[Mesa-dev] [PATCH] radeonsi: add basic glClearBufferSubData acceleration

[Ilia Mirkin]

On Wed, Nov 4, 2015 at 4:08 AM, Marek Olšák <maraeo at gmail.com> wrote:
> On Nov 4, 2015 4:02 AM, "Alex Deucher" <alexdeucher at gmail.com> wrote:
>>
>> On Tue, Nov 3, 2015 at 6:47 PM, Marek Olšák <maraeo at gmail.com> wrote:
>> > From: Marek Olšák <marek.olsak at amd.com>
>> >
>> > ---
>> >  src/gallium/drivers/radeonsi/si_blit.c | 55
>> > ++++++++++++++++++++++++++++++++++
>> >  1 file changed, 55 insertions(+)
>> >
>> > diff --git a/src/gallium/drivers/radeonsi/si_blit.c
>> > b/src/gallium/drivers/radeonsi/si_blit.c
>> > index fce014a..e934146 100644
>> > --- a/src/gallium/drivers/radeonsi/si_blit.c
>> > +++ b/src/gallium/drivers/radeonsi/si_blit.c
>> > @@ -731,9 +731,64 @@ static void si_flush_resource(struct pipe_context
>> > *ctx,
>> >         }
>> >  }
>> >
>> > +static void si_pipe_clear_buffer(struct pipe_context *ctx,
>> > +                                struct pipe_resource *dst,
>> > +                                unsigned offset, unsigned size,
>> > +                                const void *clear_value,
>> > +                                int clear_value_size)
>> > +{
>> > +       struct si_context *sctx = (struct si_context*)ctx;
>> > +       const uint32_t *u32 = clear_value;
>> > +       unsigned i;
>> > +       bool clear_value_fits_dword = true;
>> > +       uint8_t *map;
>> > +
>> > +       if (clear_value_size > 4)
>> > +               for (i = 1; i < clear_value_size / 4; i++)
>> > +                       if (u32[0] != u32[i]) {
>> > +                               clear_value_fits_dword = false;
>> > +                               break;
>> > +                       }
>> > +
>> > +       /* Use CP DMA for the simple case. */
>> > +       if (offset % 4 == 0 && size % 4 == 0 && clear_value_fits_dword)
>> > {
>> > +               uint32_t value = u32[0];
>> > +
>> > +               switch (clear_value_size) {
>> > +               case 1:
>> > +                       value &= 0xff;
>> > +                       value |= (value << 8) | (value << 16) | (value
>> > << 24);
>> > +                       break;
>> > +               case 2:
>> > +                       value &= 0xffff;
>> > +                       value |= value << 16;
>> > +                       break;
>> > +               }
>> > +
>> > +               sctx->b.clear_buffer(ctx, dst, offset, size, value,
>> > false);
>> > +               return;
>> > +       }
>> > +
>> > +       /* TODO: use a compute shader for other cases. */
>>
>> What about using SDMA?  It supports byte aligned constant fills at
>> least on CIK+.
>
> I think CP DMA supports byte aligned constant fills as well, I just need to
> test it. The bigger problem is 64 and 128 bit fills. Those can only be done
> with a shader AFAIK.

Just to briefly interject, don't forget abotu 96-bit fills for
tbo_rgb32 buffers. I assume this won't really matter if you're using a
shader.

Cheers,

  -ilia