Improve capture performance on amd/intel when using monitor capture: capture display frame using purely vaapi unless the monitor is rotated

7 files changed, 248 insertions, 21 deletions

diff --git a/include/egl.h b/include/egl.h
index 61890b4..3fdbf48 100644
--- a/include/egl.h
+++ b/include/egl.h
@@ -31,6 +31,7 @@ typedef void* EGLSurface;
 typedef void* EGLContext;
 typedef void* EGLClientBuffer;
 typedef void* EGLImage;
+typedef void* EGLImageKHR;
 typedef void *GLeglImageOES;
 typedef void (*__eglMustCastToProperFunctionPointerType)(void);
 typedef struct __GLXFBConfigRec *GLXFBConfig;
@@ -138,6 +139,8 @@ typedef void(*__GLXextFuncPtr)(void);
 #define GL_COMPILE_STATUS                       0x8B81
 #define GL_LINK_STATUS                          0x8B82
 
+typedef unsigned int (*FUNC_eglExportDMABUFImageQueryMESA)(EGLDisplay dpy, EGLImageKHR image, int *fourcc, int *num_planes, uint64_t *modifiers);
+typedef unsigned int (*FUNC_eglExportDMABUFImageMESA)(EGLDisplay dpy, EGLImageKHR image, int *fds, int32_t *strides, int32_t *offsets);
 typedef void (*FUNC_glEGLImageTargetTexture2DOES)(unsigned int target, GLeglImageOES image);
 typedef GLXContext (*FUNC_glXCreateContextAttribsARB)(Display *dpy, GLXFBConfig config, GLXContext share_context, Bool direct, const int *attrib_list);
 typedef void (*FUNC_glXSwapIntervalEXT)(Display * dpy, GLXDrawable drawable, int interval);
@@ -235,6 +238,8 @@ struct gsr_egl {
     unsigned int (*eglBindAPI)(unsigned int api);
     __eglMustCastToProperFunctionPointerType (*eglGetProcAddress)(const char *procname);
 
+    FUNC_eglExportDMABUFImageQueryMESA eglExportDMABUFImageQueryMESA;
+    FUNC_eglExportDMABUFImageMESA eglExportDMABUFImageMESA;
     FUNC_glEGLImageTargetTexture2DOES glEGLImageTargetTexture2DOES;
     FUNC_eglQueryDisplayAttribEXT eglQueryDisplayAttribEXT;
     FUNC_eglQueryDeviceStringEXT eglQueryDeviceStringEXT;
diff --git a/include/utils.h b/include/utils.h
index 99503e2..6d58db7 100644
--- a/include/utils.h
+++ b/include/utils.h
@@ -7,6 +7,9 @@
 #include <stdbool.h>
 #include <stdint.h>
 
+typedef struct AVCodecContext AVCodecContext;
+typedef struct AVFrame AVFrame;
+
 typedef struct {
     const char *name;
     int name_len;
@@ -44,5 +47,7 @@ int create_directory_recursive(char *path);
 
 /* |img_attr| needs to be at least 44 in size */
 void setup_dma_buf_attrs(intptr_t *img_attr, uint32_t format, uint32_t width, uint32_t height, const int *fds, const uint32_t *offsets, const uint32_t *pitches, const uint64_t *modifiers, int num_planes, bool use_modifier);
+bool video_codec_context_is_vaapi(AVCodecContext *video_codec_context);
+bool vaapi_copy_drm_planes_to_video_surface(AVCodecContext *video_codec_context, AVFrame *video_frame, int x, int y, uint32_t format, uint32_t width, uint32_t height, const int *fds, const uint32_t *offsets, const uint32_t *pitches, const uint64_t *modifiers, int num_planes);
 
 #endif /* GSR_UTILS_H */
diff --git a/src/capture/kms.c b/src/capture/kms.c
index 9287a8b..c36add4 100644
--- a/src/capture/kms.c
+++ b/src/capture/kms.c
@@ -49,6 +49,9 @@ typedef struct {
 
     bool is_x11;
     gsr_cursor x11_cursor;
+
+    AVCodecContext *video_codec_context;
+    bool performance_error_shown;
 } gsr_capture_kms;
 
 static void gsr_capture_kms_cleanup_kms_fds(gsr_capture_kms *self) {
@@ -194,6 +197,8 @@ static int gsr_capture_kms_start(gsr_capture *cap, AVCodecContext *video_codec_c
 
     frame->width = video_codec_context->width;
     frame->height = video_codec_context->height;
+
+    self->video_codec_context = video_codec_context;
     return 0;
 }
 
@@ -495,12 +500,6 @@ static int gsr_capture_kms_capture(gsr_capture *cap, AVFrame *frame, gsr_color_c
     if(drm_fd->has_hdr_metadata && self->params.hdr && hdr_metadata_is_supported_format(&drm_fd->hdr_metadata))
         gsr_kms_set_hdr_metadata(self, drm_fd);
 
-    EGLImage image = gsr_capture_kms_create_egl_image_with_fallback(self, drm_fd);
-    if(image) {
-        gsr_capture_kms_bind_image_to_input_texture_with_fallback(self, image);
-        self->params.egl->eglDestroyImage(self->params.egl->egl_display, image);
-    }
-
     const float texture_rotation = monitor_rotation_to_radians(self->monitor_rotation);
     const int target_x = max_int(0, frame->width / 2 - self->capture_size.x / 2);
     const int target_y = max_int(0, frame->height / 2 - self->capture_size.y / 2);
@@ -509,13 +508,40 @@ static int gsr_capture_kms_capture(gsr_capture *cap, AVFrame *frame, gsr_color_c
     if(!capture_is_combined_plane)
         capture_pos = (vec2i){drm_fd->x, drm_fd->y};
 
-    self->params.egl->glFlush();
-    self->params.egl->glFinish();
+    if(!self->performance_error_shown && self->monitor_rotation != GSR_MONITOR_ROT_0 && video_codec_context_is_vaapi(self->video_codec_context)) {
+        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");
+    }
 
-    gsr_color_conversion_draw(color_conversion, self->external_texture_fallback ? self->external_input_texture_id : self->input_texture_id,
-        (vec2i){target_x, target_y}, self->capture_size,
-        capture_pos, self->capture_size,
-        texture_rotation, self->external_texture_fallback);
+    /* Fast opengl free path */
+    if(self->monitor_rotation == GSR_MONITOR_ROT_0 && video_codec_context_is_vaapi(self->video_codec_context)) {
+        int fds[4];
+        uint32_t offsets[4];
+        uint32_t pitches[4];
+        uint64_t modifiers[4];
+        for(int i = 0; i < drm_fd->num_dma_bufs; ++i) {
+            fds[i] = drm_fd->dma_buf[i].fd;
+            offsets[i] = drm_fd->dma_buf[i].offset;
+            pitches[i] = drm_fd->dma_buf[i].pitch;
+            modifiers[i] = drm_fd->modifier;
+        }
+        vaapi_copy_drm_planes_to_video_surface(self->video_codec_context, frame, capture_pos.x, capture_pos.y, drm_fd->pixel_format, drm_fd->width, drm_fd->height, fds, offsets, pitches, modifiers, drm_fd->num_dma_bufs);
+    } else {
+        EGLImage image = gsr_capture_kms_create_egl_image_with_fallback(self, drm_fd);
+        if(image) {
+            gsr_capture_kms_bind_image_to_input_texture_with_fallback(self, image);
+            self->params.egl->eglDestroyImage(self->params.egl->egl_display, image);
+        }
+
+        self->params.egl->glFlush();
+        self->params.egl->glFinish();
+
+        gsr_color_conversion_draw(color_conversion, self->external_texture_fallback ? self->external_input_texture_id : self->input_texture_id,
+            (vec2i){target_x, target_y}, self->capture_size,
+            capture_pos, self->capture_size,
+            texture_rotation, self->external_texture_fallback);
+    }
 
     if(self->params.record_cursor) {
         gsr_kms_response_item *cursor_drm_fd = find_cursor_drm_if_on_monitor(self, drm_fd->connector_id, capture_is_combined_plane);
diff --git a/src/egl.c b/src/egl.c
index b4e3902..05d6680 100644
--- a/src/egl.c
+++ b/src/egl.c
@@ -388,11 +388,23 @@ static bool gsr_egl_load_egl(gsr_egl *self, void *library) {
 }
 
 static bool gsr_egl_proc_load_egl(gsr_egl *self) {
+    self->eglExportDMABUFImageQueryMESA = (FUNC_eglExportDMABUFImageQueryMESA)self->eglGetProcAddress("eglExportDMABUFImageQueryMESA");
+    self->eglExportDMABUFImageMESA = (FUNC_eglExportDMABUFImageMESA)self->eglGetProcAddress("eglExportDMABUFImageMESA");
     self->glEGLImageTargetTexture2DOES = (FUNC_glEGLImageTargetTexture2DOES)self->eglGetProcAddress("glEGLImageTargetTexture2DOES");
     self->eglQueryDisplayAttribEXT = (FUNC_eglQueryDisplayAttribEXT)self->eglGetProcAddress("eglQueryDisplayAttribEXT");
     self->eglQueryDeviceStringEXT = (FUNC_eglQueryDeviceStringEXT)self->eglGetProcAddress("eglQueryDeviceStringEXT");
     self->eglQueryDmaBufModifiersEXT = (FUNC_eglQueryDmaBufModifiersEXT)self->eglGetProcAddress("eglQueryDmaBufModifiersEXT");
 
+    if(!self->eglExportDMABUFImageQueryMESA) {
+        fprintf(stderr, "gsr error: gsr_egl_load failed: could not find eglExportDMABUFImageQueryMESA\n");
+        return false;
+    }
+
+    if(!self->eglExportDMABUFImageMESA) {
+        fprintf(stderr, "gsr error: gsr_egl_load failed: could not find eglExportDMABUFImageMESA\n");
+        return false;
+    }
+
     if(!self->glEGLImageTargetTexture2DOES) {
         fprintf(stderr, "gsr error: gsr_egl_load failed: could not find glEGLImageTargetTexture2DOES\n");
         return false;
diff --git a/src/encoder/video/vaapi.c b/src/encoder/video/vaapi.c
index 7a2abfc..03218cb 100644
--- a/src/encoder/video/vaapi.c
+++ b/src/encoder/video/vaapi.c
@@ -41,8 +41,7 @@ static bool gsr_video_encoder_vaapi_setup_context(gsr_video_encoder_vaapi *self,
         return false;
     }
 
-    AVHWFramesContext *hw_frame_context =
-        (AVHWFramesContext *)frame_context->data;
+    AVHWFramesContext *hw_frame_context = (AVHWFramesContext*)frame_context->data;
     hw_frame_context->width = video_codec_context->width;
     hw_frame_context->height = video_codec_context->height;
     hw_frame_context->sw_format = self->params.color_depth == GSR_COLOR_DEPTH_10_BITS ? AV_PIX_FMT_P010LE : AV_PIX_FMT_NV12;
@@ -51,7 +50,7 @@ static bool gsr_video_encoder_vaapi_setup_context(gsr_video_encoder_vaapi *self,
 
     //hw_frame_context->initial_pool_size = 20;
 
-    AVVAAPIDeviceContext *vactx =((AVHWDeviceContext*)self->device_ctx->data)->hwctx;
+    AVVAAPIDeviceContext *vactx = ((AVHWDeviceContext*)self->device_ctx->data)->hwctx;
     self->va_dpy = vactx->display;
 
     if (av_hwframe_ctx_init(frame_context) < 0) {
diff --git a/src/main.cpp b/src/main.cpp
index 7687b07..f2526f8 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -519,13 +519,13 @@ static AVCodecContext *create_video_codec_context(AVPixelFormat pix_fmt,
         if(codec_context->codec_id == AV_CODEC_ID_AV1 || codec_context->codec_id == AV_CODEC_ID_H264 || codec_context->codec_id == AV_CODEC_ID_HEVC) {
             switch(video_quality) {
                 case VideoQuality::MEDIUM:
-                    codec_context->global_quality = 160 * quality_multiply;
+                    codec_context->global_quality = 150 * quality_multiply;
                     break;
                 case VideoQuality::HIGH:
-                    codec_context->global_quality = 130 * quality_multiply;
+                    codec_context->global_quality = 120 * quality_multiply;
                     break;
                 case VideoQuality::VERY_HIGH:
-                    codec_context->global_quality = 110 * quality_multiply;
+                    codec_context->global_quality = 100 * quality_multiply;
                     break;
                 case VideoQuality::ULTRA:
                     codec_context->global_quality = 90 * quality_multiply;
@@ -2753,6 +2753,11 @@ int main(int argc, char **argv) {
         }
     }
 
+    if(wayland && is_monitor_capture) {
+        fprintf(stderr, "gsr warning: it's not possible to sync video to recorded monitor exactly on wayland when recording a monitor."
+            " If you experience stutter in the video then record with portal capture option instead (-w portal) or use X11 instead\n");
+    }
+
     // TODO: Fix constant framerate not working properly on amd/intel because capture framerate gets locked to the same framerate as
     // game framerate, which doesn't work well when you need to encode multiple duplicate frames (AMD/Intel is slow at encoding!).
     // It also appears to skip audio frames on nvidia wayland? why? that should be fine, but it causes video stuttering because of audio/video sync.
@@ -3442,8 +3447,8 @@ int main(int argc, char **argv) {
         const double frame_sleep_fps = 1.0 / update_fps;
         const double sleep_time = frame_sleep_fps - (frame_end - frame_start);
         if(sleep_time > 0.0) {
-           if(damaged)
-               av_usleep(sleep_time * 1000.0 * 1000.0);
+            if(damaged)
+                av_usleep(sleep_time * 1000.0 * 1000.0);
             else
                 av_usleep(2 * 1000.0); // 2 milliseconds
         }
diff --git a/src/utils.c b/src/utils.c
index 3e4138a..2b1b07c 100644
--- a/src/utils.c
+++ b/src/utils.c
@@ -12,9 +12,11 @@
 
 #include <xf86drmMode.h>
 #include <xf86drm.h>
-
 #include <X11/Xatom.h>
 #include <X11/extensions/Xrandr.h>
+#include <va/va_drmcommon.h>
+#include <libavcodec/avcodec.h>
+#include <libavutil/hwcontext_vaapi.h>
 
 double clock_get_monotonic_seconds(void) {
     struct timespec ts;
@@ -624,3 +626,176 @@ void setup_dma_buf_attrs(intptr_t *img_attr, uint32_t format, uint32_t width, ui
     img_attr[img_attr_index++] = EGL_NONE;
     assert(img_attr_index <= 44);
 }
+
+static VADisplay video_codec_context_get_vaapi_display(AVCodecContext *video_codec_context) {
+    AVHWFramesContext *hw_frame_context = (AVHWFramesContext*)video_codec_context->hw_frames_ctx->data;
+    AVHWDeviceContext *device_context = (AVHWDeviceContext*)hw_frame_context->device_ctx;
+    if(device_context->type == AV_HWDEVICE_TYPE_VAAPI) {
+        AVVAAPIDeviceContext *vactx = device_context->hwctx;
+        return vactx->display;
+    }
+    return NULL;
+}
+
+bool video_codec_context_is_vaapi(AVCodecContext *video_codec_context) {
+    AVHWFramesContext *hw_frame_context = (AVHWFramesContext*)video_codec_context->hw_frames_ctx->data;
+    AVHWDeviceContext *device_context = (AVHWDeviceContext*)hw_frame_context->device_ctx;
+    return device_context->type == AV_HWDEVICE_TYPE_VAAPI;
+}
+
+bool vaapi_copy_drm_planes_to_video_surface(AVCodecContext *video_codec_context, AVFrame *video_frame, int x, int y, uint32_t format, uint32_t width, uint32_t height, const int *fds, const uint32_t *offsets, const uint32_t *pitches, const uint64_t *modifiers, int num_planes) {
+    VAConfigID config_id = 0;
+    VAContextID context_id = 0;
+    VASurfaceID input_surface = 0;
+    VABufferID buffer_id = 0;
+    bool success = true;
+
+    VADisplay va_dpy = video_codec_context_get_vaapi_display(video_codec_context);
+    if(!va_dpy) {
+        success = false;
+        goto done;
+    }
+
+    VAStatus va_status = vaCreateConfig(va_dpy, VAProfileNone, VAEntrypointVideoProc, NULL, 0, &config_id);
+    if(va_status != VA_STATUS_SUCCESS) {
+        fprintf(stderr, "gsr error: vaapi_copy_drm_planes_to_video_surface: vaCreateConfig failed, error: %d\n", va_status);
+        success = false;
+        goto done;
+    }
+
+    VASurfaceID target_surface_id = (uintptr_t)video_frame->data[3];
+    va_status = vaCreateContext(va_dpy, config_id, width, height, VA_PROGRESSIVE, &target_surface_id, 1, &context_id);
+    if(va_status != VA_STATUS_SUCCESS) {
+        fprintf(stderr, "gsr error: vaapi_copy_drm_planes_to_video_surface: vaCreateContext failed, error: %d\n", va_status);
+        success = false;
+        goto done;
+    }
+
+    VADRMPRIMESurfaceDescriptor buf = {0};
+    buf.fourcc = format;//VA_FOURCC_BGRX; // TODO: VA_FOURCC_BGRA, VA_FOURCC_X2R10G10B10
+    buf.width = width;
+    buf.height = height;
+    buf.num_objects = num_planes;
+    buf.num_layers = 1;
+    buf.layers[0].drm_format = format;
+    buf.layers[0].num_planes = buf.num_objects;
+    for(int i = 0; i < num_planes; ++i) {
+        buf.objects[i].fd = fds[i];
+        buf.objects[i].size = height * pitches[i]; // TODO:
+        buf.objects[i].drm_format_modifier = modifiers[i];
+
+        buf.layers[0].object_index[i] = i;
+        buf.layers[0].offset[i] = offsets[i];
+        buf.layers[0].pitch[i] = pitches[i];
+    }
+
+    VASurfaceAttrib attribs[2] = {0};
+    attribs[0].type = VASurfaceAttribMemoryType;
+    attribs[0].flags = VA_SURFACE_ATTRIB_SETTABLE;
+    attribs[0].value.type = VAGenericValueTypeInteger;
+    attribs[0].value.value.i = VA_SURFACE_ATTRIB_MEM_TYPE_DRM_PRIME_2;
+    attribs[1].type = VASurfaceAttribExternalBufferDescriptor;
+    attribs[1].flags = VA_SURFACE_ATTRIB_SETTABLE;
+    attribs[1].value.type = VAGenericValueTypePointer;
+    attribs[1].value.value.p = &buf;
+    
+    // TODO: RT_FORMAT with 10 bit/hdr, VA_RT_FORMAT_RGB32_10
+    va_status = vaCreateSurfaces(va_dpy, VA_RT_FORMAT_RGB32, width, height, &input_surface, 1, attribs, 2);
+    if(va_status != VA_STATUS_SUCCESS) {
+        fprintf(stderr, "gsr error: vaapi_copy_drm_planes_to_video_surface: vaCreateSurfaces failed, error: %d\n", va_status);
+        success = false;
+        goto done;
+    }
+
+    // TODO:
+    const VARectangle output_region = {
+        .x = x,
+        .y = y,
+        .width = width,
+        .height = height
+    };
+
+    // Copying a surface to another surface will automatically perform the color conversion. Thanks vaapi!
+    VAProcPipelineParameterBuffer params = {0};
+    params.surface = input_surface;
+    params.surface_region = NULL;
+    params.output_region = &output_region;
+    params.output_background_color = 0;
+    params.filter_flags = VA_FRAME_PICTURE;
+    params.pipeline_flags = VA_PROC_PIPELINE_FAST;
+
+    params.input_color_properties.colour_primaries = 1;
+    params.input_color_properties.transfer_characteristics = 1;
+    params.input_color_properties.matrix_coefficients = 1;
+    params.surface_color_standard = VAProcColorStandardBT709; // TODO:
+    params.input_color_properties.color_range = video_frame->color_range == AVCOL_RANGE_JPEG ? VA_SOURCE_RANGE_FULL : VA_SOURCE_RANGE_REDUCED;
+
+    params.output_color_properties.colour_primaries = 1;
+    params.output_color_properties.transfer_characteristics = 1;
+    params.output_color_properties.matrix_coefficients = 1;
+    params.output_color_standard = VAProcColorStandardBT709; // TODO:
+    params.output_color_properties.color_range = video_frame->color_range == AVCOL_RANGE_JPEG ? VA_SOURCE_RANGE_FULL : VA_SOURCE_RANGE_REDUCED;
+
+    params.processing_mode = VAProcPerformanceMode;
+
+    // VAProcPipelineCaps pipeline_caps = {0};
+    // va_status = vaQueryVideoProcPipelineCaps(self->va_dpy,
+    //                                    self->context_id,
+    //                                    NULL, 0,
+    //                                    &pipeline_caps);
+    // if(va_status == VA_STATUS_SUCCESS) {
+    //     fprintf(stderr, "pipeline_caps: %u, %u\n", (unsigned int)pipeline_caps.rotation_flags, pipeline_caps.blend_flags);
+    // }
+
+    // TODO: params.output_hdr_metadata
+
+    // TODO:
+    // if (first surface to render)
+    //     pipeline_param->output_background_color = 0xff000000; // black
+
+    va_status = vaCreateBuffer(va_dpy, context_id, VAProcPipelineParameterBufferType, sizeof(params), 1, &params, &buffer_id);
+    if(va_status != VA_STATUS_SUCCESS) {
+        fprintf(stderr, "gsr error: vaapi_copy_drm_planes_to_video_surface: vaCreateBuffer failed, error: %d\n", va_status);
+        success = false;
+        goto done;
+    }
+
+    va_status = vaBeginPicture(va_dpy, context_id, target_surface_id);
+    if(va_status != VA_STATUS_SUCCESS) {
+        fprintf(stderr, "gsr error: vaapi_copy_drm_planes_to_video_surface: vaBeginPicture failed, error: %d\n", va_status);
+        success = false;
+        goto done;
+    }
+
+    va_status = vaRenderPicture(va_dpy, context_id, &buffer_id, 1);
+    if(va_status != VA_STATUS_SUCCESS) {
+        vaEndPicture(va_dpy, context_id);
+        fprintf(stderr, "gsr error: vaapi_copy_drm_planes_to_video_surface: vaRenderPicture failed, error: %d\n", va_status);
+        success = false;
+        goto done;
+    }
+
+    va_status = vaEndPicture(va_dpy, context_id);
+    if(va_status != VA_STATUS_SUCCESS) {
+        fprintf(stderr, "gsr error: vaapi_copy_drm_planes_to_video_surface: vaEndPicture failed, error: %d\n", va_status);
+        success = false;
+        goto done;
+    }
+
+    //vaSyncBuffer(self->va_dpy, self->buffer_id, 1000 * 1000);
+
+    done:
+    if(buffer_id)
+        vaDestroyBuffer(va_dpy, buffer_id);
+
+    if(input_surface)
+        vaDestroySurfaces(va_dpy, &input_surface, 1);
+
+    if(context_id)
+        vaDestroyContext(va_dpy, context_id);
+
+    if(config_id)
+        vaDestroyConfig(va_dpy, config_id);
+
+    return success;
+}