> Cc: Yu, Xiangliang
> Subject: [PATCH 20/23] drm/amdgpu/mxgpu:
Hi Harry,
On 14 December 2016 at 16:26, Harry Wentland wrote:
> They are still used all over the place (e.g.
> dc/dce110/dce110_resource.c:413).
>
> We should at least do an spatch to use kzalloc/krealloc/kfree across the
> board if the wrappers are an issue.
>
> NAKed
>
33 matches
Mail list logo