-overlapping-replay > -overlap-replay

README.md

@@ -36,7 +36,8 @@ When recording Legend of Zelda Breath of the Wild at 4k, fps drops from 30 to 7
 When recording GTA V at 4k on highest settings, fps drops from 60 to 23 when using obs-nvfbc + nvenc, however when using this screen recorder the fps only drops to 58.\
 GPU Screen Recorder also produces much smoother videos than OBS when GPU utilization is close to 100%, see comparison here: [https://www.youtube.com/watch?v=zfj4sNVLLLg](https://www.youtube.com/watch?v=zfj4sNVLLLg).\
 GPU Screen Recorder has much better performance than OBS Studio even with version 30.2 that does "zero-copy" recording and encoding, see: [https://www.youtube.com/watch?v=jdroRjibsDw](https://www.youtube.com/watch?v=jdroRjibsDw).\
-It is recommended to save the video to a SSD because of the large file size, which a slow HDD might not be fast enough to handle. Using variable framerate mode (-fm vfr) which is the default is also recommended as this reduces encoding load. Ultra quality is also overkill most of the time, very high (the default) or lower quality is usually enough.
+It is recommended to save the video to a SSD because of the large file size, which a slow HDD might not be fast enough to handle. Using variable framerate mode (-fm vfr) which is the default is also recommended as this reduces encoding load. Ultra quality is also overkill most of the time, very high (the default) or lower quality is usually enough.\
+Note that recording on AMD can have some performance issues on Wayland in the recording itself when recording without desktop portal unless your mesa version is 25.0.0 or greater.
 ## Note about optimal performance on NVIDIA
 NVIDIA driver has a "feature" (read: bug) where it will downclock memory transfer rate when a program uses cuda (or nvenc, which uses cuda), such as GPU Screen Recorder. To work around this bug, GPU Screen Recorder can overclock your GPU memory transfer rate to it's normal optimal level.\
 To enable overclocking for optimal performance use the `-oc` option when running GPU Screen Recorder. You also need to have "Coolbits" NVIDIA X setting set to "12" to enable overclocking. You can automatically add this option if you run `sudo nvidia-xconfig --cool-bits=12` and then reboot your computer.\
@@ -62,6 +63,7 @@ Here are some known unofficial packages:
 * Nix: [NixOS wiki](https://wiki.nixos.org/wiki/Gpu-screen-recorder)
 * openSUSE: [openSUSE software repository](https://software.opensuse.org/package/gpu-screen-recorder)
 * Fedora: [Copr](https://copr.fedorainfracloud.org/coprs/brycensranch/gpu-screen-recorder-git/)
+* OpenMandriva: [gpu-screen-recorder](https://github.com/OpenMandrivaAssociation/gpu-screen-recorder/tree/master)
 
 # Dependencies
 GPU Screen Recorder uses meson build system so you need to install `meson` to build GPU Screen Recorder.
@@ -178,5 +180,5 @@ You have to either record in hdr mode (-k `hevc_hdr` or -k `av1_hdr` option) to
 You can record with desktop portal option (`-w portal`) instead which ignores night light, if you are ok with recording without HDR.
 ## Kdenlive says that the video is not usable for editing because it has variable frame rate
 To fix this you can either record the video in .mkv format or constant frame rate (-fm cfr).
-## Colors look incorrect when recording HDR with hevc_hdr/av1_hdr
-The latest version of KDE Plasma breaks HDR for recording applications. Wayland in general doesn't properly support recording HDR yet. Use desktop portal option (`-w portal`) for now to turn HDR recording into SDR.
+## Colors look incorrect when recording HDR (with hevc_hdr/av1_hdr) or using an ICC profile
+The latest version of KDE Plasma breaks HDR and ICC profiles for recording applications. Wayland in general doesn't properly support recording HDR yet. Use desktop portal option (`-w portal`) for now to turn HDR recording into SDR and to be able to record with correct colors when using an ICC profile.

TODO

@@ -69,8 +69,6 @@ Exit if X11/Wayland killed (if drm plane dead or something?)
 
 Use SRC_W and SRC_H for screen plane instead of crtc_w and crtc_h.
 
-Make it possible to select which /dev/dri/card* to use, but that requires opengl to also use the same card. Not sure if that is possible for amd, intel and nvidia without using vulkan instead.
-
 Test if p2 state can be worked around by using pure nvenc api and overwriting cuInit/cuCtxCreate* to not do anything. Cuda might be loaded when using nvenc but it might not be used, with certain record options? (such as h264 p5).
     nvenc uses cuda when using b frames and rgb->yuv conversion, so convert the image ourselves instead.-
 
@@ -207,4 +205,19 @@ Use different exit codes for different errors. Use one for invalid -w option, an
 Ffmpeg fixed black bars in videos on amd when using hevc and when recording at some resolutions, such as 1080p:
     https://github.com/FFmpeg/FFmpeg/commit/bcfbf2bac8f9eeeedc407b40596f5c7aaa0d5b47
     https://github.com/FFmpeg/FFmpeg/commit/d0facac679faf45d3356dff2e2cb382580d7a521
-    Disable gpu screen recorder black bar handling when using hevc on amd when the libavcodec version is the one that comes after those commits.
+    Disable gpu screen recorder black bar handling when using hevc on amd when the libavcodec version is the one that comes after those commits.
+
+Use opengl compute shader instead of graphics shader. This might allow for better performance when games are using 100% of graphics unit which might fix issue with 100% gpu usage causing gpu screen recorder to run slow when not using vaapi to convert rgb to nv12(?).
+
+Always disable prime run/dri prime and list all monitors to record from from all cards.
+    Do this instead of adding an option to choose which gpu to use.
+    On X11 the primary gpu will always have the framebuffer for all monitors combined.
+        Use randr to list all monitors and always record and encode with the primary gpu.
+    On Wayland each gpu will have its own list of monitors with framebuffers.
+        Iterate through all cards with drm and list all monitors with associated framebuffers and when choosing a monitor to record
+        automatically use the associated gpu card.
+
+Allow flv av1 if recent ffmpeg version and streaming to youtube (and twitch?) and for custom services.
+Use explicit sync in pipewire video code: https://docs.pipewire.org/page_dma_buf.html.
+
+Support vaapi rotation. Support for it is added in mesa here: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/32919.

include/egl.h

@@ -277,7 +277,7 @@ struct gsr_egl {
     unsigned char (*glUnmapBuffer)(unsigned int target);
 };
 
-bool gsr_egl_load(gsr_egl *self, gsr_window *window, bool is_monitor_capture);
+bool gsr_egl_load(gsr_egl *self, gsr_window *window, bool is_monitor_capture, bool enable_debug);
 void gsr_egl_unload(gsr_egl *self);
 
 /* Does opengl swap with egl or glx, depending on which one is active */

include/utils.h

@@ -50,7 +50,7 @@ drm_connector_type_count* drm_connector_types_get_index(drm_connector_type_count
 uint32_t monitor_identifier_from_type_and_count(int monitor_type_index, int monitor_type_count);
 
 bool gl_get_gpu_info(gsr_egl *egl, gsr_gpu_info *info);
-bool gl_driver_version_greater_than(const gsr_egl *egl, int major, int minor, int patch);
+bool gl_driver_version_greater_than(const gsr_gpu_info *gpu_info, int major, int minor, int patch);
 
 bool try_card_has_valid_plane(const char *card_path);
 /* |output| should be at least 128 bytes in size */

kms/client/kms_client.c

@@ -11,6 +11,7 @@
 #include <sys/socket.h>
 #include <sys/un.h>
 #include <sys/wait.h>
+#include <poll.h>
 #include <sys/stat.h>
 #include <sys/capability.h>
 
@@ -318,17 +319,14 @@ int gsr_kms_client_init(gsr_kms_client *self, const char *card_path) {
     }
 
     fprintf(stderr, "gsr info: gsr_kms_client_init: waiting for server to connect\n");
+    struct pollfd poll_fd = {
+        .fd = self->initial_socket_fd,
+        .events = POLLIN,
+        .revents = 0
+    };
     for(;;) {
-        struct timeval tv;
-        fd_set rfds;
-        FD_ZERO(&rfds);
-        FD_SET(self->initial_socket_fd, &rfds);
-
-        tv.tv_sec = 0;
-        tv.tv_usec = 100 * 1000; // 100 ms
-
-        int select_res = select(1 + self->initial_socket_fd, &rfds, NULL, NULL, &tv);
-        if(select_res > 0) {
+        int poll_res = poll(&poll_fd, 1, 100);
+        if(poll_res > 0 && (poll_fd.revents & POLLIN)) {
             socklen_t sock_len = 0;
             self->initial_client_fd = accept(self->initial_socket_fd, (struct sockaddr*)&remote_addr, &sock_len);
             if(self->initial_client_fd == -1) {

meson.build

@@ -1,4 +1,4 @@
-project('gpu-screen-recorder', ['c', 'cpp'], version : '5.0.0', default_options : ['warning_level=2'])
+project('gpu-screen-recorder', ['c', 'cpp'], version : '5.0.2', default_options : ['warning_level=2'])
 
 add_project_arguments('-Wshadow', language : ['c', 'cpp'])
 if get_option('buildtype') == 'debug'

project.conf

@@ -1,7 +1,7 @@
 [package]
 name = "gpu-screen-recorder"
 type = "executable"
-version = "5.0.0"
+version = "5.0.2"
 platforms = ["posix"]
 
 [config]

src/capture/kms.c

@@ -18,6 +18,8 @@
 #include <libavutil/mastering_display_metadata.h>
 #include <libavformat/avformat.h>
 
+#define FIND_CRTC_BY_NAME_TIMEOUT_SECONDS 2.0
+
 #define HDMI_STATIC_METADATA_TYPE1 0
 #define HDMI_EOTF_SMPTE_ST2084 2
 
@@ -56,6 +58,7 @@ typedef struct {
     AVCodecContext *video_codec_context;
     bool performance_error_shown;
     bool fast_path_failed;
+    bool mesa_supports_compute_only_vaapi_copy;
 
     //int drm_fd;
     //uint64_t prev_sequence;
@@ -63,6 +66,8 @@ typedef struct {
 
     vec2i prev_target_pos;
     vec2i prev_plane_size;
+
+    double last_time_monitor_check;
 } gsr_capture_kms;
 
 static void gsr_capture_kms_cleanup_kms_fds(gsr_capture_kms *self) {
@@ -227,14 +232,17 @@ static int gsr_capture_kms_start(gsr_capture *cap, AVCodecContext *video_codec_c
         video_codec_context->height = FFALIGN(self->params.output_resolution.y, 2);
     }
 
-    self->fast_path_failed = self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && !gl_driver_version_greater_than(self->params.egl, 24, 0, 9);
+    self->fast_path_failed = self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && !gl_driver_version_greater_than(&self->params.egl->gpu_info, 24, 0, 9);
     if(self->fast_path_failed)
         fprintf(stderr, "gsr warning: gsr_capture_kms_start: your amd driver (mesa) version is known to be buggy (<= version 24.0.9), falling back to opengl copy\n");
 
+    self->mesa_supports_compute_only_vaapi_copy = self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && gl_driver_version_greater_than(&self->params.egl->gpu_info, 24, 3, 6);
+
     frame->width = video_codec_context->width;
     frame->height = video_codec_context->height;
 
     self->video_codec_context = video_codec_context;
+    self->last_time_monitor_check = clock_get_monotonic_seconds();
     return 0;
 }
 
@@ -428,7 +436,7 @@ static gsr_kms_response_item* find_monitor_drm(gsr_capture_kms *self, bool *capt
     }
 
     // Will never happen on wayland unless the target monitor has been disconnected
-    if(!drm_fd) {
+    if(!drm_fd && self->is_x11) {
         drm_fd = find_largest_drm(&self->kms_response);
         *capture_is_combined_plane = true;
     }
@@ -555,6 +563,55 @@ static void gsr_capture_kms_update_capture_size_change(gsr_capture_kms *self, gs
     }
 }
 
+static void gsr_capture_kms_update_connector_ids(gsr_capture_kms *self) {
+    const double now = clock_get_monotonic_seconds();
+    if(now - self->last_time_monitor_check < FIND_CRTC_BY_NAME_TIMEOUT_SECONDS)
+        return;
+
+    self->last_time_monitor_check = now;
+    /* TODO: Assume for now that there is only 1 framebuffer for all monitors and it doesn't change */
+    if(self->is_x11)
+        return;
+
+    self->monitor_id.num_connector_ids = 0;
+    const gsr_connection_type connection_type = self->is_x11 ? GSR_CONNECTION_X11 : GSR_CONNECTION_DRM;
+    // MonitorCallbackUserdata monitor_callback_userdata = {
+    //     &self->monitor_id,
+    //     self->params.display_to_capture, strlen(self->params.display_to_capture),
+    //     0,
+    // };
+    // for_each_active_monitor_output(self->params.egl->window, self->params.egl->card_path, connection_type, monitor_callback, &monitor_callback_userdata);
+
+    gsr_monitor monitor;
+    if(!get_monitor_by_name(self->params.egl, connection_type, self->params.display_to_capture, &monitor)) {
+        fprintf(stderr, "gsr error: gsr_capture_kms_update_connector_ids: failed to find monitor by name \"%s\"\n", self->params.display_to_capture);
+        return;
+    }
+
+    self->monitor_id.num_connector_ids = 1;
+    self->monitor_id.connector_ids[0] = monitor.connector_id;
+
+    monitor.name = self->params.display_to_capture;
+    self->monitor_rotation = drm_monitor_get_display_server_rotation(self->params.egl->window, &monitor);
+
+    self->capture_pos = monitor.pos;
+    /* Monitor size is already rotated on x11 when the monitor is rotated, no need to apply it ourselves */
+    if(self->is_x11)
+        self->capture_size = monitor.size;
+    else
+        self->capture_size = rotate_capture_size_if_rotated(self, monitor.size);
+}
+
+static void gsr_capture_kms_fail_fast_path_if_not_fast(gsr_capture_kms *self, uint32_t pixel_format) {
+    const uint8_t pixel_format_color_depth_1 = (pixel_format >> 16) & 0xFF;
+    if(!self->fast_path_failed && self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && !self->mesa_supports_compute_only_vaapi_copy && (pixel_format_color_depth_1 == '3' || pixel_format_color_depth_1 == '4')) {
+        self->fast_path_failed = true;
+        fprintf(stderr, "gsr warning: gsr_capture_kms_capture: the monitor you are recording is in 10/12-bit color format and your mesa version is <= 24.3.6, composition will be used."
+            " If you experience performance problems in the video then record on a single window on X11 or use portal capture option instead or disable 10/12-bit color option in your desktop environment settings,"
+            " or try to record the monitor on X11 instead (if you aren't already doing that) or update your mesa version.\n");
+    }
+}
+
 static int gsr_capture_kms_capture(gsr_capture *cap, AVFrame *frame, gsr_color_conversion *color_conversion) {
     gsr_capture_kms *self = cap->priv;
 
@@ -574,6 +631,8 @@ static int gsr_capture_kms_capture(gsr_capture *cap, AVFrame *frame, gsr_color_c
         return -1;
     }
 
+    gsr_capture_kms_update_connector_ids(self);
+
     bool capture_is_combined_plane = false;
     const gsr_kms_response_item *drm_fd = find_monitor_drm(self, &capture_is_combined_plane);
     if(!drm_fd) {
@@ -586,10 +645,13 @@ static int gsr_capture_kms_capture(gsr_capture *cap, AVFrame *frame, gsr_color_c
 
     if(!self->performance_error_shown && self->monitor_rotation != GSR_MONITOR_ROT_0 && video_codec_context_is_vaapi(self->video_codec_context) && self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD) {
         self->performance_error_shown = true;
-        fprintf(stderr,"gsr warning: gsr_capture_kms_capture: the monitor you are recording is rotated, composition will have to be used."
-            " If you are experience performance problems in the video then record a single window on X11 or use portal capture option instead\n");
+        self->fast_path_failed = true;
+        fprintf(stderr, "gsr warning: gsr_capture_kms_capture: the monitor you are recording is rotated, composition will have to be used."
+            " If you experience performance problems in the video then record a single window on X11 or use portal capture option instead\n");
     }
 
+    gsr_capture_kms_fail_fast_path_if_not_fast(self, drm_fd->pixel_format);
+
     self->capture_size = rotate_capture_size_if_rotated(self, (vec2i){ drm_fd->src_w, drm_fd->src_h });
 
     const bool is_scaled = self->params.output_resolution.x > 0 && self->params.output_resolution.y > 0;

src/capture/portal.c

@@ -27,6 +27,7 @@ typedef struct {
 
     AVCodecContext *video_codec_context;
     bool fast_path_failed;
+    bool mesa_supports_compute_only_vaapi_copy;
 } gsr_capture_portal;
 
 static void gsr_capture_portal_cleanup_plane_fds(gsr_capture_portal *self) {
@@ -310,10 +311,12 @@ static int gsr_capture_portal_start(gsr_capture *cap, AVCodecContext *video_code
         video_codec_context->height = FFALIGN(self->params.output_resolution.y, 2);
     }
 
-    self->fast_path_failed = self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && !gl_driver_version_greater_than(self->params.egl, 24, 0, 9);
+    self->fast_path_failed = self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && !gl_driver_version_greater_than(&self->params.egl->gpu_info, 24, 0, 9);
     if(self->fast_path_failed)
         fprintf(stderr, "gsr warning: gsr_capture_kms_start: your amd driver (mesa) version is known to be buggy (<= version 24.0.9), falling back to opengl copy\n");
 
+    self->mesa_supports_compute_only_vaapi_copy = self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && gl_driver_version_greater_than(&self->params.egl->gpu_info, 24, 3, 6);
+
     frame->width = video_codec_context->width;
     frame->height = video_codec_context->height;
 
@@ -325,6 +328,16 @@ static int max_int(int a, int b) {
     return a > b ? a : b;
 }
 
+static void gsr_capture_portal_fail_fast_path_if_not_fast(gsr_capture_portal *self, uint32_t pixel_format) {
+    const uint8_t pixel_format_color_depth_1 = (pixel_format >> 16) & 0xFF;
+    if(!self->fast_path_failed && self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && !self->mesa_supports_compute_only_vaapi_copy && (pixel_format_color_depth_1 == '3' || pixel_format_color_depth_1 == '4')) {
+        self->fast_path_failed = true;
+        fprintf(stderr, "gsr warning: gsr_capture_kms_capture: the monitor you are recording is in 10/12-bit color format and your mesa version is <= 24.3.6, composition will be used."
+            " If you experience performance problems in the video then record on a single window on X11 instead or disable 10/12-bit color option in your desktop environment settings,"
+            " or try to record the monitor on X11 instead (if you aren't already doing that) or update your mesa version.\n");
+    }
+}
+
 static int gsr_capture_portal_capture(gsr_capture *cap, AVFrame *frame, gsr_color_conversion *color_conversion) {
     (void)frame;
     (void)color_conversion;
@@ -346,6 +359,8 @@ static int gsr_capture_portal_capture(gsr_capture *cap, AVFrame *frame, gsr_colo
         return 0;
     }
 
+    gsr_capture_portal_fail_fast_path_if_not_fast(self, pipewire_fourcc);
+
     const bool is_scaled = self->params.output_resolution.x > 0 && self->params.output_resolution.y > 0;
     vec2i output_size = is_scaled ? self->params.output_resolution : self->capture_size;
     output_size = scale_keep_aspect_ratio(self->capture_size, output_size);

src/capture/xcomposite.c

@@ -124,7 +124,7 @@ static int gsr_capture_xcomposite_start(gsr_capture *cap, AVCodecContext *video_
         video_codec_context->height = FFALIGN(self->params.output_resolution.y, 2);
     }
 
-    self->fast_path_failed = self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && !gl_driver_version_greater_than(self->params.egl, 24, 0, 9);
+    self->fast_path_failed = self->params.egl->gpu_info.vendor == GSR_GPU_VENDOR_AMD && !gl_driver_version_greater_than(&self->params.egl->gpu_info, 24, 0, 9);
     if(self->fast_path_failed)
         fprintf(stderr, "gsr warning: gsr_capture_kms_start: your amd driver (mesa) version is known to be buggy (<= version 24.0.9), falling back to opengl copy\n");
 

src/egl.c

@@ -344,25 +344,18 @@ static bool gsr_egl_load_gl(gsr_egl *self, void *library) {
     return true;
 }
 
-// #define GL_DEBUG_TYPE_ERROR               0x824C
-// static void debug_callback( unsigned int source,
-//                  unsigned int type,
-//                  unsigned int id,
-//                  unsigned int severity,
-//                  int length,
-//                  const char* message,
-//                  const void* userParam )
-// {
-//     (void)source;
-//     (void)id;
-//     (void)length;
-//     (void)userParam;
-//   fprintf( stderr, "GL CALLBACK: %s type = 0x%x, severity = 0x%x, message = %s\n",
-//            ( type == GL_DEBUG_TYPE_ERROR ? "** GL ERROR **" : "" ),
-//             type, severity, message );
-// }
+#define GL_DEBUG_TYPE_ERROR               0x824C
+#define GL_DEBUG_SEVERITY_NOTIFICATION    0x826B
+static void debug_callback(unsigned int source, unsigned int type, unsigned int id, unsigned int severity, int length, const char* message, const void* userParam) {
+    (void)source;
+    (void)id;
+    (void)length;
+    (void)userParam;
+    if(severity != GL_DEBUG_SEVERITY_NOTIFICATION)
+        fprintf(stderr, "gsr info: gl callback: %s type = 0x%x, severity = 0x%x, message = %s\n", type == GL_DEBUG_TYPE_ERROR ? "** GL ERROR **" : "", type, severity, message);
+}
 
-bool gsr_egl_load(gsr_egl *self, gsr_window *window, bool is_monitor_capture) {
+bool gsr_egl_load(gsr_egl *self, gsr_window *window, bool is_monitor_capture, bool enable_debug) {
     memset(self, 0, sizeof(gsr_egl));
     self->context_type = GSR_GL_CONTEXT_TYPE_EGL;
     self->window = window;
@@ -418,8 +411,10 @@ bool gsr_egl_load(gsr_egl *self, gsr_window *window, bool is_monitor_capture) {
     self->glEnable(GL_BLEND);
     self->glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
 
-    //self->glEnable(GL_DEBUG_OUTPUT);
-    //self->glDebugMessageCallback(debug_callback, NULL);
+    if(enable_debug) {
+        self->glEnable(GL_DEBUG_OUTPUT);
+        self->glDebugMessageCallback(debug_callback, NULL);
+    }
 
     return true;
 

src/main.cpp

@@ -1069,7 +1069,7 @@ static void open_video_hardware(AVCodecContext *codec_context, VideoQuality vide
 static void usage_header() {
     const bool inside_flatpak = getenv("FLATPAK_ID") != NULL;
     const char *program_name = inside_flatpak ? "flatpak run --command=gpu-screen-recorder com.dec05eba.gpu_screen_recorder" : "gpu-screen-recorder";
-    printf("usage: %s -w <window_id|monitor|focused|portal> [-c <container_format>] [-s WxH] -f <fps> [-a <audio_input>] [-q <quality>] [-r <replay_buffer_size_sec>] [-k h264|hevc|av1|vp8|vp9|hevc_hdr|av1_hdr|hevc_10bit|av1_10bit] [-ac aac|opus|flac] [-ab <bitrate>] [-oc yes|no] [-fm cfr|vfr|content] [-bm auto|qp|vbr|cbr] [-cr limited|full] [-df yes|no] [-sc <script_path>] [-cursor yes|no] [-keyint <value>] [-restore-portal-session yes|no] [-portal-session-token-filepath filepath] [-encoder gpu|cpu] [-o <output_file>] [--list-capture-options [card_path] [vendor]] [--list-audio-devices] [--list-application-audio] [-v yes|no] [--version] [-h|--help]\n", program_name);
+    printf("usage: %s -w <window_id|monitor|focused|portal> [-c <container_format>] [-s WxH] -f <fps> [-a <audio_input>] [-q <quality>] [-r <replay_buffer_size_sec>] [-overlap-replay yes|no] [-k h264|hevc|av1|vp8|vp9|hevc_hdr|av1_hdr|hevc_10bit|av1_10bit] [-ac aac|opus|flac] [-ab <bitrate>] [-oc yes|no] [-fm cfr|vfr|content] [-bm auto|qp|vbr|cbr] [-cr limited|full] [-df yes|no] [-sc <script_path>] [-cursor yes|no] [-keyint <value>] [-restore-portal-session yes|no] [-portal-session-token-filepath filepath] [-encoder gpu|cpu] [-o <output_file>] [--list-capture-options [card_path] [vendor]] [--list-audio-devices] [--list-application-audio] [-v yes|no] [-gl-debug yes|no] [--version] [-h|--help]\n", program_name);
     fflush(stdout);
 }
 
@@ -1130,6 +1130,12 @@ static void usage_full() {
     printf("        Note that the video data is stored in RAM, so don't use too long replay buffer time and use constant bitrate option (-bm cbr) to prevent RAM usage from going too high in busy scenes.\n");
     printf("        Optional, disabled by default.\n");
     printf("\n");
+    printf("  -overlap-replay\n");
+    printf("        Should replays overlap. For example if this is set to 'yes' and replay time (-r) is set to 60 seconds and a replay is saved once then the first replay video is 60 seconds long\n");
+    printf("        and if a replay is saved 10 seconds later then the second replay video will also be 60 seconds long and contain 50 seconds of the previous video as well.\n");
+    printf("        If this is set to 'no' then after a replay is saved the replay buffer data is cleared and the second replay will start from that point onward.\n");
+    printf("        Optional, set to 'yes' by default.\n");
+    printf("\n");
     printf("  -k    Video codec to use. Should be either 'auto', 'h264', 'hevc', 'av1', 'vp8', 'vp9', 'hevc_hdr', 'av1_hdr', 'hevc_10bit' or 'av1_10bit'.\n");
     printf("        Optional, set to 'auto' by default which defaults to 'h264'. Forcefully set to 'h264' if the file container type is 'flv'.\n");
     printf("        'hevc_hdr' and 'av1_hdr' option is not available on X11 nor when using the portal capture option.\n");
@@ -1229,6 +1235,9 @@ static void usage_full() {
     printf("\n");
     printf("  -v    Prints fps and damage info once per second. Optional, set to 'yes' by default.\n");
     printf("\n");
+    printf("  -gl-debug\n");
+    printf("        Print opengl debug output. Optional, set to 'no' by default.\n");
+    printf("\n");
     printf("  -h, --help\n");
     printf("        Show this help.\n");
     printf("\n");
@@ -1894,7 +1903,7 @@ static bool is_xwayland(Display *display) {
 
 static bool is_using_prime_run() {
     const char *prime_render_offload = getenv("__NV_PRIME_RENDER_OFFLOAD");
-    return prime_render_offload && strcmp(prime_render_offload, "1") == 0;
+    return (prime_render_offload && strcmp(prime_render_offload, "1") == 0) || getenv("DRI_PRIME");
 }
 
 static void disable_prime_run() {
@@ -1902,6 +1911,7 @@ static void disable_prime_run() {
     unsetenv("__NV_PRIME_RENDER_OFFLOAD_PROVIDER");
     unsetenv("__GLX_VENDOR_LIBRARY_NAME");
     unsetenv("__VK_LAYER_NV_optimus");
+    unsetenv("DRI_PRIME");
 }
 
 static gsr_window* gsr_window_create(Display *display, bool wayland) {
@@ -2121,7 +2131,7 @@ static void info_command() {
     }
 
     gsr_egl egl;
-    if(!gsr_egl_load(&egl, window, false)) {
+    if(!gsr_egl_load(&egl, window, false, false)) {
         fprintf(stderr, "gsr error: failed to load opengl\n");
         _exit(22);
     }
@@ -2144,6 +2154,7 @@ static void info_command() {
         puts("is_steam_deck|yes");
     else
         puts("is_steam_deck|no");
+    printf("gsr_version|%s\n", GSR_VERSION);
     puts("section=gpu_info");
     list_gpu_info(&egl);
     puts("section=video_codecs");
@@ -2233,7 +2244,7 @@ static void list_capture_options_command(const char *card_path, gsr_gpu_vendor v
         list_supported_capture_options(window, card_path, true);
     } else {
         gsr_egl egl;
-        if(!gsr_egl_load(&egl, window, false)) {
+        if(!gsr_egl_load(&egl, window, false, false)) {
             fprintf(stderr, "gsr error: failed to load opengl\n");
             _exit(1);
         }
@@ -3051,6 +3062,7 @@ int main(int argc, char **argv) {
         { "-q", Arg { {}, true, false } },
         { "-o", Arg { {}, true, false } },
         { "-r", Arg { {}, true, false } },
+        { "-overlap-replay", Arg { {}, true, false } },
         { "-k", Arg { {}, true, false } },
         { "-ac", Arg { {}, true, false } },
         { "-ab", Arg { {}, true, false } },
@@ -3059,6 +3071,7 @@ int main(int argc, char **argv) {
         { "-bm", Arg { {}, true, false } },
         { "-pixfmt", Arg { {}, true, false } },
         { "-v", Arg { {}, true, false } },
+        { "-gl-debug", Arg { {}, true, false } },
         { "-df", Arg { {}, true, false } },
         { "-sc", Arg { {}, true, false } },
         { "-cr", Arg { {}, true, false } },
@@ -3072,17 +3085,17 @@ int main(int argc, char **argv) {
     for(int i = 1; i < argc; i += 2) {
         auto it = args.find(argv[i]);
         if(it == args.end()) {
-            fprintf(stderr, "Invalid argument '%s'\n", argv[i]);
+            fprintf(stderr, "Error: invalid argument '%s'\n", argv[i]);
             usage();
         }
 
         if(!it->second.values.empty() && !it->second.list) {
-            fprintf(stderr, "Expected argument '%s' to only be specified once\n", argv[i]);
+            fprintf(stderr, "Error: expected argument '%s' to only be specified once\n", argv[i]);
             usage();
         }
 
         if(i + 1 >= argc) {
-            fprintf(stderr, "Missing value for argument '%s'\n", argv[i]);
+            fprintf(stderr, "Error: missing value for argument '%s'\n", argv[i]);
             usage();
         }
 
@@ -3091,7 +3104,7 @@ int main(int argc, char **argv) {
 
     for(auto &it : args) {
         if(!it.second.optional && !it.second.value()) {
-            fprintf(stderr, "Missing argument '%s'\n", it.first.c_str());
+            fprintf(stderr, "Error: missing argument '%s'\n", it.first.c_str());
             usage();
         }
     }
@@ -3226,6 +3239,20 @@ int main(int argc, char **argv) {
         usage();
     }
 
+    bool gl_debug = false;
+    const char *gl_debug_str = args["-gl-debug"].value();
+    if(!gl_debug_str)
+        gl_debug_str = "no";
+
+    if(strcmp(gl_debug_str, "yes") == 0) {
+        gl_debug = true;
+    } else if(strcmp(gl_debug_str, "no") == 0) {
+        gl_debug = false;
+    } else {
+        fprintf(stderr, "Error: -gl-debug should either be either 'yes' or 'no', got: '%s'\n", gl_debug_str);
+        usage();
+    }
+
     bool record_cursor = true;
     const char *record_cursor_str = args["-cursor"].value();
     if(!record_cursor_str)
@@ -3362,6 +3389,20 @@ int main(int argc, char **argv) {
         replay_buffer_size_secs += std::ceil(keyint); // Add a few seconds to account of lost packets because of non-keyframe packets skipped
     }
 
+    bool overlap_replay = true;
+    const char *overlap_replay_str = args["-overlap-replay"].value();
+    if(!overlap_replay_str)
+        overlap_replay_str = "yes";
+
+    if(strcmp(overlap_replay_str, "yes") == 0) {
+        overlap_replay = true;
+    } else if(strcmp(overlap_replay_str, "no") == 0) {
+        overlap_replay = false;
+    } else {
+        fprintf(stderr, "Error: -overlap-replap should either be either 'yes' or 'no', got: '%s'\n", overlap_replay_str);
+        usage();
+    }
+
     std::string window_str = args["-w"].value();
     const bool is_portal_capture = strcmp(window_str.c_str(), "portal") == 0;
 
@@ -3408,7 +3449,7 @@ int main(int argc, char **argv) {
 
     const bool is_monitor_capture = strcmp(window_str.c_str(), "focused") != 0 && !is_portal_capture && contains_non_hex_number(window_str.c_str());
     gsr_egl egl;
-    if(!gsr_egl_load(&egl, window, is_monitor_capture)) {
+    if(!gsr_egl_load(&egl, window, is_monitor_capture, gl_debug)) {
         fprintf(stderr, "gsr error: failed to load opengl\n");
         _exit(1);
     }
@@ -3847,6 +3888,7 @@ int main(int argc, char **argv) {
     std::mutex audio_filter_mutex;
 
     const double record_start_time = clock_get_monotonic_seconds();
+    std::atomic<double> replay_start_time(record_start_time);
     std::deque<std::shared_ptr<PacketData>> frame_data_queue;
     bool frames_erased = false;
 
@@ -3965,7 +4007,7 @@ int main(int argc, char **argv) {
                                 ret = avcodec_send_frame(audio_track.codec_context, audio_device.frame);
                                 if(ret >= 0) {
                                     // TODO: Move to separate thread because this could write to network (for example when livestreaming)
-                                    receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, audio_device.frame->pts, av_format_context, record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset);
+                                    receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, audio_device.frame->pts, av_format_context, replay_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset);
                                 } else {
                                     fprintf(stderr, "Failed to encode audio!\n");
                                 }
@@ -3997,7 +4039,7 @@ int main(int argc, char **argv) {
                             ret = avcodec_send_frame(audio_track.codec_context, audio_device.frame);
                             if(ret >= 0) {
                                 // TODO: Move to separate thread because this could write to network (for example when livestreaming)
-                                receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, audio_device.frame->pts, av_format_context, record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset);
+                                receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, audio_device.frame->pts, av_format_context, replay_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset);
                             } else {
                                 fprintf(stderr, "Failed to encode audio!\n");
                             }
@@ -4031,7 +4073,7 @@ int main(int argc, char **argv) {
                             err = avcodec_send_frame(audio_track.codec_context, aframe);
                             if(err >= 0){
                                 // TODO: Move to separate thread because this could write to network (for example when livestreaming)
-                                receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, aframe->pts, av_format_context, record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset);
+                                receive_frames(audio_track.codec_context, audio_track.stream_index, audio_track.stream, aframe->pts, av_format_context, replay_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset);
                             } else {
                                 fprintf(stderr, "Failed to encode audio!\n");
                             }
@@ -4177,7 +4219,7 @@ int main(int argc, char **argv) {
                 if(ret == 0) {
                     // TODO: Move to separate thread because this could write to network (for example when livestreaming)
                     receive_frames(video_codec_context, VIDEO_STREAM_INDEX, video_stream, video_frame->pts, av_format_context,
-                        record_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset);
+                        replay_start_time, frame_data_queue, replay_buffer_size_secs, frames_erased, write_output_mutex, paused_time_offset);
                 } else {
                     fprintf(stderr, "Error: avcodec_send_frame failed, error: %s\n", av_error_to_string(ret));
                 }
@@ -4206,8 +4248,14 @@ int main(int argc, char **argv) {
             fflush(stdout);
             if(recording_saved_script)
                 run_recording_saved_script_async(recording_saved_script, save_replay_output_filepath.c_str(), "replay");
+
             std::lock_guard<std::mutex> lock(write_output_mutex);
             save_replay_packets.clear();
+            if(!overlap_replay) {
+                frame_data_queue.clear();
+                frames_erased = true;
+                replay_start_time = clock_get_monotonic_seconds() - paused_time_offset;
+            }
         }
 
         if(save_replay == 1 && !save_replay_thread.valid() && replay_buffer_size_secs != -1) {

src/pipewire_audio.c

@@ -371,7 +371,7 @@ void gsr_pipewire_audio_deinit(gsr_pipewire_audio *self) {
 
 static struct pw_properties* gsr_pipewire_create_null_audio_sink(const char *name) {
     char props_str[512];
-    snprintf(props_str, sizeof(props_str), "{ factory.name=support.null-audio-sink node.name=\"%s\" media.class=Audio/Sink object.linger=false audio.position=[FL FR] monitor.channel-volumes=true monitor.passthrough=true adjust_time=0 slaves=\"\" }", name);
+    snprintf(props_str, sizeof(props_str), "{ factory.name=support.null-audio-sink node.name=\"%s\" media.class=Audio/Sink object.linger=false audio.position=[FL FR] monitor.channel-volumes=true monitor.passthrough=true adjust_time=0 node.description=gsr-app-sink slaves=\"\" }", name);
     struct pw_properties *props = pw_properties_new_string(props_str);
     if(!props) {
         fprintf(stderr, "gsr error: gsr_pipewire_create_null_audio_sink: failed to create virtual sink properties\n");

src/utils.c

@@ -417,8 +417,8 @@ static bool version_greater_than(int major, int minor, int patch, int other_majo
     return (major > other_major) || (major == other_major && minor > other_minor) || (major == other_major && minor == other_minor && patch > other_patch);
 }
 
-bool gl_driver_version_greater_than(const gsr_egl *egl, int major, int minor, int patch) {
-    return version_greater_than(egl->gpu_info.driver_major, egl->gpu_info.driver_minor, egl->gpu_info.driver_patch, major, minor, patch);
+bool gl_driver_version_greater_than(const gsr_gpu_info *gpu_info, int major, int minor, int patch) {
+    return version_greater_than(gpu_info->driver_major, gpu_info->driver_minor, gpu_info->driver_patch, major, minor, patch);
 }
 
 bool try_card_has_valid_plane(const char *card_path) {