Allow linking to CUDA dynamically instead of dlopen()ing it at runtime

configure

@@ -179,6 +179,7 @@ Individual component options:
 External library support:
   --enable-avisynth        enable reading of AviSynth script files [no]
   --enable-bzlib           enable bzlib [autodetect]
+  --enable-cuda            enable dynamically linked CUDA [no]
   --enable-frei0r          enable frei0r video filtering
   --enable-gnutls          enable gnutls [no]
   --enable-libbs2b         enable bs2b DSP library [no]
@@ -1238,6 +1239,7 @@ EXAMPLE_LIST="
 EXTERNAL_LIBRARY_LIST="
     avisynth
     bzlib
+    cuda
     frei0r
     gnutls
     libbs2b
@@ -4043,6 +4045,7 @@ die_license_disabled gpl libxavs
 die_license_disabled gpl libxvid
 die_license_disabled gpl x11grab
 
+die_license_disabled nonfree cuda
 die_license_disabled nonfree libfaac
 die_license_disabled nonfree libfdk_aac
 die_license_disabled nonfree nvenc
@@ -4523,6 +4526,7 @@ done
 enabled avisynth          && { check_lib2 "avisynth/avisynth_c.h windows.h" LoadLibrary ||
                                check_lib2 "avxsynth/avxsynth_c.h dlfcn.h" dlopen -ldl   ||
                                die "ERROR: LoadLibrary/dlopen not found, or avisynth header not found"; }
+enabled cuda              && check_lib cuda.h cuInit -lcuda
 enabled frei0r            && { check_header frei0r.h || die "ERROR: frei0r.h header not found"; }
 enabled gnutls            && require_pkg_config gnutls gnutls/gnutls.h gnutls_global_init
 enabled libbs2b           && require_pkg_config libbs2b bs2b.h bs2b_open
@@ -4617,7 +4621,7 @@ if enabled libdc1394; then
 fi
 
 if enabled nvenc; then
-    check_header cuda.h || die "ERROR: cuda.h not found.";
+    enabled cuda || check_header cuda.h || die "ERROR: cuda.h not found.";
     check_header nvEncodeAPI.h || die "ERROR: nvEncodeAPI.h not found.";
     check_cpp_condition nvEncodeAPI.h "NVENCAPI_MAJOR_VERSION >= 5" ||
         die "ERROR: NVENC API version 4 or older is not supported";

libavcodec/nvenc.c

@@ -148,6 +148,16 @@ static av_cold int nvenc_load_libraries(AVCodecContext *avctx)
     PNVENCODEAPICREATEINSTANCE nvenc_create_instance;
     NVENCSTATUS err;
 
+#if CONFIG_CUDA
+    nvel->cu_init                      = cuInit;
+    nvel->cu_device_get_count          = cuDeviceGetCount;
+    nvel->cu_device_get                = cuDeviceGet;
+    nvel->cu_device_get_name           = cuDeviceGetName;
+    nvel->cu_device_compute_capability = cuDeviceComputeCapability;
+    nvel->cu_ctx_create                = cuCtxCreate_v2;
+    nvel->cu_ctx_pop_current           = cuCtxPopCurrent_v2;
+    nvel->cu_ctx_destroy               = cuCtxDestroy_v2;
+#else
     LOAD_LIBRARY(nvel->cuda, CUDA_LIBNAME);
 
     LOAD_SYMBOL(nvel->cu_init, nvel->cuda, "cuInit");
@@ -159,6 +169,7 @@ static av_cold int nvenc_load_libraries(AVCodecContext *avctx)
     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_destroy, nvel->cuda, "cuCtxDestroy_v2");
+#endif
 
     LOAD_LIBRARY(nvel->nvenc, NVENC_LIBNAME);
 
@@ -859,8 +870,10 @@ av_cold int ff_nvenc_encode_close(AVCodecContext *avctx)
     if (ctx->nvel.nvenc)
         dlclose(ctx->nvel.nvenc);
 
+#if !CONFIG_CUDA
     if (ctx->nvel.cuda)
         dlclose(ctx->nvel.cuda);
+#endif
 
     return 0;
 }

libavcodec/nvenc.h

@@ -22,6 +22,8 @@
 #include <cuda.h>
 #include <nvEncodeAPI.h>
 
+#include "config.h"
+
 #include "libavutil/fifo.h"
 #include "libavutil/opt.h"
 
@@ -47,7 +49,9 @@ typedef NVENCSTATUS (NVENCAPI *PNVENCODEAPICREATEINSTANCE)(NV_ENCODE_API_FUNCTIO
 
 typedef struct NVENCLibraryContext
 {
+#if !CONFIG_CUDA
     void *cuda;
+#endif
     void *nvenc;
 
     PCUINIT cu_init;