From 6b745769b1fd15ba03edad3fe381abb745bf9907 Mon Sep 17 00:00:00 2001
From: wm4 <wm4@nowhere>
Date: Sat, 28 Oct 2017 19:59:08 +0200
Subject: vd_lavc: add support for nvdec hwaccel

See manpage additions.

(In ffmpeg-mpv and Libav, this is still called "cuvid". Libav won't work
yet, because it has no frame params support yet, but this could get
fixed soon.)
---
 DOCS/man/options.rst          |  8 ++++++++
 options/options.c             |  2 ++
 video/decode/vd_lavc.c        | 15 +++++++++++++++
 video/hwdec.h                 |  2 ++
 video/out/gpu/hwdec.c         |  2 ++
 video/out/opengl/hwdec_cuda.c | 18 +++++++++++++++++-
 6 files changed, 46 insertions(+), 1 deletion(-)

diff --git a/DOCS/man/options.rst b/DOCS/man/options.rst
index 5c4882883a..7cd399d4f1 100644
--- a/DOCS/man/options.rst
+++ b/DOCS/man/options.rst
@@ -695,6 +695,8 @@ Video
     :rpi-copy:  copies video back to system RAM (Raspberry Pi only)
     :cuda:      requires ``--vo=gpu`` (Any platform CUDA is available)
     :cuda-copy: copies video back to system RAM (Any platform CUDA is available)
+    :nvdec:     requires ``--vo=gpu`` (Any platform CUDA is available)
+    :nvdec-copy: copies video back to system RAM (Any platform CUDA is available)
     :crystalhd: copies video back to system RAM (Any platform supported by hardware)
     :rkmpp:     requires ``--vo=gpu`` (some RockChip devices only)
 
@@ -723,6 +725,12 @@ Video
     deinterlacing. ``cuda`` should always be preferred unless the ``gpu``
     vo is not being used or filters are required.
 
+    ``nvdec`` is a newer implementation of CUVID/CUDA decoding, which uses the
+    FFmpeg decoders for file parsing. Experimental, is known not to correctly
+    check whether decoding is supported by the hardware at all. Deinterlacing
+    is not supported. Since this uses FFmpeg's codec parsers, it is expected
+    that this generally causes fewer issues than ``cuda``. Requires ffmpeg-mpv.
+
     Most video filters will not work with hardware decoding as they are
     primarily implemented on the CPU. Some exceptions are ``vdpaupp``,
     ``vdpaurb`` and ``vavpp``. See `VIDEO FILTERS`_ for more details.
diff --git a/options/options.c b/options/options.c
index 8e940be347..bac9b3bcaf 100644
--- a/options/options.c
+++ b/options/options.c
@@ -117,6 +117,8 @@ const struct m_opt_choice_alternatives mp_hwdec_names[] = {
     {"mediacodec-copy",HWDEC_MEDIACODEC_COPY},
     {"cuda",        HWDEC_CUDA},
     {"cuda-copy",   HWDEC_CUDA_COPY},
+    {"nvdec",       HWDEC_NVDEC},
+    {"nvdec-copy",  HWDEC_NVDEC_COPY},
     {"crystalhd",   HWDEC_CRYSTALHD},
     {0}
 };
diff --git a/video/decode/vd_lavc.c b/video/decode/vd_lavc.c
index c07c1590fe..6f09908090 100644
--- a/video/decode/vd_lavc.c
+++ b/video/decode/vd_lavc.c
@@ -158,6 +158,19 @@ static const struct vd_lavc_hwdec mp_vd_lavc_rkmpp = {
 };
 
 #if HAVE_CUDA_HWACCEL
+static const struct vd_lavc_hwdec mp_vd_lavc_nvdec = {
+    .type = HWDEC_NVDEC,
+    .image_format = IMGFMT_CUDA,
+    .generic_hwaccel = true,
+    .set_hwframes = true,
+};
+static const struct vd_lavc_hwdec mp_vd_lavc_nvdec_copy = {
+    .type = HWDEC_NVDEC_COPY,
+    .image_format = IMGFMT_CUDA,
+    .generic_hwaccel = true,
+    .set_hwframes = true,
+    .copying = true,
+};
 static const struct vd_lavc_hwdec mp_vd_lavc_cuda = {
     .type = HWDEC_CUDA,
     .image_format = IMGFMT_CUDA,
@@ -272,6 +285,8 @@ static const struct vd_lavc_hwdec *const hwdec_list[] = {
     &mp_vd_lavc_mediacodec_copy,
 #endif
 #if HAVE_CUDA_HWACCEL
+    &mp_vd_lavc_nvdec,
+    &mp_vd_lavc_nvdec_copy,
     &mp_vd_lavc_cuda,
     &mp_vd_lavc_cuda_copy,
 #endif
diff --git a/video/hwdec.h b/video/hwdec.h
index 81ba4794b2..2b89c3247c 100644
--- a/video/hwdec.h
+++ b/video/hwdec.h
@@ -26,6 +26,8 @@ enum hwdec_type {
     HWDEC_MEDIACODEC_COPY,
     HWDEC_CUDA,
     HWDEC_CUDA_COPY,
+    HWDEC_NVDEC,
+    HWDEC_NVDEC_COPY,
     HWDEC_CRYSTALHD,
     HWDEC_RKMPP,
 };
diff --git a/video/out/gpu/hwdec.c b/video/out/gpu/hwdec.c
index 9eefe3e287..d88dc5e779 100644
--- a/video/out/gpu/hwdec.c
+++ b/video/out/gpu/hwdec.c
@@ -35,6 +35,7 @@ extern const struct ra_hwdec_driver ra_hwdec_d3d11eglrgb;
 extern const struct ra_hwdec_driver ra_hwdec_dxva2gldx;
 extern const struct ra_hwdec_driver ra_hwdec_dxva2;
 extern const struct ra_hwdec_driver ra_hwdec_cuda;
+extern const struct ra_hwdec_driver ra_hwdec_cuda_nvdec;
 extern const struct ra_hwdec_driver ra_hwdec_rpi_overlay;
 #if HAVE_DRMPRIME && HAVE_DRM
 extern const struct ra_hwdec_driver ra_hwdec_drmprime_drm;
@@ -65,6 +66,7 @@ static const struct ra_hwdec_driver *const mpgl_hwdec_drivers[] = {
 #endif
 #if HAVE_CUDA_HWACCEL
     &ra_hwdec_cuda,
+    &ra_hwdec_cuda_nvdec,
 #endif
 #if HAVE_RPI
     &ra_hwdec_rpi_overlay,
diff --git a/video/out/opengl/hwdec_cuda.c b/video/out/opengl/hwdec_cuda.c
index d9c4c199f1..5aefed106d 100644
--- a/video/out/opengl/hwdec_cuda.c
+++ b/video/out/opengl/hwdec_cuda.c
@@ -160,7 +160,7 @@ static int cuda_init(struct ra_hwdec *hw)
         goto error;
 
     p->hwctx = (struct mp_hwdec_ctx) {
-        .type = HWDEC_CUDA,
+        .type = hw->driver->api,
         .ctx = p->decode_ctx,
         .av_device_ref = hw_device_ctx,
     };
@@ -340,3 +340,19 @@ const struct ra_hwdec_driver ra_hwdec_cuda = {
         .unmap = mapper_unmap,
     },
 };
+
+const struct ra_hwdec_driver ra_hwdec_cuda_nvdec = {
+    .name = "cuda-nvdec",
+    .api = HWDEC_NVDEC,
+    .imgfmts = {IMGFMT_CUDA, 0},
+    .priv_size = sizeof(struct priv_owner),
+    .init = cuda_init,
+    .uninit = cuda_uninit,
+    .mapper = &(const struct ra_hwdec_mapper_driver){
+        .priv_size = sizeof(struct priv),
+        .init = mapper_init,
+        .uninit = mapper_uninit,
+        .map = mapper_map,
+        .unmap = mapper_unmap,
+    },
+};
-- 
cgit v1.2.3