From 07915b12273a36bc7f104a5f3fc949a407d243dc Mon Sep 17 00:00:00 2001 From: Philip Langdale Date: Sun, 15 Apr 2018 09:06:34 -0700 Subject: vo_gpu: hwdec: Use ffnvcodec to load CUDA symbols The CUDA dynamic loader was broken out of ffmpeg into its own repo and package. This gives us an opportunity to re-use it in mpv and remove our custom loader logic. --- video/out/opengl/hwdec_cuda.c | 78 +++++++++++++++++++++++++------------------ 1 file changed, 45 insertions(+), 33 deletions(-) (limited to 'video/out/opengl/hwdec_cuda.c') diff --git a/video/out/opengl/hwdec_cuda.c b/video/out/opengl/hwdec_cuda.c index 1a7df2020a..f80c14500a 100644 --- a/video/out/opengl/hwdec_cuda.c +++ b/video/out/opengl/hwdec_cuda.c @@ -27,8 +27,7 @@ * when decoding 10bit streams (there is some hardware dithering going on). */ -#include "cuda_dynamic.h" - +#include #include #include @@ -39,6 +38,7 @@ struct priv_owner { struct mp_hwdec_ctx hwctx; + CudaFunctions *cu; CUcontext display_ctx; CUcontext decode_ctx; }; @@ -56,13 +56,15 @@ static int check_cu(struct ra_hwdec *hw, CUresult err, const char *func) const char *err_name; const char *err_string; + struct priv_owner *p = hw->priv; + MP_TRACE(hw, "Calling %s\n", func); if (err == CUDA_SUCCESS) return 0; - cuGetErrorName(err, &err_name); - cuGetErrorString(err, &err_string); + p->cu->cuGetErrorName(err, &err_name); + p->cu->cuGetErrorString(err, &err_string); MP_ERR(hw, "%s failed", func); if (err_name && err_string) @@ -82,6 +84,7 @@ static int cuda_init(struct ra_hwdec *hw) unsigned int device_count; int ret = 0; struct priv_owner *p = hw->priv; + CudaFunctions *cu; if (!ra_is_gl(hw->ra)) return -1; @@ -92,24 +95,25 @@ static int cuda_init(struct ra_hwdec *hw) return -1; } - bool loaded = cuda_load(); - if (!loaded) { + ret = cuda_load_functions(&p->cu, NULL); + if (ret != 0) { MP_VERBOSE(hw, "Failed to load CUDA symbols\n"); return -1; } + cu = p->cu; - ret = CHECK_CU(cuInit(0)); + ret = CHECK_CU(cu->cuInit(0)); if (ret < 0) goto error; // Allocate display context - ret = CHECK_CU(cuGLGetDevices(&device_count, &display_dev, 1, - CU_GL_DEVICE_LIST_ALL)); + ret = CHECK_CU(cu->cuGLGetDevices(&device_count, &display_dev, 1, + CU_GL_DEVICE_LIST_ALL)); if (ret < 0) goto error; - ret = CHECK_CU(cuCtxCreate(&p->display_ctx, CU_CTX_SCHED_BLOCKING_SYNC, - display_dev)); + ret = CHECK_CU(cu->cuCtxCreate(&p->display_ctx, CU_CTX_SCHED_BLOCKING_SYNC, + display_dev)); if (ret < 0) goto error; @@ -121,7 +125,7 @@ static int cuda_init(struct ra_hwdec *hw) if (decode_dev_idx > -1) { CUdevice decode_dev; - ret = CHECK_CU(cuDeviceGet(&decode_dev, decode_dev_idx)); + ret = CHECK_CU(cu->cuDeviceGet(&decode_dev, decode_dev_idx)); if (ret < 0) goto error; @@ -129,12 +133,12 @@ static int cuda_init(struct ra_hwdec *hw) MP_INFO(hw, "Using separate decoder and display devices\n"); // Pop the display context. We won't use it again during init() - ret = CHECK_CU(cuCtxPopCurrent(&dummy)); + ret = CHECK_CU(cu->cuCtxPopCurrent(&dummy)); if (ret < 0) goto error; - ret = CHECK_CU(cuCtxCreate(&p->decode_ctx, CU_CTX_SCHED_BLOCKING_SYNC, - decode_dev)); + ret = CHECK_CU(cu->cuCtxCreate(&p->decode_ctx, CU_CTX_SCHED_BLOCKING_SYNC, + decode_dev)); if (ret < 0) goto error; } @@ -155,7 +159,7 @@ static int cuda_init(struct ra_hwdec *hw) goto error; } - ret = CHECK_CU(cuCtxPopCurrent(&dummy)); + ret = CHECK_CU(cu->cuCtxPopCurrent(&dummy)); if (ret < 0) goto error; @@ -168,7 +172,7 @@ static int cuda_init(struct ra_hwdec *hw) error: av_buffer_unref(&hw_device_ctx); - CHECK_CU(cuCtxPopCurrent(&dummy)); + CHECK_CU(cu->cuCtxPopCurrent(&dummy)); return -1; } @@ -176,15 +180,18 @@ static int cuda_init(struct ra_hwdec *hw) static void cuda_uninit(struct ra_hwdec *hw) { struct priv_owner *p = hw->priv; + CudaFunctions *cu = p->cu; hwdec_devices_remove(hw->devs, &p->hwctx); av_buffer_unref(&p->hwctx.av_device_ref); if (p->decode_ctx && p->decode_ctx != p->display_ctx) - CHECK_CU(cuCtxDestroy(p->decode_ctx)); + CHECK_CU(cu->cuCtxDestroy(p->decode_ctx)); if (p->display_ctx) - CHECK_CU(cuCtxDestroy(p->display_ctx)); + CHECK_CU(cu->cuCtxDestroy(p->display_ctx)); + + cuda_free_functions(&p->cu); } #undef CHECK_CU @@ -195,6 +202,7 @@ static int mapper_init(struct ra_hwdec_mapper *mapper) struct priv_owner *p_owner = mapper->owner->priv; struct priv *p = mapper->priv; CUcontext dummy; + CudaFunctions *cu = p_owner->cu; int ret = 0, eret = 0; p->display_ctx = p_owner->display_ctx; @@ -212,7 +220,7 @@ static int mapper_init(struct ra_hwdec_mapper *mapper) return -1; } - ret = CHECK_CU(cuCtxPushCurrent(p->display_ctx)); + ret = CHECK_CU(cu->cuCtxPushCurrent(p->display_ctx)); if (ret < 0) return ret; @@ -239,27 +247,27 @@ static int mapper_init(struct ra_hwdec_mapper *mapper) GLenum target; ra_gl_get_raw_tex(mapper->ra, mapper->tex[n], &texture, &target); - ret = CHECK_CU(cuGraphicsGLRegisterImage(&p->cu_res[n], texture, target, - CU_GRAPHICS_REGISTER_FLAGS_WRITE_DISCARD)); + ret = CHECK_CU(cu->cuGraphicsGLRegisterImage(&p->cu_res[n], texture, target, + CU_GRAPHICS_REGISTER_FLAGS_WRITE_DISCARD)); if (ret < 0) goto error; - ret = CHECK_CU(cuGraphicsMapResources(1, &p->cu_res[n], 0)); + ret = CHECK_CU(cu->cuGraphicsMapResources(1, &p->cu_res[n], 0)); if (ret < 0) goto error; - ret = CHECK_CU(cuGraphicsSubResourceGetMappedArray(&p->cu_array[n], p->cu_res[n], - 0, 0)); + ret = CHECK_CU(cu->cuGraphicsSubResourceGetMappedArray(&p->cu_array[n], p->cu_res[n], + 0, 0)); if (ret < 0) goto error; - ret = CHECK_CU(cuGraphicsUnmapResources(1, &p->cu_res[n], 0)); + ret = CHECK_CU(cu->cuGraphicsUnmapResources(1, &p->cu_res[n], 0)); if (ret < 0) goto error; } error: - eret = CHECK_CU(cuCtxPopCurrent(&dummy)); + eret = CHECK_CU(cu->cuCtxPopCurrent(&dummy)); if (eret < 0) return eret; @@ -269,17 +277,19 @@ static int mapper_init(struct ra_hwdec_mapper *mapper) static void mapper_uninit(struct ra_hwdec_mapper *mapper) { struct priv *p = mapper->priv; + struct priv_owner *p_owner = mapper->owner->priv; + CudaFunctions *cu = p_owner->cu; CUcontext dummy; // Don't bail if any CUDA calls fail. This is all best effort. - CHECK_CU(cuCtxPushCurrent(p->display_ctx)); + CHECK_CU(cu->cuCtxPushCurrent(p->display_ctx)); for (int n = 0; n < 4; n++) { if (p->cu_res[n] > 0) - CHECK_CU(cuGraphicsUnregisterResource(p->cu_res[n])); + CHECK_CU(cu->cuGraphicsUnregisterResource(p->cu_res[n])); p->cu_res[n] = 0; ra_tex_free(mapper->ra, &mapper->tex[n]); } - CHECK_CU(cuCtxPopCurrent(&dummy)); + CHECK_CU(cu->cuCtxPopCurrent(&dummy)); } static void mapper_unmap(struct ra_hwdec_mapper *mapper) @@ -289,10 +299,12 @@ static void mapper_unmap(struct ra_hwdec_mapper *mapper) static int mapper_map(struct ra_hwdec_mapper *mapper) { struct priv *p = mapper->priv; + struct priv_owner *p_owner = mapper->owner->priv; + CudaFunctions *cu = p_owner->cu; CUcontext dummy; int ret = 0, eret = 0; - ret = CHECK_CU(cuCtxPushCurrent(p->display_ctx)); + ret = CHECK_CU(cu->cuCtxPushCurrent(p->display_ctx)); if (ret < 0) return ret; @@ -308,14 +320,14 @@ static int mapper_map(struct ra_hwdec_mapper *mapper) mapper->tex[n]->params.format->pixel_size, .Height = mp_image_plane_h(&p->layout, n), }; - ret = CHECK_CU(cuMemcpy2D(&cpy)); + ret = CHECK_CU(cu->cuMemcpy2D(&cpy)); if (ret < 0) goto error; } error: - eret = CHECK_CU(cuCtxPopCurrent(&dummy)); + eret = CHECK_CU(cu->cuCtxPopCurrent(&dummy)); if (eret < 0) return eret; -- cgit v1.2.3