From: Rodrigo Vivi <rodrigo.vivi@intel.com>
To: "Matthew Brost" <matthew.brost@intel.com>,
"José Roberto de Souza" <jose.souza@intel.com>,
"Francois Dugast" <francois.dugast@intel.com>
Cc: Francois Dugast <francois.dugast@intel.com>,
intel-xe@lists.freedesktop.org
Subject: Re: [Intel-xe] [PATCH v2 03/14] drm/xe: Make DRM_XE_DEVICE_QUERY_ENGINES future proof
Date: Wed, 29 Nov 2023 11:54:51 -0500 [thread overview]
Message-ID: <ZWds2xD8bXkBy-t0@intel.com> (raw)
In-Reply-To: <ZWZY/zQQzVeWV1PG@DUT025-TGLU.fm.intel.com>
On Tue, Nov 28, 2023 at 09:17:51PM +0000, Matthew Brost wrote:
> On Wed, Nov 22, 2023 at 02:38:22PM +0000, Francois Dugast wrote:
> > From: José Roberto de Souza <jose.souza@intel.com>
> >
> > We have at least 2 future features(OA and future media engines
> > capabilities) that will require Xe to provide more information about
> > engines to UMDs.
> >
> > But this information should not just be added to
> > drm_xe_engine_class_instance for a couple of reasons:
> > - drm_xe_engine_class_instance is used as input to other structs/uAPIs
> > and those uAPIs don't care about any of these future new engine fields
> > - those new fields are useless information after initialization for
> > some UMDs, so it should not need to carry that around
> >
> > So here my proposal is to make DRM_XE_DEVICE_QUERY_ENGINES return an
> > array of drm_xe_query_engine_info that contain
> > drm_xe_engine_class_instance and 3 u64s to be used for future features.
>
> There is 5 u64 in this patch.
my bad, sorry. just trying to prepare for the many capabilities and other
stuff we will likely need. But likely 3 would be enough.
>
> Anyways agree this better but what if we improve this a bit more making
> each __drm_xe_query_engine_info a dynamic size.
hmmm... I think I like this idea. pahole seems to have a good output
with our other existent dynamic array.
>
> e.g.
> struct __drm_xe_query_engine_info { /* Bad name, just for example */
> /** @instance: The @drm_xe_engine_class_instance */
> struct drm_xe_engine_class_instance instance;
>
> /** @reserved: Reserved, dynamic size */
> __u64 reserved[0];
I believe [] instead of [0] is the preferred way nowadays.
> };
>
> struct drm_xe_query_engine_info {
> /** @num_engines: Number engines */
> __u64 num_engines;
> /** @engine_info: Array of engine info */
> struct __drm_xe_query_engine_info engine_info[0];
> };
>
> The size of __drm_xe_query_engine_info then change and can be calculated
> to be:
>
> (returned size of query - sizoeof(__u64)) / drm_xe_query_engine_info.num_engines
>
> We then should be able to add additional fields to
> __drm_xe_query_engine_info indefinitely.
>
> What does everyone think? We likely can apply this idea to other
> queries too as it makes sense.
>
> Matt
>
> >
> > Reference OA:
> > https://patchwork.freedesktop.org/patch/558362/?series=121084&rev=6
> >
> > Cc: Francois Dugast <francois.dugast@intel.com>
> > Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
> > Signed-off-by: José Roberto de Souza <jose.souza@intel.com>
> > Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
> > [Rodrigo Rebased]
> > Signed-off-by: Francois Dugast <francois.dugast@intel.com>
> > ---
> > drivers/gpu/drm/xe/xe_query.c | 15 ++++++++-------
> > include/uapi/drm/xe_drm.h | 24 +++++++++++++++++++++++-
> > 2 files changed, 31 insertions(+), 8 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/xe/xe_query.c b/drivers/gpu/drm/xe/xe_query.c
> > index 61a7d92b7e88..0cbfeaeb1330 100644
> > --- a/drivers/gpu/drm/xe/xe_query.c
> > +++ b/drivers/gpu/drm/xe/xe_query.c
> > @@ -53,7 +53,7 @@ static size_t calc_hw_engine_info_size(struct xe_device *xe)
> > i++;
> > }
> >
> > - return i * sizeof(struct drm_xe_engine_class_instance);
> > + return i * sizeof(struct drm_xe_query_engine_info);
> > }
> >
> > typedef u64 (*__ktime_func_t)(void);
> > @@ -186,9 +186,9 @@ static int query_engines(struct xe_device *xe,
> > struct drm_xe_device_query *query)
> > {
> > size_t size = calc_hw_engine_info_size(xe);
> > - struct drm_xe_engine_class_instance __user *query_ptr =
> > + struct drm_xe_query_engine_info __user *query_ptr =
> > u64_to_user_ptr(query->data);
> > - struct drm_xe_engine_class_instance *hw_engine_info;
> > + struct drm_xe_query_engine_info *hw_engine_info;
> > struct xe_hw_engine *hwe;
> > enum xe_hw_engine_id id;
> > struct xe_gt *gt;
> > @@ -211,12 +211,13 @@ static int query_engines(struct xe_device *xe,
> > if (xe_hw_engine_is_reserved(hwe))
> > continue;
> >
> > - hw_engine_info[i].engine_class =
> > + hw_engine_info[i].instance.engine_class =
> > xe_to_user_engine_class[hwe->class];
> > - hw_engine_info[i].engine_instance =
> > + hw_engine_info[i].instance.engine_instance =
> > hwe->logical_instance;
> > - hw_engine_info[i].gt_id = gt->info.id;
> > - hw_engine_info[i].pad = 0;
> > + hw_engine_info[i].instance.gt_id = gt->info.id;
> > + hw_engine_info[i].instance.pad = 0;
> > + memset(hw_engine_info->reserved, 0, sizeof(hw_engine_info->reserved));
> >
> > i++;
> > }
> > diff --git a/include/uapi/drm/xe_drm.h b/include/uapi/drm/xe_drm.h
> > index 1bdd20d3c4a8..c80e03b61489 100644
> > --- a/include/uapi/drm/xe_drm.h
> > +++ b/include/uapi/drm/xe_drm.h
> > @@ -124,7 +124,14 @@ struct xe_user_extension {
> > #define DRM_IOCTL_XE_EXEC_QUEUE_GET_PROPERTY DRM_IOWR(DRM_COMMAND_BASE + DRM_XE_EXEC_QUEUE_GET_PROPERTY, struct drm_xe_exec_queue_get_property)
> > #define DRM_IOCTL_XE_WAIT_USER_FENCE DRM_IOWR(DRM_COMMAND_BASE + DRM_XE_WAIT_USER_FENCE, struct drm_xe_wait_user_fence)
> >
> > -/** struct drm_xe_engine_class_instance - instance of an engine class */
> > +/**
> > + * struct drm_xe_engine_class_instance - instance of an engine class
> > + *
> > + * It is returned as part of the @drm_xe_query_engine_info, but it also is
> > + * used as the input of engine selection for both @drm_xe_exec_queue_create
> > + * and @drm_xe_query_engine_cycles
> > + *
> > + */
> > struct drm_xe_engine_class_instance {
> > #define DRM_XE_ENGINE_CLASS_RENDER 0
> > #define DRM_XE_ENGINE_CLASS_COPY 1
> > @@ -145,6 +152,21 @@ struct drm_xe_engine_class_instance {
> > __u16 pad;
> > };
> >
> > +/**
> > + * struct drm_xe_query_engine_info - describe hardware engine
> > + *
> > + * If a query is made with a struct @drm_xe_device_query where .query
> > + * is equal to %DRM_XE_DEVICE_QUERY_ENGINES, then the reply uses an array of
> > + * struct @drm_xe_query_engine_info in .data.
> > + */
> > +struct drm_xe_query_engine_info {
> > + /** @instance: The @drm_xe_engine_class_instance */
> > + struct drm_xe_engine_class_instance instance;
> > +
> > + /** @reserved: Reserved */
> > + __u64 reserved[5];
> > +};
> > +
> > /**
> > * enum drm_xe_memory_class - Supported memory classes.
> > */
> > --
> > 2.34.1
> >
next prev parent reply other threads:[~2023-11-29 16:55 UTC|newest]
Thread overview: 49+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-11-22 14:38 [Intel-xe] [PATCH v2 00/14] uAPI Alignment - Cleanup and future proof Francois Dugast
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 01/14] drm/xe: Extend drm_xe_vm_bind_op Francois Dugast
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 02/14] drm/xe/uapi: Separate bo_create placement from flags Francois Dugast
2023-11-29 19:36 ` Welty, Brian
2023-11-29 20:41 ` Rodrigo Vivi
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 03/14] drm/xe: Make DRM_XE_DEVICE_QUERY_ENGINES future proof Francois Dugast
2023-11-28 21:17 ` Matthew Brost
2023-11-29 16:54 ` Rodrigo Vivi [this message]
2023-11-29 12:35 ` Matthew Brost
2023-11-29 20:04 ` Souza, Jose
2023-11-29 22:52 ` Rodrigo Vivi
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 04/14] drm/xe/uapi: Reject bo creation of unaligned size Francois Dugast
2023-11-24 18:15 ` Souza, Jose
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 05/14] drm/xe/uapi: Align on a common way to return arrays (memory regions) Francois Dugast
2023-11-24 18:19 ` Souza, Jose
2023-11-28 20:51 ` Rodrigo Vivi
2023-11-29 12:33 ` Francois Dugast
2023-11-30 20:53 ` Dixit, Ashutosh
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 06/14] drm/xe/uapi: Align on a common way to return arrays (gt) Francois Dugast
2023-11-28 20:51 ` Rodrigo Vivi
2023-11-29 18:30 ` Matt Roper
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 07/14] drm/xe/uapi: Align on a common way to return arrays (engines) Francois Dugast
2023-11-28 20:56 ` Rodrigo Vivi
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 08/14] drm/xe/uapi: Split xe_sync types from flags Francois Dugast
2023-11-28 21:19 ` Matthew Brost
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 09/14] drm/xe/uapi: Kill tile_mask Francois Dugast
2023-11-29 9:07 ` Matthew Brost
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 10/14] drm/xe/uapi: Crystal Reference Clock updates Francois Dugast
2023-11-24 18:38 ` Souza, Jose
2023-11-29 14:08 ` Francois Dugast
2023-11-29 18:33 ` Souza, Jose
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 11/14] drm/xe/uapi: Remove bogus engine list from the wait_user_fence IOCTL Francois Dugast
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 12/14] drm/xe/uapi: Add Tile ID information to the GT info query Francois Dugast
2023-11-24 18:45 ` Souza, Jose
2023-11-27 14:08 ` Francois Dugast
2023-11-27 14:20 ` Souza, Jose
2023-11-29 18:33 ` Souza, Jose
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 13/14] drm/xe/uapi: Fix various struct padding for 64b alignment Francois Dugast
2023-11-29 17:02 ` Souza, Jose
2023-11-29 17:39 ` Francois Dugast
2023-11-22 14:38 ` [Intel-xe] [PATCH v2 14/14] drm/xe/uapi: Move xe_exec after xe_exec_queue Francois Dugast
2023-11-29 18:34 ` Souza, Jose
2023-11-23 14:14 ` [Intel-xe] ✓ CI.Patch_applied: success for uAPI Alignment - Cleanup and future proof (rev5) Patchwork
2023-11-23 14:14 ` [Intel-xe] ✗ CI.checkpatch: warning " Patchwork
2023-11-23 14:15 ` [Intel-xe] ✓ CI.KUnit: success " Patchwork
2023-11-23 14:23 ` [Intel-xe] ✓ CI.Build: " Patchwork
2023-11-23 14:23 ` [Intel-xe] ✗ CI.Hooks: failure " Patchwork
2023-11-23 14:24 ` [Intel-xe] ✓ CI.checksparse: success " Patchwork
2023-11-23 15:01 ` [Intel-xe] ✗ CI.BAT: failure " Patchwork
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=ZWds2xD8bXkBy-t0@intel.com \
--to=rodrigo.vivi@intel.com \
--cc=francois.dugast@intel.com \
--cc=intel-xe@lists.freedesktop.org \
--cc=jose.souza@intel.com \
--cc=matthew.brost@intel.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox