Video: add Video Filter Threads setting

CPU video filters always ran on one worker per detected core
(RARCH_SOFTFILTER_THREADS_AUTO). That suits the heavy filters (NTSC,
NTSC CRT, 2xBR) but over-parallelises the light ones: at 256x224 EPX
is ~26% slower on eight workers than on one, and LQ2x/2xSaI lose
their four-worker gain at eight.

video_filter_threads (Settings > Video > Output, under Video Filter;
advanced) picks the count: 0 is Automatic, the previous behaviour and
the default; 1-16 forces that many workers. Changing it while a filter
is running rebuilds the filter on the new count at once; with no filter
loaded the value applies at the next load.

video_driver_init_filter() now frees a live filter before building the
new one. Re-init while content was running (the new setting, and also
a video filter path change) previously overwrote state_filter and
state_buffer, leaking the old filter, its worker pool and its output
buffer.

tools/settings_table.reference is regenerated; it had drifted on
master (HDR, custom viewport and window size rows) independently of
this change.

config.def.h

@@ -514,6 +514,11 @@
 #define MINIMUM_SWAP_INTERVAL 1
 #define MAXIMUM_SWAP_INTERVAL 16
 
+/* Worker threads a CPU video filter runs on; 0 (Automatic) uses one
+ * per detected CPU core. */
+#define DEFAULT_VIDEO_FILTER_THREADS 0
+#define MAXIMUM_VIDEO_FILTER_THREADS 16
+
 /* Threaded video: the core runs on one thread and the video driver
  * presents on another. Off by default, as it has always been; the
  * Switch keeps its own default. When it is on, hardware-rendered cores

configuration.h

@@ -258,6 +258,7 @@ typedef struct settings
       unsigned video_scale_integer_axis;
       unsigned video_scale_integer_scaling;
       unsigned video_max_swapchain_images;
+      unsigned video_filter_threads;
       unsigned video_swap_interval;
       unsigned video_hard_sync_frames;
       unsigned video_frame_delay;

gfx/video_driver.c

@@ -2358,6 +2358,11 @@ void video_driver_init_filter(enum retro_pixel_format colfmt_int,
    video_thread_wait_idle();
 #endif
 
+   /* Re-init (a new filter path, a new thread count) replaces the live
+    * filter: release it and its output buffer rather than orphan them */
+   if (video_st->state_filter)
+      video_driver_filter_free();
+
    if (video_driver_is_hw_context())
    {
       RARCH_WARN("[Video] Cannot use CPU filters when hardware rendering is used.\n");
@@ -2366,7 +2371,7 @@ void video_driver_init_filter(enum retro_pixel_format colfmt_int,
 
    if (!(video_st->state_filter = rarch_softfilter_new(
          settings->paths.path_softfilter_plugin,
-         RARCH_SOFTFILTER_THREADS_AUTO, colfmt, dims)))
+         settings->uints.video_filter_threads, colfmt, dims)))
    {
       RARCH_ERR("[Video] Failed to load filter.\n");
       return;

menu/cbs/menu_cbs_sublabel.c

@@ -2320,6 +2320,9 @@ int menu_cbs_init_bind_sublabel(menu_file_list_cbs_t *cbs,
       { MENU_ENUM_LABEL_CHEEVOS_PASSWORD, MENU_ENUM_SUBLABEL_CHEEVOS_PASSWORD },
       { MENU_ENUM_LABEL_VIDEO_FILTER, MENU_ENUM_SUBLABEL_VIDEO_FILTER },
       { MENU_ENUM_LABEL_VIDEO_FILTER_REMOVE, MENU_ENUM_SUBLABEL_VIDEO_FILTER_REMOVE },
+#ifdef HAVE_VIDEO_FILTER
+      { MENU_ENUM_LABEL_VIDEO_FILTER_THREADS, MENU_ENUM_SUBLABEL_VIDEO_FILTER_THREADS },
+#endif
       { MENU_ENUM_LABEL_VIDEO_CROP_OVERSCAN, MENU_ENUM_SUBLABEL_VIDEO_CROP_OVERSCAN },
       { MENU_ENUM_LABEL_VIDEO_SMOOTH, MENU_ENUM_SUBLABEL_VIDEO_SMOOTH },
       { MENU_ENUM_LABEL_VIDEO_FONT_ENABLE, MENU_ENUM_SUBLABEL_VIDEO_FONT_ENABLE },

menu/menu_displaylist.c

@@ -9532,6 +9532,11 @@ unsigned menu_displaylist_build_list(
                      PARSE_ONLY_PATH, false) == 0)
                count++;
 
+            if (MENU_DISPLAYLIST_PARSE_SETTINGS_ENUM(list,
+                     MENU_ENUM_LABEL_VIDEO_FILTER_THREADS,
+                     PARSE_ONLY_UINT, false) == 0)
+               count++;
+
             if (*settings->paths.path_softfilter_plugin)
                if (menu_entries_append(list,
                      msg_hash_to_str(MENU_ENUM_LABEL_VALUE_VIDEO_FILTER_REMOVE),

menu/menu_setting.c

@@ -7070,6 +7070,18 @@ static size_t setting_get_string_representation_video_swap_interval(rarch_settin
    return snprintf(s, len, "%u", *setting->value.target.unsigned_integer);
 }
 
+#ifdef HAVE_VIDEO_FILTER
+static size_t setting_get_string_representation_uint_video_filter_threads(
+      rarch_setting_t *setting, char *s, size_t len)
+{
+   if (!setting)
+      return 0;
+   if (*setting->value.target.unsigned_integer == 0)
+      return strlcpy(s, msg_hash_to_str(MENU_ENUM_LABEL_VALUE_AUTO), len);
+   return snprintf(s, len, "%u", *setting->value.target.unsigned_integer);
+}
+#endif
+
 static size_t setting_get_string_representation_black_frame_insertion(rarch_setting_t *setting,
       char *s, size_t len)
 {
@@ -9694,6 +9706,14 @@ static void general_write_handler(rarch_setting_t *setting)
                   log_dir);
          }
          break;
+#ifdef HAVE_VIDEO_FILTER
+      case MENU_ENUM_LABEL_VIDEO_FILTER_THREADS:
+         /* Rebuild a running filter on the new worker count; with no
+          * filter loaded the value is picked up at the next load. */
+         if (video_state_get_ptr()->state_filter)
+            command_event(CMD_EVENT_VIDEO_FILTER_INIT, NULL);
+         break;
+#endif
       case MENU_ENUM_LABEL_VIDEO_SMOOTH:
       case MENU_ENUM_LABEL_VIDEO_CTX_SCALING:
 #if defined(DINGUX)

msg_hash_lbl_str.h

@@ -1630,6 +1630,7 @@
 #define MENU_ENUM_LABEL_VIDEO_FILTER_STR "video_filter"
 #define MENU_ENUM_LABEL_VIDEO_FILTER_DIR_STR "video_filter_dir"
 #define MENU_ENUM_LABEL_VIDEO_FILTER_ENABLE_STR "filter_enable"
+#define MENU_ENUM_LABEL_VIDEO_FILTER_THREADS_STR "video_filter_threads"
 #define MENU_ENUM_LABEL_VIDEO_FONT_ENABLE_STR "video_font_enable"
 #define MENU_ENUM_LABEL_VIDEO_FONT_PATH_STR "video_font_path"
 #define MENU_ENUM_LABEL_VIDEO_FONT_SIZE_STR "video_font_size"

settings/settings_def_video_filter_rotation.h

@@ -26,4 +26,9 @@ S_BOOL(video_filter_enable, VIDEO_FILTER_ENABLE,
       "Video Filter Enable",
       "Apply Video Filter. Is a hint that does not necessarily have to be honored by the video driver.")
 #endif
+S_UINT(video_filter_threads, VIDEO_FILTER_THREADS,
+      "video_filter_threads",
+      DEFAULT_VIDEO_FILTER_THREADS, SD_FLAG_ADVANCED, SDESC_RANGE_MINMAX, CMD_EVENT_NONE, 0, MAXIMUM_VIDEO_FILTER_THREADS, 1, 0, setting_action_ok_uint, setting_get_string_representation_uint_video_filter_threads,
+      "Video Filter Threads",
+      "How many threads a CPU video filter may run on. 'Automatic' uses one per CPU core. Light filters can run slower when spread over many threads, so a lower count may be faster for them; heavy filters such as NTSC gain the most from more threads. A running filter picks up the change straight away.")
 #endif

tools/settings_table.reference

@@ -224,6 +224,8 @@ bool	pause_on_disconnect	1	0
 bool	video_frame_delay_auto	1	0
 bool	audio_fastpath_s16	1	0
 bool	audio_resampler_hq_oversampling	1	0
+bool	video_hdr_use_display_peak	1	0
+bool	video_hdr_send_luminance	1	0
 bool	video_window_custom_size_enable	1	0
 bool	input_overlay_auto_scale	1	1
 bool	video_font_enable	1	1
@@ -405,6 +407,10 @@ bool	netplay_request_device_p13	1	0
 bool	netplay_request_device_p14	1	0
 bool	netplay_request_device_p15	1	0
 bool	netplay_request_device_p16	1	0
+int	custom_viewport_x	0	0
+int	custom_viewport_y	0	0
+int	video_windowed_position_x	1	0
+int	video_windowed_position_y	1	0
 int	crt_switch_center_adjust	0	0
 int	crt_switch_porch_adjust	0	0
 int	crt_switch_vertical_adjust	0	0
@@ -436,12 +442,10 @@ uint	microphone_latency	0	0
 uint	microphone_rate	1	48000
 uint	custom_viewport_width	0	0
 uint	custom_viewport_height	0	0
-uint	custom_viewport_x	0	0
-uint	custom_viewport_y	0	0
-uint	video_windowed_position_x	1	0
-uint	video_windowed_position_y	1	0
 uint	video_windowed_position_width	1	1280
 uint	video_windowed_position_height	1	720
+uint	video_window_auto_width_max	1	1920
+uint	video_window_auto_height_max	1	1080
 uint	video_fse_negotiation	1	VIDEO_FSE_RELAXED
 uint	video_fullscreen_x	1	0
 uint	video_fullscreen_y	1	0
@@ -577,8 +581,6 @@ uint	microphone_block_frames	1	0
 uint	menu_thumbnail_preview_threads	1	1
 uint	menu_file_browser_extension_display	1	MENU_FILE_BROWSER_EXTENSION_DISPLAY_ALWAYS
 uint	video_scale	1	3
-uint	video_window_auto_width_max	1	1920
-uint	video_window_auto_height_max	1	1080
 uint	video_window_opacity	1	100
 uint	video_stream_port	1	56400
 uint	video_record_scale_factor	1	1
@@ -889,6 +891,8 @@ bool	pause_on_disconnect	1	0
 bool	video_frame_delay_auto	1	0
 bool	audio_fastpath_s16	1	0
 bool	audio_resampler_hq_oversampling	1	0
+bool	video_hdr_use_display_peak	1	0
+bool	video_hdr_send_luminance	1	0
 bool	video_window_custom_size_enable	1	0
 bool	input_overlay_auto_scale	1	0
 bool	video_font_enable	1	1
@@ -1069,6 +1073,10 @@ bool	netplay_request_device_p13	1	0
 bool	netplay_request_device_p14	1	0
 bool	netplay_request_device_p15	1	0
 bool	netplay_request_device_p16	1	0
+int	custom_viewport_x	0	0
+int	custom_viewport_y	0	0
+int	video_windowed_position_x	1	0
+int	video_windowed_position_y	1	0
 int	crt_switch_center_adjust	0	0
 int	crt_switch_porch_adjust	0	0
 int	crt_switch_vertical_adjust	0	0
@@ -1099,12 +1107,10 @@ uint	microphone_latency	0	0
 uint	microphone_rate	1	48000
 uint	custom_viewport_width	0	0
 uint	custom_viewport_height	0	0
-uint	custom_viewport_x	0	0
-uint	custom_viewport_y	0	0
-uint	video_windowed_position_x	1	0
-uint	video_windowed_position_y	1	0
 uint	video_windowed_position_width	1	1280
 uint	video_windowed_position_height	1	720
+uint	video_window_auto_width_max	1	1920
+uint	video_window_auto_height_max	1	1080
 uint	video_fse_negotiation	1	VIDEO_FSE_RELAXED
 uint	video_fullscreen_x	1	0
 uint	video_fullscreen_y	1	0
@@ -1238,8 +1244,6 @@ uint	microphone_block_frames	1	0
 uint	menu_thumbnail_preview_threads	1	1
 uint	menu_file_browser_extension_display	1	MENU_FILE_BROWSER_EXTENSION_DISPLAY_ALWAYS
 uint	video_scale	1	3
-uint	video_window_auto_width_max	1	1920
-uint	video_window_auto_height_max	1	1080
 uint	video_window_opacity	1	100
 uint	video_stream_port	1	56400
 uint	video_record_scale_factor	1	1
@@ -1559,6 +1563,8 @@ bool	pause_on_disconnect	1	0
 bool	video_frame_delay_auto	1	0
 bool	audio_fastpath_s16	1	0
 bool	audio_resampler_hq_oversampling	1	0
+bool	video_hdr_use_display_peak	1	0
+bool	video_hdr_send_luminance	1	0
 bool	video_window_custom_size_enable	1	0
 bool	input_overlay_auto_scale	1	0
 bool	video_font_enable	1	1
@@ -1740,6 +1746,10 @@ bool	netplay_request_device_p13	1	0
 bool	netplay_request_device_p14	1	0
 bool	netplay_request_device_p15	1	0
 bool	netplay_request_device_p16	1	0
+int	custom_viewport_x	0	0
+int	custom_viewport_y	0	0
+int	video_windowed_position_x	1	0
+int	video_windowed_position_y	1	0
 int	crt_switch_center_adjust	0	0
 int	crt_switch_porch_adjust	0	0
 int	crt_switch_vertical_adjust	0	0
@@ -1771,12 +1781,14 @@ uint	microphone_latency	0	0
 uint	microphone_rate	1	48000
 uint	custom_viewport_width	0	0
 uint	custom_viewport_height	0	0
-uint	custom_viewport_x	0	0
-uint	custom_viewport_y	0	0
-uint	video_windowed_position_x	1	0
-uint	video_windowed_position_y	1	0
 uint	video_windowed_position_width	1	1280
 uint	video_windowed_position_height	1	720
+uint	video_window_auto_width_max	1	1920
+uint	video_window_auto_height_max	1	1080
+uint	desktop_menu_window_x	1	0
+uint	desktop_menu_window_y	1	0
+uint	desktop_menu_window_width	1	0
+uint	desktop_menu_window_height	1	0
 uint	video_fse_negotiation	1	VIDEO_FSE_RELAXED
 uint	video_fullscreen_x	1	0
 uint	video_fullscreen_y	1	0
@@ -1844,10 +1856,6 @@ uint	desktop_menu_icon_view_zoom	1	50
 uint	desktop_menu_all_playlists_list_max_count	1	0
 uint	desktop_menu_all_playlists_grid_max_count	1	0
 uint	desktop_menu_theme	1	0
-uint	desktop_menu_window_x	1	0
-uint	desktop_menu_window_y	1	0
-uint	desktop_menu_window_width	1	0
-uint	desktop_menu_window_height	1	0
 uint	ai_service_source_lang	1	0
 uint	ai_service_target_lang	1	0
 uint	video_scale_integer_axis	1	0
@@ -1925,8 +1933,6 @@ uint	microphone_block_frames	1	0
 uint	menu_thumbnail_preview_threads	1	1
 uint	menu_file_browser_extension_display	1	MENU_FILE_BROWSER_EXTENSION_DISPLAY_ALWAYS
 uint	video_scale	1	3
-uint	video_window_auto_width_max	1	1920
-uint	video_window_auto_height_max	1	1080
 uint	video_window_opacity	1	100
 uint	video_stream_port	1	56400
 uint	video_record_scale_factor	1	1