From: "Christian König" <christian.koenig@amd.com>
To: Shashank Sharma <shashank.sharma@amd.com>, amd-gfx@lists.freedesktop.org
Cc: alexander.deucher@amd.com, amaranath.somalapuram@amd.com
Subject: Re: [PATCH v3 3/5] drm/amdgpu: set GPU workload via ctx IOCTL
Date: Tue, 27 Sep 2022 08:09:56 +0200 [thread overview]
Message-ID: <b7ad5684-1d73-e4da-7019-18db5ad16a62@amd.com> (raw)
In-Reply-To: <20220926214014.5450-4-shashank.sharma@amd.com>
Am 26.09.22 um 23:40 schrieb Shashank Sharma:
> This patch adds new IOCTL flags in amdgpu_context_IOCTL to set
> GPU workload profile. These calls will allow a user to switch
> to a GPU power profile which might be better suitable to its
> workload type. The currently supported workload types are:
> "None": Default workload profile
> "3D": Workload profile for 3D rendering work
> "Video": Workload profile for Media/Encode/Decode work
> "VR": Workload profile for VR rendering work
> "Compute": Workload profile for Compute work
>
> The workload hint flag is saved in GPU context, and then its
> applied when we actually run the job.
>
> V3: Create only set_workload interface, there is no need for
> get_workload (Christian)
>
> Cc: Alex Deucher <alexander.deucher@amd.com>
> Signed-off-by: Shashank Sharma <shashank.sharma@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
> ---
> drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c | 42 +++++++++++++++++++++++--
> drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h | 1 +
> 2 files changed, 41 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
> index 8ee4e8491f39..937c294f8d84 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c
> @@ -27,6 +27,7 @@
> #include "amdgpu.h"
> #include "amdgpu_sched.h"
> #include "amdgpu_ras.h"
> +#include "amdgpu_ctx_workload.h"
> #include <linux/nospec.h>
>
> #define to_amdgpu_ctx_entity(e) \
> @@ -328,7 +329,7 @@ static int amdgpu_ctx_init(struct amdgpu_ctx_mgr *mgr, int32_t priority,
> return r;
>
> ctx->stable_pstate = current_stable_pstate;
> -
> + ctx->workload_mode = AMDGPU_CTX_WORKLOAD_HINT_NONE;
> return 0;
> }
>
> @@ -633,11 +634,34 @@ static int amdgpu_ctx_stable_pstate(struct amdgpu_device *adev,
> return r;
> }
>
> +static int amdgpu_ctx_set_workload_profile(struct amdgpu_device *adev,
> + struct amdgpu_fpriv *fpriv, uint32_t id,
> + u32 workload_hint)
> +{
> + struct amdgpu_ctx *ctx;
> + struct amdgpu_ctx_mgr *mgr;
> +
> + if (!fpriv)
> + return -EINVAL;
> +
> + mgr = &fpriv->ctx_mgr;
> + mutex_lock(&mgr->lock);
> + ctx = idr_find(&mgr->ctx_handles, id);
> + if (!ctx) {
> + mutex_unlock(&mgr->lock);
> + return -EINVAL;
> + }
> +
> + ctx->workload_mode = workload_hint;
> + mutex_unlock(&mgr->lock);
> + return 0;
> +}
> +
> int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
> struct drm_file *filp)
> {
> int r;
> - uint32_t id, stable_pstate;
> + uint32_t id, stable_pstate, wl_hint;
> int32_t priority;
>
> union drm_amdgpu_ctx *args = data;
> @@ -681,6 +705,20 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data,
> return -EINVAL;
> r = amdgpu_ctx_stable_pstate(adev, fpriv, id, true, &stable_pstate);
> break;
> + case AMDGPU_CTX_OP_SET_WORKLOAD_PROFILE:
> + if (args->in.flags & ~AMDGPU_CTX_WORKLOAD_HINT_MASK)
> + return -EINVAL;
> + wl_hint = args->in.flags & AMDGPU_CTX_WORKLOAD_HINT_MASK;
> + if (wl_hint > AMDGPU_CTX_WORKLOAD_HINT_MAX)
> + return -EINVAL;
> + r = amdgpu_ctx_set_workload_profile(adev, fpriv, id, wl_hint);
> + if (r)
> + DRM_ERROR("Failed to set workload profile to %s\n",
> + amdgpu_workload_profile_name(wl_hint));
> + else
> + DRM_DEBUG_DRIVER("Workload profile set to %s\n",
> + amdgpu_workload_profile_name(wl_hint));
> + break;
> default:
> return -EINVAL;
> }
> diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
> index cc7c8afff414..6c8032c3291a 100644
> --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
> +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.h
> @@ -58,6 +58,7 @@ struct amdgpu_ctx {
> unsigned long ras_counter_ce;
> unsigned long ras_counter_ue;
> uint32_t stable_pstate;
> + uint32_t workload_mode;
> };
>
> struct amdgpu_ctx_mgr {
next prev parent reply other threads:[~2022-09-27 6:10 UTC|newest]
Thread overview: 76+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-09-26 21:40 [PATCH v3 0/5] GPU workload hints for better performance Shashank Sharma
2022-09-26 21:40 ` [PATCH v3 1/5] drm/amdgpu: add UAPI for workload hints to ctx ioctl Shashank Sharma
2022-09-27 6:07 ` Christian König
2022-09-27 14:28 ` Felix Kuehling
2023-03-21 3:05 ` Marek Olšák
2023-03-21 13:00 ` Sharma, Shashank
2023-03-21 13:54 ` Christian König
2023-03-22 14:05 ` Marek Olšák
2023-03-22 14:08 ` Christian König
2023-03-22 14:24 ` Marek Olšák
2023-03-22 14:29 ` Christian König
2023-03-22 14:36 ` Marek Olšák
2023-03-22 14:52 ` Alex Deucher
2023-03-22 15:11 ` Marek Olšák
2023-03-22 14:38 ` Sharma, Shashank
2022-09-26 21:40 ` [PATCH v3 2/5] drm/amdgpu: add new functions to set GPU power profile Shashank Sharma
2022-09-27 2:14 ` Quan, Evan
2022-09-27 7:29 ` Sharma, Shashank
2022-09-27 9:29 ` Quan, Evan
2022-09-27 10:00 ` Sharma, Shashank
2022-09-27 6:08 ` Christian König
2022-09-27 9:58 ` Lazar, Lijo
2022-09-27 11:41 ` Sharma, Shashank
2022-09-27 12:10 ` Lazar, Lijo
2022-09-27 12:23 ` Sharma, Shashank
2022-09-27 12:39 ` Lazar, Lijo
2022-09-27 12:53 ` Sharma, Shashank
2022-09-27 13:29 ` Lazar, Lijo
2022-09-27 13:47 ` Sharma, Shashank
2022-09-27 14:00 ` Lazar, Lijo
2022-09-27 14:20 ` Sharma, Shashank
2022-09-27 14:34 ` Lazar, Lijo
2022-09-27 14:50 ` Sharma, Shashank
2022-09-27 15:20 ` Felix Kuehling
2022-09-26 21:40 ` [PATCH v3 3/5] drm/amdgpu: set GPU workload via ctx IOCTL Shashank Sharma
2022-09-27 6:09 ` Christian König [this message]
2022-09-26 21:40 ` [PATCH v3 4/5] drm/amdgpu: switch GPU workload profile Shashank Sharma
2022-09-27 6:11 ` Christian König
2022-09-27 10:03 ` Lazar, Lijo
2022-09-27 11:47 ` Sharma, Shashank
2022-09-27 12:20 ` Lazar, Lijo
2022-09-27 12:25 ` Sharma, Shashank
2022-09-27 16:33 ` Michel Dänzer
2022-09-27 17:06 ` Sharma, Shashank
2022-09-27 17:29 ` Michel Dänzer
2022-09-26 21:40 ` [PATCH v3 5/5] drm/amdgpu: switch workload context to/from compute Shashank Sharma
2022-09-27 6:12 ` Christian König
2022-09-27 14:48 ` Felix Kuehling
2022-09-27 14:58 ` Sharma, Shashank
2022-09-27 15:23 ` Felix Kuehling
2022-09-27 15:38 ` Sharma, Shashank
2022-09-27 20:40 ` Alex Deucher
2022-09-28 7:05 ` Lazar, Lijo
2022-09-28 8:56 ` Sharma, Shashank
2022-09-28 9:00 ` Sharma, Shashank
2022-09-28 21:51 ` Alex Deucher
2022-09-29 8:48 ` Sharma, Shashank
2022-09-29 11:10 ` Lazar, Lijo
2022-09-29 13:20 ` Sharma, Shashank
2022-09-29 13:37 ` Lazar, Lijo
2022-09-29 14:00 ` Sharma, Shashank
2022-09-29 14:14 ` Lazar, Lijo
2022-09-29 14:40 ` Sharma, Shashank
2022-09-29 18:32 ` Alex Deucher
2022-09-30 5:08 ` Lazar, Lijo
2022-09-30 8:37 ` Sharma, Shashank
2022-09-30 9:13 ` Lazar, Lijo
2022-09-30 9:22 ` Sharma, Shashank
2022-09-30 9:54 ` Lazar, Lijo
2022-09-30 10:09 ` Sharma, Shashank
2022-09-29 18:07 ` Felix Kuehling
2022-09-30 4:46 ` Lazar, Lijo
2022-09-27 16:24 ` [PATCH v3 0/5] GPU workload hints for better performance Michel Dänzer
2022-09-27 16:59 ` Sharma, Shashank
2022-09-27 17:13 ` Michel Dänzer
2022-09-27 17:25 ` Sharma, Shashank
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=b7ad5684-1d73-e4da-7019-18db5ad16a62@amd.com \
--to=christian.koenig@amd.com \
--cc=alexander.deucher@amd.com \
--cc=amaranath.somalapuram@amd.com \
--cc=amd-gfx@lists.freedesktop.org \
--cc=shashank.sharma@amd.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.