[Mesa-dev] [PATCH 23/26] i965: Restructure fast clear eligibility decision

[Jason Ekstrand]

On Mon, Oct 31, 2016 at 2:38 PM, Jason Ekstrand <jason at jlekstrand.net>
wrote:

> On Tue, Oct 11, 2016 at 12:26 PM, Topi Pohjolainen <
> topi.pohjolainen at gmail.com> wrote:
>
>> From: Ben Widawsky <ben at bwidawsk.net>
>>
>> Signed-off-by: Ben Widawsky <benjamin.widawsky at intel.com>
>> ---
>>  src/mesa/drivers/dri/i965/intel_mipmap_tree.c | 51
>> ++++++++++++++++++---------
>>  1 file changed, 34 insertions(+), 17 deletions(-)
>>
>> diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
>> b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
>> index f51392f..a41a654 100644
>> --- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
>> +++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
>> @@ -224,33 +224,50 @@ intel_miptree_supports_non_msrt_fast_clear(struct
>> brw_context *brw,
>>        return false;
>>     }
>>
>> +   /* Handle the hardware restrictions...
>> +    *
>> +    * All GENs have the following restriction: "MCS buffer for non-MSRT
>> is
>> +    * supported only for RT formats 32bpp, 64bpp, and 128bpp."
>> +    *
>> +    * From the HSW PRM Volume 7: 3D-Media-GPGPU, page 652: (Color Clear
>> of
>> +    * Non-MultiSampler Render Target Restrictions) Support is for
>> non-mip-mapped
>> +    * and non-array surface types only.
>> +    *
>> +    * From the BDW PRM Volume 7: 3D-Media-GPGPU, page 649: (Color Clear
>> of
>> +    * Non-MultiSampler Render Target Restriction). Mip-mapped and arrayed
>> +    * surfaces are supported with MCS buffer layout with these
>> alignments in the
>> +    * RT space: Horizontal Alignment = 256 and Vertical Alignment = 128.
>> +    *
>> +    * Skylake and above (docs are currently unpublished) are similar to
>> BDW with
>>
>
> Heh... Old patch.  We should put in the real PRM citation.
>
>
>> +    * looser restrictions.
>> +    */
>> +   const bool mip_mapped = mt->first_level != 0 || mt->last_level != 0;
>> +   const bool arrayed = mt->physical_depth0 != 1;
>> +
>>     if (mt->cpp != 4 && mt->cpp != 8 && mt->cpp != 16)
>>        return false;
>> -   if (mt->first_level != 0 || mt->last_level != 0) {
>> -      if (brw->gen >= 8) {
>> +
>> +   if (mip_mapped || arrayed) {
>> +      /* Multisample surfaces with the CMS layout are not layered
>> surfaces, yet
>> +       * still have physical_depth0 > 1. Assert that we don't
>> accidentally
>> +       * reject a multisampled surface here. We should have rejected it
>> earlier
>> +       * by explicitly checking the sample count.
>> +       */
>> +      if (arrayed)
>> +         assert(mt->num_samples <= 1);
>> +
>> +      if (mip_mapped && brw->gen >= 8) {
>>           perf_debug("Multi-LOD fast clear - giving up (%dx%dx%d).\n",
>>                      mt->logical_width0, mt->logical_height0,
>> mt->last_level);
>> +         return false;
>>
>
Also, this is wrong.  We don't want to return false only on gen8+, we want
to warn only on gen8+ and return false always.

The later patch fixes it to be correct (return false on gen7-) but I'd like
it to be correct in the mean-time.


>        }
>>
>> -      return false;
>> -   }
>> -
>> -   /* Check for layered surfaces. */
>> -   if (mt->physical_depth0 != 1) {
>> -       /* Multisample surfaces with the CMS layout are not layered
>> surfaces,
>> -        * yet still have physical_depth0 > 1. Assert that we don't
>> -        * accidentally reject a multisampled surface here. We should have
>> -        * rejected it earlier by explicitly checking the sample count.
>> -        */
>> -      assert(mt->num_samples <= 1);
>> -
>> -      if (brw->gen >= 8) {
>> +      if (arrayed && brw->gen >= 8) {
>>           perf_debug("Layered fast clear - giving up. (%dx%d%d)\n",
>>                      mt->logical_width0, mt->logical_height0,
>>                      mt->physical_depth0);
>> +         return false;
>>        }
>> -
>> -      return false;
>>     }
>>
>>     /* There's no point in using an MCS buffer if the surface isn't in a
>> --
>> 2.5.5
>>
>> _______________________________________________
>> mesa-dev mailing list
>> mesa-dev at lists.freedesktop.org
>> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20161031/25799fae/attachment.html>