[Freedreno] [PATCH 12/12] drm/msm/adreno: Switch to chip-id for identifying GPU

[Rob Clark]

On Wed, Jul 26, 2023 at 2:38 PM Dmitry Baryshkov
<dmitry.baryshkov at linaro.org> wrote:
>
> On 27/07/2023 00:37, Rob Clark wrote:
> > On Thu, Jul 6, 2023 at 8:45 PM Dmitry Baryshkov
> > <dmitry.baryshkov at linaro.org> wrote:
> >>
> >> On 07/07/2023 00:10, Rob Clark wrote:
> >>> From: Rob Clark <robdclark at chromium.org>
> >>>
> >>>    /* Helper for formating the chip_id in the way that userspace tools like
> >>>     * crashdec expect.
> >>>     */
> >>>    #define ADRENO_CHIPID_FMT "u.%u.%u.%u"
> >>> -#define ADRENO_CHIPID_ARGS(_r) (_r).core, (_r).major, (_r).minor, (_r).patchid
> >>> +#define ADRENO_CHIPID_ARGS(_c) \
> >>> +     (((_c) >> 24) & 0xff), \
> >>> +     (((_c) >> 16) & 0xff), \
> >>> +     (((_c) >> 8)  & 0xff), \
> >>> +     ((_c) & 0xff)
> >>
> >> So, we still have some meaning for chipid?
> >
> > Only enough to do the inverse of what userspace does when parsing
> > devcoredump to construct chip-id.  Basically it is just a different
> > way to represent a 32b #
>
> What about passing it in the direct form? The macro adds assumptions.

It is uabi

I wouldn't call it adding assumptions as much as just a funny way to
format a number.  In some cases it might work out to something that
vaguely resembles a marketing name (6.3.0.2), in other cases it won't
(12.34.56.78).. it's just formatting a 32b # to match the way existing
userspace parses it

BR,
-R