From b9fa7f2fa25ee37c87077bce468c95e48fce5c18 Mon Sep 17 00:00:00 2001 From: dec05eba Date: Fri, 5 Jul 2024 01:59:04 +0200 Subject: Separate video encoding method from capture method With this instead of kms_cuda/kms_vaapi/kms_software and xcomposite_cuda/xcomposite_vaapi/xcomposite_software there is now only kms and xcomposite. --- src/encoder/video/cuda.c | 236 +++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 236 insertions(+) create mode 100644 src/encoder/video/cuda.c (limited to 'src/encoder/video/cuda.c') diff --git a/src/encoder/video/cuda.c b/src/encoder/video/cuda.c new file mode 100644 index 0000000..2568bc7 --- /dev/null +++ b/src/encoder/video/cuda.c @@ -0,0 +1,236 @@ +#include "../../../include/encoder/video/cuda.h" +#include "../../../include/egl.h" +#include "../../../include/cuda.h" + +#include +#include + +#include + +typedef struct { + gsr_video_encoder_cuda_params params; + + unsigned int target_textures[2]; + + gsr_cuda cuda; + CUgraphicsResource cuda_graphics_resources[2]; + CUarray mapped_arrays[2]; + CUstream cuda_stream; +} gsr_video_encoder_cuda; + +static bool gsr_video_encoder_cuda_setup_context(gsr_video_encoder_cuda *self, AVCodecContext *video_codec_context) { + AVBufferRef *device_ctx = av_hwdevice_ctx_alloc(AV_HWDEVICE_TYPE_CUDA); + if(!device_ctx) { + fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_context failed: failed to create hardware device context\n"); + return false; + } + + AVHWDeviceContext *hw_device_context = (AVHWDeviceContext*)device_ctx->data; + AVCUDADeviceContext *cuda_device_context = (AVCUDADeviceContext*)hw_device_context->hwctx; + cuda_device_context->cuda_ctx = self->cuda.cu_ctx; + if(av_hwdevice_ctx_init(device_ctx) < 0) { + fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_context failed: failed to create hardware device context\n"); + av_buffer_unref(&device_ctx); + return false; + } + + AVBufferRef *frame_context = av_hwframe_ctx_alloc(device_ctx); + if(!frame_context) { + fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_context failed: failed to create hwframe context\n"); + av_buffer_unref(&device_ctx); + return false; + } + + 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.hdr ? AV_PIX_FMT_P010LE : AV_PIX_FMT_NV12; + hw_frame_context->format = video_codec_context->pix_fmt; + hw_frame_context->device_ref = device_ctx; + hw_frame_context->device_ctx = (AVHWDeviceContext*)device_ctx->data; + + if (av_hwframe_ctx_init(frame_context) < 0) { + fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_context failed: failed to initialize hardware frame context " + "(note: ffmpeg version needs to be > 4.0)\n"); + av_buffer_unref(&device_ctx); + //av_buffer_unref(&frame_context); + return false; + } + + self->cuda_stream = cuda_device_context->stream; + video_codec_context->hw_device_ctx = av_buffer_ref(device_ctx); + video_codec_context->hw_frames_ctx = av_buffer_ref(frame_context); + return true; +} + +static unsigned int gl_create_texture(gsr_egl *egl, int width, int height, int internal_format, unsigned int format) { + unsigned int texture_id = 0; + egl->glGenTextures(1, &texture_id); + egl->glBindTexture(GL_TEXTURE_2D, texture_id); + egl->glTexImage2D(GL_TEXTURE_2D, 0, internal_format, width, height, 0, format, GL_UNSIGNED_BYTE, NULL); + + egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE); + egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE); + egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); + egl->glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); + + egl->glBindTexture(GL_TEXTURE_2D, 0); + return texture_id; +} + +static bool cuda_register_opengl_texture(gsr_cuda *cuda, CUgraphicsResource *cuda_graphics_resource, CUarray *mapped_array, unsigned int texture_id) { + CUresult res; + res = cuda->cuGraphicsGLRegisterImage(cuda_graphics_resource, texture_id, GL_TEXTURE_2D, CU_GRAPHICS_REGISTER_FLAGS_NONE); + if (res != CUDA_SUCCESS) { + const char *err_str = "unknown"; + cuda->cuGetErrorString(res, &err_str); + fprintf(stderr, "gsr error: cuda_register_opengl_texture: cuGraphicsGLRegisterImage failed, error: %s, texture " "id: %u\n", err_str, texture_id); + return false; + } + + res = cuda->cuGraphicsResourceSetMapFlags(*cuda_graphics_resource, CU_GRAPHICS_MAP_RESOURCE_FLAGS_NONE); + res = cuda->cuGraphicsMapResources(1, cuda_graphics_resource, 0); + + res = cuda->cuGraphicsSubResourceGetMappedArray(mapped_array, *cuda_graphics_resource, 0, 0); + return true; +} + +static bool gsr_video_encoder_cuda_setup_textures(gsr_video_encoder_cuda *self, AVCodecContext *video_codec_context, AVFrame *frame) { + const int res = av_hwframe_get_buffer(video_codec_context->hw_frames_ctx, frame, 0); + if(res < 0) { + fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_textures: av_hwframe_get_buffer failed: %d\n", res); + return false; + } + + const unsigned int internal_formats_nv12[2] = { GL_R8, GL_RG8 }; + const unsigned int internal_formats_p010[2] = { GL_R16, GL_RG16 }; + const unsigned int formats[2] = { GL_RED, GL_RG }; + const int div[2] = {1, 2}; // divide UV texture size by 2 because chroma is half size + + for(int i = 0; i < 2; ++i) { + self->target_textures[i] = gl_create_texture(self->params.egl, video_codec_context->width / div[i], video_codec_context->height / div[i], !self->params.hdr ? internal_formats_nv12[i] : internal_formats_p010[i], formats[i]); + if(self->target_textures[i] == 0) { + fprintf(stderr, "gsr error: gsr_video_encoder_cuda_setup_textures: failed to create opengl texture\n"); + return false; + } + + if(!cuda_register_opengl_texture(&self->cuda, &self->cuda_graphics_resources[i], &self->mapped_arrays[i], self->target_textures[i])) { + return false; + } + } + + return true; +} + +static void gsr_video_encoder_cuda_stop(gsr_video_encoder_cuda *self, AVCodecContext *video_codec_context); + +static bool gsr_video_encoder_cuda_start(gsr_video_encoder *encoder, AVCodecContext *video_codec_context, AVFrame *frame) { + gsr_video_encoder_cuda *encoder_cuda = encoder->priv; + + // TODO: Force set overclock to false if wayland + if(!gsr_cuda_load(&encoder_cuda->cuda, encoder_cuda->params.egl->x11.dpy, encoder_cuda->params.overclock)) { + fprintf(stderr, "gsr error: gsr_video_encoder_cuda_start: failed to load cuda\n"); + gsr_video_encoder_cuda_stop(encoder_cuda, video_codec_context); + return false; + } + + if(!gsr_video_encoder_cuda_setup_context(encoder_cuda, video_codec_context)) { + gsr_video_encoder_cuda_stop(encoder_cuda, video_codec_context); + return false; + } + + if(!gsr_video_encoder_cuda_setup_textures(encoder_cuda, video_codec_context, frame)) { + gsr_video_encoder_cuda_stop(encoder_cuda, video_codec_context); + return false; + } + + return true; +} + +void gsr_video_encoder_cuda_stop(gsr_video_encoder_cuda *self, AVCodecContext *video_codec_context) { + self->params.egl->glDeleteTextures(2, self->target_textures); + self->target_textures[0] = 0; + self->target_textures[1] = 0; + + if(video_codec_context->hw_device_ctx) + av_buffer_unref(&video_codec_context->hw_device_ctx); + if(video_codec_context->hw_frames_ctx) + av_buffer_unref(&video_codec_context->hw_frames_ctx); + + if(self->cuda.cu_ctx) { + for(int i = 0; i < 2; ++i) { + if(self->cuda_graphics_resources[i]) { + self->cuda.cuGraphicsUnmapResources(1, &self->cuda_graphics_resources[i], 0); + self->cuda.cuGraphicsUnregisterResource(self->cuda_graphics_resources[i]); + self->cuda_graphics_resources[i] = 0; + } + } + } + + gsr_cuda_unload(&self->cuda); +} + +static void gsr_video_encoder_cuda_copy_textures_to_frame(gsr_video_encoder *encoder, AVFrame *frame) { + gsr_video_encoder_cuda *encoder_cuda = encoder->priv; + const int div[2] = {1, 2}; // divide UV texture size by 2 because chroma is half size + for(int i = 0; i < 2; ++i) { + CUDA_MEMCPY2D memcpy_struct; + memcpy_struct.srcXInBytes = 0; + memcpy_struct.srcY = 0; + memcpy_struct.srcMemoryType = CU_MEMORYTYPE_ARRAY; + + memcpy_struct.dstXInBytes = 0; + memcpy_struct.dstY = 0; + memcpy_struct.dstMemoryType = CU_MEMORYTYPE_DEVICE; + + memcpy_struct.srcArray = encoder_cuda->mapped_arrays[i]; + memcpy_struct.srcPitch = frame->width / div[i]; + memcpy_struct.dstDevice = (CUdeviceptr)frame->data[i]; + memcpy_struct.dstPitch = frame->linesize[i]; + memcpy_struct.WidthInBytes = frame->width * (encoder_cuda->params.hdr ? 2 : 1); + memcpy_struct.Height = frame->height / div[i]; + // TODO: Remove this copy if possible + encoder_cuda->cuda.cuMemcpy2DAsync_v2(&memcpy_struct, encoder_cuda->cuda_stream); + } + + // TODO: needed? + encoder_cuda->cuda.cuStreamSynchronize(encoder_cuda->cuda_stream); +} + +static void gsr_video_encoder_cuda_get_textures(gsr_video_encoder *encoder, unsigned int *textures, int *num_textures, gsr_destination_color *destination_color) { + gsr_video_encoder_cuda *encoder_cuda = encoder->priv; + textures[0] = encoder_cuda->target_textures[0]; + textures[1] = encoder_cuda->target_textures[1]; + *num_textures = 2; + *destination_color = encoder_cuda->params.hdr ? GSR_DESTINATION_COLOR_P010 : GSR_DESTINATION_COLOR_NV12; +} + +static void gsr_video_encoder_cuda_destroy(gsr_video_encoder *encoder, AVCodecContext *video_codec_context) { + gsr_video_encoder_cuda_stop(encoder->priv, video_codec_context); + free(encoder->priv); + free(encoder); +} + +gsr_video_encoder* gsr_video_encoder_cuda_create(const gsr_video_encoder_cuda_params *params) { + gsr_video_encoder *encoder = calloc(1, sizeof(gsr_video_encoder)); + if(!encoder) + return NULL; + + gsr_video_encoder_cuda *encoder_cuda = calloc(1, sizeof(gsr_video_encoder_cuda)); + if(!encoder_cuda) { + free(encoder); + return NULL; + } + + encoder_cuda->params = *params; + + *encoder = (gsr_video_encoder) { + .start = gsr_video_encoder_cuda_start, + .copy_textures_to_frame = gsr_video_encoder_cuda_copy_textures_to_frame, + .get_textures = gsr_video_encoder_cuda_get_textures, + .destroy = gsr_video_encoder_cuda_destroy, + .priv = encoder_cuda + }; + + return encoder; +} -- cgit v1.2.3