anv: Slow clear if fast-clear cost is not mitigated

Fast-clears require expensive flushes beforehand and afterwards. The
cost of flushes are decreased in a series of back-to-back fast-clears as
no extra fast-clear flushes are required in between them. If the ratio
of a command buffer's recorded back-to-back fast clears over independent
fast-clears falls below 1/2, prevent that command buffer from recording
any further fast-clears.

Averaging two runs of our Factorio trace on an A750 shows a +14.37%
improvement in FPS.

Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32984>

src/intel/vulkan/anv_image.c

@@ -3564,6 +3564,14 @@ anv_can_fast_clear_color(const struct anv_cmd_buffer *cmd_buffer,
       return false;
    }
 
+   if (cmd_buffer->num_independent_clears >= 16 &&
+       cmd_buffer->num_independent_clears >
+       cmd_buffer->num_dependent_clears * 2) {
+      anv_perf_warn(VK_LOG_OBJS(&image->vk.base),
+                    "Not enough back-to-back fast-clears. Slow clearing.");
+      return false;
+   }
+
    return true;
 }
 

src/intel/vulkan/anv_private.h

@@ -4249,6 +4249,10 @@ struct anv_cmd_buffer {
 
    struct anv_cmd_state                         state;
 
+   /* Fast-clear statistics. */
+   uint64_t                                     num_dependent_clears;
+   uint64_t                                     num_independent_clears;
+
    struct anv_address                           return_addr;
 
    /* Set by SetPerformanceMarkerINTEL, written into queries by CmdBeginQuery */

src/intel/vulkan/genX_cmd_buffer.c

@@ -3108,6 +3108,14 @@ genX(cmd_buffer_update_color_aux_op)(struct anv_cmd_buffer *cmd_buffer,
    } else {
       cmd_buffer->state.color_aux_op = next_aux_op;
    }
+
+   if (next_aux_op == ISL_AUX_OP_FAST_CLEAR) {
+      if (aux_op_clears(last_aux_op)) {
+         cmd_buffer->num_dependent_clears++;
+      } else {
+         cmd_buffer->num_independent_clears++;
+      }
+   }
 }
 
 static void