radv: Implement radv_flush_before_query_copy to workaround UE Vulkan bugs.

Cc: mesa-stable
Closes: https://gitlab.freedesktop.org/mesa/mesa/-/issues/5740

Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14208>
(cherry picked from commit abc4eda846)

.pick_status.json

@@ -886,7 +886,7 @@
         "description": "radv: Implement radv_flush_before_query_copy to workaround UE Vulkan bugs.",
         "nominated": true,
         "nomination_type": 0,
-        "resolution": 0,
+        "resolution": 1,
         "main_sha": null,
         "because_sha": null
     },

src/amd/vulkan/00-radv-defaults.conf

@@ -52,6 +52,10 @@ Application bugs worked around in this file:
             <option name="radv_disable_tc_compat_htile_general" value="true" />
         </engine>
 
+        <engine engine_name_match="^UnrealEngine"> <!-- UE4/5 Vulkan RHI -->
+            <option name="radv_flush_before_query_copy" value="true" />
+        </engine>
+
         <!-- Game workarounds -->
         <application name="Shadow Of The Tomb Raider (Native)" application_name_match="ShadowOfTheTomb">
             <option name="radv_report_llvm9_version_string" value="true" />

src/amd/vulkan/radv_device.c

@@ -1038,6 +1038,7 @@ static const driOptionDescription radv_dri_options[] = {
       DRI_CONF_RADV_DISABLE_ANISO_SINGLE_LEVEL(false)
       DRI_CONF_RADV_DISABLE_SINKING_LOAD_INPUT_FS(false)
       DRI_CONF_RADV_DGC(false)
+      DRI_CONF_RADV_FLUSH_BEFORE_QUERY_COPY(false)
    DRI_CONF_SECTION_END
 };
 // clang-format on
@@ -1086,6 +1087,9 @@ radv_init_dri_options(struct radv_instance *instance)
 
    instance->disable_sinking_load_input_fs =
       driQueryOptionb(&instance->dri_options, "radv_disable_sinking_load_input_fs");
+
+   instance->flush_before_query_copy =
+      driQueryOptionb(&instance->dri_options, "radv_flush_before_query_copy");
 }
 
 VKAPI_ATTR VkResult VKAPI_CALL

src/amd/vulkan/radv_private.h

@@ -370,6 +370,7 @@ struct radv_instance {
    bool disable_aniso_single_level;
    bool zero_vram;
    bool disable_sinking_load_input_fs;
+   bool flush_before_query_copy;
 };
 
 VkResult radv_init_wsi(struct radv_physical_device *physical_device);

src/amd/vulkan/radv_query.c

@@ -1501,6 +1501,12 @@ radv_CmdCopyQueryPoolResults(VkCommandBuffer commandBuffer, VkQueryPool queryPoo
    radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, pool->bo);
    radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, dst_buffer->bo);
 
+   /* Workaround engines that forget to properly specify WAIT_BIT because some driver implicitly
+    * synchronizes before query copy.
+    */
+   if (cmd_buffer->device->instance->flush_before_query_copy)
+      cmd_buffer->state.flush_bits |= cmd_buffer->active_query_flush_bits;
+
    /* From the Vulkan spec 1.1.108:
     *
     * "vkCmdCopyQueryPoolResults is guaranteed to see the effect of

src/util/driconf.h

@@ -582,6 +582,11 @@
    DRI_CONF_OPT_B(radv_dgc, def, \
                   "Expose an experimental implementation of VK_NV_device_generated_commands")
 
+#define DRI_CONF_RADV_FLUSH_BEFORE_QUERY_COPY(def) \
+  DRI_CONF_OPT_B( \
+      radv_flush_before_query_copy, def, \
+      "Wait for timestamps to be written before a query copy command")
+
 /**
  * \brief ANV specific configuration options
  */