Merge commit '21f7cd4acd8dc4b4796b55966dd015cb037164d8'

* commit '21f7cd4acd8dc4b4796b55966dd015cb037164d8':
  lavfi: add a filter for uploading normal frames to CUDA

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

5 files changed, 247 insertions, 0 deletions

diff --git a/configure b/configure
index 857b68313f..e1ead49f78 100755
--- a/configure
+++ b/configure
@@ -2872,6 +2872,7 @@ fspp_filter_deps="gpl"
 geq_filter_deps="gpl"
 histeq_filter_deps="gpl"
 hqdn3d_filter_deps="gpl"
+hwupload_cuda_filter_deps="cuda"
 interlace_filter_deps="gpl"
 kerndeint_filter_deps="gpl"
 ladspa_filter_deps="ladspa dlopen"
diff --git a/doc/filters.texi b/doc/filters.texi
index 32042d013e..616d69409d 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -7940,6 +7940,18 @@ A floating point number which specifies chroma temporal strength. It defaults to
 @var{luma_tmp}*@var{chroma_spatial}/@var{luma_spatial}.
 @end table
 
+@anchor{hwupload_cuda}
+@section hwupload_cuda
+
+Upload system memory frames to a CUDA device.
+
+It accepts the following optional parameters:
+
+@table @option
+@item device
+The number of the CUDA device to use
+@end table
+
 @section hqx
 
 Apply a high-quality magnification filter designed for pixel art. This filter
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index a8469fd224..1d3d0c6061 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -175,6 +175,7 @@ OBJS-$(CONFIG_HISTEQ_FILTER)                 += vf_histeq.o
 OBJS-$(CONFIG_HISTOGRAM_FILTER)              += vf_histogram.o
 OBJS-$(CONFIG_HQDN3D_FILTER)                 += vf_hqdn3d.o
 OBJS-$(CONFIG_HQX_FILTER)                    += vf_hqx.o
+OBJS-$(CONFIG_HWUPLOAD_CUDA_FILTER)          += vf_hwupload_cuda.o
 OBJS-$(CONFIG_HSTACK_FILTER)                 += vf_stack.o framesync.o
 OBJS-$(CONFIG_HUE_FILTER)                    += vf_hue.o
 OBJS-$(CONFIG_IDET_FILTER)                   += vf_idet.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index 3163831a8b..25fe0c16f5 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -196,6 +196,7 @@ void avfilter_register_all(void)
     REGISTER_FILTER(HISTOGRAM,      histogram,      vf);
     REGISTER_FILTER(HQDN3D,         hqdn3d,         vf);
     REGISTER_FILTER(HQX,            hqx,            vf);
+    REGISTER_FILTER(HWUPLOAD_CUDA,  hwupload_cuda,  vf);
     REGISTER_FILTER(HSTACK,         hstack,         vf);
     REGISTER_FILTER(HUE,            hue,            vf);
     REGISTER_FILTER(IDET,           idet,           vf);
diff --git a/libavfilter/vf_hwupload_cuda.c b/libavfilter/vf_hwupload_cuda.c
new file mode 100644
index 0000000000..6f4176f212
--- /dev/null
+++ b/libavfilter/vf_hwupload_cuda.c
@@ -0,0 +1,232 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#include "libavutil/buffer.h"
+#include "libavutil/hwcontext.h"
+#include "libavutil/hwcontext_cuda.h"
+#include "libavutil/log.h"
+#include "libavutil/opt.h"
+
+#include "avfilter.h"
+#include "formats.h"
+#include "internal.h"
+#include "video.h"
+
+typedef struct CudaUploadContext {
+    const AVClass *class;
+    int device_idx;
+
+    AVBufferRef *hwdevice;
+    AVBufferRef *hwframe;
+} CudaUploadContext;
+
+static void cudaupload_ctx_free(AVHWDeviceContext *ctx)
+{
+    AVCUDADeviceContext *hwctx = ctx->hwctx;
+    cuCtxDestroy(hwctx->cuda_ctx);
+}
+
+static av_cold int cudaupload_init(AVFilterContext *ctx)
+{
+    CudaUploadContext *s = ctx->priv;
+
+    AVHWDeviceContext   *device_ctx;
+    AVCUDADeviceContext *device_hwctx;
+    CUdevice device;
+    CUcontext cuda_ctx = NULL, dummy;
+    CUresult err;
+    int ret;
+
+    err = cuInit(0);
+    if (err != CUDA_SUCCESS) {
+        av_log(ctx, AV_LOG_ERROR, "Could not initialize the CUDA driver API\n");
+        return AVERROR_UNKNOWN;
+    }
+
+    err = cuDeviceGet(&device, s->device_idx);
+    if (err != CUDA_SUCCESS) {
+        av_log(ctx, AV_LOG_ERROR, "Could not get the device number %d\n", s->device_idx);
+        return AVERROR_UNKNOWN;
+    }
+
+    err = cuCtxCreate(&cuda_ctx, 0, device);
+    if (err != CUDA_SUCCESS) {
+        av_log(ctx, AV_LOG_ERROR, "Error creating a CUDA context\n");
+        return AVERROR_UNKNOWN;
+    }
+
+    cuCtxPopCurrent(&dummy);
+
+    s->hwdevice = av_hwdevice_ctx_alloc(AV_HWDEVICE_TYPE_CUDA);
+    if (!s->hwdevice) {
+        cuCtxDestroy(cuda_ctx);
+        return AVERROR(ENOMEM);
+    }
+
+    device_ctx       = (AVHWDeviceContext*)s->hwdevice->data;
+    device_ctx->free = cudaupload_ctx_free;
+
+    device_hwctx = device_ctx->hwctx;
+    device_hwctx->cuda_ctx = cuda_ctx;
+
+    ret = av_hwdevice_ctx_init(s->hwdevice);
+    if (ret < 0)
+        return ret;
+
+    return 0;
+}
+
+static av_cold void cudaupload_uninit(AVFilterContext *ctx)
+{
+    CudaUploadContext *s = ctx->priv;
+
+    av_buffer_unref(&s->hwframe);
+    av_buffer_unref(&s->hwdevice);
+}
+
+static int cudaupload_query_formats(AVFilterContext *ctx)
+{
+    static const enum AVPixelFormat input_pix_fmts[] = {
+        AV_PIX_FMT_NV12, AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV444P,
+        AV_PIX_FMT_NONE,
+    };
+    static const enum AVPixelFormat output_pix_fmts[] = {
+        AV_PIX_FMT_CUDA, AV_PIX_FMT_NONE,
+    };
+    AVFilterFormats *in_fmts  = ff_make_format_list(input_pix_fmts);
+    AVFilterFormats *out_fmts = ff_make_format_list(output_pix_fmts);
+
+    ff_formats_ref(in_fmts,  &ctx->inputs[0]->out_formats);
+    ff_formats_ref(out_fmts, &ctx->outputs[0]->in_formats);
+
+    return 0;
+}
+
+static int cudaupload_config_output(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    AVFilterLink *inlink = ctx->inputs[0];
+    CudaUploadContext *s = ctx->priv;
+
+    AVHWFramesContext *hwframe_ctx;
+    int ret;
+
+    av_buffer_unref(&s->hwframe);
+    s->hwframe = av_hwframe_ctx_alloc(s->hwdevice);
+    if (!s->hwframe)
+        return AVERROR(ENOMEM);
+
+    hwframe_ctx            = (AVHWFramesContext*)s->hwframe->data;
+    hwframe_ctx->format    = AV_PIX_FMT_CUDA;
+    hwframe_ctx->sw_format = inlink->format;
+    hwframe_ctx->width     = FFALIGN(inlink->w, 16);
+    hwframe_ctx->height    = FFALIGN(inlink->h, 16);
+
+    ret = av_hwframe_ctx_init(s->hwframe);
+    if (ret < 0)
+        return ret;
+
+    outlink->hw_frames_ctx = av_buffer_ref(s->hwframe);
+    if (!outlink->hw_frames_ctx)
+        return AVERROR(ENOMEM);
+
+    return 0;
+}
+
+static int cudaupload_filter_frame(AVFilterLink *link, AVFrame *in)
+{
+    AVFilterContext   *ctx = link->dst;
+    CudaUploadContext   *s = ctx->priv;
+
+    AVFrame *out = NULL;
+    int ret;
+
+    out = av_frame_alloc();
+    if (!out) {
+        ret = AVERROR(ENOMEM);
+        goto fail;
+    }
+
+    ret = av_hwframe_get_buffer(s->hwframe, out, 0);
+    if (ret < 0)
+        goto fail;
+
+    out->width  = in->width;
+    out->height = in->height;
+
+    ret = av_hwframe_transfer_data(out, in, 0);
+    if (ret < 0) {
+        av_log(ctx, AV_LOG_ERROR, "Error transferring data to the GPU\n");
+        goto fail;
+    }
+
+    ret = av_frame_copy_props(out, in);
+    if (ret < 0)
+        goto fail;
+
+    av_frame_free(&in);
+
+    return ff_filter_frame(ctx->outputs[0], out);
+fail:
+    av_frame_free(&in);
+    av_frame_free(&out);
+    return ret;
+}
+
+#define OFFSET(x) offsetof(CudaUploadContext, x)
+#define FLAGS (AV_OPT_FLAG_FILTERING_PARAM | AV_OPT_FLAG_VIDEO_PARAM)
+static const AVOption cudaupload_options[] = {
+    { "device", "Number of the device to use", OFFSET(device_idx), AV_OPT_TYPE_INT, { .i64 = 0 }, .flags = FLAGS },
+    { NULL },
+};
+
+AVFILTER_DEFINE_CLASS(cudaupload)
+
+static const AVFilterPad cudaupload_inputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .filter_frame = cudaupload_filter_frame,
+    },
+    { NULL }
+};
+
+static const AVFilterPad cudaupload_outputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .config_props = cudaupload_config_output,
+    },
+    { NULL }
+};
+
+AVFilter ff_vf_hwupload_cuda = {
+    .name        = "hwupload_cuda",
+    .description = NULL_IF_CONFIG_SMALL("Upload a system memory frame to a CUDA device."),
+
+    .init      = cudaupload_init,
+    .uninit    = cudaupload_uninit,
+
+    .query_formats = cudaupload_query_formats,
+
+    .priv_size  = sizeof(CudaUploadContext),
+    .priv_class = &cudaupload_class,
+
+    .inputs    = cudaupload_inputs,
+    .outputs   = cudaupload_outputs,
+};