[Mesa-dev] [PATCH 3/3] nvc0: add ARB_clear_texture support

[Samuel Pitoiset]

On 11/09/2015 09:03 PM, Ilia Mirkin wrote:
> On Mon, Nov 9, 2015 at 2:58 PM, Samuel Pitoiset
> <samuel.pitoiset at gmail.com> wrote:
>>
>>
>> On 11/09/2015 07:40 PM, Ilia Mirkin wrote:
>>>
>>> Signed-off-by: Ilia Mirkin <imirkin at alum.mit.edu>
>>> ---
>>>    docs/GL3.txt                                    |  2 +-
>>>    docs/relnotes/11.1.0.html                       |  1 +
>>>    src/gallium/drivers/nouveau/nvc0/nvc0_screen.c  |  2 +-
>>>    src/gallium/drivers/nouveau/nvc0/nvc0_surface.c | 82
>>> +++++++++++++++++++++++++
>>>    4 files changed, 85 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/docs/GL3.txt b/docs/GL3.txt
>>> index 7abdcd8..da0ffca 100644
>>> --- a/docs/GL3.txt
>>> +++ b/docs/GL3.txt
>>> @@ -177,7 +177,7 @@ GL 4.4, GLSL 4.40:
>>>
>>>      GL_MAX_VERTEX_ATTRIB_STRIDE                          DONE (all
>>> drivers)
>>>      GL_ARB_buffer_storage                                DONE (i965, nv50,
>>> nvc0, r600, radeonsi)
>>> -  GL_ARB_clear_texture                                 DONE (i965)
>>> (gallium - in progress, VMware)
>>> +  GL_ARB_clear_texture                                 DONE (i965, nvc0)
>>>      GL_ARB_enhanced_layouts                              in progress
>>> (Timothy)
>>>      - compile-time constant expressions                  in progress
>>>      - explicit byte offsets for blocks                   in progress
>>> diff --git a/docs/relnotes/11.1.0.html b/docs/relnotes/11.1.0.html
>>> index 11fbdff..33fd0b8 100644
>>> --- a/docs/relnotes/11.1.0.html
>>> +++ b/docs/relnotes/11.1.0.html
>>> @@ -46,6 +46,7 @@ Note: some of the new features are only available with
>>> certain drivers.
>>>    <ul>
>>>    <li>GL_ARB_arrays_of_arrays on i965</li>
>>>    <li>GL_ARB_blend_func_extended on freedreno (a3xx)</li>
>>> +<li>GL_ARB_clear_texture on nvc0</li>
>>>    <li>GL_ARB_copy_image on nv50, nvc0, radeonsi</li>
>>>    <li>GL_ARB_gpu_shader_fp64 on r600 for Cypress/Cayman/Aruba chips</li>
>>>    <li>GL_ARB_gpu_shader5 on r600 for Evergreen and later chips</li>
>>> diff --git a/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c
>>> b/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c
>>> index f2e3bf0..fbeec7f 100644
>>> --- a/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c
>>> +++ b/src/gallium/drivers/nouveau/nvc0/nvc0_screen.c
>>> @@ -182,6 +182,7 @@ nvc0_screen_get_param(struct pipe_screen *pscreen,
>>> enum pipe_cap param)
>>>       case PIPE_CAP_COPY_BETWEEN_COMPRESSED_AND_PLAIN_FORMATS:
>>>       case PIPE_CAP_FORCE_PERSAMPLE_INTERP:
>>>       case PIPE_CAP_SHAREABLE_SHADERS:
>>> +   case PIPE_CAP_CLEAR_TEXTURE:
>>>          return 1;
>>>       case PIPE_CAP_SEAMLESS_CUBE_MAP_PER_TEXTURE:
>>>          return (class_3d >= NVE4_3D_CLASS) ? 1 : 0;
>>> @@ -204,7 +205,6 @@ nvc0_screen_get_param(struct pipe_screen *pscreen,
>>> enum pipe_cap param)
>>>       case PIPE_CAP_VERTEXID_NOBASE:
>>>       case PIPE_CAP_RESOURCE_FROM_USER_MEMORY:
>>>       case PIPE_CAP_DEVICE_RESET_STATUS_QUERY:
>>> -   case PIPE_CAP_CLEAR_TEXTURE:
>>>          return 0;
>>>
>>>       case PIPE_CAP_VENDOR_ID:
>>> diff --git a/src/gallium/drivers/nouveau/nvc0/nvc0_surface.c
>>> b/src/gallium/drivers/nouveau/nvc0/nvc0_surface.c
>>> index 5f47bad..3ae9943 100644
>>> --- a/src/gallium/drivers/nouveau/nvc0/nvc0_surface.c
>>> +++ b/src/gallium/drivers/nouveau/nvc0/nvc0_surface.c
>>> @@ -319,6 +319,8 @@ nvc0_clear_render_target(struct pipe_context *pipe,
>>>          PUSH_DATA(push, dst->u.tex.first_layer + sf->depth);
>>>          PUSH_DATA(push, mt->layer_stride >> 2);
>>>          PUSH_DATA(push, dst->u.tex.first_layer);
>>> +
>>> +      IMMED_NVC0(push, NVC0_3D(MULTISAMPLE_MODE), mt->ms_mode);
>>>       } else {
>>>          if (res->base.target == PIPE_BUFFER) {
>>>             PUSH_DATA(push, 262144);
>>> @@ -540,6 +542,7 @@ nvc0_clear_depth_stencil(struct pipe_context *pipe,
>>>       PUSH_DATA (push, (unk << 16) | (dst->u.tex.first_layer + sf->depth));
>>>       BEGIN_NVC0(push, NVC0_3D(ZETA_BASE_LAYER), 1);
>>>       PUSH_DATA (push, dst->u.tex.first_layer);
>>> +   IMMED_NVC0(push, NVC0_3D(MULTISAMPLE_MODE), mt->ms_mode);
>>>
>>>       BEGIN_NIC0(push, NVC0_3D(CLEAR_BUFFERS), sf->depth);
>>>       for (z = 0; z < sf->depth; ++z) {
>>> @@ -550,6 +553,84 @@ nvc0_clear_depth_stencil(struct pipe_context *pipe,
>>>       nvc0->dirty |= NVC0_NEW_FRAMEBUFFER;
>>>    }
>>>
>>> +static void
>>> +nvc0_clear_texture(struct pipe_context *pipe,
>>> +                   struct pipe_resource *res,
>>> +                   unsigned level,
>>> +                   const struct pipe_box *box,
>>> +                   const void *data)
>>> +{
>>> +   struct nv50_miptree *mt = nv50_miptree(res);
>>> +   struct nv50_surface sf = {{{0}}};
>>
>>
>> I'm just curious about this, does '= {}' is not enough?
>
> I wanted to be *really* sure it got initialized... figured 3 sets was enough :)
>
> But seriously -- allegedly some compilers don't like that. I can't be
> bothered to check on the actual situation, so I'm including the 0 in
> there. And gcc wanted more {} since the first field was a struct whose
> first field was a struct, etc.

Okay, I think it's *really* initialized. :)
But maybe, a memset() call could be better (really doesn't matter).

Anyway, I'm not quite familiar with that part of the driver to add a Rb 
but it looks fine. I hope you did a full piglit run this time. ;)

>
>    -ilia
>