[FFmpeg-cvslog] nvenc: Explicitly push the cuda context on encoding

[Luca Barbato]

ffmpeg | branch: master | Luca Barbato <lu_zero at gentoo.org> | Fri Jan 13 10:53:35 2017 +0100| [fb59f87ce72035b940c3f5045884098b9324e1b2] | committer: Luca Barbato

nvenc: Explicitly push the cuda context on encoding

Make sure that NVENC does not misbehave if other cuda usages happen
in the application.

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=fb59f87ce72035b940c3f5045884098b9324e1b2
---

 libavcodec/nvenc.c | 7 +++++++
 libavcodec/nvenc.h | 2 ++
 2 files changed, 9 insertions(+)

diff --git a/libavcodec/nvenc.c b/libavcodec/nvenc.c
index 4054e43eec..fe7b7f40f0 100644
--- a/libavcodec/nvenc.c
+++ b/libavcodec/nvenc.c
@@ -178,6 +178,7 @@ static av_cold int nvenc_load_libraries(AVCodecContext *avctx)
     nvel->cu_device_compute_capability = cuDeviceComputeCapability;
     nvel->cu_ctx_create                = cuCtxCreate_v2;
     nvel->cu_ctx_pop_current           = cuCtxPopCurrent_v2;
+    nvel->cu_ctx_push_current          = cuCtxPushCurrent_v2;
     nvel->cu_ctx_destroy               = cuCtxDestroy_v2;
 #else
     LOAD_LIBRARY(nvel->cuda, CUDA_LIBNAME);
@@ -190,6 +191,7 @@ static av_cold int nvenc_load_libraries(AVCodecContext *avctx)
                 "cuDeviceComputeCapability");
     LOAD_SYMBOL(nvel->cu_ctx_create, nvel->cuda, "cuCtxCreate_v2");
     LOAD_SYMBOL(nvel->cu_ctx_pop_current, nvel->cuda, "cuCtxPopCurrent_v2");
+    LOAD_SYMBOL(nvel->cu_ctx_push_current, nvel->cuda, "cuCtxPushCurrent_v2");
     LOAD_SYMBOL(nvel->cu_ctx_destroy, nvel->cuda, "cuCtxDestroy_v2");
 #endif
 
@@ -1522,9 +1524,11 @@ int ff_nvenc_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
                           const AVFrame *frame, int *got_packet)
 {
     NVENCContext *ctx               = avctx->priv_data;
+    NVENCLibraryContext *nvel       = &ctx->nvel;
     NV_ENCODE_API_FUNCTION_LIST *nv = &ctx->nvel.nvenc_funcs;
     NV_ENC_PIC_PARAMS params        = { 0 };
     NVENCFrame         *nvenc_frame = NULL;
+    CUcontext dummy;
     int enc_ret, ret;
 
     params.version = NV_ENC_PIC_PARAMS_VER;
@@ -1570,7 +1574,10 @@ int ff_nvenc_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
         params.encodePicFlags = NV_ENC_PIC_FLAG_EOS;
     }
 
+    nvel->cu_ctx_push_current(ctx->cu_context);
     enc_ret = nv->nvEncEncodePicture(ctx->nvenc_ctx, &params);
+    nvel->cu_ctx_pop_current(&dummy);
+
     if (enc_ret != NV_ENC_SUCCESS &&
         enc_ret != NV_ENC_ERR_NEED_MORE_INPUT)
         return nvenc_print_error(avctx, enc_ret, "Error encoding the frame");
diff --git a/libavcodec/nvenc.h b/libavcodec/nvenc.h
index dfd03b5ebd..6df45480a9 100644
--- a/libavcodec/nvenc.h
+++ b/libavcodec/nvenc.h
@@ -66,6 +66,7 @@ typedef CUresult(CUDAAPI *PCUDEVICEGETNAME)(char *name, int len, CUdevice dev);
 typedef CUresult(CUDAAPI *PCUDEVICECOMPUTECAPABILITY)(int *major, int *minor, CUdevice dev);
 typedef CUresult(CUDAAPI *PCUCTXCREATE)(CUcontext *pctx, unsigned int flags, CUdevice dev);
 typedef CUresult(CUDAAPI *PCUCTXPOPCURRENT)(CUcontext *pctx);
+typedef CUresult(CUDAAPI *PCUCTXPUSHCURRENT)(CUcontext ctx);
 typedef CUresult(CUDAAPI *PCUCTXDESTROY)(CUcontext ctx);
 
 typedef NVENCSTATUS (NVENCAPI *PNVENCODEAPICREATEINSTANCE)(NV_ENCODE_API_FUNCTION_LIST *functionList);
@@ -84,6 +85,7 @@ typedef struct NVENCLibraryContext
     PCUDEVICECOMPUTECAPABILITY cu_device_compute_capability;
     PCUCTXCREATE cu_ctx_create;
     PCUCTXPOPCURRENT cu_ctx_pop_current;
+    PCUCTXPUSHCURRENT cu_ctx_push_current;
     PCUCTXDESTROY cu_ctx_destroy;
 
     NV_ENCODE_API_FUNCTION_LIST nvenc_funcs;