gallium/u_threaded,st/mesa: add a merged set_vertex_elements_and_buffers call

Setting vertex elements before vertex buffers is a new requirement of gallium.

This is the only way to set the vertex elements state after vertex buffers
in st/mesa while setting the state before vertex buffers in tc_batch_execute.

A new TC call is added to set both vertex elements and vertex buffers.
Vertex buffers are filled by st/mesa first, and then the vertex elements
state is set in the same call. When TC calls it, it binds vertex elements
before vertex buffers.

Reviewed-By: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Acked-by: Pierre-Eric Pelloux-Prayer <pierre-eric.pelloux-prayer@amd.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/27736>

src/gallium/auxiliary/util/u_threaded_context.c

@@ -2110,6 +2110,54 @@ tc_add_set_vertex_buffers_call(struct pipe_context *_pipe, unsigned count)
    return p->slot;
 }
 
+struct tc_vertex_elements_and_buffers {
+   struct tc_call_base base;
+   uint8_t count;
+   void *velems_state;
+   struct pipe_vertex_buffer slot[0]; /* more will be allocated if needed */
+};
+
+static uint16_t
+tc_call_set_vertex_elements_and_buffers(struct pipe_context *pipe, void *call)
+{
+   struct tc_vertex_elements_and_buffers *p =
+      (struct tc_vertex_elements_and_buffers *)call;
+   unsigned count = p->count;
+
+   for (unsigned i = 0; i < count; i++)
+      tc_assert(!p->slot[i].is_user_buffer);
+
+   if (p->velems_state)
+      pipe->bind_vertex_elements_state(pipe, p->velems_state);
+
+   pipe->set_vertex_buffers(pipe, count, p->slot);
+   return p->base.num_slots;
+}
+
+/**
+ * Same as tc_add_set_vertex_buffers_call, but the caller should call
+ * tc_set_vertex_elements_for_call with the return value to set the vertex
+ * elements state. The vertex elements state will be bound before vertex
+ * buffers.
+ */
+struct pipe_vertex_buffer *
+tc_add_set_vertex_elements_and_buffers_call(struct pipe_context *_pipe,
+                                            unsigned count)
+{
+   struct threaded_context *tc = threaded_context(_pipe);
+
+   /* We don't need to unbind trailing buffers because we never touch bindings
+    * after num_vertex_buffers.
+    */
+   tc->num_vertex_buffers = count;
+
+   struct tc_vertex_elements_and_buffers *p =
+      tc_add_slot_based_call(tc, TC_CALL_set_vertex_elements_and_buffers,
+                             tc_vertex_elements_and_buffers, count);
+   p->count = count;
+   return p->slot;
+}
+
 struct tc_stream_outputs {
    struct tc_call_base base;
    uint8_t count;

src/gallium/auxiliary/util/u_threaded_context.h

@@ -712,6 +712,10 @@ tc_add_draw_single_call(struct pipe_context *_pipe,
 struct pipe_vertex_buffer *
 tc_add_set_vertex_buffers_call(struct pipe_context *_pipe, unsigned count);
 
+struct pipe_vertex_buffer *
+tc_add_set_vertex_elements_and_buffers_call(struct pipe_context *_pipe,
+                                            unsigned count);
+
 void
 tc_draw_vbo(struct pipe_context *_pipe, const struct pipe_draw_info *info,
             unsigned drawid_offset,
@@ -834,6 +838,19 @@ tc_track_vertex_buffer(struct pipe_context *_pipe, unsigned index,
    }
 }
 
+/**
+ * "buffers" must be a result of tc_add_set_vertex_elements_and_buffers_call.
+ * This sets the vertex elements state for it. It will be bound before vertex
+ * buffers.
+ */
+static inline void
+tc_set_vertex_elements_for_call(struct pipe_vertex_buffer *buffers,
+                                void *state)
+{
+   void **ptr = (void**)buffers;
+   ptr[-1] = state;
+}
+
 #ifdef __cplusplus
 }
 #endif

src/gallium/auxiliary/util/u_threaded_context_calls.h

@@ -10,6 +10,7 @@
  */
 CALL(draw_single)
 CALL(set_vertex_buffers)
+CALL(set_vertex_elements_and_buffers)
 CALL(set_constant_buffer)
 CALL(bind_rasterizer_state)
 CALL(buffer_unmap)

src/mesa/state_tracker/st_atom_array.cpp

@@ -432,7 +432,10 @@ st_update_array_templ(struct st_context *st,
       /* Add up to 1 vertex buffer for zero-stride vertex attribs. */
       num_vbuffers_tc += ALLOW_ZERO_STRIDE_ATTRIBS &&
                          inputs_read & ~enabled_arrays;
-      vbuffer = tc_add_set_vertex_buffers_call(st->pipe, num_vbuffers_tc);
+      vbuffer = UPDATE_VELEMS ?
+         tc_add_set_vertex_elements_and_buffers_call(st->pipe,
+                                                     num_vbuffers_tc) :
+         tc_add_set_vertex_buffers_call(st->pipe, num_vbuffers_tc);
    } else {
       vbuffer = vbuffer_local;
    }
@@ -464,7 +467,8 @@ st_update_array_templ(struct st_context *st,
 
       /* Set vertex buffers and elements. */
       if (FILL_TC_SET_VB) {
-         cso_set_vertex_elements(cso, &velements);
+         void *state = cso_get_vertex_elements_for_bind(cso, &velements);
+         tc_set_vertex_elements_for_call(vbuffer, state);
       } else {
          cso_set_vertex_buffers_and_elements(cso, &velements, num_vbuffers,
                                              uses_user_vertex_buffers, vbuffer);