st/mesa: start adding support for hw atomics atom. (v2)

This adds a new atom that calls the new driver API to
bind buffers containing hw atomics.

v2: fixup bindings for sparse buffers. (mareko/nha)
don't bind buffer atomics when hw atomics are enabled.
use NewAtomicBuffer (mareko)

Tested-By: Gert Wollny <gw.fossdev@gmail.com>
Reviewed-by: Marek Olšák <marek.olsak@amd.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>

src/mesa/state_tracker/st_atom_atomicbuf.c

@@ -46,7 +46,7 @@ st_bind_atomics(struct st_context *st, struct gl_program *prog,
 {
    unsigned i;
 
-   if (!prog || !st->pipe->set_shader_buffers)
+   if (!prog || !st->pipe->set_shader_buffers || st->has_hw_atomics)
       return;
 
    for (i = 0; i < prog->sh.data->NumAtomicBuffers; i++) {
@@ -128,3 +128,37 @@ st_bind_cs_atomics(struct st_context *st)
 
    st_bind_atomics(st, prog, PIPE_SHADER_COMPUTE);
 }
+
+void
+st_bind_hw_atomic_buffers(struct st_context *st)
+{
+   struct pipe_shader_buffer buffers[PIPE_MAX_HW_ATOMIC_BUFFERS];
+   int i;
+
+   if (!st->has_hw_atomics)
+      return;
+
+   for (i = 0; i < st->ctx->Const.MaxAtomicBufferBindings; i++) {
+      struct gl_buffer_binding *binding = &st->ctx->AtomicBufferBindings[i];
+      struct st_buffer_object *st_obj = st_buffer_object(binding->BufferObject);
+      struct pipe_shader_buffer *sb = &buffers[i];
+
+      if (st_obj && st_obj->buffer) {
+        sb->buffer = st_obj->buffer;
+        sb->buffer_offset = binding->Offset;
+        sb->buffer_size = st_obj->buffer->width0 - binding->Offset;
+
+        /* AutomaticSize is FALSE if the buffer was set with BindBufferRange.
+         * Take the minimum just to be sure.
+         */
+        if (!binding->AutomaticSize)
+          sb->buffer_size = MIN2(sb->buffer_size, (unsigned) binding->Size);
+      } else {
+        sb->buffer = NULL;
+        sb->buffer_offset = 0;
+        sb->buffer_size = 0;
+      }
+   }
+
+   st->pipe->set_hw_atomic_buffers(st->pipe, 0, st->ctx->Const.MaxAtomicBufferBindings, buffers);
+}

src/mesa/state_tracker/st_atom_list.h

@@ -66,6 +66,8 @@ ST_STATE(ST_NEW_GS_SSBOS, st_bind_gs_ssbos)
 ST_STATE(ST_NEW_PIXEL_TRANSFER, st_update_pixel_transfer)
 ST_STATE(ST_NEW_TESS_STATE, st_update_tess)
 
+ST_STATE(ST_NEW_HW_ATOMICS, st_bind_hw_atomic_buffers)
+
 /* this must be done after the vertex program update */
 ST_STATE(ST_NEW_VERTEX_ARRAYS, st_update_array)
 

src/mesa/state_tracker/st_cb_bufferobjects.c

@@ -348,7 +348,7 @@ bufferobj_data(struct gl_context *ctx,
    if (st_obj->Base.UsageHistory & USAGE_TEXTURE_BUFFER)
       ctx->NewDriverState |= ST_NEW_SAMPLER_VIEWS | ST_NEW_IMAGE_UNITS;
    if (st_obj->Base.UsageHistory & USAGE_ATOMIC_COUNTER_BUFFER)
-      ctx->NewDriverState |= ST_NEW_ATOMIC_BUFFER;
+      ctx->NewDriverState |= ctx->DriverFlags.NewAtomicBuffer;
 
    return GL_TRUE;
 }

src/mesa/state_tracker/st_context.c

@@ -405,6 +405,10 @@ st_create_context_priv( struct gl_context *ctx, struct pipe_context *pipe,
    st->has_multi_draw_indirect =
       screen->get_param(screen, PIPE_CAP_MULTI_DRAW_INDIRECT);
 
+   st->has_hw_atomics =
+      screen->get_shader_param(screen, PIPE_SHADER_FRAGMENT,
+                               PIPE_SHADER_CAP_MAX_HW_ATOMIC_COUNTERS) ? true : false;
+
    /* GL limits and extensions */
    st_init_limits(pipe->screen, &ctx->Const, &ctx->Extensions);
    st_init_extensions(pipe->screen, &ctx->Const,
@@ -497,6 +501,9 @@ static void st_init_driver_flags(struct st_context *st)
 
    /* Shader resources */
    f->NewTextureBuffer = ST_NEW_SAMPLER_VIEWS;
+   if (st->has_hw_atomics)
+      f->NewAtomicBuffer = ST_NEW_HW_ATOMICS;
+   else
       f->NewAtomicBuffer = ST_NEW_ATOMIC_BUFFER;
    f->NewShaderStorageBuffer = ST_NEW_STORAGE_BUFFER;
    f->NewImageUnits = ST_NEW_IMAGE_UNITS;

src/mesa/state_tracker/st_context.h

@@ -129,6 +129,7 @@ struct st_context
    boolean invalidate_on_gl_viewport;
    boolean draw_needs_minmax_index;
    boolean vertex_array_out_of_memory;
+   boolean has_hw_atomics;
 
    /* Some state is contained in constant objects.
     * Other state is just parameter values.