draw/gs: reverse the polarity of the invocation/prims execution

The current code runs primitives per invocation, but the spec wants
invocations per primitive. However it means having to flush
after each invocation to get correct XFB behaviour

Fixes:
GTF-GL41.gtf40.GL3Tests.transform_feedback3.transform_feedback3_geometry_instanced

Reviewed-by: Roland Scheidegger <sroland@vmware.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/5555>

diff --git a/src/gallium/auxiliary/draw/draw_gs.c b/src/gallium/auxiliary/draw/draw_gs.c
index f648a96db7b785f3d564cb081764247458514680..2d11430b274af2b5c83f7e089a82a7dbe8e00968 100644 (file)
--- a/src/gallium/auxiliary/draw/draw_gs.c
+++ b/src/gallium/auxiliary/draw/draw_gs.c
@@ -69,7 +69,7 @@ draw_gs_get_input_index(int semantic, int index,
 static inline boolean
 draw_gs_should_flush(struct draw_geometry_shader *shader)
 {
-   return (shader->fetched_prim_count == shader->vector_length);
+   return (shader->fetched_prim_count == shader->vector_length || shader->num_invocations > 1);
 }
 
 /*#define DEBUG_OUTPUTS 1*/
@@ -192,15 +192,8 @@ static void tgsi_gs_prepare(struct draw_geometry_shader *shader,
                             const unsigned constants_size[PIPE_MAX_CONSTANT_BUFFERS])
 {
    struct tgsi_exec_machine *machine = shader->machine;
-   int j;
    tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
                                   constants, constants_size);
-
-   if (shader->info.uses_invocationid) {
-      unsigned i = machine->SysSemanticToIndex[TGSI_SEMANTIC_INVOCATIONID];
-      for (j = 0; j < TGSI_QUAD_SIZE; j++)
-         machine->SystemValue[i].xyzw[0].i[j] = shader->invocation_id;
-   }
 }
 
 static void tgsi_gs_run(struct draw_geometry_shader *shader,
@@ -210,6 +203,12 @@ static void tgsi_gs_run(struct draw_geometry_shader *shader,
    struct tgsi_exec_machine *machine = shader->machine;
    int i;
 
+   if (shader->info.uses_invocationid) {
+      unsigned i = machine->SysSemanticToIndex[TGSI_SEMANTIC_INVOCATIONID];
+      for (int j = 0; j < TGSI_QUAD_SIZE; j++)
+         machine->SystemValue[i].xyzw[0].i[j] = shader->invocation_id;
+   }
+
    /* run interpreter */
    tgsi_exec_machine_run(machine, 0);
 
@@ -437,10 +436,13 @@ static void gs_flush(struct draw_geometry_shader *shader)
    debug_assert(input_primitives > 0 &&
                 input_primitives <= 4);
 
-   shader->run(shader, input_primitives, out_prim_count);
-   for (i = 0; i < shader->num_vertex_streams; i++) {
-      shader->fetch_outputs(shader, i, out_prim_count[i],
-                            &shader->stream[i].tmp_output);
+   for (unsigned invocation = 0; invocation < shader->num_invocations; invocation++) {
+      shader->invocation_id = invocation;
+      shader->run(shader, input_primitives, out_prim_count);
+      for (i = 0; i < shader->num_vertex_streams; i++) {
+         shader->fetch_outputs(shader, i, out_prim_count[i],
+                               &shader->stream[i].tmp_output);
+      }
    }
 
 #if 0
@@ -592,7 +594,6 @@ int draw_geometry_shader_run(struct draw_geometry_shader *shader,
     * overflown vertices into some area where they won't harm anyone */
    unsigned total_verts_per_buffer = shader->primitive_boundary *
       num_in_primitives;
-   unsigned invocation;
    int i;
    //Assume at least one primitive
    max_out_prims = MAX2(max_out_prims, 1);
@@ -666,26 +667,22 @@ int draw_geometry_shader_run(struct draw_geometry_shader *shader,
    }
 #endif
 
-   for (invocation = 0; invocation < shader->num_invocations; invocation++) {
-      shader->invocation_id = invocation;
+   shader->prepare(shader, constants, constants_size);
 
-      shader->prepare(shader, constants, constants_size);
+   if (input_prim->linear)
+      gs_run(shader, input_prim, input_verts,
+             output_prims, output_verts);
+   else
+      gs_run_elts(shader, input_prim, input_verts,
+                  output_prims, output_verts);
 
-      if (input_prim->linear)
-         gs_run(shader, input_prim, input_verts,
-                output_prims, output_verts);
-      else
-         gs_run_elts(shader, input_prim, input_verts,
-                     output_prims, output_verts);
-
-      /* Flush the remaining primitives. Will happen if
-       * num_input_primitives % 4 != 0
-       */
-      if (shader->fetched_prim_count > 0) {
-         gs_flush(shader);
-      }
-      debug_assert(shader->fetched_prim_count == 0);
+   /* Flush the remaining primitives. Will happen if
+    * num_input_primitives % 4 != 0
+    */
+   if (shader->fetched_prim_count > 0) {
+      gs_flush(shader);
    }
+   debug_assert(shader->fetched_prim_count == 0);
 
    /* Update prim_info:
     */