anv: move generation shader return instruction to last draw lane

If we dispatch exactly a multiple of 8192 items, there is additional
lane left to generate the jump instruction.

Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Fixes: c950fe97a0 ("anv: implement generated (indexed) indirect draws")
Reviewed-by: Ivan Briano <ivan.briano@intel.com>
Tested-by: Felix DeGrood <felix.j.degrood@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/25361>

src/intel/vulkan/shaders/common_generated_draws.glsl

@@ -131,3 +131,18 @@ void write_MI_BATCH_BUFFER_START(uint write_offset,
    commands[write_offset + 1] = uint(addr & 0xffffffff);
    commands[write_offset + 2] = uint(addr >> 32);
 }
+
+void end_generated_draws(uint cmd_idx, uint draw_id, uint draw_count)
+{
+   uint _3dprim_dw_size = (flags >> 16) & 0xff;
+   /* We can have an indirect draw count = 0. */
+   uint last_draw_id = draw_count == 0 ? 0 : (min(draw_count, max_draw_count) - 1);
+   uint jump_offset = draw_count == 0 ? 0 : _3dprim_dw_size;
+
+   if (draw_id == last_draw_id && draw_count < max_draw_count) {
+      /* Only write a jump forward in the batch if we have fewer elements than
+       * the max draw count.
+       */
+      write_MI_BATCH_BUFFER_START(cmd_idx + jump_offset, end_addr);
+   }
+}

src/intel/vulkan/shaders/gfx11_generated_draws.glsl

@@ -76,10 +76,7 @@ void main()
                                     first_instance,
                                     draw_id);
       }
-   } else if (draw_id == draw_count && draw_id < max_draw_count) {
-      /* Only write a jump forward in the batch if we have fewer elements than
-       * the max draw count.
-       */
-      write_MI_BATCH_BUFFER_START(cmd_idx, end_addr);
    }
+
+   end_generated_draws(cmd_idx, draw_id, draw_count);
 }

src/intel/vulkan/shaders/gfx9_generated_draws.glsl

@@ -135,10 +135,7 @@ void main()
                            first_instance,
                            0 /* base_vertex_location */);
       }
-   } else if (draw_id == draw_count && draw_id < max_draw_count) {
-      /* Only write a jump forward in the batch if we have fewer elements than
-       * the max draw count.
-       */
-      write_MI_BATCH_BUFFER_START(cmd_idx, end_addr);
    }
+
+   end_generated_draws(cmd_idx, draw_id, draw_count);
 }