gstreamer/subprojects/gst-plugins-bad/sys/nvcodec/gstnvh264encoder.cpp

1918 lines
62 KiB
C++
Raw Normal View History

nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
/* GStreamer
* Copyright (C) 2022 Seungha Yang <seungha@centricular.com>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library 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
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
* Boston, MA 02110-1301, USA.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include "gstnvh264encoder.h"
#include <gst/codecparsers/gsth264parser.h>
#include <gst/pbutils/codec-utils.h>
#include <string>
#include <set>
#include <string.h>
GST_DEBUG_CATEGORY_STATIC (gst_nv_h264_encoder_debug);
#define GST_CAT_DEFAULT gst_nv_h264_encoder_debug
static GTypeClass *parent_class = NULL;
typedef struct
{
gint max_bframes;
gint ratecontrol_modes;
gint field_encoding;
gint monochrome;
gint fmo;
gint qpelmv;
gint bdirect_mode;
gint cabac;
gint adaptive_transform;
gint stereo_mvc;
gint temoral_layers;
gint hierarchical_pframes;
gint hierarchical_bframes;
gint level_max;
gint level_min;
gint separate_colour_plane;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
gint width_max;
gint height_max;
gint temporal_svc;
gint dyn_res_change;
gint dyn_bitrate_change;
gint dyn_force_constqp;
gint dyn_rcmode_change;
gint subframe_readback;
gint constrained_encoding;
gint intra_refresh;
gint custom_vbv_buf_size;
gint dynamic_slice_mode;
gint ref_pic_invalidation;
gint preproc_support;
gint async_encoding_support;
gint mb_num_max;
gint mb_per_sec_max;
gint yuv444_encode;
gint lossless_encode;
gint meonly_mode;
gint lookahead;
gint temporal_aq;
gint supports_10bit_encode;
gint num_max_ltr_frames;
gint weighted_prediction;
gint bframe_ref_mode;
gint emphasis_level_map;
gint width_min;
gint height_min;
gint multiple_ref_frames;
} GstNvH264EncoderDeviceCaps;
typedef struct
{
GstCaps *sink_caps;
GstCaps *src_caps;
guint cuda_device_id;
gint64 adapter_luid;
gboolean d3d11_mode;
GstNvH264EncoderDeviceCaps dev_caps;
} GstNvH264EncoderClassData;
enum
{
PROP_0,
PROP_ADAPTER_LUID,
PROP_CUDA_DEVICE_ID,
/* init params */
PROP_PRESET,
PROP_WEIGHTED_PRED,
/* encoding config */
PROP_GOP_SIZE,
PROP_B_FRAMES,
/* rate-control params */
PROP_RATE_CONTROL,
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
PROP_QP_I,
PROP_QP_P,
PROP_QP_B,
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
PROP_BITRATE,
PROP_MAX_BITRATE,
PROP_VBV_BUFFER_SIZE,
PROP_RC_LOOKAHEAD,
PROP_I_ADAPT,
PROP_B_ADAPT,
PROP_SPATIAL_AQ,
PROP_TEMPORAL_AQ,
PROP_ZERO_REORDER_DELAY,
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
PROP_NON_REF_P,
PROP_STRICT_GOP,
PROP_AQ_STRENGTH,
PROP_MIN_QP_I,
PROP_MIN_QP_P,
PROP_MIN_QP_B,
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
PROP_MAX_QP_I,
PROP_MAX_QP_P,
PROP_MAX_QP_B,
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
PROP_CONST_QUALITY,
/* h264 specific */
PROP_AUD,
PROP_CABAC,
PROP_REPEAT_SEQUENCE_HEADER,
};
#define DEFAULT_PRESET GST_NV_ENCODER_PRESET_DEFAULT
#define DEFAULT_WEIGHTED_PRED FALSE
#define DEFAULT_GOP_SIZE 30
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
#define DEFAULT_B_FRAMES 0
#define DEFAULT_RATE_CONTROL GST_NV_ENCODER_RC_MODE_VBR
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
#define DEFAULT_QP -1
#define DEFAULT_BITRATE 0
#define DEFAULT_MAX_BITRATE 0
#define DEFAULT_VBV_BUFFER_SIZE 0
#define DEFAULT_RC_LOOKAHEAD 0
#define DEFAULT_I_ADAPT FALSE
#define DEFAULT_B_ADAPT FALSE
#define DEFAULT_SPATIAL_AQ FALSE
#define DEFAULT_TEMPORAL_AQ FALSE
#define DEFAULT_ZERO_REORDER_DELAY FALSE
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
#define DEFAULT_NON_REF_P FALSE
#define DEFAULT_STRICT_GOP FALSE
#define DEFAULT_AQ_STRENGTH FALSE
#define DEFAULT_CONST_QUALITY 0
#define DEFAULT_AUD TRUE
#define DEFAULT_REPEAT_SEQUENCE_HEADER FALSE
typedef struct _GstNvH264Encoder
{
GstNvEncoder parent;
GMutex prop_lock;
gboolean init_param_updated;
gboolean rc_param_updated;
gboolean bitrate_updated;
gboolean packetized;
GstH264NalParser *parser;
/* Properties */
GstNvEncoderPreset preset;
gboolean weighted_pred;
gint gop_size;
guint bframes;
GstNvEncoderRCMode rc_mode;
gint qp_i;
gint qp_p;
gint qp_b;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
guint bitrate;
guint max_bitrate;
guint vbv_buffer_size;
guint rc_lookahead;
gboolean i_adapt;
gboolean b_adapt;
gboolean spatial_aq;
gboolean temporal_aq;
gboolean zero_reorder_delay;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
gboolean non_ref_p;
gboolean strict_gop;
guint aq_strength;
gint min_qp_i;
gint min_qp_p;
gint min_qp_b;
gint max_qp_i;
gint max_qp_p;
gint max_qp_b;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
gdouble const_quality;
gboolean aud;
gboolean cabac;
gboolean repeat_sequence_header;
} GstNvH264Encoder;
typedef struct _GstNvH264EncoderClass
{
GstNvEncoderClass parent_class;
GstNvH264EncoderDeviceCaps dev_caps;
guint cuda_device_id;
gint64 adapter_luid;
gboolean d3d11_mode;
} GstNvH264EncoderClass;
#define GST_NV_H264_ENCODER(object) ((GstNvH264Encoder *) (object))
#define GST_NV_H264_ENCODER_GET_CLASS(object) \
(G_TYPE_INSTANCE_GET_CLASS ((object),G_TYPE_FROM_INSTANCE (object),GstNvH264EncoderClass))
static void gst_nv_h264_encoder_finalize (GObject * object);
static void gst_nv_h264_encoder_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec);
static void gst_nv_h264_encoder_get_property (GObject * object, guint prop_id,
GValue * value, GParamSpec * pspec);
static GstCaps *gst_nv_h264_encoder_getcaps (GstVideoEncoder * encoder,
GstCaps * filter);
static gboolean gst_nv_h264_encoder_set_format (GstNvEncoder * encoder,
GstVideoCodecState * state, gpointer session,
NV_ENC_INITIALIZE_PARAMS * init_params, NV_ENC_CONFIG * config);
static gboolean gst_nv_h264_encoder_set_output_state (GstNvEncoder * encoder,
GstVideoCodecState * state, gpointer session);
static GstBuffer *gst_nv_h264_encoder_create_output_buffer (GstNvEncoder *
encoder, NV_ENC_LOCK_BITSTREAM * bitstream);
static GstNvEncoderReconfigure
gst_nv_h264_encoder_check_reconfigure (GstNvEncoder * encoder,
NV_ENC_CONFIG * config);
static void
gst_nv_h264_encoder_class_init (GstNvH264EncoderClass * klass, gpointer data)
{
GObjectClass *object_class = G_OBJECT_CLASS (klass);
GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
GstVideoEncoderClass *videoenc_class = GST_VIDEO_ENCODER_CLASS (klass);
GstNvEncoderClass *nvenc_class = GST_NV_ENCODER_CLASS (klass);
GstNvH264EncoderClassData *cdata = (GstNvH264EncoderClassData *) data;
GstNvH264EncoderDeviceCaps *dev_caps = &cdata->dev_caps;
GParamFlags param_flags = (GParamFlags) (G_PARAM_READWRITE |
GST_PARAM_MUTABLE_PLAYING | G_PARAM_STATIC_STRINGS);
GParamFlags conditional_param_flags = (GParamFlags) (G_PARAM_READWRITE |
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
GST_PARAM_CONDITIONALLY_AVAILABLE | GST_PARAM_MUTABLE_PLAYING |
G_PARAM_STATIC_STRINGS);
parent_class = (GTypeClass *) g_type_class_peek_parent (klass);
object_class->finalize = gst_nv_h264_encoder_finalize;
object_class->set_property = gst_nv_h264_encoder_set_property;
object_class->get_property = gst_nv_h264_encoder_get_property;
if (cdata->d3d11_mode) {
g_object_class_install_property (object_class, PROP_ADAPTER_LUID,
g_param_spec_int64 ("adapter-luid", "Adapter LUID",
"DXGI Adapter LUID (Locally Unique Identifier) of associated GPU",
G_MININT64, G_MAXINT64, cdata->adapter_luid,
(GParamFlags) (G_PARAM_READABLE | G_PARAM_STATIC_STRINGS)));
} else {
g_object_class_install_property (object_class, PROP_CUDA_DEVICE_ID,
g_param_spec_uint ("cuda-device-id", "CUDA Device ID",
"CUDA device ID of associated GPU",
0, G_MAXINT, cdata->cuda_device_id,
(GParamFlags) (G_PARAM_READABLE | G_PARAM_STATIC_STRINGS)));
}
g_object_class_install_property (object_class, PROP_PRESET,
g_param_spec_enum ("preset", "Encoding Preset",
"Encoding Preset", GST_TYPE_NV_ENCODER_PRESET,
DEFAULT_PRESET, param_flags));
if (dev_caps->weighted_prediction) {
g_object_class_install_property (object_class, PROP_WEIGHTED_PRED,
g_param_spec_boolean ("weighted-pred", "Weighted Pred",
"Enables Weighted Prediction", DEFAULT_WEIGHTED_PRED,
conditional_param_flags));
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
}
g_object_class_install_property (object_class, PROP_GOP_SIZE,
g_param_spec_int ("gop-size", "GOP size",
"Number of frames between intra frames (-1 = infinite)",
-1, G_MAXINT, DEFAULT_GOP_SIZE, param_flags));
if (dev_caps->max_bframes > 0) {
g_object_class_install_property (object_class, PROP_B_FRAMES,
g_param_spec_uint ("b-frames", "B-Frames",
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
"Number of B-frames between I and P", 0, dev_caps->max_bframes,
DEFAULT_B_FRAMES, conditional_param_flags));
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
}
g_object_class_install_property (object_class, PROP_RATE_CONTROL,
g_param_spec_enum ("rate-control", "Rate Control", "Rate Control Method",
GST_TYPE_NV_ENCODER_RC_MODE, DEFAULT_RATE_CONTROL, param_flags));
g_object_class_install_property (object_class, PROP_QP_I,
g_param_spec_int ("qp-i", "QP I",
"Constant QP value for I frame (-1 = default)", -1, 51,
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
DEFAULT_QP, param_flags));
g_object_class_install_property (object_class, PROP_QP_P,
g_param_spec_int ("qp-p", "QP P",
"Constant QP value for P frame (-1 = default)", -1, 51,
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
DEFAULT_QP, param_flags));
g_object_class_install_property (object_class, PROP_QP_B,
g_param_spec_int ("qp-b", "QP B",
"Constant QP value for B frame (-1 = default)", -1, 51,
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
DEFAULT_QP, param_flags));
g_object_class_install_property (object_class, PROP_BITRATE,
g_param_spec_uint ("bitrate", "Bitrate",
"Bitrate in kbit/sec (0 = automatic)", 0, 2000 * 1024,
DEFAULT_BITRATE, param_flags));
g_object_class_install_property (object_class, PROP_MAX_BITRATE,
g_param_spec_uint ("max-bitrate", "Max Bitrate",
"Maximum Bitrate in kbit/sec (ignored in CBR mode)", 0, 2000 * 1024,
DEFAULT_MAX_BITRATE, param_flags));
if (dev_caps->custom_vbv_buf_size) {
g_object_class_install_property (object_class,
PROP_VBV_BUFFER_SIZE,
g_param_spec_uint ("vbv-buffer-size", "VBV Buffer Size",
"VBV(HRD) Buffer Size in kbits (0 = NVENC default)",
0, G_MAXUINT, DEFAULT_VBV_BUFFER_SIZE, conditional_param_flags));
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
}
if (dev_caps->lookahead) {
g_object_class_install_property (object_class, PROP_RC_LOOKAHEAD,
g_param_spec_uint ("rc-lookahead", "Rate Control Lookahead",
"Number of frames for frame type lookahead",
0, 32, DEFAULT_RC_LOOKAHEAD, conditional_param_flags));
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
g_object_class_install_property (object_class, PROP_I_ADAPT,
g_param_spec_boolean ("i-adapt", "I Adapt",
"Enable adaptive I-frame insert when lookahead is enabled",
DEFAULT_I_ADAPT, conditional_param_flags));
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
if (dev_caps->max_bframes > 0) {
g_object_class_install_property (object_class, PROP_B_ADAPT,
g_param_spec_boolean ("b-adapt", "B Adapt",
"Enable adaptive B-frame insert when lookahead is enabled",
DEFAULT_B_ADAPT, conditional_param_flags));
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
}
}
g_object_class_install_property (object_class, PROP_SPATIAL_AQ,
g_param_spec_boolean ("spatial-aq", "Spatial AQ",
"Spatial Adaptive Quantization", DEFAULT_SPATIAL_AQ, param_flags));
if (dev_caps->temporal_aq) {
g_object_class_install_property (object_class, PROP_TEMPORAL_AQ,
g_param_spec_boolean ("temporal-aq", "Temporal AQ",
"Temporal Adaptive Quantization", DEFAULT_TEMPORAL_AQ,
conditional_param_flags));
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
}
g_object_class_install_property (object_class, PROP_ZERO_REORDER_DELAY,
g_param_spec_boolean ("zero-reorder-delay", "Zero Reorder Delay",
"Zero latency operation (i.e., num_reorder_frames = 0)",
DEFAULT_ZERO_REORDER_DELAY, param_flags));
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
g_object_class_install_property (object_class, PROP_NON_REF_P,
g_param_spec_boolean ("nonref-p", "Nonref P",
"Automatic insertion of non-reference P-frames", DEFAULT_NON_REF_P,
param_flags));
g_object_class_install_property (object_class, PROP_STRICT_GOP,
g_param_spec_boolean ("strict-gop", "Strict GOP",
"Minimize GOP-to-GOP rate fluctuations", DEFAULT_STRICT_GOP,
param_flags));
g_object_class_install_property (object_class, PROP_AQ_STRENGTH,
g_param_spec_uint ("aq-strength", "AQ Strength",
"Adaptive Quantization Strength when spatial-aq is enabled"
" from 1 (low) to 15 (aggressive), (0 = autoselect)",
0, 15, DEFAULT_AQ_STRENGTH, param_flags));
g_object_class_install_property (object_class, PROP_MIN_QP_I,
g_param_spec_int ("min-qp-i", "Min QP I",
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
"Minimum QP value for I frame, (-1 = disabled)", -1, 51,
DEFAULT_QP, param_flags));
g_object_class_install_property (object_class, PROP_MIN_QP_P,
g_param_spec_int ("min-qp-p", "Min QP P",
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
"Minimum QP value for P frame, (-1 = automatic)", -1, 51,
DEFAULT_QP, param_flags));
g_object_class_install_property (object_class, PROP_MIN_QP_B,
g_param_spec_int ("min-qp-b", "Min QP B",
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
"Minimum QP value for B frame, (-1 = automatic)", -1, 51,
DEFAULT_QP, param_flags));
g_object_class_install_property (object_class, PROP_MAX_QP_I,
g_param_spec_int ("max-qp-i", "Max QP I",
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
"Maximum QP value for I frame, (-1 = disabled)", -1, 51,
DEFAULT_QP, param_flags));
g_object_class_install_property (object_class, PROP_MAX_QP_P,
g_param_spec_int ("max-qp-p", "Max QP P",
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
"Maximum QP value for P frame, (-1 = automatic)", -1, 51,
DEFAULT_QP, param_flags));
g_object_class_install_property (object_class, PROP_MAX_QP_B,
g_param_spec_int ("max-qp-b", "Max QP B",
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
"Maximum QP value for B frame, (-1 = automatic)", -1, 51,
DEFAULT_QP, param_flags));
g_object_class_install_property (object_class, PROP_CONST_QUALITY,
g_param_spec_double ("const-quality", "Constant Quality",
"Target Constant Quality level for VBR mode (0 = automatic)",
0, 51, DEFAULT_CONST_QUALITY, param_flags));
g_object_class_install_property (object_class, PROP_AUD,
g_param_spec_boolean ("aud", "AUD",
"Use AU (Access Unit) delimiter", DEFAULT_AUD, param_flags));
if (dev_caps->cabac) {
g_object_class_install_property (object_class, PROP_CABAC,
g_param_spec_boolean ("cabac", "CABAC",
"Enable CABAC entropy coding", TRUE, conditional_param_flags));
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
}
g_object_class_install_property (object_class, PROP_REPEAT_SEQUENCE_HEADER,
g_param_spec_boolean ("repeat-sequence-header", "Repeat Sequence Header",
"Insert sequence headers (SPS/PPS) per IDR",
DEFAULT_REPEAT_SEQUENCE_HEADER, param_flags));
if (cdata->d3d11_mode) {
gst_element_class_set_metadata (element_class,
"NVENC H.264 Video Encoder Direct3D11 Mode",
"Codec/Encoder/Video/Hardware",
"Encode H.264 video streams using NVCODEC API Direct3D11 Mode",
"Seungha Yang <seungha@centricular.com>");
} else {
gst_element_class_set_metadata (element_class,
"NVENC H.264 Video Encoder CUDA Mode",
"Codec/Encoder/Video/Hardware",
"Encode H.264 video streams using NVCODEC API CUDA Mode",
"Seungha Yang <seungha@centricular.com>");
}
gst_element_class_add_pad_template (element_class,
gst_pad_template_new ("sink", GST_PAD_SINK, GST_PAD_ALWAYS,
cdata->sink_caps));
gst_element_class_add_pad_template (element_class,
gst_pad_template_new ("src", GST_PAD_SRC, GST_PAD_ALWAYS,
cdata->src_caps));
videoenc_class->getcaps = GST_DEBUG_FUNCPTR (gst_nv_h264_encoder_getcaps);
nvenc_class->set_format = GST_DEBUG_FUNCPTR (gst_nv_h264_encoder_set_format);
nvenc_class->set_output_state =
GST_DEBUG_FUNCPTR (gst_nv_h264_encoder_set_output_state);
nvenc_class->create_output_buffer =
GST_DEBUG_FUNCPTR (gst_nv_h264_encoder_create_output_buffer);
nvenc_class->check_reconfigure =
GST_DEBUG_FUNCPTR (gst_nv_h264_encoder_check_reconfigure);
klass->dev_caps = cdata->dev_caps;
klass->cuda_device_id = cdata->cuda_device_id;
klass->adapter_luid = cdata->adapter_luid;
klass->d3d11_mode = cdata->d3d11_mode;
gst_caps_unref (cdata->sink_caps);
gst_caps_unref (cdata->src_caps);
g_free (cdata);
}
static void
gst_nv_h264_encoder_init (GstNvH264Encoder * self)
{
GstNvH264EncoderClass *klass = GST_NV_H264_ENCODER_GET_CLASS (self);
g_mutex_init (&self->prop_lock);
self->preset = DEFAULT_PRESET;
self->weighted_pred = DEFAULT_WEIGHTED_PRED;
self->gop_size = DEFAULT_GOP_SIZE;
self->bframes = DEFAULT_B_FRAMES;
self->rc_mode = DEFAULT_RATE_CONTROL;
self->qp_i = DEFAULT_QP;
self->qp_p = DEFAULT_QP;
self->qp_b = DEFAULT_QP;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
self->bitrate = DEFAULT_BITRATE;
self->max_bitrate = DEFAULT_MAX_BITRATE;
self->vbv_buffer_size = DEFAULT_VBV_BUFFER_SIZE;
self->rc_lookahead = DEFAULT_RC_LOOKAHEAD;
self->i_adapt = DEFAULT_I_ADAPT;
self->b_adapt = DEFAULT_B_ADAPT;
self->spatial_aq = DEFAULT_SPATIAL_AQ;
self->temporal_aq = DEFAULT_TEMPORAL_AQ;
self->zero_reorder_delay = DEFAULT_ZERO_REORDER_DELAY;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
self->non_ref_p = DEFAULT_NON_REF_P;
self->strict_gop = DEFAULT_STRICT_GOP;
self->aq_strength = DEFAULT_AQ_STRENGTH;
self->min_qp_i = DEFAULT_QP;
self->min_qp_p = DEFAULT_QP;
self->min_qp_b = DEFAULT_QP;
self->max_qp_i = DEFAULT_QP;
self->max_qp_p = DEFAULT_QP;
self->max_qp_b = DEFAULT_QP;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
self->const_quality = DEFAULT_CONST_QUALITY;
self->aud = DEFAULT_AUD;
if (klass->dev_caps.cabac)
self->cabac = TRUE;
self->repeat_sequence_header = DEFAULT_REPEAT_SEQUENCE_HEADER;
self->parser = gst_h264_nal_parser_new ();
if (klass->d3d11_mode) {
gst_nv_encoder_set_dxgi_adapter_luid (GST_NV_ENCODER (self),
klass->adapter_luid);
} else {
gst_nv_encoder_set_cuda_device_id (GST_NV_ENCODER (self),
klass->cuda_device_id);
}
}
static void
gst_nv_h264_encoder_finalize (GObject * object)
{
GstNvH264Encoder *self = GST_NV_H264_ENCODER (object);
g_mutex_clear (&self->prop_lock);
gst_h264_nal_parser_free (self->parser);
G_OBJECT_CLASS (parent_class)->finalize (object);
}
typedef enum
{
UPDATE_INIT_PARAM,
UPDATE_RC_PARAM,
UPDATE_BITRATE,
} PropUpdateLevel;
static void
update_boolean (GstNvH264Encoder * self, gboolean * old_val,
const GValue * new_val, PropUpdateLevel level)
{
gboolean val = g_value_get_boolean (new_val);
if (*old_val == val)
return;
*old_val = val;
switch (level) {
case UPDATE_INIT_PARAM:
self->init_param_updated = TRUE;
break;
case UPDATE_RC_PARAM:
self->rc_param_updated = TRUE;
break;
case UPDATE_BITRATE:
self->bitrate_updated = TRUE;
break;
}
}
static void
update_int (GstNvH264Encoder * self, gint * old_val,
const GValue * new_val, PropUpdateLevel level)
{
gint val = g_value_get_int (new_val);
if (*old_val == val)
return;
*old_val = val;
switch (level) {
case UPDATE_INIT_PARAM:
self->init_param_updated = TRUE;
break;
case UPDATE_RC_PARAM:
self->rc_param_updated = TRUE;
break;
case UPDATE_BITRATE:
self->bitrate_updated = TRUE;
break;
}
}
static void
update_uint (GstNvH264Encoder * self, guint * old_val,
const GValue * new_val, PropUpdateLevel level)
{
guint val = g_value_get_uint (new_val);
if (*old_val == val)
return;
*old_val = val;
switch (level) {
case UPDATE_INIT_PARAM:
self->init_param_updated = TRUE;
break;
case UPDATE_RC_PARAM:
self->rc_param_updated = TRUE;
break;
case UPDATE_BITRATE:
self->bitrate_updated = TRUE;
break;
}
}
static void
update_double (GstNvH264Encoder * self, gdouble * old_val,
const GValue * new_val, PropUpdateLevel level)
{
gdouble val = g_value_get_double (new_val);
if (*old_val == val)
return;
*old_val = val;
switch (level) {
case UPDATE_INIT_PARAM:
self->init_param_updated = TRUE;
break;
case UPDATE_RC_PARAM:
self->rc_param_updated = TRUE;
break;
case UPDATE_BITRATE:
self->bitrate_updated = TRUE;
break;
}
}
static void
gst_nv_h264_encoder_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec)
{
GstNvH264Encoder *self = GST_NV_H264_ENCODER (object);
g_mutex_lock (&self->prop_lock);
switch (prop_id) {
case PROP_PRESET:{
GstNvEncoderPreset preset = (GstNvEncoderPreset) g_value_get_enum (value);
if (preset != self->preset) {
self->preset = preset;
self->init_param_updated = TRUE;
}
break;
}
case PROP_WEIGHTED_PRED:
update_boolean (self, &self->weighted_pred, value, UPDATE_INIT_PARAM);
break;
case PROP_GOP_SIZE:
update_int (self, &self->gop_size, value, UPDATE_INIT_PARAM);
break;
case PROP_B_FRAMES:
update_uint (self, &self->bframes, value, UPDATE_INIT_PARAM);
break;
case PROP_RATE_CONTROL:{
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
GstNvEncoderRCMode mode = (GstNvEncoderRCMode) g_value_get_enum (value);
if (mode != self->rc_mode) {
self->rc_mode = mode;
self->rc_param_updated = TRUE;
}
break;
}
case PROP_QP_I:
update_int (self, &self->qp_i, value, UPDATE_RC_PARAM);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_QP_P:
update_int (self, &self->qp_p, value, UPDATE_RC_PARAM);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_QP_B:
update_int (self, &self->qp_b, value, UPDATE_RC_PARAM);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_BITRATE:
update_uint (self, &self->bitrate, value, UPDATE_BITRATE);
break;
case PROP_MAX_BITRATE:
update_uint (self, &self->max_bitrate, value, UPDATE_BITRATE);
break;
case PROP_VBV_BUFFER_SIZE:
update_uint (self, &self->vbv_buffer_size, value, UPDATE_RC_PARAM);
break;
case PROP_RC_LOOKAHEAD:
/* rc-lookahead update requires pool size change */
update_uint (self, &self->rc_lookahead, value, UPDATE_INIT_PARAM);
break;
case PROP_I_ADAPT:
update_boolean (self, &self->i_adapt, value, UPDATE_RC_PARAM);
break;
case PROP_B_ADAPT:
update_boolean (self, &self->b_adapt, value, UPDATE_RC_PARAM);
break;
case PROP_SPATIAL_AQ:
update_boolean (self, &self->spatial_aq, value, UPDATE_RC_PARAM);
break;
case PROP_TEMPORAL_AQ:
update_boolean (self, &self->temporal_aq, value, UPDATE_RC_PARAM);
break;
case PROP_ZERO_REORDER_DELAY:
update_boolean (self, &self->zero_reorder_delay, value, UPDATE_RC_PARAM);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_NON_REF_P:
update_boolean (self, &self->non_ref_p, value, UPDATE_RC_PARAM);
break;
case PROP_STRICT_GOP:
update_boolean (self, &self->strict_gop, value, UPDATE_RC_PARAM);
break;
case PROP_AQ_STRENGTH:
update_uint (self, &self->aq_strength, value, UPDATE_RC_PARAM);
break;
case PROP_MIN_QP_I:
update_int (self, &self->min_qp_i, value, UPDATE_RC_PARAM);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_MIN_QP_P:
update_int (self, &self->min_qp_p, value, UPDATE_RC_PARAM);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_MIN_QP_B:
update_int (self, &self->min_qp_b, value, UPDATE_RC_PARAM);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_MAX_QP_I:
update_int (self, &self->min_qp_i, value, UPDATE_RC_PARAM);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_MAX_QP_P:
update_int (self, &self->min_qp_p, value, UPDATE_RC_PARAM);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_MAX_QP_B:
update_int (self, &self->min_qp_b, value, UPDATE_RC_PARAM);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_CONST_QUALITY:
update_double (self, &self->const_quality, value, UPDATE_RC_PARAM);
break;
case PROP_AUD:
update_boolean (self, &self->aud, value, UPDATE_INIT_PARAM);
break;
case PROP_CABAC:
update_boolean (self, &self->cabac, value, UPDATE_INIT_PARAM);
break;
case PROP_REPEAT_SEQUENCE_HEADER:
update_boolean (self,
&self->repeat_sequence_header, value, UPDATE_INIT_PARAM);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
g_mutex_unlock (&self->prop_lock);
}
static void
gst_nv_h264_encoder_get_property (GObject * object, guint prop_id,
GValue * value, GParamSpec * pspec)
{
GstNvH264Encoder *self = GST_NV_H264_ENCODER (object);
GstNvH264EncoderClass *klass = GST_NV_H264_ENCODER_GET_CLASS (self);
switch (prop_id) {
case PROP_ADAPTER_LUID:
g_value_set_int64 (value, klass->adapter_luid);
break;
case PROP_CUDA_DEVICE_ID:
g_value_set_uint (value, klass->cuda_device_id);
break;
case PROP_PRESET:
g_value_set_enum (value, self->preset);
break;
case PROP_WEIGHTED_PRED:
g_value_set_boolean (value, self->weighted_pred);
break;
case PROP_GOP_SIZE:
g_value_set_int (value, self->gop_size);
break;
case PROP_B_FRAMES:
g_value_set_uint (value, self->bframes);
break;
case PROP_RATE_CONTROL:
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
g_value_set_enum (value, self->rc_mode);
break;
case PROP_QP_I:
g_value_set_int (value, self->qp_i);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_QP_P:
g_value_set_int (value, self->qp_p);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_QP_B:
g_value_set_int (value, self->qp_b);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_BITRATE:
g_value_set_uint (value, self->bitrate);
break;
case PROP_MAX_BITRATE:
g_value_set_uint (value, self->max_bitrate);
break;
case PROP_VBV_BUFFER_SIZE:
g_value_set_uint (value, self->vbv_buffer_size);
break;
case PROP_RC_LOOKAHEAD:
g_value_set_uint (value, self->rc_lookahead);
break;
case PROP_I_ADAPT:
g_value_set_boolean (value, self->i_adapt);
break;
case PROP_B_ADAPT:
g_value_set_boolean (value, self->b_adapt);
break;
case PROP_SPATIAL_AQ:
g_value_set_boolean (value, self->spatial_aq);
break;
case PROP_TEMPORAL_AQ:
g_value_set_boolean (value, self->temporal_aq);
break;
case PROP_ZERO_REORDER_DELAY:
g_value_set_boolean (value, self->zero_reorder_delay);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_NON_REF_P:
g_value_set_boolean (value, self->non_ref_p);
break;
case PROP_STRICT_GOP:
g_value_set_boolean (value, self->strict_gop);
break;
case PROP_AQ_STRENGTH:
g_value_set_uint (value, self->aq_strength);
break;
case PROP_MIN_QP_I:
g_value_set_int (value, self->min_qp_i);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_MIN_QP_P:
g_value_set_int (value, self->min_qp_p);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_MIN_QP_B:
g_value_set_int (value, self->min_qp_b);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_MAX_QP_I:
g_value_set_int (value, self->max_qp_i);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_MAX_QP_P:
g_value_set_int (value, self->max_qp_p);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_MAX_QP_B:
g_value_set_int (value, self->max_qp_b);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
break;
case PROP_CONST_QUALITY:
g_value_set_double (value, self->const_quality);
break;
case PROP_AUD:
g_value_set_boolean (value, self->aud);
break;
case PROP_CABAC:
g_value_set_boolean (value, self->cabac);
break;
case PROP_REPEAT_SEQUENCE_HEADER:
g_value_set_boolean (value, self->repeat_sequence_header);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
static void
gst_nv_h264_encoder_get_downstream_profiles_and_format (GstNvH264Encoder * self,
std::set < std::string > &downstream_profiles, gboolean * packetized)
{
GstCaps *allowed_caps;
GstStructure *s;
const gchar *stream_format;
allowed_caps = gst_pad_get_allowed_caps (GST_VIDEO_ENCODER_SRC_PAD (self));
if (!allowed_caps || gst_caps_is_empty (allowed_caps) ||
gst_caps_is_any (allowed_caps)) {
gst_clear_caps (&allowed_caps);
return;
}
for (guint i = 0; i < gst_caps_get_size (allowed_caps); i++) {
const GValue *profile_value;
const gchar *profile;
s = gst_caps_get_structure (allowed_caps, i);
profile_value = gst_structure_get_value (s, "profile");
if (!profile_value)
continue;
if (GST_VALUE_HOLDS_LIST (profile_value)) {
for (guint j = 0; j < gst_value_list_get_size (profile_value); j++) {
const GValue *p = gst_value_list_get_value (profile_value, j);
if (!G_VALUE_HOLDS_STRING (p))
continue;
profile = g_value_get_string (p);
if (profile)
downstream_profiles.insert (profile);
}
} else if (G_VALUE_HOLDS_STRING (profile_value)) {
profile = g_value_get_string (profile_value);
if (profile)
downstream_profiles.insert (profile);
}
}
if (packetized) {
*packetized = FALSE;
allowed_caps = gst_caps_fixate (allowed_caps);
s = gst_caps_get_structure (allowed_caps, 0);
stream_format = gst_structure_get_string (s, "stream-format");
if (g_strcmp0 (stream_format, "avc") == 0)
*packetized = TRUE;
}
gst_caps_unref (allowed_caps);
}
static GstCaps *
gst_nv_h264_encoder_getcaps (GstVideoEncoder * encoder, GstCaps * filter)
{
GstNvH264Encoder *self = GST_NV_H264_ENCODER (encoder);
GstNvH264EncoderClass *klass = GST_NV_H264_ENCODER_GET_CLASS (self);
GstCaps *allowed_caps;
GstCaps *template_caps;
GstCaps *filtered_caps;
GstCaps *supported_caps;
std::set < std::string > downstream_profiles;
std::set < std::string > allowed_formats;
gboolean profile_support_interlaced = FALSE;
gst_nv_h264_encoder_get_downstream_profiles_and_format (self,
downstream_profiles, NULL);
GST_DEBUG_OBJECT (self, "Downstream specified %" G_GSIZE_FORMAT " profiles",
downstream_profiles.size ());
if (downstream_profiles.size () == 0)
return gst_video_encoder_proxy_getcaps (encoder, NULL, filter);
/* *INDENT-OFF* */
for (const auto &iter: downstream_profiles) {
if (iter == "high" || iter == "main") {
profile_support_interlaced = TRUE;
}
if (iter == "high-4:4:4") {
profile_support_interlaced = TRUE;
allowed_formats.insert("Y444");
} else {
allowed_formats.insert("NV12");
}
}
/* *INDENT-ON* */
GST_DEBUG_OBJECT (self, "Downstream %s support interlaced format",
profile_support_interlaced ? "can" : "cannot");
template_caps = gst_pad_get_pad_template_caps (encoder->sinkpad);
allowed_caps = gst_caps_copy (template_caps);
if (klass->dev_caps.field_encoding == 0 || !profile_support_interlaced) {
gst_caps_set_simple (allowed_caps, "interlace-mode", G_TYPE_STRING,
"progressive", NULL);
}
GValue formats = G_VALUE_INIT;
g_value_init (&formats, GST_TYPE_LIST);
/* *INDENT-OFF* */
for (const auto &iter: allowed_formats) {
GValue val = G_VALUE_INIT;
g_value_init (&val, G_TYPE_STRING);
g_value_set_string (&val, iter.c_str());
gst_value_list_append_and_take_value (&formats, &val);
}
/* *INDENT-ON* */
gst_caps_set_value (allowed_caps, "format", &formats);
g_value_unset (&formats);
filtered_caps = gst_caps_intersect_full (template_caps, allowed_caps,
GST_CAPS_INTERSECT_FIRST);
supported_caps = gst_video_encoder_proxy_getcaps (encoder,
filtered_caps, filter);
gst_caps_unref (filtered_caps);
gst_caps_unref (allowed_caps);
gst_caps_unref (template_caps);
GST_DEBUG_OBJECT (self, "Returning %" GST_PTR_FORMAT, supported_caps);
return supported_caps;
}
static gboolean
gst_nv_h264_encoder_set_format (GstNvEncoder * encoder,
GstVideoCodecState * state, gpointer session,
NV_ENC_INITIALIZE_PARAMS * init_params, NV_ENC_CONFIG * config)
{
GstNvH264Encoder *self = GST_NV_H264_ENCODER (encoder);
GstNvH264EncoderClass *klass = GST_NV_H264_ENCODER_GET_CLASS (self);
GstNvH264EncoderDeviceCaps *dev_caps = &klass->dev_caps;
NV_ENC_RC_PARAMS *rc_params;
GstVideoInfo *info = &state->info;
NVENCSTATUS status;
NV_ENC_PRESET_CONFIG preset_config = { 0, };
gint dar_n, dar_d;
GstNvEncoderRCMode rc_mode;
NV_ENC_CONFIG_H264 *h264_config;
NV_ENC_CONFIG_H264_VUI_PARAMETERS *vui;
std::set < std::string > downstream_profiles;
GUID selected_profile = NV_ENC_CODEC_PROFILE_AUTOSELECT_GUID;
gboolean downstream_supports_bframe = FALSE;
gboolean bframe_aborted = FALSE;
self->packetized = FALSE;
gst_nv_h264_encoder_get_downstream_profiles_and_format (self,
downstream_profiles, &self->packetized);
if (downstream_profiles.empty ()) {
GST_ERROR_OBJECT (self, "Unable to get downstream profile");
return FALSE;
}
if (GST_VIDEO_INFO_IS_INTERLACED (info)) {
downstream_profiles.erase ("progressive-high");
downstream_profiles.erase ("constrained-high");
downstream_profiles.erase ("constrained-baseline");
downstream_profiles.erase ("baseline");
if (downstream_profiles.empty ()) {
GST_ERROR_OBJECT (self,
"None of downstream profile supports interlaced encoding");
return FALSE;
}
}
if (GST_VIDEO_INFO_FORMAT (info) == GST_VIDEO_FORMAT_Y444) {
if (downstream_profiles.find ("high-4:4:4") == downstream_profiles.end ()) {
GST_ERROR_OBJECT (self, "Downstream does not support 4:4:4 profile");
return FALSE;
} else {
selected_profile = NV_ENC_H264_PROFILE_HIGH_444_GUID;
downstream_supports_bframe = TRUE;
}
} else {
/* *INDENT-OFF* */
for (const auto &iter: downstream_profiles) {
if (iter == "high" || iter == "main" || iter == "progressive-high") {
downstream_supports_bframe = TRUE;
}
}
/* *INDENT-ON* */
}
g_mutex_lock (&self->prop_lock);
init_params->version = gst_nvenc_get_initialize_params_version ();
init_params->encodeGUID = NV_ENC_CODEC_H264_GUID;
init_params->encodeWidth = GST_VIDEO_INFO_WIDTH (info);
init_params->maxEncodeWidth = GST_VIDEO_INFO_WIDTH (info);
init_params->encodeHeight = GST_VIDEO_INFO_HEIGHT (info);
init_params->maxEncodeHeight = GST_VIDEO_INFO_HEIGHT (info);
init_params->enablePTD = TRUE;
if (dev_caps->async_encoding_support)
init_params->enableEncodeAsync = 1;
if (info->fps_d > 0 && info->fps_n > 0) {
init_params->frameRateNum = info->fps_n;
init_params->frameRateDen = info->fps_d;
} else {
init_params->frameRateNum = 0;
init_params->frameRateDen = 1;
}
init_params->enableWeightedPrediction = self->weighted_pred;
if (gst_util_fraction_multiply (GST_VIDEO_INFO_WIDTH (info),
GST_VIDEO_INFO_HEIGHT (info), GST_VIDEO_INFO_PAR_N (info),
GST_VIDEO_INFO_PAR_D (info), &dar_n, &dar_d) && dar_n > 0
&& dar_d > 0) {
init_params->darWidth = dar_n;
init_params->darHeight = dar_d;
}
if (GST_VIDEO_INFO_IS_INTERLACED (info) && dev_caps->field_encoding > 0) {
switch (GST_VIDEO_INFO_INTERLACE_MODE (info)) {
case GST_VIDEO_INTERLACE_MODE_INTERLEAVED:
case GST_VIDEO_INTERLACE_MODE_MIXED:
config->frameFieldMode = NV_ENC_PARAMS_FRAME_FIELD_MODE_FIELD;
preset_config.presetCfg.frameFieldMode =
NV_ENC_PARAMS_FRAME_FIELD_MODE_FIELD;
break;
default:
break;
}
}
gst_nv_encoder_preset_to_guid (self->preset, &init_params->presetGUID);
preset_config.version = gst_nvenc_get_preset_config_version ();
preset_config.presetCfg.version = gst_nvenc_get_config_version ();
status = NvEncGetEncodePresetConfig (session, NV_ENC_CODEC_H264_GUID,
init_params->presetGUID, &preset_config);
if (status != NV_ENC_SUCCESS) {
GST_ERROR_OBJECT (self, "Failed to get preset config %"
GST_NVENC_STATUS_FORMAT, GST_NVENC_STATUS_ARGS (status));
g_mutex_unlock (&self->prop_lock);
return FALSE;
}
*config = preset_config.presetCfg;
if (self->gop_size < 0) {
config->gopLength = NVENC_INFINITE_GOPLENGTH;
config->frameIntervalP = 1;
} else if (self->gop_size > 0) {
config->gopLength = self->gop_size;
/* frameIntervalP
* 0: All Intra frames
* 1: I/P only
* 2: IBP
* 3: IBBP
*/
config->frameIntervalP = 1;
if (self->bframes > 0 && !downstream_supports_bframe) {
GST_WARNING_OBJECT (self,
"B-frame was enabled but downstream profile does not support it");
bframe_aborted = TRUE;
self->bframes = 0;
}
config->frameIntervalP = self->bframes + 1;
} else {
/* gop size == 0 means all intra frames */
config->gopLength = 1;
config->frameIntervalP = 0;
}
rc_params = &config->rcParams;
rc_mode = self->rc_mode;
if (self->bitrate)
rc_params->averageBitRate = self->bitrate * 1024;
if (self->max_bitrate)
rc_params->maxBitRate = self->max_bitrate * 1024;
if (self->vbv_buffer_size)
rc_params->vbvBufferSize = self->vbv_buffer_size * 1024;
if (self->min_qp_i >= 0) {
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
rc_params->enableMinQP = TRUE;
rc_params->minQP.qpIntra = self->min_qp_i;
if (self->min_qp_p >= 0) {
rc_params->minQP.qpInterP = self->min_qp_p;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
} else {
rc_params->minQP.qpInterP = rc_params->minQP.qpIntra;
}
if (self->min_qp_b >= 0) {
rc_params->minQP.qpInterB = self->min_qp_b;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
} else {
rc_params->minQP.qpInterB = rc_params->minQP.qpInterP;
}
}
if (self->max_qp_i >= 0) {
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
rc_params->enableMaxQP = TRUE;
rc_params->maxQP.qpIntra = self->max_qp_i;
if (self->max_qp_p >= 0) {
rc_params->maxQP.qpInterP = self->max_qp_p;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
} else {
rc_params->maxQP.qpInterP = rc_params->maxQP.qpIntra;
}
if (self->max_qp_b >= 0) {
rc_params->maxQP.qpInterB = self->max_qp_b;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
} else {
rc_params->maxQP.qpInterB = rc_params->maxQP.qpInterP;
}
}
if (rc_mode == GST_NV_ENCODER_RC_MODE_CONSTQP) {
if (self->qp_i >= 0)
rc_params->constQP.qpIntra = self->qp_i;
if (self->qp_p >= 0)
rc_params->constQP.qpInterP = self->qp_p;
if (self->qp_p >= 0)
rc_params->constQP.qpInterB = self->qp_b;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
}
rc_params->rateControlMode = gst_nv_encoder_rc_mode_to_native (rc_mode);
if (self->spatial_aq) {
rc_params->enableAQ = TRUE;
rc_params->aqStrength = self->aq_strength;
}
rc_params->enableTemporalAQ = self->temporal_aq;
if (self->rc_lookahead) {
rc_params->enableLookahead = 1;
rc_params->lookaheadDepth = self->rc_lookahead;
rc_params->disableIadapt = !self->i_adapt;
rc_params->disableBadapt = !self->b_adapt;
}
rc_params->strictGOPTarget = self->strict_gop;
rc_params->enableNonRefP = self->non_ref_p;
rc_params->zeroReorderDelay = self->zero_reorder_delay;
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
if (self->const_quality) {
guint scaled = (gint) (self->const_quality * 256.0);
rc_params->targetQuality = (guint8) (scaled >> 8);
rc_params->targetQualityLSB = (guint8) (scaled & 0xff);
}
self->init_param_updated = FALSE;
self->bitrate_updated = FALSE;
self->rc_param_updated = FALSE;
if (selected_profile == NV_ENC_CODEC_PROFILE_AUTOSELECT_GUID &&
config->frameIntervalP > 1) {
if (downstream_profiles.find ("main") != downstream_profiles.end ()) {
selected_profile = NV_ENC_H264_PROFILE_MAIN_GUID;
} else if (downstream_profiles.find ("high") != downstream_profiles.end ()) {
selected_profile = NV_ENC_H264_PROFILE_HIGH_GUID;
} else if (downstream_profiles.find ("progressive-high") !=
downstream_profiles.end ()) {
selected_profile = NV_ENC_H264_PROFILE_PROGRESSIVE_HIGH_GUID;
}
}
/* Pick the first profile */
if (selected_profile == NV_ENC_CODEC_PROFILE_AUTOSELECT_GUID) {
if (*downstream_profiles.begin () == "baseline" ||
*downstream_profiles.begin () == "constrained-baseline") {
selected_profile = NV_ENC_H264_PROFILE_BASELINE_GUID;
} else if (*downstream_profiles.begin () == "main") {
selected_profile = NV_ENC_H264_PROFILE_MAIN_GUID;
} else if (*downstream_profiles.begin () == "progressive-high") {
selected_profile = NV_ENC_H264_PROFILE_PROGRESSIVE_HIGH_GUID;
} else if (*downstream_profiles.begin () == "constrained-high") {
selected_profile = NV_ENC_H264_PROFILE_CONSTRAINED_HIGH_GUID;
}
}
config->profileGUID = selected_profile;
h264_config = &config->encodeCodecConfig.h264Config;
vui = &h264_config->h264VUIParameters;
h264_config->level = NV_ENC_LEVEL_AUTOSELECT;
h264_config->chromaFormatIDC = 1;
if (selected_profile == NV_ENC_H264_PROFILE_HIGH_444_GUID)
h264_config->chromaFormatIDC = 3;
h264_config->idrPeriod = config->gopLength;
h264_config->outputAUD = self->aud;
if (self->repeat_sequence_header) {
h264_config->disableSPSPPS = 0;
h264_config->repeatSPSPPS = 1;
} else {
if (self->packetized)
h264_config->disableSPSPPS = 1;
else
h264_config->disableSPSPPS = 0;
}
if (dev_caps->cabac && selected_profile != NV_ENC_H264_PROFILE_BASELINE_GUID) {
if (self->cabac)
h264_config->entropyCodingMode = NV_ENC_H264_ENTROPY_CODING_MODE_CABAC;
else
h264_config->entropyCodingMode = NV_ENC_H264_ENTROPY_CODING_MODE_CAVLC;
} else {
h264_config->entropyCodingMode = NV_ENC_H264_ENTROPY_CODING_MODE_AUTOSELECT;
}
vui->videoSignalTypePresentFlag = 1;
/* Unspecified */
vui->videoFormat = 5;
if (info->colorimetry.range == GST_VIDEO_COLOR_RANGE_0_255) {
vui->videoFullRangeFlag = 1;
} else {
vui->videoFullRangeFlag = 0;
}
vui->colourDescriptionPresentFlag = 1;
vui->colourMatrix = gst_video_color_matrix_to_iso (info->colorimetry.matrix);
vui->colourPrimaries =
gst_video_color_primaries_to_iso (info->colorimetry.primaries);
vui->transferCharacteristics =
gst_video_transfer_function_to_iso (info->colorimetry.transfer);
g_mutex_unlock (&self->prop_lock);
if (bframe_aborted)
g_object_notify (G_OBJECT (self), "b-frames");
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
return TRUE;
}
static gboolean
gst_nv_h264_encoder_set_output_state (GstNvEncoder * encoder,
GstVideoCodecState * state, gpointer session)
{
GstNvH264Encoder *self = GST_NV_H264_ENCODER (encoder);
GstVideoCodecState *output_state;
NV_ENC_SEQUENCE_PARAM_PAYLOAD seq_params = { 0, };
guint8 spspps[1024];
guint32 seq_size;
GstCaps *caps;
const gchar *profile_from_sps;
NVENCSTATUS status;
std::set < std::string > downstream_profiles;
std::string caps_str;
GstTagList *tags;
GstBuffer *codec_data = NULL;
GstH264NalUnit sps_nalu, pps_nalu;
GstH264ParserResult rst;
caps_str = "video/x-h264, alignment = (string) au";
gst_nv_h264_encoder_get_downstream_profiles_and_format (self,
downstream_profiles, NULL);
seq_params.version = gst_nvenc_get_sequence_param_payload_version ();
seq_params.inBufferSize = sizeof (spspps);
seq_params.spsppsBuffer = &spspps;
seq_params.outSPSPPSPayloadSize = &seq_size;
status = NvEncGetSequenceParams (session, &seq_params);
if (status != NV_ENC_SUCCESS) {
GST_ERROR_OBJECT (self, "Failed to get sequence header, status %"
GST_NVENC_STATUS_FORMAT, GST_NVENC_STATUS_ARGS (status));
return FALSE;
}
rst = gst_h264_parser_identify_nalu (self->parser,
spspps, 0, seq_size, &sps_nalu);
if (rst != GST_H264_PARSER_OK) {
GST_ERROR_OBJECT (self, "Failed to identify SPS nal");
return FALSE;
}
if (sps_nalu.size < 4) {
GST_ERROR_OBJECT (self, "Too small sps nal size %d", sps_nalu.size);
return FALSE;
}
rst = gst_h264_parser_identify_nalu_unchecked (self->parser,
spspps, sps_nalu.offset + sps_nalu.size, seq_size, &pps_nalu);
if (rst != GST_H264_PARSER_OK && self->packetized) {
GST_ERROR_OBJECT (self, "Failed to identify PPS nal, %d", rst);
return FALSE;
}
if (self->packetized) {
GstMapInfo info;
guint8 *data;
guint8 profile_idc, profile_comp, level_idc;
const guint nal_length_size = 4;
const guint num_sps = 1;
const guint num_pps = 1;
data = sps_nalu.data + sps_nalu.offset + sps_nalu.header_bytes;
profile_idc = data[0];
profile_comp = data[1];
level_idc = data[2];
/* 5: configuration version, profile, compatibility, level, nal length
* 1: num sps
* 2: sps size bytes
* sizeof (sps)
* 1: num pps
* 2: pps size bytes
* sizeof (pps)
*
* -> 11 + sps_size + pps_size
*/
codec_data = gst_buffer_new_and_alloc (11 + sps_nalu.size + pps_nalu.size);
gst_buffer_map (codec_data, &info, GST_MAP_WRITE);
data = (guint8 *) info.data;
data[0] = 1;
data[1] = profile_idc;
data[2] = profile_comp;
data[3] = level_idc;
data[4] = 0xfc | (nal_length_size - 1);
data[5] = 0xe0 | num_sps;
data += 6;
GST_WRITE_UINT16_BE (data, sps_nalu.size);
data += 2;
memcpy (data, sps_nalu.data + sps_nalu.offset, sps_nalu.size);
data += sps_nalu.size;
data[0] = num_pps;
data++;
GST_WRITE_UINT16_BE (data, pps_nalu.size);
data += 2;
memcpy (data, pps_nalu.data + pps_nalu.offset, pps_nalu.size);
gst_buffer_unmap (codec_data, &info);
}
profile_from_sps =
gst_codec_utils_h264_get_profile (sps_nalu.data + sps_nalu.offset +
sps_nalu.header_bytes, 3);
if (!profile_from_sps) {
GST_WARNING_OBJECT (self, "Failed to parse profile from SPS");
} else if (!downstream_profiles.empty ()) {
if (downstream_profiles.find (profile_from_sps) !=
downstream_profiles.end ()) {
caps_str += ", profile = (string) " + std::string (profile_from_sps);
} else if (downstream_profiles.find ("baseline") !=
downstream_profiles.end () &&
strcmp (profile_from_sps, "constrained-baseline") == 0) {
caps_str += ", profile = (string) baseline";
} else if (downstream_profiles.find ("constrained-baseline") !=
downstream_profiles.end () &&
strcmp (profile_from_sps, "constrained-baseline") == 0) {
caps_str += ", profile = (string) constrained-baseline";
}
} else {
caps_str += ", profile = (string) " + std::string (profile_from_sps);
}
if (self->packetized) {
caps_str += ", stream-format = (string) avc";
} else {
caps_str += ", stream-format = (string) byte-stream";
}
caps = gst_caps_from_string (caps_str.c_str ());
if (self->packetized) {
gst_caps_set_simple (caps, "codec_data", GST_TYPE_BUFFER, codec_data, NULL);
gst_buffer_unref (codec_data);
}
output_state = gst_video_encoder_set_output_state (GST_VIDEO_ENCODER (self),
caps, state);
GST_INFO_OBJECT (self, "Output caps: %" GST_PTR_FORMAT, output_state->caps);
gst_video_codec_state_unref (output_state);
tags = gst_tag_list_new_empty ();
gst_tag_list_add (tags, GST_TAG_MERGE_REPLACE, GST_TAG_ENCODER,
"nvh264encoder", NULL);
gst_video_encoder_merge_tags (GST_VIDEO_ENCODER (encoder),
tags, GST_TAG_MERGE_REPLACE);
gst_tag_list_unref (tags);
return TRUE;
}
static GstBuffer *
gst_nv_h264_encoder_create_output_buffer (GstNvEncoder * encoder,
NV_ENC_LOCK_BITSTREAM * bitstream)
{
GstNvH264Encoder *self = GST_NV_H264_ENCODER (encoder);
GstBuffer *buffer;
GstH264ParserResult rst;
GstH264NalUnit nalu;
if (!self->packetized) {
return gst_buffer_new_memdup (bitstream->bitstreamBufferPtr,
bitstream->bitstreamSizeInBytes);
}
buffer = gst_buffer_new ();
rst = gst_h264_parser_identify_nalu (self->parser,
(guint8 *) bitstream->bitstreamBufferPtr, 0,
bitstream->bitstreamSizeInBytes, &nalu);
if (rst == GST_H264_PARSER_NO_NAL_END)
rst = GST_H264_PARSER_OK;
while (rst == GST_H264_PARSER_OK) {
GstMemory *mem;
guint8 *data;
data = (guint8 *) g_malloc0 (nalu.size + 4);
GST_WRITE_UINT32_BE (data, nalu.size);
memcpy (data + 4, nalu.data + nalu.offset, nalu.size);
mem = gst_memory_new_wrapped ((GstMemoryFlags) 0, data, nalu.size + 4,
0, nalu.size + 4, data, (GDestroyNotify) g_free);
gst_buffer_append_memory (buffer, mem);
rst = gst_h264_parser_identify_nalu (self->parser,
(guint8 *) bitstream->bitstreamBufferPtr, nalu.offset + nalu.size,
bitstream->bitstreamSizeInBytes, &nalu);
if (rst == GST_H264_PARSER_NO_NAL_END)
rst = GST_H264_PARSER_OK;
}
return buffer;
}
static GstNvEncoderReconfigure
gst_nv_h264_encoder_check_reconfigure (GstNvEncoder * encoder,
NV_ENC_CONFIG * config)
{
GstNvH264Encoder *self = GST_NV_H264_ENCODER (encoder);
GstNvEncoderReconfigure reconfig = GST_NV_ENCODER_RECONFIGURE_NONE;
/* Dynamic RC param update is not tested, do soft-reconfigure only for
* bitrate update */
g_mutex_lock (&self->prop_lock);
if (self->init_param_updated || self->rc_param_updated) {
reconfig = GST_NV_ENCODER_RECONFIGURE_FULL;
goto done;
}
if (self->bitrate_updated) {
GstNvH264EncoderClass *klass = GST_NV_H264_ENCODER_GET_CLASS (self);
if (klass->dev_caps.dyn_bitrate_change > 0) {
config->rcParams.averageBitRate = self->bitrate * 1024;
config->rcParams.maxBitRate = self->max_bitrate * 1024;
reconfig = GST_NV_ENCODER_RECONFIGURE_BITRATE;
} else {
reconfig = GST_NV_ENCODER_RECONFIGURE_FULL;
}
}
done:
self->init_param_updated = FALSE;
self->rc_param_updated = FALSE;
self->bitrate_updated = FALSE;
g_mutex_unlock (&self->prop_lock);
return reconfig;
}
static GstNvH264EncoderClassData *
gst_nv_h264_encoder_create_class_data (GstObject * device, gpointer session,
gboolean d3d11_mode)
{
NVENCSTATUS status;
GstNvH264EncoderDeviceCaps dev_caps = { 0, };
NV_ENC_CAPS_PARAM caps_param = { 0, };
GUID profile_guids[16];
NV_ENC_BUFFER_FORMAT input_formats[16];
guint32 profile_guid_count = 0;
guint32 input_format_count = 0;
std::string sink_caps_str;
std::string src_caps_str;
std::string format_str;
std::set < std::string > formats;
std::set < std::string > profiles;
std::string profile_str;
std::string resolution_str;
GstNvH264EncoderClassData *cdata;
GstCaps *sink_caps;
GstCaps *system_caps;
status = NvEncGetEncodeProfileGUIDs (session, NV_ENC_CODEC_H264_GUID,
profile_guids, G_N_ELEMENTS (profile_guids), &profile_guid_count);
if (status != NV_ENC_SUCCESS || profile_guid_count == 0) {
GST_WARNING_OBJECT (device, "Unable to get supported profiles");
return NULL;
}
status = NvEncGetInputFormats (session, NV_ENC_CODEC_H264_GUID, input_formats,
G_N_ELEMENTS (input_formats), &input_format_count);
if (status != NV_ENC_SUCCESS || input_format_count == 0) {
GST_WARNING_OBJECT (device, "Unable to get supported input formats");
return NULL;
}
caps_param.version = gst_nvenc_get_caps_param_version ();
#define CHECK_CAPS(to_query,val,default_val) G_STMT_START { \
gint _val; \
caps_param.capsToQuery = to_query; \
status = NvEncGetEncodeCaps (session, NV_ENC_CODEC_H264_GUID, &caps_param, \
&_val); \
if (status != NV_ENC_SUCCESS) { \
GST_WARNING_OBJECT (device, "Unable to query %s, status: %" \
GST_NVENC_STATUS_FORMAT, G_STRINGIFY (to_query), \
GST_NVENC_STATUS_ARGS (status)); \
val = default_val; \
} else { \
GST_DEBUG_OBJECT (device, "%s: %d", G_STRINGIFY (to_query), _val); \
val = _val; \
} \
} G_STMT_END
CHECK_CAPS (NV_ENC_CAPS_NUM_MAX_BFRAMES, dev_caps.max_bframes, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORTED_RATECONTROL_MODES,
dev_caps.ratecontrol_modes, NV_ENC_PARAMS_RC_VBR);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_FIELD_ENCODING, dev_caps.field_encoding, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_MONOCHROME, dev_caps.monochrome, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_FMO, dev_caps.fmo, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_QPELMV, dev_caps.qpelmv, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_BDIRECT_MODE, dev_caps.bdirect_mode, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_CABAC, dev_caps.cabac, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_ADAPTIVE_TRANSFORM,
dev_caps.adaptive_transform, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_STEREO_MVC, dev_caps.stereo_mvc, 0);
CHECK_CAPS (NV_ENC_CAPS_NUM_MAX_TEMPORAL_LAYERS, dev_caps.temoral_layers, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_HIERARCHICAL_PFRAMES,
dev_caps.hierarchical_pframes, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_HIERARCHICAL_BFRAMES,
dev_caps.hierarchical_bframes, 0);
CHECK_CAPS (NV_ENC_CAPS_LEVEL_MAX, dev_caps.level_max, 0);
CHECK_CAPS (NV_ENC_CAPS_LEVEL_MIN, dev_caps.level_min, 0);
CHECK_CAPS (NV_ENC_CAPS_SEPARATE_COLOUR_PLANE,
dev_caps.separate_colour_plane, 0);
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
CHECK_CAPS (NV_ENC_CAPS_WIDTH_MAX, dev_caps.width_max, 4096);
CHECK_CAPS (NV_ENC_CAPS_HEIGHT_MAX, dev_caps.height_max, 4096);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_TEMPORAL_SVC, dev_caps.temporal_svc, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_DYN_RES_CHANGE, dev_caps.dyn_res_change, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_DYN_BITRATE_CHANGE,
dev_caps.dyn_bitrate_change, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_DYN_FORCE_CONSTQP,
dev_caps.dyn_force_constqp, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_DYN_RCMODE_CHANGE,
dev_caps.dyn_rcmode_change, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_SUBFRAME_READBACK,
dev_caps.subframe_readback, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_CONSTRAINED_ENCODING,
dev_caps.constrained_encoding, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_INTRA_REFRESH, dev_caps.intra_refresh, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_CUSTOM_VBV_BUF_SIZE,
dev_caps.custom_vbv_buf_size, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_DYNAMIC_SLICE_MODE,
dev_caps.dynamic_slice_mode, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_REF_PIC_INVALIDATION,
dev_caps.ref_pic_invalidation, 0);
CHECK_CAPS (NV_ENC_CAPS_PREPROC_SUPPORT, dev_caps.preproc_support, 0);
/* NOTE: Async is Windows only */
#ifdef G_OS_WIN32
CHECK_CAPS (NV_ENC_CAPS_ASYNC_ENCODE_SUPPORT,
dev_caps.async_encoding_support, 0);
#endif
CHECK_CAPS (NV_ENC_CAPS_MB_NUM_MAX, dev_caps.mb_num_max, 0);
CHECK_CAPS (NV_ENC_CAPS_MB_PER_SEC_MAX, dev_caps.mb_per_sec_max, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_YUV444_ENCODE, dev_caps.yuv444_encode, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_LOSSLESS_ENCODE, dev_caps.lossless_encode, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_MEONLY_MODE, dev_caps.meonly_mode, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_LOOKAHEAD, dev_caps.lookahead, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_TEMPORAL_AQ, dev_caps.temporal_aq, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_10BIT_ENCODE,
dev_caps.supports_10bit_encode, 0);
CHECK_CAPS (NV_ENC_CAPS_NUM_MAX_LTR_FRAMES, dev_caps.num_max_ltr_frames, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_WEIGHTED_PREDICTION,
dev_caps.weighted_prediction, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_BFRAME_REF_MODE, dev_caps.bframe_ref_mode, 0);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_EMPHASIS_LEVEL_MAP,
dev_caps.emphasis_level_map, 0);
CHECK_CAPS (NV_ENC_CAPS_WIDTH_MIN, dev_caps.width_min, 16);
CHECK_CAPS (NV_ENC_CAPS_HEIGHT_MIN, dev_caps.height_min, 16);
CHECK_CAPS (NV_ENC_CAPS_SUPPORT_MULTIPLE_REF_FRAMES,
dev_caps.multiple_ref_frames, 0);
#undef CHECK_CAPS
for (guint32 i = 0; i < input_format_count; i++) {
switch (input_formats[i]) {
case NV_ENC_BUFFER_FORMAT_NV12:
formats.insert ("NV12");
break;
case NV_ENC_BUFFER_FORMAT_YUV444:
if (dev_caps.yuv444_encode)
formats.insert ("Y444");
break;
default:
break;
}
}
if (formats.empty ()) {
GST_WARNING_OBJECT (device, "Empty supported input format");
return NULL;
}
#define APPEND_STRING(dst,set,str) G_STMT_START { \
if (set.find(str) != set.end()) { \
if (!first) \
dst += ", "; \
dst += str; \
first = false; \
} \
} G_STMT_END
if (formats.size () == 1) {
format_str = "format = (string) " + *(formats.begin ());
} else {
bool first = true;
format_str = "format = (string) { ";
APPEND_STRING (format_str, formats, "NV12");
APPEND_STRING (format_str, formats, "Y444");
format_str += " }";
}
for (guint32 i = 0; i < profile_guid_count; i++) {
if (profile_guids[i] == NV_ENC_H264_PROFILE_BASELINE_GUID) {
profiles.insert ("baseline");
profiles.insert ("constrained-baseline");
} else if (profile_guids[i] == NV_ENC_H264_PROFILE_MAIN_GUID) {
profiles.insert ("main");
} else if (profile_guids[i] == NV_ENC_H264_PROFILE_HIGH_GUID) {
profiles.insert ("high");
} else if (profile_guids[i] == NV_ENC_H264_PROFILE_HIGH_444_GUID) {
profiles.insert ("high-4:4:4");
} else if (profile_guids[i] == NV_ENC_H264_PROFILE_PROGRESSIVE_HIGH_GUID) {
profiles.insert ("progressive-high");
} else if (profile_guids[i] == NV_ENC_H264_PROFILE_CONSTRAINED_HIGH_GUID) {
profiles.insert ("constrained-high");
}
}
if (profiles.empty ()) {
GST_WARNING_OBJECT (device, "Empty supported h264 profile");
return NULL;
}
if (profiles.size () == 1) {
profile_str = "profile = (string) " + *(profiles.begin ());
} else {
bool first = true;
profile_str = "profile = (string) { ";
APPEND_STRING (profile_str, profiles, "main");
APPEND_STRING (profile_str, profiles, "high");
APPEND_STRING (profile_str, profiles, "progressive-high");
APPEND_STRING (profile_str, profiles, "constrained-high");
APPEND_STRING (profile_str, profiles, "constrained-baseline");
APPEND_STRING (profile_str, profiles, "baseline");
APPEND_STRING (profile_str, profiles, "high-4:4:4");
profile_str += " }";
}
#undef APPEND_STRING
resolution_str = "width = (int) [ " +
std::to_string (GST_ROUND_UP_16 (dev_caps.width_min))
+ ", " + std::to_string (dev_caps.width_max) + " ]";
resolution_str += ", height = (int) [ " +
std::to_string (GST_ROUND_UP_16 (dev_caps.height_min))
+ ", " + std::to_string (dev_caps.height_max) + " ]";
sink_caps_str = "video/x-raw, " + format_str + ", " + resolution_str;
if (dev_caps.field_encoding > 0) {
sink_caps_str += ", interlace-mode = (string) { interleaved, mixed }";
} else {
sink_caps_str += ", interlace-mode = (string) progressive";
}
src_caps_str = "video/x-h264, " + resolution_str + ", " + profile_str +
", stream-format = (string) { avc, byte-stream }, alignment = (string) au";
system_caps = gst_caps_from_string (sink_caps_str.c_str ());
sink_caps = gst_caps_copy (system_caps);
#ifdef GST_CUDA_HAS_D3D
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
if (d3d11_mode) {
gst_caps_set_features (sink_caps, 0,
gst_caps_features_new (GST_CAPS_FEATURE_MEMORY_D3D11_MEMORY, NULL));
} else
#endif
{
gst_caps_set_features (sink_caps, 0,
gst_caps_features_new (GST_CAPS_FEATURE_MEMORY_CUDA_MEMORY, NULL));
}
gst_caps_append (sink_caps, system_caps);
cdata = g_new0 (GstNvH264EncoderClassData, 1);
cdata->sink_caps = sink_caps;
cdata->src_caps = gst_caps_from_string (src_caps_str.c_str ());
cdata->dev_caps = dev_caps;
cdata->d3d11_mode = d3d11_mode;
if (d3d11_mode)
g_object_get (device, "adapter-luid", &cdata->adapter_luid, NULL);
else
g_object_get (device, "cuda-device-id", &cdata->cuda_device_id, NULL);
GST_MINI_OBJECT_FLAG_SET (cdata->sink_caps,
GST_MINI_OBJECT_FLAG_MAY_BE_LEAKED);
GST_MINI_OBJECT_FLAG_SET (cdata->src_caps,
GST_MINI_OBJECT_FLAG_MAY_BE_LEAKED);
return cdata;
}
void
gst_nv_h264_encoder_register_cuda (GstPlugin * plugin, GstCudaContext * context,
guint rank)
{
NV_ENC_OPEN_ENCODE_SESSION_EX_PARAMS session_params = { 0, };
gpointer session;
NVENCSTATUS status;
GstNvH264EncoderClassData *cdata;
GST_DEBUG_CATEGORY_INIT (gst_nv_h264_encoder_debug, "nvh264encoder", 0,
"nvh264encoder");
session_params.version =
gst_nvenc_get_open_encode_session_ex_params_version ();
session_params.deviceType = NV_ENC_DEVICE_TYPE_CUDA;
session_params.device = gst_cuda_context_get_handle (context);
session_params.apiVersion = gst_nvenc_get_api_version ();
status = NvEncOpenEncodeSessionEx (&session_params, &session);
if (status != NV_ENC_SUCCESS) {
GST_WARNING_OBJECT (context, "Failed to open session");
return;
}
cdata = gst_nv_h264_encoder_create_class_data (GST_OBJECT (context), session,
FALSE);
NvEncDestroyEncoder (session);
if (!cdata)
return;
GType type;
gchar *type_name;
gchar *feature_name;
GTypeInfo type_info = {
sizeof (GstNvH264EncoderClass),
NULL,
NULL,
(GClassInitFunc) gst_nv_h264_encoder_class_init,
NULL,
cdata,
sizeof (GstNvH264Encoder),
0,
(GInstanceInitFunc) gst_nv_h264_encoder_init,
};
type_name = g_strdup ("GstNvCudaH264Enc");
feature_name = g_strdup ("nvcudah264enc");
gint index = 0;
while (g_type_from_name (type_name)) {
index++;
g_free (type_name);
g_free (feature_name);
type_name = g_strdup_printf ("GstNvCudaH264Device%dEnc", index);
feature_name = g_strdup_printf ("nvcudah264device%denc", index);
}
type = g_type_register_static (GST_TYPE_NV_ENCODER, type_name,
&type_info, (GTypeFlags) 0);
if (rank > 0 && index != 0)
rank--;
if (!gst_element_register (plugin, feature_name, rank, type))
GST_WARNING ("Failed to register plugin '%s'", type_name);
g_free (type_name);
g_free (feature_name);
}
#ifdef GST_CUDA_HAS_D3D
nvcodec: Add new Direct3D11/CUDA mode encoder implementation Adding new encoder elements nvd3d11{h264,h265}enc for Direct3D11 input support and re-written nvcuda{h264,h265}enc elements. Newly writeen elements have some differences compared with old nv{h264,h265}enc including non-backward compatible changes. * RGBA is not a supported input format any more: New elements will support only YUV formats to avoid implicit conversion done by hardware. Ideally it should be done by upstream element in order to have more control on it. Moreover, RGBA support can cause redundant RGBA -> YUV conversion if multiple encoders are used for the same RGBA input * Subsampled planar format support is dropped: I420 and YV12 format are not supported formats for Direct3D11. Although it's supported in CUDA mode, it's not a hardware friendly memory layout and it will waste GPU memory since UV planes will have large padding due to the memory layout requirement of NVENC. * GL support is dropped: Similar to the RGBA case, GL support in encoder would be suboptimal if GL input is used by multiple encoders, because each encoder will copy GL memory into CUDA memory. Upstream cudaupload element can be used for GL <-> CUDA interop instead. * No more pre-allocation of encoder input surfaces. New implementation will use input CUDA memory without copy (zero-copy) or will copy into a NVENC's input buffer struct in case of system memory input. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1997>
2022-03-15 19:58:16 +00:00
void
gst_nv_h264_encoder_register_d3d11 (GstPlugin * plugin, GstD3D11Device * device,
guint rank)
{
NV_ENC_OPEN_ENCODE_SESSION_EX_PARAMS session_params = { 0, };
gpointer session;
NVENCSTATUS status;
GstNvH264EncoderClassData *cdata;
GST_DEBUG_CATEGORY_INIT (gst_nv_h264_encoder_debug, "nvh264encoder", 0,
"nvh264encoder");
session_params.version =
gst_nvenc_get_open_encode_session_ex_params_version ();
session_params.deviceType = NV_ENC_DEVICE_TYPE_DIRECTX;
session_params.device = gst_d3d11_device_get_device_handle (device);
session_params.apiVersion = gst_nvenc_get_api_version ();
status = NvEncOpenEncodeSessionEx (&session_params, &session);
if (status != NV_ENC_SUCCESS) {
GST_WARNING_OBJECT (device, "Failed to open session");
return;
}
cdata = gst_nv_h264_encoder_create_class_data (GST_OBJECT (device), session,
TRUE);
NvEncDestroyEncoder (session);
if (!cdata)
return;
GType type;
gchar *type_name;
gchar *feature_name;
GTypeInfo type_info = {
sizeof (GstNvH264EncoderClass),
NULL,
NULL,
(GClassInitFunc) gst_nv_h264_encoder_class_init,
NULL,
cdata,
sizeof (GstNvH264Encoder),
0,
(GInstanceInitFunc) gst_nv_h264_encoder_init,
};
type_name = g_strdup ("GstNvD3D11H264Enc");
feature_name = g_strdup ("nvd3d11h264enc");
gint index = 0;
while (g_type_from_name (type_name)) {
index++;
g_free (type_name);
g_free (feature_name);
type_name = g_strdup_printf ("GstNvD3D11H264Device%dEnc", index);
feature_name = g_strdup_printf ("nvd3d11h264device%denc", index);
}
type = g_type_register_static (GST_TYPE_NV_ENCODER, type_name,
&type_info, (GTypeFlags) 0);
if (rank > 0 && index != 0)
rank--;
if (!gst_element_register (plugin, feature_name, rank, type))
GST_WARNING ("Failed to register plugin '%s'", type_name);
g_free (type_name);
g_free (feature_name);
}
#endif