[Mesa-dev] [PATCH 00/26] GLSL memory diet

Ian Romanick idr at freedesktop.org
Tue Sep 9 12:40:14 PDT 2014


On 09/04/2014 11:14 PM, Matt Turner wrote:
> On Fri, Jul 18, 2014 at 4:35 PM, Matt Turner <mattst88 at gmail.com> wrote:
>> On Tue, Jul 15, 2014 at 10:37 AM, Matt Turner <mattst88 at gmail.com> wrote:
>>> On Mon, Jul 14, 2014 at 3:48 PM, Ian Romanick <idr at freedesktop.org> wrote:
>>>> Most of these patches have been sent to the list already in one form or
>>>> another.  There are a few changes, removals, and additions.  The series
>>>> has also been re-ordered.
>>>>
>>>> - The extra memory accounting code has been removed.  This was suggested
>>>>   by Ken.  Instead, all memory usage data is from Valgrind massiv.
>>>>
>>>> - The "store short names where padding would be" code is still there,
>>>>   but many of the patches to shorten the names of temporaries has been
>>>>   removed.  Those patches became unnecessary because...
>>>>
>>>> - In release builds (when MESA_GLSL is also unset), all temporaries get
>>>>   the name "compiler_temp" that is in static storage.
>>>>
>>>> - A small set of names that frequently occur in shader-db are kept in
>>>>   static storage.  When a variable is created with one of these names,
>>>>   no additional storage is allocated.  We could probably give local
>>>>   variables and post-linking globals the same treatment as temporaries,
>>>>   but a lot of the frequently occuring names (that don't fit in padding)
>>>>   are uniforms or shader inputs.  These need to keep their names.
>>>>
>>>> I suspect the last six patches will be contentious.  I'd like to get the
>>>> first 20 (or at least the first two!) reviewed and landed sooner... then
>>>> we can debate the last few more leisurely.
>>>
>>> The first 9 (so far) are
>>>
>>> Reviewed-by: Matt Turner <mattst88 at gmail.com>
>>
>> As are 10-20.
>>
>> Patch 22 should just mark the enum as packed, and then you can drop
>> patch 21. With that, patch 22 is also R-b.
> 
> So you never committed any of these twenty patches I reviewed? Looks
> like Ken committed three of them this week without my R-b attached.

Yes, I think that happened. :(  I think I got caught up fixing all the
UBO layout bugs that I just forgot about all this work.



More information about the mesa-dev mailing list