[Intel-gfx] [PATCH 0/6] drm/i915: start splitting off runtime device info

[Jani Nikula]

On Wed, 02 Jan 2019, Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com> wrote:
> On 02/01/2019 09:13, Jani Nikula wrote:
>> On Tue, 01 Jan 2019, Chris Wilson <chris at chris-wilson.co.uk> wrote:
>>> Quoting Jani Nikula (2018-12-31 14:56:40)
>>>> The mkwrite_device_info removal series [1] seems to have stalled. I'm
>>>> trying to nudge things forward a bit with this series. This isn't near
>>>> as complete as Tvrtko's work, but does some of the prep work I wanted,
>>>> specifically using INTEL_INFO() and RUNTIME_INFO() to access the
>>>> fields. There are obviously conflicts, but mostly I think this should
>>>> make the rest of Tvrtko's work easier, not harder.
>>>>
>>>> BR,
>>>> Jani.
>>>>
>>>>
>>>> [1] https://patchwork.freedesktop.org/series/52381/
>>>>
>>>> Cc: Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com>
>>>>
>>>>
>>>> Jani Nikula (6):
>>>>    drm/i915: start moving runtime device info to a separate struct
>>>>    drm/i915/reg: abstract display_mmio_offset access
>>>>    drm/i915: pass dev_priv to intel_device_info_runtime_init()
>>>>    drm/i915: always use INTEL_INFO() to access device info
>>>>    drm/i915: drop intel_device_info_dump()
>>>>    drm/i915: rename dev_priv info to __info to avoid usage
>>>
>>> Looked ok, and didn't see anything odd compared to my own attempts.
>>> Reviewed-by: Chris Wilson <chris at chris-wilson.co.uk>
>> 
>> Thanks for the review.
>> 
>> Tvrtko, I'd still like to get your ack before I go ahead and push
>> these. Also, do you think you'll have time in the near future to pick up
>> your series, or shall I?
>
> I missed Chris had already reviewed it. Yeah, series looks fine to me.

Thanks for your reviews too, pushed the series.

> With regards to the second part, after your work what I think would be 
> left from mine is:
>
>   * move some more fields/flags to runtime info
>   * subplatform/devid consolidation
>   * looking at how to wean selftests off modifying the gen field

Once we make dev_priv->info a pointer to the static const structs, I
think the selftests can make a copy, and point dev_priv->info at that
instead, and go wild. There might be a slight chicken and egg issue to
deal with in the patch series ordering though to keep it clean.

>   * maybe introducting INTEL_SSEU to avoid many RUNTIME_INFO(...)->sseu
>
> If you are in a hurry and have time you can take over, or if you are 
> just in a hurry I might be able to do within a week or two.

I'll eyeball it a bit, and let you know if I do anything to avoid
duplicating the effort.

BR,
Jani.


-- 
Jani Nikula, Intel Open Source Graphics Center