[Mesa-dev] [PATCH 1/4] r600g: avoid redundant DB register updates

[Marek Olšák]

On Thu, Apr 27, 2017 at 4:57 PM, Emil Velikov <emil.l.velikov at gmail.com> wrote:
> On 27 April 2017 at 11:52, Constantine Kharlamov <hi-angel at yandex.ru> wrote:
>> 27.04.2017, 13:26, "Emil Velikov" <emil.l.velikov at gmail.com>:
>>> Hi Constantine,
>>>
>>> I'm not familiar with r600, so mostly a drive-by idea/nit.
>>
>> I'm not familiar with graphics :P Well, perhaps a little now.
>>
>>> On 25 April 2017 at 12:59, Constantine Kharlamov <Hi-Angel at yandex.ru> wrote:
>>>
>>>>  @@ -1746,45 +1747,47 @@ static void evergreen_emit_framebuffer_state(struct r600_context *rctx, struct r
>>>>                  radeon_set_context_reg(cs, R_028E50_CB_COLOR8_INFO + (i - 8) * 0x1C, 0);
>>>>
>>>>          /* ZS buffer. */
>>>>  - if (state->zsbuf) {
>>>
>>>>  + if (rctx->framebuffer.dirty_zsbuf) {
>>>>  + if (state->zsbuf) {
>>>
>>> You can fold both if statements in a single one, leaving the
>>> indentation level as-is.
>>> As a bonus the resulting patch will be a lot smaller ;-)
>>>
>>> Same can be done across the patch.
>>>
>>> -Emil
>>
>> Btw it's how it's done in radeonsi, but I can't do the same in r600g because of the drm_minor check. The original code goes like:
>>
>>         if (state->zsbuf) {
>>                 /* … */
>>         } else if (rctx->screen->b.info.drm_minor >= 18) {
>>                 /* … */
>>         }
>>
> I've missed the drm_version check, thanks for the wake-up call.
>
> Thinking out loud:
> Last year we bumped to the kernel requirement to 3.2 (released in 2011).
> Worth doing a similar one this year, say 3.10 (2013)? It should drop
> ~1/3 of the drm_minor checks.

There is no maintenance or other reason to do so as far as I know.

Marek