Intel-XE Archive on lore.kernel.org
 help / color / mirror / Atom feed
From: Gustavo Sousa <gustavo.sousa@intel.com>
To: Jani Nikula <jani.nikula@intel.com>,
	<intel-gfx@lists.freedesktop.org>,
	<intel-xe@lists.freedesktop.org>
Cc: "Ville Syrjälä" <ville.syrjala@linux.intel.com>
Subject: Re: [PATCH 2/3] drm/i915/display: Add infra to reduce global state funcs boilerplate
Date: Fri, 20 Dec 2024 10:54:17 -0300	[thread overview]
Message-ID: <173470285774.2440.7585312979612500951@intel.com> (raw)
In-Reply-To: <87o716pwle.fsf@intel.com>

Quoting Jani Nikula (2024-12-20 05:50:05-03:00)
>On Thu, 19 Dec 2024, Gustavo Sousa <gustavo.sousa@intel.com> wrote:
>> If we look at how the members of struct intel_global_state_funcs, we see
>> a common pattern repeating itself. Let's add the necessary
>> infra-structure to allow reducing the boilerplate. We do that by
>> adding common generic implementations for each member and adding a macro
>> INTEL_GLOBAL_STATE_DEFAULTS() to be used when initializing an instance
>> of struct intel_global_state_funcs.
>>
>> That way, a global state that does not need custom behavior can have
>> its funcs structure be initialized as in the following example,
>>
>>     static const struct intel_global_state_funcs <prefix>_funcs = {
>>            INTEL_GLOBAL_STATE_DEFAULTS(struct <prefix>_state, <base_member_name>),
>>     };
>>
>> , without the need to implementing the functions.
>>
>> That doesn't come without cost - we will need to store two size_t
>> members -, but that cost is arguably justified by the simplification
>> gained.
>>
>> In an upcoming change we will put that infra into action on existing
>> users.
>>
>> Signed-off-by: Gustavo Sousa <gustavo.sousa@intel.com>
>> ---
>>  .../gpu/drm/i915/display/intel_global_state.c | 41 ++++++++++++++++++-
>>  .../gpu/drm/i915/display/intel_global_state.h | 15 +++++++
>>  2 files changed, 54 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/i915/display/intel_global_state.c b/drivers/gpu/drm/i915/display/intel_global_state.c
>> index cbcd1e91b7be..4b4c33fa99fb 100644
>> --- a/drivers/gpu/drm/i915/display/intel_global_state.c
>> +++ b/drivers/gpu/drm/i915/display/intel_global_state.c
>> @@ -59,7 +59,10 @@ static void __intel_atomic_global_state_free(struct kref *kref)
>>  
>>          commit_put(obj_state->commit);
>>  
>> -        obj->funcs->atomic_destroy_state(obj, obj_state);
>> +        if (obj->funcs->atomic_destroy_state)
>> +                obj->funcs->atomic_destroy_state(obj, obj_state);
>> +        else
>> +                intel_atomic_global_destroy_state_common(obj, obj_state);
>>  }
>>  
>>  static void intel_atomic_global_state_put(struct intel_global_state *obj_state)
>> @@ -164,7 +167,11 @@ intel_atomic_get_global_obj_state(struct intel_atomic_state *state,
>>          index = state->num_global_objs;
>>          memset(&state->global_objs[index], 0, sizeof(*state->global_objs));
>>  
>> -        obj_state = obj->funcs->atomic_duplicate_state(obj);
>> +        if (obj->funcs->atomic_duplicate_state)
>> +                obj_state = obj->funcs->atomic_duplicate_state(obj);
>> +        else
>> +                obj_state = intel_atomic_global_duplicate_state_common(obj);
>> +
>>          if (!obj_state)
>>                  return ERR_PTR(-ENOMEM);
>>  
>> @@ -382,3 +389,33 @@ intel_atomic_global_state_commit_done(struct intel_atomic_state *state)
>>                  complete_all(&commit->done);
>>          }
>>  }
>> +
>> +struct intel_global_state *
>> +intel_atomic_global_duplicate_state_common(struct intel_global_obj *obj)
>> +{
>> +        void *state_wrapper;
>> +
>> +        if (WARN_ON(obj->funcs->state_size == 0))
>> +                return NULL;
>> +
>> +        state_wrapper = (void *)obj->state - obj->funcs->base_offset;
>> +
>> +        state_wrapper = kmemdup(state_wrapper, obj->funcs->state_size, GFP_KERNEL);
>> +        if (!state_wrapper)
>> +                return NULL;
>> +
>> +        return state_wrapper + obj->funcs->base_offset;
>> +}
>> +
>> +void intel_atomic_global_destroy_state_common(struct intel_global_obj *obj,
>> +                                              struct intel_global_state *state)
>> +{
>> +        void *state_wrapper;
>> +
>> +        if (WARN_ON(obj->funcs->state_size == 0))
>> +                return;
>> +
>> +        state_wrapper = (void *)state - obj->funcs->base_offset;
>> +
>> +        kfree(state_wrapper);
>> +}
>> diff --git a/drivers/gpu/drm/i915/display/intel_global_state.h b/drivers/gpu/drm/i915/display/intel_global_state.h
>> index 6506a8e32972..e47e007225cc 100644
>> --- a/drivers/gpu/drm/i915/display/intel_global_state.h
>> +++ b/drivers/gpu/drm/i915/display/intel_global_state.h
>> @@ -8,6 +8,8 @@
>>  
>>  #include <linux/kref.h>
>>  #include <linux/list.h>
>> +#include <linux/stddef.h>
>> +#include <linux/types.h>
>>  
>>  struct drm_i915_private;
>>  struct intel_atomic_state;
>> @@ -15,6 +17,10 @@ struct intel_global_obj;
>>  struct intel_global_state;
>>  
>>  struct intel_global_state_funcs {
>> +        /* state_size and base_offset are initialized by INTEL_GLOBAL_STATE_DEFAULTS() */
>> +        size_t state_size;
>> +        size_t base_offset;
>> +
>
>Gut feeling says these should be part of struct intel_global_state
>rather than struct intel_global_state_funcs. Keyword being "funcs".

Yeah, I kept wondering about this and even considered a possible rename
of the struct type, but thought that would be probably too much just
because of two fields that should be considered private.

>
>They would have to be passed to intel_atomic_global_obj_init() and
>initialized runtime. That's a downside.

That sounds okay to me, although I would prefer something that
would not require explicit calls to sizeof() and offsetof().

> But then you could do away with
>the funcs struct altogether when defaults are used, and pass NULL.
>
>And you could also drop INTEL_GLOBAL_STATE_DEFAULTS() which I don't find
>particularly pretty.

Yeah, passing NULL seems like a good idea as it allows one to even skip
defining the funcs structure, but I think dropping
INTEL_GLOBAL_STATE_DEFAULTS() would make this an "all or nothing" type
of stuff: if you just need to customize one behavior, you will need to
define the whole structure explicitly.

--
Gustavo Sousa

>
>BR,
>Jani.
>
>>          struct intel_global_state *(*atomic_duplicate_state)(struct intel_global_obj *obj);
>>          void (*atomic_destroy_state)(struct intel_global_obj *obj,
>>                                       struct intel_global_state *state);
>> @@ -26,6 +32,10 @@ struct intel_global_obj {
>>          const struct intel_global_state_funcs *funcs;
>>  };
>>  
>> +#define INTEL_GLOBAL_STATE_DEFAULTS(type, base_member) \
>> +        .state_size = sizeof(type), \
>> +        .base_offset = offsetof(type, base_member)
>> +
>>  #define intel_for_each_global_obj(obj, dev_priv) \
>>          list_for_each_entry(obj, &(dev_priv)->display.global.obj_list, head)
>>  
>> @@ -96,4 +106,9 @@ int intel_atomic_global_state_wait_for_dependencies(struct intel_atomic_state *s
>>  
>>  bool intel_atomic_global_state_is_serialized(struct intel_atomic_state *state);
>>  
>> +struct intel_global_state *
>> +intel_atomic_global_duplicate_state_common(struct intel_global_obj *obj);
>> +void intel_atomic_global_destroy_state_common(struct intel_global_obj *obj,
>> +                                              struct intel_global_state *state);
>> +
>>  #endif
>
>-- 
>Jani Nikula, Intel

  reply	other threads:[~2024-12-20 13:54 UTC|newest]

Thread overview: 25+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-12-19 21:48 [PATCH 0/3] drm/i915/display: Reduce global state funcs boilerplate Gustavo Sousa
2024-12-19 21:48 ` [PATCH 1/3] drm/i915/display: Do not assume zero offset when duplicating global state Gustavo Sousa
2024-12-19 22:43   ` Cavitt, Jonathan
2024-12-20  9:11   ` Ville Syrjälä
2024-12-20 13:37     ` Gustavo Sousa
2024-12-19 21:48 ` [PATCH 2/3] drm/i915/display: Add infra to reduce global state funcs boilerplate Gustavo Sousa
2024-12-19 22:44   ` Cavitt, Jonathan
2024-12-20 13:43     ` Gustavo Sousa
2024-12-20  8:50   ` Jani Nikula
2024-12-20 13:54     ` Gustavo Sousa [this message]
2024-12-20  8:51   ` Jani Nikula
2024-12-20 13:56     ` Gustavo Sousa
2024-12-20  9:23   ` Ville Syrjälä
2024-12-20 14:02     ` Gustavo Sousa
2024-12-19 21:48 ` [PATCH 3/3] drm/i915/display: Use INTEL_GLOBAL_STATE_DEFAULTS Gustavo Sousa
2024-12-19 22:45   ` Cavitt, Jonathan
2024-12-20 14:08     ` Gustavo Sousa
2024-12-19 22:51 ` ✓ CI.Patch_applied: success for drm/i915/display: Reduce global state funcs boilerplate Patchwork
2024-12-19 22:51 ` ✗ CI.checkpatch: warning " Patchwork
2024-12-19 22:53 ` ✓ CI.KUnit: success " Patchwork
2024-12-19 23:13 ` ✓ CI.Build: " Patchwork
2024-12-19 23:15 ` ✓ CI.Hooks: " Patchwork
2024-12-19 23:17 ` ✗ CI.checksparse: warning " Patchwork
2024-12-19 23:52 ` ✓ Xe.CI.BAT: success " Patchwork
2024-12-20 22:48 ` ✗ Xe.CI.Full: 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=173470285774.2440.7585312979612500951@intel.com \
    --to=gustavo.sousa@intel.com \
    --cc=intel-gfx@lists.freedesktop.org \
    --cc=intel-xe@lists.freedesktop.org \
    --cc=jani.nikula@intel.com \
    --cc=ville.syrjala@linux.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