mirrors/gstreamer
mirrors/gstreamer
1
1
Fork 0
mirror of https://gitlab.freedesktop.org/gstreamer/gstreamer.git synced 2024-12-30 12:10:37 +00:00
Code Issues 1 Projects Releases Wiki Activity
gstreamer/subprojects/gst-plugins-bad/sys/qsv/gstqsvh264enc.cpp

1925 lines
63 KiB
C++
Raw Normal View History

qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
/* GStreamer
* Copyright (C) 2021 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 "gstqsvh264enc.h"
#include <gst/base/gstbytewriter.h>
#include <gst/codecparsers/gsth264parser.h>
#include <vector>
#include <string>
#include <set>
#include <string.h>
#ifdef G_OS_WIN32
#include <gst/d3d11/gstd3d11.h>
#else
qsvencoder: Add support for VA memory Use VA allocator and buffer pool implementation for zero-copy encoding with upstream VA elements Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2030>
2022-03-25 13:04:05 +00:00
#include <gst/va/gstva.h>
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
#endif
GST_DEBUG_CATEGORY_EXTERN (gst_qsv_h264_enc_debug);
#define GST_CAT_DEFAULT gst_qsv_h264_enc_debug
typedef enum
{
GST_QSV_H264_ENC_SEI_INSERT,
GST_QSV_H264_ENC_SEI_INSERT_AND_DROP,
GST_QSV_H264_ENC_SEI_DISABLED,
} GstQsvH264EncSeiInsertMode;
#define GST_TYPE_QSV_H264_ENC_SEI_INSERT_MODE (gst_qsv_h264_enc_sei_insert_mode_get_type ())
static GType
gst_qsv_h264_enc_sei_insert_mode_get_type (void)
{
static GType sei_insert_mode_type = 0;
static const GEnumValue insert_modes[] = {
{GST_QSV_H264_ENC_SEI_INSERT, "Insert SEI", "insert"},
{GST_QSV_H264_ENC_SEI_INSERT_AND_DROP,
"Insert SEI and remove corresponding meta from output buffer",
"insert-and-drop"},
{GST_QSV_H264_ENC_SEI_DISABLED, "Disable SEI insertion", "disabled"},
{0, nullptr, nullptr}
};
if (g_once_init_enter (&sei_insert_mode_type)) {
GType type =
g_enum_register_static ("GstQsvH264EncSeiInsertMode", insert_modes);
g_once_init_leave (&sei_insert_mode_type, type);
}
return sei_insert_mode_type;
}
#define GST_TYPE_QSV_H264_ENC_RATE_CONTROL (gst_qsv_h264_enc_rate_control_get_type ())
static GType
gst_qsv_h264_enc_rate_control_get_type (void)
{
static GType rate_control_type = 0;
static const GEnumValue rate_controls[] = {
{MFX_RATECONTROL_CBR, "Constant Bitrate", "cbr"},
{MFX_RATECONTROL_VBR, "Variable Bitrate", "vbr"},
{MFX_RATECONTROL_CQP, "Constant Quantizer", "cqp"},
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
{MFX_RATECONTROL_AVBR, "Average Variable Bitrate", "avbr"},
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
{MFX_RATECONTROL_LA, "VBR with look ahead (Non HRD compliant)", "la_vbr"},
{MFX_RATECONTROL_ICQ, "Intelligent CQP", "icq"},
{MFX_RATECONTROL_VCM, "Video Conferencing Mode (Non HRD compliant)", "vcm"},
{MFX_RATECONTROL_LA_ICQ, "Intelligent CQP with LA (Non HRD compliant)",
"la_icq"},
{MFX_RATECONTROL_LA_HRD, "HRD compliant LA", "la_hrd"},
{MFX_RATECONTROL_QVBR, "VBR with CQP", "qvbr"},
{0, nullptr, nullptr}
};
if (g_once_init_enter (&rate_control_type)) {
GType type =
g_enum_register_static ("GstQsvH264EncRateControl", rate_controls);
g_once_init_leave (&rate_control_type, type);
}
return rate_control_type;
}
#define GST_TYPE_QSV_H264_ENC_RC_LOOKAHEAD_DS (gst_qsv_h264_enc_rc_lookahead_ds_get_type ())
static GType
gst_qsv_h264_enc_rc_lookahead_ds_get_type (void)
{
static GType rc_lookahead_ds_type = 0;
static const GEnumValue rc_lookahead_ds[] = {
{MFX_LOOKAHEAD_DS_UNKNOWN, "Unknown", "unknown"},
{MFX_LOOKAHEAD_DS_OFF, "Do not use down sampling", "off"},
{MFX_LOOKAHEAD_DS_2x,
"Down sample frames two times before estimation", "2x"},
{MFX_LOOKAHEAD_DS_4x,
"Down sample frames four times before estimation", "4x"},
{0, nullptr, nullptr}
};
if (g_once_init_enter (&rc_lookahead_ds_type)) {
GType type =
g_enum_register_static ("GstQsvH264EncRCLookAheadDS", rc_lookahead_ds);
g_once_init_leave (&rc_lookahead_ds_type, type);
}
return rc_lookahead_ds_type;
}
enum
{
PROP_0,
PROP_ADAPTER_LUID,
PROP_DEVICE_PATH,
PROP_CABAC,
PROP_MIN_QP_I,
PROP_MIN_QP_P,
PROP_MIN_QP_B,
PROP_MAX_QP_I,
PROP_MAX_QP_P,
PROP_MAX_QP_B,
PROP_QP_I,
PROP_QP_P,
PROP_QP_B,
PROP_GOP_SIZE,
PROP_I_FRAMES,
PROP_B_FRAMES,
PROP_REF_FRAMES,
PROP_BITRATE,
PROP_MAX_BITRATE,
PROP_RATE_CONTROL,
PROP_RC_LOOKAHEAD,
PROP_RC_LOOKAHEAD_DS,
PROP_AVBR_ACCURACY,
PROP_AVBR_CONVERGENCE,
PROP_ICQ_QUALITY,
PROP_QVBR_QUALITY,
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
PROP_DISABLE_HRD_CONFORMANCE,
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
PROP_CC_INSERT,
};
#define DEFAULT_CABAC MFX_CODINGOPTION_UNKNOWN
#define DEFAULT_QP 0
#define DEFAULT_GOP_SIZE 0
#define DEFAULT_I_FRAMES 0
#define DEFAULT_B_FRAMES 0
#define DEFAULT_REF_FRAMES 2
#define DEFAULT_BITRATE 2000
#define DEFAULT_MAX_BITRATE 0
#define DEFAULT_RATE_CONTROL MFX_RATECONTROL_CBR
#define DEFAULT_RC_LOOKAHEAD 10
#define DEFAULT_RC_LOOKAHEAD_DS MFX_LOOKAHEAD_DS_UNKNOWN
#define DEFAULT_AVBR_ACCURACY 0
#define DEFAULT_AVBR_CONVERGENCE 0
#define DEFAULT_IQC_QUALITY 0
#define DEFAULT_QVBR_QUALITY 0
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
#define DEFAULT_DISABLE_HRD_CONFORMANCE FALSE
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
#define DEFAULT_CC_INSERT GST_QSV_H264_ENC_SEI_INSERT
typedef struct _GstQsvH264EncClassData
{
GstCaps *sink_caps;
GstCaps *src_caps;
guint impl_index;
gint64 adapter_luid;
gchar *display_path;
} GstQsvH264EncClassData;
typedef struct _GstQsvH264Enc
{
GstQsvEncoder parent;
mfxExtVideoSignalInfo signal_info;
mfxExtCodingOption option;
mfxExtCodingOption2 option2;
mfxExtCodingOption3 option3;
gboolean packetized;
GstH264NalParser *parser;
mfxU16 profile;
GMutex prop_lock;
/* protected by prop_lock */
gboolean bitrate_updated;
gboolean property_updated;
/* properties */
mfxU16 cabac;
guint min_qp_i;
guint min_qp_p;
guint min_qp_b;
guint max_qp_i;
guint max_qp_p;
guint max_qp_b;
guint qp_i;
guint qp_p;
guint qp_b;
guint gop_size;
guint iframes;
guint bframes;
guint ref_frames;
guint bitrate;
guint max_bitrate;
mfxU16 rate_control;
guint rc_lookahead;
mfxU16 rc_lookahead_ds;
guint avbr_accuracy;
guint avbr_convergence;
guint icq_quality;
guint qvbr_quality;
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
gboolean disable_hrd_conformance;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
GstQsvH264EncSeiInsertMode cc_insert;
} GstQsvH264Enc;
typedef struct _GstQsvH264EncClass
{
GstQsvEncoderClass parent_class;
} GstQsvH264EncClass;
static GstElementClass *parent_class = nullptr;
#define GST_QSV_H264_ENC(object) ((GstQsvH264Enc *) (object))
#define GST_QSV_H264_ENC_GET_CLASS(object) \
(G_TYPE_INSTANCE_GET_CLASS ((object),G_TYPE_FROM_INSTANCE (object),GstQsvH264EncClass))
static void gst_qsv_h264_enc_finalize (GObject * object);
static void gst_qsv_h264_enc_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec);
static void gst_qsv_h264_enc_get_property (GObject * object, guint prop_id,
GValue * value, GParamSpec * pspec);
static gboolean gst_qsv_h264_enc_start (GstVideoEncoder * encoder);
static gboolean gst_qsv_h264_enc_transform_meta (GstVideoEncoder * encoder,
GstVideoCodecFrame * frame, GstMeta * meta);
static GstCaps *gst_qsv_h264_enc_getcaps (GstVideoEncoder * encoder,
GstCaps * filter);
static gboolean gst_qsv_h264_enc_set_format (GstQsvEncoder * encoder,
GstVideoCodecState * state, mfxVideoParam * param,
GPtrArray * extra_params);
static gboolean gst_qsv_h264_enc_set_output_state (GstQsvEncoder * encoder,
GstVideoCodecState * state, mfxSession session);
static gboolean gst_qsv_h264_enc_attach_payload (GstQsvEncoder * encoder,
GstVideoCodecFrame * frame, GPtrArray * payload);
static GstBuffer *gst_qsv_h264_enc_create_output_buffer (GstQsvEncoder *
encoder, mfxBitstream * bitstream);
static GstQsvEncoderReconfigure
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
gst_qsv_h264_enc_check_reconfigure (GstQsvEncoder * encoder, mfxSession session,
mfxVideoParam * param, GPtrArray * extra_params);
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
static void
gst_qsv_h264_enc_class_init (GstQsvH264EncClass * klass, gpointer data)
{
GObjectClass *object_class = G_OBJECT_CLASS (klass);
GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
GstVideoEncoderClass *encoder_class = GST_VIDEO_ENCODER_CLASS (klass);
GstQsvEncoderClass *qsvenc_class = GST_QSV_ENCODER_CLASS (klass);
GstQsvH264EncClassData *cdata = (GstQsvH264EncClassData *) data;
qsvenc_class->codec_id = MFX_CODEC_AVC;
qsvenc_class->impl_index = cdata->impl_index;
qsvenc_class->adapter_luid = cdata->adapter_luid;
qsv: Remove strcpy for DRM device path Simplify code. It's just one-time allocation, so don't need to worry about leaking. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1835>
2022-03-02 19:22:15 +00:00
qsvenc_class->display_path = cdata->display_path;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
object_class->finalize = gst_qsv_h264_enc_finalize;
object_class->set_property = gst_qsv_h264_enc_set_property;
object_class->get_property = gst_qsv_h264_enc_get_property;
#ifdef G_OS_WIN32
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 created device",
G_MININT64, G_MAXINT64, qsvenc_class->adapter_luid,
(GParamFlags) (GST_PARAM_CONDITIONALLY_AVAILABLE | G_PARAM_READABLE |
G_PARAM_STATIC_STRINGS)));
#else
g_object_class_install_property (object_class, PROP_DEVICE_PATH,
g_param_spec_string ("device-path", "Device Path",
"DRM device path", cdata->display_path,
(GParamFlags) (GST_PARAM_CONDITIONALLY_AVAILABLE |
G_PARAM_READABLE | G_PARAM_STATIC_STRINGS)));
#endif
g_object_class_install_property (object_class, PROP_CABAC,
g_param_spec_enum ("cabac", "Cabac", "Enables CABAC entropy coding",
GST_TYPE_QSV_CODING_OPTION, DEFAULT_CABAC,
(GParamFlags) (G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_MIN_QP_I,
g_param_spec_uint ("min-qpi", "Min QP I",
"Minimum allowed QP value for I-frame types (0: no limitations)",
0, 51, DEFAULT_QP, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_MIN_QP_P,
g_param_spec_uint ("min-qpp", "Min QP P",
"Minimum allowed QP value for P-frame types (0: no limitations)",
0, 51, DEFAULT_QP, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_MIN_QP_B,
g_param_spec_uint ("min-qpb", "Min QP B",
"Minimum allowed QP value for B-frame types (0: no limitations)",
0, 51, DEFAULT_QP, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_MAX_QP_I,
g_param_spec_uint ("max-qpi", "Max QP I",
"Maximum allowed QP value for I-frame types (0: no limitations)",
0, 51, DEFAULT_QP, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_MAX_QP_P,
g_param_spec_uint ("max-qpp", "Max QP P",
"Maximum allowed QP value for P-frame types (0: no limitations)",
0, 51, DEFAULT_QP, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_MAX_QP_B,
g_param_spec_uint ("max-qpb", "Max QP B",
"Maximum allowed QP value for B-frame types (0: no limitations)",
0, 51, DEFAULT_QP, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_QP_I,
g_param_spec_uint ("qpi", "QP I",
"Constant quantizer for I frames (0: no limitations)",
0, 51, DEFAULT_QP, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_QP_P,
g_param_spec_uint ("qpp", "QP P",
"Constant quantizer for P frames (0: no limitations)",
0, 51, DEFAULT_QP, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_QP_B,
g_param_spec_uint ("qpb", "QP B",
"Constant quantizer for B frames (0: no limitations)",
0, 51, DEFAULT_QP, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_GOP_SIZE,
g_param_spec_uint ("gop-size", "GOP Size",
"Number of pictures within a GOP (0: unspecified)",
0, G_MAXINT, DEFAULT_GOP_SIZE, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_I_FRAMES,
g_param_spec_uint ("i-frames", "I Frames",
"Number of I frames between IDR frames"
"(0: every I frame is an IDR frame)",
0, G_MAXINT, DEFAULT_I_FRAMES, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_B_FRAMES,
g_param_spec_uint ("b-frames", "B Frames",
"Number of B frames between I and P frames",
0, G_MAXINT, DEFAULT_B_FRAMES, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_REF_FRAMES,
g_param_spec_uint ("ref-frames", "Reference Frames",
"Number of reference frames (0: unspecified)",
0, 16, DEFAULT_REF_FRAMES, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_BITRATE,
g_param_spec_uint ("bitrate", "Bitrate",
"Target bitrate in kbit/sec, Ignored when selected rate-control mode "
"is constant QP variants (i.e., \"cqp\", \"icq\", and \"la_icq\")",
0, G_MAXINT, DEFAULT_BITRATE, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_MAX_BITRATE,
g_param_spec_uint ("max-bitrate", "Max Bitrate",
"Maximum bitrate in kbit/sec, Ignored when selected rate-control mode "
"is constant QP variants (i.e., \"cqp\", \"icq\", and \"la_icq\")",
0, G_MAXINT, DEFAULT_MAX_BITRATE, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_RATE_CONTROL,
g_param_spec_enum ("rate-control", "Rate Control",
"Rate Control Method", GST_TYPE_QSV_H264_ENC_RATE_CONTROL,
DEFAULT_RATE_CONTROL,
(GParamFlags) (G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_RC_LOOKAHEAD,
g_param_spec_uint ("rc-lookahead", "Rate Control Look-ahead",
"Number of frames to look ahead for Rate Control, used for "
"\"la_vbr\", \"la_icq\", and \"la_hrd\" rate-control modes",
10, 100, DEFAULT_RC_LOOKAHEAD, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_RC_LOOKAHEAD_DS,
g_param_spec_enum ("rc-lookahead-ds",
"Rate Control Look-ahead Downsampling",
"Downsampling method in look-ahead rate control",
GST_TYPE_QSV_H264_ENC_RC_LOOKAHEAD_DS, DEFAULT_RC_LOOKAHEAD_DS,
(GParamFlags) (G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_AVBR_ACCURACY,
g_param_spec_uint ("avbr-accuracy", "AVBR Accuracy",
"AVBR Accuracy in the unit of tenth of percent",
0, G_MAXUINT16, DEFAULT_AVBR_ACCURACY, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_AVBR_CONVERGENCE,
g_param_spec_uint ("avbr-convergence", "AVBR Convergence",
"AVBR Convergence in the unit of 100 frames",
0, G_MAXUINT16, DEFAULT_AVBR_ACCURACY, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_ICQ_QUALITY,
g_param_spec_uint ("icq-quality", "ICQ Quality",
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
"Intelligent Constant Quality for \"icq\" rate-control (0: default)",
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
0, 51, DEFAULT_IQC_QUALITY, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
g_object_class_install_property (object_class, PROP_QVBR_QUALITY,
g_param_spec_uint ("qvbr-quality", "QVBR Quality",
"Quality level used for \"qvbr\" rate-control mode (0: default)",
0, 51, DEFAULT_QVBR_QUALITY, (GParamFlags)
(G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
g_object_class_install_property (object_class, PROP_DISABLE_HRD_CONFORMANCE,
g_param_spec_boolean ("disable-hrd-conformance",
"Disable HRD Conformance", "Allow NAL HRD non-conformant stream",
DEFAULT_DISABLE_HRD_CONFORMANCE,
(GParamFlags) (G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
g_object_class_install_property (object_class, PROP_CC_INSERT,
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
g_param_spec_enum ("cc-insert", "Closed Caption Insert",
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
"Closed Caption Insert mode. "
"Only CEA-708 RAW format is supported for now",
GST_TYPE_QSV_H264_ENC_SEI_INSERT_MODE, DEFAULT_CC_INSERT,
(GParamFlags) (G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS)));
parent_class = (GstElementClass *) g_type_class_peek_parent (klass);
gst_element_class_set_static_metadata (element_class,
"Intel Quick Sync Video H.264 Encoder",
"Codec/Encoder/Video/Hardware",
"Intel Quick Sync Video H.264 Encoder",
"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));
encoder_class->start = GST_DEBUG_FUNCPTR (gst_qsv_h264_enc_start);
encoder_class->transform_meta =
GST_DEBUG_FUNCPTR (gst_qsv_h264_enc_transform_meta);
encoder_class->getcaps = GST_DEBUG_FUNCPTR (gst_qsv_h264_enc_getcaps);
qsvenc_class->set_format = GST_DEBUG_FUNCPTR (gst_qsv_h264_enc_set_format);
qsvenc_class->set_output_state =
GST_DEBUG_FUNCPTR (gst_qsv_h264_enc_set_output_state);
qsvenc_class->attach_payload =
GST_DEBUG_FUNCPTR (gst_qsv_h264_enc_attach_payload);
qsvenc_class->create_output_buffer =
GST_DEBUG_FUNCPTR (gst_qsv_h264_enc_create_output_buffer);
qsvenc_class->check_reconfigure =
GST_DEBUG_FUNCPTR (gst_qsv_h264_enc_check_reconfigure);
gst_caps_unref (cdata->sink_caps);
gst_caps_unref (cdata->src_caps);
g_free (cdata);
}
static void
gst_qsv_h264_enc_init (GstQsvH264Enc * self)
{
self->cabac = DEFAULT_CABAC;
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_p = DEFAULT_QP;
self->qp_i = DEFAULT_QP;
self->qp_p = DEFAULT_QP;
self->qp_b = DEFAULT_QP;
self->gop_size = DEFAULT_GOP_SIZE;
self->iframes = DEFAULT_I_FRAMES;
self->bframes = DEFAULT_B_FRAMES;
self->ref_frames = DEFAULT_REF_FRAMES;
self->bitrate = DEFAULT_BITRATE;
self->max_bitrate = DEFAULT_MAX_BITRATE;
self->rate_control = DEFAULT_RATE_CONTROL;
self->rc_lookahead = DEFAULT_RC_LOOKAHEAD;
self->rc_lookahead_ds = DEFAULT_RC_LOOKAHEAD_DS;
self->avbr_accuracy = DEFAULT_AVBR_ACCURACY;
self->avbr_convergence = DEFAULT_AVBR_CONVERGENCE;
self->icq_quality = DEFAULT_IQC_QUALITY;
self->qvbr_quality = DEFAULT_QVBR_QUALITY;
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
self->disable_hrd_conformance = DEFAULT_DISABLE_HRD_CONFORMANCE;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
self->cc_insert = DEFAULT_CC_INSERT;
g_mutex_init (&self->prop_lock);
self->parser = gst_h264_nal_parser_new ();
}
static void
gst_qsv_h264_enc_finalize (GObject * object)
{
GstQsvH264Enc *self = GST_QSV_H264_ENC (object);
g_mutex_clear (&self->prop_lock);
gst_h264_nal_parser_free (self->parser);
G_OBJECT_CLASS (parent_class)->finalize (object);
}
static void
gst_qsv_h264_enc_check_update_uint (GstQsvH264Enc * self, guint * old_val,
guint new_val, gboolean is_bitrate_param)
{
if (*old_val == new_val)
return;
g_mutex_lock (&self->prop_lock);
*old_val = new_val;
if (is_bitrate_param)
self->bitrate_updated = TRUE;
else
self->property_updated = TRUE;
g_mutex_unlock (&self->prop_lock);
}
static void
gst_qsv_h264_enc_check_update_enum (GstQsvH264Enc * self, mfxU16 * old_val,
gint new_val)
{
if (*old_val == (mfxU16) new_val)
return;
g_mutex_lock (&self->prop_lock);
*old_val = (mfxU16) new_val;
self->property_updated = TRUE;
g_mutex_unlock (&self->prop_lock);
}
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
static void
gst_qsv_h264_enc_check_update_boolean (GstQsvH264Enc * self, gboolean * old_val,
gboolean new_val)
{
if (*old_val == new_val)
return;
g_mutex_lock (&self->prop_lock);
*old_val = new_val;
self->property_updated = TRUE;
g_mutex_unlock (&self->prop_lock);
}
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
static void
gst_qsv_h264_enc_set_property (GObject * object, guint prop_id,
const GValue * value, GParamSpec * pspec)
{
GstQsvH264Enc *self = GST_QSV_H264_ENC (object);
switch (prop_id) {
case PROP_CABAC:
gst_qsv_h264_enc_check_update_enum (self, &self->cabac,
g_value_get_enum (value));
break;
/* MIN/MAX QP change requires new sequence */
case PROP_MIN_QP_I:
gst_qsv_h264_enc_check_update_uint (self, &self->min_qp_i,
g_value_get_uint (value), FALSE);
break;
case PROP_MIN_QP_P:
gst_qsv_h264_enc_check_update_uint (self, &self->min_qp_p,
g_value_get_uint (value), FALSE);
break;
case PROP_MIN_QP_B:
gst_qsv_h264_enc_check_update_uint (self, &self->min_qp_b,
g_value_get_uint (value), FALSE);
break;
case PROP_MAX_QP_I:
gst_qsv_h264_enc_check_update_uint (self, &self->max_qp_i,
g_value_get_uint (value), FALSE);
break;
case PROP_MAX_QP_P:
gst_qsv_h264_enc_check_update_uint (self, &self->max_qp_p,
g_value_get_uint (value), FALSE);
break;
case PROP_MAX_QP_B:
gst_qsv_h264_enc_check_update_uint (self, &self->max_qp_b,
g_value_get_uint (value), FALSE);
break;
case PROP_QP_I:
gst_qsv_h264_enc_check_update_uint (self, &self->qp_i,
g_value_get_uint (value), TRUE);
break;
case PROP_QP_P:
gst_qsv_h264_enc_check_update_uint (self, &self->qp_p,
g_value_get_uint (value), TRUE);
break;
case PROP_QP_B:
gst_qsv_h264_enc_check_update_uint (self, &self->qp_b,
g_value_get_uint (value), TRUE);
break;
case PROP_GOP_SIZE:
gst_qsv_h264_enc_check_update_uint (self, &self->gop_size,
g_value_get_uint (value), FALSE);
break;
case PROP_I_FRAMES:
gst_qsv_h264_enc_check_update_uint (self, &self->iframes,
g_value_get_uint (value), FALSE);
break;
case PROP_B_FRAMES:
gst_qsv_h264_enc_check_update_uint (self, &self->bframes,
g_value_get_uint (value), FALSE);
break;
case PROP_REF_FRAMES:
gst_qsv_h264_enc_check_update_uint (self, &self->ref_frames,
g_value_get_uint (value), FALSE);
break;
case PROP_BITRATE:
gst_qsv_h264_enc_check_update_uint (self, &self->bitrate,
g_value_get_uint (value), TRUE);
break;
case PROP_MAX_BITRATE:
gst_qsv_h264_enc_check_update_uint (self, &self->max_bitrate,
g_value_get_uint (value), TRUE);
break;
case PROP_RATE_CONTROL:
gst_qsv_h264_enc_check_update_enum (self, &self->rate_control,
g_value_get_enum (value));
break;
case PROP_RC_LOOKAHEAD:
gst_qsv_h264_enc_check_update_uint (self, &self->rc_lookahead,
g_value_get_uint (value), FALSE);
break;
case PROP_RC_LOOKAHEAD_DS:
gst_qsv_h264_enc_check_update_enum (self, &self->rc_lookahead_ds,
g_value_get_enum (value));
break;
case PROP_AVBR_ACCURACY:
gst_qsv_h264_enc_check_update_uint (self, &self->avbr_accuracy,
g_value_get_uint (value), FALSE);
break;
case PROP_AVBR_CONVERGENCE:
gst_qsv_h264_enc_check_update_uint (self, &self->avbr_convergence,
g_value_get_uint (value), FALSE);
break;
case PROP_ICQ_QUALITY:
gst_qsv_h264_enc_check_update_uint (self, &self->icq_quality,
g_value_get_uint (value), FALSE);
break;
case PROP_QVBR_QUALITY:
gst_qsv_h264_enc_check_update_uint (self, &self->qvbr_quality,
g_value_get_uint (value), FALSE);
break;
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
case PROP_DISABLE_HRD_CONFORMANCE:
gst_qsv_h264_enc_check_update_boolean (self,
&self->disable_hrd_conformance, g_value_get_boolean (value));
break;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
case PROP_CC_INSERT:
/* This property is unrelated to encoder-reset */
self->cc_insert = (GstQsvH264EncSeiInsertMode) g_value_get_enum (value);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
static void
gst_qsv_h264_enc_get_property (GObject * object, guint prop_id, GValue * value,
GParamSpec * pspec)
{
GstQsvH264Enc *self = GST_QSV_H264_ENC (object);
GstQsvEncoderClass *klass = GST_QSV_ENCODER_GET_CLASS (self);
switch (prop_id) {
case PROP_ADAPTER_LUID:
g_value_set_int64 (value, klass->adapter_luid);
break;
case PROP_DEVICE_PATH:
g_value_set_string (value, klass->display_path);
break;
case PROP_CABAC:
g_value_set_enum (value, self->cabac);
break;
case PROP_MIN_QP_I:
g_value_set_uint (value, self->min_qp_i);
break;
case PROP_MIN_QP_P:
g_value_set_uint (value, self->min_qp_p);
break;
case PROP_MIN_QP_B:
g_value_set_uint (value, self->min_qp_b);
break;
case PROP_MAX_QP_I:
g_value_set_uint (value, self->max_qp_i);
break;
case PROP_MAX_QP_P:
g_value_set_uint (value, self->max_qp_p);
break;
case PROP_MAX_QP_B:
g_value_set_uint (value, self->max_qp_b);
break;
case PROP_QP_I:
g_value_set_uint (value, self->qp_i);
break;
case PROP_QP_P:
g_value_set_uint (value, self->qp_p);
break;
case PROP_QP_B:
g_value_set_uint (value, self->qp_b);
break;
case PROP_GOP_SIZE:
g_value_set_uint (value, self->gop_size);
break;
case PROP_I_FRAMES:
g_value_set_uint (value, self->iframes);
break;
case PROP_B_FRAMES:
g_value_set_uint (value, self->bframes);
break;
case PROP_REF_FRAMES:
g_value_set_uint (value, self->ref_frames);
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_RATE_CONTROL:
g_value_set_enum (value, self->rate_control);
break;
case PROP_RC_LOOKAHEAD:
g_value_set_uint (value, self->rc_lookahead);
break;
case PROP_RC_LOOKAHEAD_DS:
g_value_set_enum (value, self->rc_lookahead_ds);
break;
case PROP_AVBR_ACCURACY:
g_value_set_uint (value, self->avbr_accuracy);
break;
case PROP_AVBR_CONVERGENCE:
g_value_set_uint (value, self->avbr_convergence);
break;
case PROP_ICQ_QUALITY:
g_value_set_uint (value, self->icq_quality);
break;
case PROP_QVBR_QUALITY:
g_value_set_uint (value, self->qvbr_quality);
break;
case PROP_CC_INSERT:
g_value_set_enum (value, self->cc_insert);
break;
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
case PROP_DISABLE_HRD_CONFORMANCE:
g_value_set_boolean (value, self->disable_hrd_conformance);
break;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
static gboolean
gst_qsv_h264_enc_start (GstVideoEncoder * encoder)
{
/* To avoid negative DTS when B frame is enabled */
gst_video_encoder_set_min_pts (encoder, GST_SECOND * 60 * 60 * 1000);
return TRUE;
}
static gboolean
gst_qsv_h264_enc_transform_meta (GstVideoEncoder * encoder,
GstVideoCodecFrame * frame, GstMeta * meta)
{
GstQsvH264Enc *self = GST_QSV_H264_ENC (encoder);
GstVideoCaptionMeta *cc_meta;
/* We need to handle only case CC meta should be dropped */
if (self->cc_insert != GST_QSV_H264_ENC_SEI_INSERT_AND_DROP)
goto out;
if (meta->info->api != GST_VIDEO_CAPTION_META_API_TYPE)
goto out;
cc_meta = (GstVideoCaptionMeta *) meta;
if (cc_meta->caption_type != GST_VIDEO_CAPTION_TYPE_CEA708_RAW)
goto out;
/* Don't copy this meta into output buffer */
return FALSE;
out:
return GST_VIDEO_ENCODER_CLASS (parent_class)->transform_meta (encoder,
frame, meta);
}
static GstCaps *
gst_qsv_h264_enc_getcaps (GstVideoEncoder * encoder, GstCaps * filter)
{
GstQsvH264Enc *self = GST_QSV_H264_ENC (encoder);
GstCaps *allowed_caps;
GstCaps *template_caps;
GstCaps *supported_caps;
std::set < std::string > downstream_profiles;
allowed_caps = gst_pad_get_allowed_caps (encoder->srcpad);
/* Shouldn't be any or empty though, just return template caps in this case */
if (!allowed_caps || gst_caps_is_empty (allowed_caps) ||
gst_caps_is_any (allowed_caps)) {
gst_clear_caps (&allowed_caps);
return gst_video_encoder_proxy_getcaps (encoder, nullptr, filter);
}
/* Check if downstream specified profile explicitly, then filter out
* incompatible interlaced field */
for (guint i = 0; i < gst_caps_get_size (allowed_caps); i++) {
const GValue *profile_value;
const gchar *profile;
GstStructure *s;
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);
}
}
GST_DEBUG_OBJECT (self, "Downstream specified %" G_GSIZE_FORMAT " profiles",
downstream_profiles.size ());
/* Caps returned by gst_pad_get_allowed_caps() should hold profile field
* already */
if (downstream_profiles.size () == 0) {
GST_WARNING_OBJECT (self,
"Allowed caps holds no profile field %" GST_PTR_FORMAT, allowed_caps);
qsvencoder: Clear caps after debug print Otherwise debug function will print (null) for alread cleared pointer Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2012>
2022-03-23 17:52:45 +00:00
gst_clear_caps (&allowed_caps);
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
return gst_video_encoder_proxy_getcaps (encoder, nullptr, filter);
}
qsvencoder: Clear caps after debug print Otherwise debug function will print (null) for alread cleared pointer Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2012>
2022-03-23 17:52:45 +00:00
gst_clear_caps (&allowed_caps);
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
/* Profile allows interlaced? */
/* *INDENT-OFF* */
gboolean can_support_interlaced = FALSE;
for (const auto &iter: downstream_profiles) {
if (iter == "high" || iter == "main") {
can_support_interlaced = TRUE;
break;
}
}
/* *INDENT-ON* */
GST_DEBUG_OBJECT (self, "Downstream %s support interlaced format",
can_support_interlaced ? "can" : "cannot");
if (can_support_interlaced) {
/* No special handling is needed */
return gst_video_encoder_proxy_getcaps (encoder, nullptr, filter);
}
template_caps = gst_pad_get_pad_template_caps (encoder->sinkpad);
template_caps = gst_caps_make_writable (template_caps);
gst_caps_set_simple (template_caps, "interlace-mode", G_TYPE_STRING,
"progressive", nullptr);
supported_caps = gst_video_encoder_proxy_getcaps (encoder,
template_caps, filter);
gst_caps_unref (template_caps);
GST_DEBUG_OBJECT (self, "Returning %" GST_PTR_FORMAT, supported_caps);
return supported_caps;
}
typedef struct
{
mfxU16 profile;
const gchar *profile_str;
} H264Profile;
static const H264Profile profile_map[] = {
{MFX_PROFILE_AVC_HIGH, "high"},
{MFX_PROFILE_AVC_MAIN, "main"},
{MFX_PROFILE_AVC_CONSTRAINED_BASELINE, "constrained-baseline"},
{MFX_PROFILE_AVC_PROGRESSIVE_HIGH, "progressive-high"},
{MFX_PROFILE_AVC_CONSTRAINED_HIGH, "constrained-high"},
{MFX_PROFILE_AVC_BASELINE, "baseline"}
};
static const gchar *
gst_qsv_h264_profile_to_string (mfxU16 profile)
{
for (guint i = 0; i < G_N_ELEMENTS (profile_map); i++) {
if (profile_map[i].profile == profile)
return profile_map[i].profile_str;
}
return nullptr;
}
static mfxU16
gst_qsv_h264_profile_string_to_value (const gchar * profile_str)
{
for (guint i = 0; i < G_N_ELEMENTS (profile_map); i++) {
if (g_strcmp0 (profile_map[i].profile_str, profile_str) == 0)
return profile_map[i].profile;
}
return MFX_PROFILE_UNKNOWN;
}
static void
gst_qsv_h264_enc_init_extra_params (GstQsvH264Enc * self)
{
memset (&self->signal_info, 0, sizeof (mfxExtVideoSignalInfo));
memset (&self->option, 0, sizeof (mfxExtCodingOption));
memset (&self->option2, 0, sizeof (mfxExtCodingOption2));
memset (&self->option3, 0, sizeof (mfxExtCodingOption3));
self->signal_info.Header.BufferId = MFX_EXTBUFF_VIDEO_SIGNAL_INFO;
self->signal_info.Header.BufferSz = sizeof (mfxExtVideoSignalInfo);
self->option.Header.BufferId = MFX_EXTBUFF_CODING_OPTION;
self->option.Header.BufferSz = sizeof (mfxExtCodingOption);
self->option2.Header.BufferId = MFX_EXTBUFF_CODING_OPTION2;
self->option2.Header.BufferSz = sizeof (mfxExtCodingOption2);
self->option3.Header.BufferId = MFX_EXTBUFF_CODING_OPTION3;
self->option3.Header.BufferSz = sizeof (mfxExtCodingOption3);
}
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
static void
gst_qsv_h264_enc_set_bitrate (GstQsvH264Enc * self, mfxVideoParam * param)
{
guint max_val;
guint multiplier;
switch (param->mfx.RateControlMethod) {
case MFX_RATECONTROL_CBR:
multiplier = (self->bitrate + 0x10000) / 0x10000;
param->mfx.TargetKbps = param->mfx.MaxKbps = self->bitrate / multiplier;
param->mfx.BRCParamMultiplier = (mfxU16) multiplier;
break;
case MFX_RATECONTROL_VBR:
case MFX_RATECONTROL_VCM:
case MFX_RATECONTROL_QVBR:
max_val = MAX (self->bitrate, self->max_bitrate);
multiplier = (max_val + 0x10000) / 0x10000;
param->mfx.TargetKbps = self->bitrate / multiplier;
param->mfx.MaxKbps = self->max_bitrate / multiplier;
param->mfx.BRCParamMultiplier = (mfxU16) multiplier;
break;
case MFX_RATECONTROL_CQP:
param->mfx.QPI = self->qp_i;
param->mfx.QPP = self->qp_p;
param->mfx.QPB = self->qp_b;
break;
case MFX_RATECONTROL_AVBR:
multiplier = (self->bitrate + 0x10000) / 0x10000;
param->mfx.TargetKbps = self->bitrate / multiplier;
param->mfx.Accuracy = self->avbr_accuracy;
param->mfx.Convergence = self->avbr_convergence;
param->mfx.BRCParamMultiplier = (mfxU16) multiplier;
break;
case MFX_RATECONTROL_LA:
multiplier = (self->bitrate + 0x10000) / 0x10000;
param->mfx.TargetKbps = self->bitrate / multiplier;
param->mfx.BRCParamMultiplier = (mfxU16) multiplier;
break;
case MFX_RATECONTROL_LA_HRD:
max_val = MAX (self->bitrate, self->max_bitrate);
multiplier = (max_val + 0x10000) / 0x10000;
param->mfx.TargetKbps = self->bitrate / multiplier;
param->mfx.MaxKbps = self->max_bitrate / multiplier;
param->mfx.BRCParamMultiplier = (mfxU16) multiplier;
break;
case MFX_RATECONTROL_ICQ:
case MFX_RATECONTROL_LA_ICQ:
param->mfx.ICQQuality = self->icq_quality;
break;
default:
GST_WARNING_OBJECT (self,
"Unhandled rate-control method %d", self->rate_control);
break;
}
}
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
static gboolean
gst_qsv_h264_enc_set_format (GstQsvEncoder * encoder,
GstVideoCodecState * state, mfxVideoParam * param, GPtrArray * extra_params)
{
GstQsvH264Enc *self = GST_QSV_H264_ENC (encoder);
GstCaps *allowed_caps, *fixated_caps;
std::set < std::string > downstream_profiles;
std::set < std::string > tmp;
guint bframes = self->bframes;
mfxU16 cabac = self->cabac;
std::string profile_str;
mfxU16 mfx_profile = MFX_PROFILE_UNKNOWN;
GstVideoInfo *info = &state->info;
mfxExtVideoSignalInfo *signal_info = nullptr;
mfxExtCodingOption *option;
mfxExtCodingOption2 *option2;
mfxExtCodingOption3 *option3;
GstStructure *s;
const gchar *stream_format;
mfxFrameInfo *frame_info;
frame_info = &param->mfx.FrameInfo;
/* QSV specific alignment requirement:
* width/height should be multiple of 16, and for interlaced encoding,
* height should be multiple of 32 */
frame_info->Width = GST_ROUND_UP_16 (info->width);
if (GST_VIDEO_INFO_IS_INTERLACED (info)) {
frame_info->Height = GST_ROUND_UP_32 (info->height);
switch (GST_VIDEO_INFO_FIELD_ORDER (info)) {
case GST_VIDEO_FIELD_ORDER_TOP_FIELD_FIRST:
frame_info->PicStruct = MFX_PICSTRUCT_FIELD_TFF;
break;
case GST_VIDEO_FIELD_ORDER_BOTTOM_FIELD_FIRST:
frame_info->PicStruct = MFX_PICSTRUCT_FIELD_BFF;
break;
default:
frame_info->PicStruct = MFX_PICSTRUCT_UNKNOWN;
break;
}
} else {
frame_info->Height = GST_ROUND_UP_16 (info->height);
frame_info->PicStruct = MFX_PICSTRUCT_PROGRESSIVE;
}
/* QSV wouldn't be happy with this size, increase */
if (frame_info->Width == 16)
frame_info->Width = 32;
if (frame_info->Height == 16)
frame_info->Height = 32;
frame_info->CropW = info->width;
frame_info->CropH = info->height;
if (GST_VIDEO_INFO_FPS_N (info) > 0 && GST_VIDEO_INFO_FPS_D (info) > 0) {
frame_info->FrameRateExtN = GST_VIDEO_INFO_FPS_N (info);
frame_info->FrameRateExtD = GST_VIDEO_INFO_FPS_D (info);
} else {
/* HACK: Same as x264enc */
frame_info->FrameRateExtN = 25;
frame_info->FrameRateExtD = 1;
}
frame_info->AspectRatioW = GST_VIDEO_INFO_PAR_N (info);
frame_info->AspectRatioH = GST_VIDEO_INFO_PAR_D (info);
/* TODO: update for non 4:2:0 formats. Currently NV12 only */
frame_info->ChromaFormat = MFX_CHROMAFORMAT_YUV420;
switch (GST_VIDEO_INFO_FORMAT (info)) {
case GST_VIDEO_FORMAT_NV12:
frame_info->FourCC = MFX_FOURCC_NV12;
frame_info->BitDepthLuma = 8;
frame_info->BitDepthChroma = 8;
break;
default:
GST_ERROR_OBJECT (self, "Unexpected format %s",
gst_video_format_to_string (GST_VIDEO_INFO_FORMAT (info)));
return FALSE;
}
allowed_caps = gst_pad_get_allowed_caps (GST_VIDEO_ENCODER_SRC_PAD (encoder));
if (!allowed_caps) {
GST_WARNING_OBJECT (self, "Failed to get allowed caps");
return FALSE;
}
gst_qsv_h264_enc_init_extra_params (self);
option = &self->option;
option2 = &self->option2;
option3 = &self->option3;
self->packetized = FALSE;
fixated_caps = gst_caps_fixate (gst_caps_copy (allowed_caps));
s = gst_caps_get_structure (fixated_caps, 0);
stream_format = gst_structure_get_string (s, "stream-format");
if (g_strcmp0 (stream_format, "avc") == 0)
self->packetized = TRUE;
gst_caps_unref (fixated_caps);
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);
}
}
GST_DEBUG_OBJECT (self, "Downstream supports %" G_GSIZE_FORMAT " profiles",
downstream_profiles.size ());
/* Prune incompatible profiles */
if ((param->mfx.FrameInfo.PicStruct & MFX_PICSTRUCT_PROGRESSIVE) == 0) {
/* Interlaced, only main and high profiles are allowed */
downstream_profiles.erase (gst_qsv_h264_profile_to_string
(MFX_PROFILE_AVC_CONSTRAINED_BASELINE));
downstream_profiles.erase (gst_qsv_h264_profile_to_string
(MFX_PROFILE_AVC_PROGRESSIVE_HIGH));
downstream_profiles.erase (gst_qsv_h264_profile_to_string
(MFX_PROFILE_AVC_CONSTRAINED_HIGH));
downstream_profiles.erase (gst_qsv_h264_profile_to_string
(MFX_PROFILE_AVC_BASELINE));
}
if (downstream_profiles.empty ()) {
GST_WARNING_OBJECT (self, "No compatible profile was detected");
gst_clear_caps (&allowed_caps);
return FALSE;
}
if (bframes > 0) {
/* baseline and constrained-high doesn't support bframe */
tmp = downstream_profiles;
tmp.erase (gst_qsv_h264_profile_to_string
(MFX_PROFILE_AVC_CONSTRAINED_BASELINE));
tmp.erase (gst_qsv_h264_profile_to_string
(MFX_PROFILE_AVC_CONSTRAINED_HIGH));
tmp.erase (gst_qsv_h264_profile_to_string (MFX_PROFILE_AVC_BASELINE));
if (tmp.empty ()) {
GST_WARNING_OBJECT (self, "None of downstream profile supports bframes");
bframes = 0;
tmp = downstream_profiles;
}
downstream_profiles = tmp;
}
if (cabac == MFX_CODINGOPTION_ON) {
/* baseline doesn't support cabac */
tmp = downstream_profiles;
tmp.erase (gst_qsv_h264_profile_to_string
(MFX_PROFILE_AVC_CONSTRAINED_BASELINE));
tmp.erase (gst_qsv_h264_profile_to_string (MFX_PROFILE_AVC_BASELINE));
if (tmp.empty ()) {
GST_WARNING_OBJECT (self, "None of downstream profile supports cabac");
cabac = MFX_CODINGOPTION_OFF;
tmp = downstream_profiles;
}
downstream_profiles = tmp;
}
/* we have our preference order */
/* *INDENT-OFF* */
for (guint i = 0; i < G_N_ELEMENTS (profile_map); i++) {
auto it = downstream_profiles.find (profile_map[i].profile_str);
if (it != downstream_profiles.end ()) {
profile_str = *it;
break;
}
}
/* *INDENT-ON* */
if (profile_str.empty ()) {
GST_WARNING_OBJECT (self, "Failed to determine profile");
return FALSE;
}
GST_DEBUG_OBJECT (self, "Selected profile %s", profile_str.c_str ());
mfx_profile = gst_qsv_h264_profile_string_to_value (profile_str.c_str ());
gst_clear_caps (&allowed_caps);
if (cabac == MFX_CODINGOPTION_UNKNOWN) {
switch (mfx_profile) {
case MFX_PROFILE_AVC_CONSTRAINED_BASELINE:
case MFX_PROFILE_AVC_BASELINE:
cabac = MFX_CODINGOPTION_OFF;
break;
default:
cabac = MFX_CODINGOPTION_ON;
break;
}
}
g_mutex_lock (&self->prop_lock);
param->mfx.CodecId = MFX_CODEC_AVC;
param->mfx.CodecProfile = mfx_profile;
param->mfx.GopRefDist = bframes + 1;
param->mfx.GopPicSize = self->gop_size;
param->mfx.IdrInterval = self->iframes;
param->mfx.RateControlMethod = self->rate_control;
param->mfx.NumRefFrame = self->ref_frames;
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
gst_qsv_h264_enc_set_bitrate (self, param);
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
/* Write signal info only when upstream caps contains valid colorimetry,
* because derived default colorimetry in gst_video_info_from_caps() tends to
* very wrong in various cases, and it's even worse than "unknown" */
if (state->caps) {
GstStructure *s = gst_caps_get_structure (state->caps, 0);
GstVideoColorimetry cinfo;
const gchar *str;
str = gst_structure_get_string (s, "colorimetry");
if (str && gst_video_colorimetry_from_string (&cinfo, str)) {
signal_info = &self->signal_info;
/* 0: Component, 1: PAL, 2: NTSC, 3: SECAM, 4: MAC, 5: Unspecified */
signal_info->VideoFormat = 5;
if (cinfo.range == GST_VIDEO_COLOR_RANGE_0_255)
signal_info->VideoFullRange = 1;
else
signal_info->VideoFullRange = 0;
signal_info->ColourDescriptionPresent = 1;
signal_info->ColourPrimaries =
gst_video_color_primaries_to_iso (cinfo.primaries);
signal_info->TransferCharacteristics =
gst_video_transfer_function_to_iso (cinfo.transfer);
signal_info->MatrixCoefficients =
gst_video_color_matrix_to_iso (cinfo.matrix);
}
}
if (cabac == MFX_CODINGOPTION_OFF)
option->CAVLC = MFX_CODINGOPTION_ON;
else
option->CAVLC = MFX_CODINGOPTION_OFF;
/* TODO: property ? */
option->AUDelimiter = MFX_CODINGOPTION_ON;
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
if (self->disable_hrd_conformance) {
option->NalHrdConformance = MFX_CODINGOPTION_OFF;
option->VuiVclHrdParameters = MFX_CODINGOPTION_OFF;
}
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
/* Enables PicTiming SEI by default */
option->PicTimingSEI = MFX_CODINGOPTION_ON;
/* VUI is useful in various cases, so we don't want to disable it */
option2->DisableVUI = MFX_CODINGOPTION_OFF;
/* Do not repeat PPS */
option2->RepeatPPS = MFX_CODINGOPTION_OFF;
if (param->mfx.RateControlMethod == MFX_RATECONTROL_LA ||
param->mfx.RateControlMethod == MFX_RATECONTROL_LA_HRD ||
param->mfx.RateControlMethod == MFX_RATECONTROL_LA_ICQ) {
option2->LookAheadDS = self->rc_lookahead_ds;
option2->LookAheadDepth = self->rc_lookahead;
}
option2->MinQPI = self->min_qp_i;
option2->MinQPP = self->min_qp_p;
option2->MinQPB = self->min_qp_b;
option2->MaxQPI = self->max_qp_i;
option2->MaxQPP = self->max_qp_p;
option2->MaxQPB = self->max_qp_b;
/* QSV wants MFX_B_REF_PYRAMID when more than 1 b-frame is enabled */
if (param->mfx.GopRefDist > 2)
option2->BRefType = MFX_B_REF_PYRAMID;
/* Upstream specified framerate, we will believe it's fixed framerate */
if (GST_VIDEO_INFO_FPS_N (info) > 0 && GST_VIDEO_INFO_FPS_D (info) > 0) {
option2->FixedFrameRate = MFX_CODINGOPTION_ON;
option3->TimingInfoPresent = MFX_CODINGOPTION_ON;
}
if (param->mfx.RateControlMethod == MFX_RATECONTROL_QVBR)
option3->QVBRQuality = self->qvbr_quality;
if (signal_info)
g_ptr_array_add (extra_params, signal_info);
g_ptr_array_add (extra_params, option);
g_ptr_array_add (extra_params, option2);
g_ptr_array_add (extra_params, option3);
param->ExtParam = (mfxExtBuffer **) extra_params->pdata;
param->NumExtParam = extra_params->len;
self->bitrate_updated = FALSE;
self->property_updated = FALSE;
g_mutex_unlock (&self->prop_lock);
return TRUE;
}
static gboolean
gst_qsv_h264_enc_set_output_state (GstQsvEncoder * encoder,
GstVideoCodecState * state, mfxSession session)
{
GstQsvH264Enc *self = GST_QSV_H264_ENC (encoder);
GstCaps *caps;
GstTagList *tags;
GstVideoCodecState *out_state;
guint bitrate, max_bitrate;
guint multiplier = 1;
mfxVideoParam param;
const gchar *profile_str;
mfxStatus status;
mfxExtCodingOptionSPSPPS sps_pps;
mfxExtBuffer *ext_buffers[1];
mfxU8 sps[1024];
mfxU8 pps[1024];
GstBuffer *codec_data = nullptr;
memset (&param, 0, sizeof (mfxVideoParam));
memset (&sps_pps, 0, sizeof (mfxExtCodingOptionSPSPPS));
if (self->packetized) {
sps_pps.Header.BufferId = MFX_EXTBUFF_CODING_OPTION_SPSPPS;
sps_pps.Header.BufferSz = sizeof (mfxExtCodingOptionSPSPPS);
sps_pps.SPSBuffer = sps;
sps_pps.SPSBufSize = sizeof (sps);
sps_pps.PPSBuffer = pps;
sps_pps.PPSBufSize = sizeof (pps);
ext_buffers[0] = (mfxExtBuffer *) & sps_pps;
param.NumExtParam = 1;
param.ExtParam = ext_buffers;
}
status = MFXVideoENCODE_GetVideoParam (session, &param);
if (status < MFX_ERR_NONE) {
GST_ERROR_OBJECT (self, "Failed to get video param %d (%s)",
QSV_STATUS_ARGS (status));
return FALSE;
} else if (status != MFX_ERR_NONE) {
GST_WARNING_OBJECT (self, "GetVideoParam returned warning %d (%s)",
QSV_STATUS_ARGS (status));
}
if (self->packetized) {
GstH264ParserResult rst;
GstH264NalUnit sps_nalu, pps_nalu;
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;
rst = gst_h264_parser_identify_nalu_unchecked (self->parser,
sps, 0, sps_pps.SPSBufSize, &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;
}
qsvh264enc: Fix profile and level setting in codec data The profile field in SPS is located after header byte Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2012>
2022-03-23 17:43:43 +00:00
data = sps_nalu.data + sps_nalu.offset + sps_nalu.header_bytes;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
profile_idc = data[0];
profile_comp = data[1];
level_idc = data[2];
rst = gst_h264_parser_identify_nalu_unchecked (self->parser,
pps, 0, sps_pps.PPSBufSize, &pps_nalu);
if (rst != GST_H264_PARSER_OK) {
GST_ERROR_OBJECT (self, "Failed to identify PPS nal");
return FALSE;
}
/* 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);
}
caps = gst_caps_from_string ("video/x-h264, alignment = (string) au");
profile_str = gst_qsv_h264_profile_to_string (param.mfx.CodecProfile);
if (profile_str)
gst_caps_set_simple (caps, "profile", G_TYPE_STRING, profile_str, nullptr);
if (self->packetized) {
gst_caps_set_simple (caps, "stream-format", G_TYPE_STRING, "avc",
"codec_data", GST_TYPE_BUFFER, codec_data, nullptr);
gst_buffer_unref (codec_data);
} else {
gst_caps_set_simple (caps, "stream-format", G_TYPE_STRING, "byte-stream",
nullptr);
}
out_state = gst_video_encoder_set_output_state (GST_VIDEO_ENCODER (encoder),
caps, state);
gst_video_codec_state_unref (out_state);
tags = gst_tag_list_new_empty ();
gst_tag_list_add (tags, GST_TAG_MERGE_REPLACE, GST_TAG_ENCODER, "qsvh264enc",
nullptr);
if (param.mfx.BRCParamMultiplier > 0)
multiplier = param.mfx.BRCParamMultiplier;
switch (param.mfx.RateControlMethod) {
case MFX_RATECONTROL_CQP:
case MFX_RATECONTROL_ICQ:
case MFX_RATECONTROL_LA_ICQ:
/* We don't know target/max bitrate in this case */
break;
default:
max_bitrate = (guint) param.mfx.MaxKbps * multiplier;
bitrate = (guint) param.mfx.TargetKbps * multiplier;
if (bitrate > 0) {
gst_tag_list_add (tags, GST_TAG_MERGE_REPLACE,
GST_TAG_NOMINAL_BITRATE, bitrate * 1000, nullptr);
}
if (max_bitrate > 0) {
gst_tag_list_add (tags, GST_TAG_MERGE_REPLACE,
GST_TAG_MAXIMUM_BITRATE, max_bitrate * 1000, nullptr);
}
break;
}
gst_video_encoder_merge_tags (GST_VIDEO_ENCODER (encoder),
tags, GST_TAG_MERGE_REPLACE);
gst_tag_list_unref (tags);
return TRUE;
}
static gboolean
gst_qsv_h264_enc_foreach_caption_meta (GstBuffer * buffer, GstMeta ** meta,
GPtrArray * payload)
{
GstVideoCaptionMeta *cc_meta;
GstByteWriter br;
guint payload_size;
guint extra_size;
mfxPayload *p;
if ((*meta)->info->api != GST_VIDEO_CAPTION_META_API_TYPE)
return TRUE;
cc_meta = (GstVideoCaptionMeta *) (*meta);
if (cc_meta->caption_type != GST_VIDEO_CAPTION_TYPE_CEA708_RAW)
return TRUE;
/* QSV requires full sei_message() structure */
/* 1 byte contry_code + 10 bytes CEA-708 specific data + caption data */
payload_size = 11 + cc_meta->size;
extra_size = payload_size / 255;
/* 1 byte SEI type + 1 byte SEI payload size (+ extra) + payload data */
gst_byte_writer_init_with_size (&br, 2 + extra_size + payload_size, FALSE);
/* SEI type */
gst_byte_writer_put_uint8 (&br, 4);
/* SEI payload size */
while (payload_size >= 0xff) {
gst_byte_writer_put_uint8 (&br, 0xff);
payload_size -= 0xff;
}
gst_byte_writer_put_uint8 (&br, payload_size);
/* 8-bits itu_t_t35_country_code */
gst_byte_writer_put_uint8 (&br, 181);
/* 16-bits itu_t_t35_provider_code */
gst_byte_writer_put_uint8 (&br, 0);
gst_byte_writer_put_uint8 (&br, 49);
/* 32-bits ATSC_user_identifier */
gst_byte_writer_put_uint8 (&br, 'G');
gst_byte_writer_put_uint8 (&br, 'A');
gst_byte_writer_put_uint8 (&br, '9');
gst_byte_writer_put_uint8 (&br, '4');
/* 8-bits ATSC1_data_user_data_type_code */
gst_byte_writer_put_uint8 (&br, 3);
/* 8-bits:
* 1 bit process_em_data_flag (0)
* 1 bit process_cc_data_flag (1)
* 1 bit additional_data_flag (0)
* 5-bits cc_count
*/
gst_byte_writer_put_uint8 (&br, ((cc_meta->size / 3) & 0x1f) | 0x40);
/* 8 bits em_data, unused */
gst_byte_writer_put_uint8 (&br, 255);
gst_byte_writer_put_data (&br, cc_meta->data, cc_meta->size);
/* 8 marker bits */
gst_byte_writer_put_uint8 (&br, 255);
p = g_new0 (mfxPayload, 1);
p->BufSize = gst_byte_writer_get_pos (&br);
p->NumBit = p->BufSize * 8;
p->Type = 4;
p->Data = gst_byte_writer_reset_and_get_data (&br);
g_ptr_array_add (payload, p);
return TRUE;
}
static gboolean
gst_qsv_h264_enc_attach_payload (GstQsvEncoder * encoder,
GstVideoCodecFrame * frame, GPtrArray * payload)
{
GstQsvH264Enc *self = GST_QSV_H264_ENC (encoder);
if (self->cc_insert == GST_QSV_H264_ENC_SEI_DISABLED)
return TRUE;
gst_buffer_foreach_meta (frame->input_buffer,
(GstBufferForeachMetaFunc) gst_qsv_h264_enc_foreach_caption_meta,
payload);
return TRUE;
}
static GstBuffer *
gst_qsv_h264_enc_create_output_buffer (GstQsvEncoder * encoder,
mfxBitstream * bitstream)
{
GstQsvH264Enc *self = GST_QSV_H264_ENC (encoder);
GstBuffer *buf;
GstH264ParserResult rst;
GstH264NalUnit nalu;
if (!self->packetized) {
buf = gst_buffer_new_memdup (bitstream->Data + bitstream->DataOffset,
bitstream->DataLength);
} else {
buf = gst_buffer_new ();
rst = gst_h264_parser_identify_nalu (self->parser,
bitstream->Data + bitstream->DataOffset, 0, bitstream->DataLength,
&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 (buf, mem);
rst = gst_h264_parser_identify_nalu (self->parser,
bitstream->Data + bitstream->DataOffset, nalu.offset + nalu.size,
bitstream->DataLength, &nalu);
if (rst == GST_H264_PARSER_NO_NAL_END)
rst = GST_H264_PARSER_OK;
}
}
/* This buffer must be the end of a frame boundary */
GST_BUFFER_FLAG_SET (buf, GST_BUFFER_FLAG_MARKER);
return buf;
}
static GstQsvEncoderReconfigure
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
gst_qsv_h264_enc_check_reconfigure (GstQsvEncoder * encoder, mfxSession session,
mfxVideoParam * param, GPtrArray * extra_params)
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
{
GstQsvH264Enc *self = GST_QSV_H264_ENC (encoder);
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
GstQsvEncoderReconfigure ret = GST_QSV_ENCODER_RECONFIGURE_NONE;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
g_mutex_lock (&self->prop_lock);
if (self->property_updated) {
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
ret = GST_QSV_ENCODER_RECONFIGURE_FULL;
goto done;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
}
if (self->bitrate_updated) {
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
mfxStatus status;
mfxExtEncoderResetOption reset_opt;
reset_opt.Header.BufferId = MFX_EXTBUFF_ENCODER_RESET_OPTION;
reset_opt.Header.BufferSz = sizeof (mfxExtEncoderResetOption);
reset_opt.StartNewSequence = MFX_CODINGOPTION_UNKNOWN;
gst_qsv_h264_enc_set_bitrate (self, param);
g_ptr_array_add (extra_params, &reset_opt);
param->ExtParam = (mfxExtBuffer **) extra_params->pdata;
param->NumExtParam = extra_params->len;
status = MFXVideoENCODE_Query (session, param, param);
g_ptr_array_remove_index (extra_params, extra_params->len - 1);
param->NumExtParam = extra_params->len;
if (status != MFX_ERR_NONE) {
GST_WARNING_OBJECT (self, "MFXVideoENCODE_Query returned %d (%s)",
QSV_STATUS_ARGS (status));
ret = GST_QSV_ENCODER_RECONFIGURE_FULL;
} else {
if (reset_opt.StartNewSequence == MFX_CODINGOPTION_OFF) {
GST_DEBUG_OBJECT (self, "Can update without new sequence");
ret = GST_QSV_ENCODER_RECONFIGURE_BITRATE;
} else {
GST_DEBUG_OBJECT (self, "Need new sequence");
ret = GST_QSV_ENCODER_RECONFIGURE_FULL;
}
}
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
}
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
done:
self->property_updated = FALSE;
self->bitrate_updated = FALSE;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
g_mutex_unlock (&self->prop_lock);
qsvencoder: Add support for dynamic bitrate update ... and add more encoding options. QSV API supports dynamic bitrate change without IDR insertion. That's more efficient way of runtime encoding option update than starting from new sequence with IDR per bitrate option change. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2039>
2022-03-27 14:27:54 +00:00
return ret;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
}
typedef struct
{
guint width;
guint height;
} Resolution;
void
gst_qsv_h264_enc_register (GstPlugin * plugin, guint rank, guint impl_index,
GstObject * device, mfxSession session)
{
mfxStatus status;
mfxVideoParam param;
mfxInfoMFX *mfx;
static const Resolution resolutions_to_check[] = {
{1280, 720}, {1920, 1088}, {2560, 1440}, {3840, 2160}, {4096, 2160},
{7680, 4320}, {8192, 4320}
};
std::vector < mfxU16 > supported_profiles;
Resolution max_resolution;
bool supports_interlaced = false;
memset (&param, 0, sizeof (mfxVideoParam));
memset (&max_resolution, 0, sizeof (Resolution));
param.AsyncDepth = 4;
param.IOPattern = MFX_IOPATTERN_IN_VIDEO_MEMORY;
mfx = &param.mfx;
mfx->CodecId = MFX_CODEC_AVC;
mfx->FrameInfo.Width = GST_ROUND_UP_16 (320);
mfx->FrameInfo.Height = GST_ROUND_UP_16 (240);
mfx->FrameInfo.CropW = 320;
mfx->FrameInfo.CropH = 240;
mfx->FrameInfo.FrameRateExtN = 30;
mfx->FrameInfo.FrameRateExtD = 1;
mfx->FrameInfo.AspectRatioW = 1;
mfx->FrameInfo.AspectRatioH = 1;
mfx->FrameInfo.ChromaFormat = MFX_CHROMAFORMAT_YUV420;
mfx->FrameInfo.FourCC = MFX_FOURCC_NV12;
mfx->FrameInfo.BitDepthLuma = 8;
mfx->FrameInfo.BitDepthChroma = 8;
mfx->FrameInfo.PicStruct = MFX_PICSTRUCT_PROGRESSIVE;
/* Check supported profiles */
for (guint i = 0; i < G_N_ELEMENTS (profile_map); i++) {
mfx->CodecProfile = profile_map[i].profile;
if (MFXVideoENCODE_Query (session, &param, &param) != MFX_ERR_NONE)
continue;
supported_profiles.push_back (profile_map[i].profile);
}
if (supported_profiles.empty ()) {
GST_INFO ("Device doesn't support H.264 encoding");
return;
}
mfx->CodecProfile = supported_profiles[0];
/* Check max-resolution */
for (guint i = 0; i < G_N_ELEMENTS (resolutions_to_check); i++) {
mfx->FrameInfo.Width = GST_ROUND_UP_16 (resolutions_to_check[i].width);
mfx->FrameInfo.Height = GST_ROUND_UP_16 (resolutions_to_check[i].height);
mfx->FrameInfo.CropW = resolutions_to_check[i].width;
mfx->FrameInfo.CropH = resolutions_to_check[i].height;
if (MFXVideoENCODE_Query (session, &param, &param) != MFX_ERR_NONE)
break;
max_resolution.width = resolutions_to_check[i].width;
max_resolution.height = resolutions_to_check[i].height;
}
GST_INFO ("Maximum supported resolution: %dx%d",
max_resolution.width, max_resolution.height);
/* TODO: check supported rate-control methods and expose only supported
* methods, since the device might not be able to support some of them */
/* Check interlaced encoding */
/* *INDENT-OFF* */
for (const auto &iter: supported_profiles) {
if (iter == MFX_PROFILE_AVC_MAIN ||
iter == MFX_PROFILE_AVC_HIGH) {
/* Make sure non-lowpower mode, otherwise QSV will not accept
* interlaced format during Query() */
mfx->LowPower = MFX_CODINGOPTION_UNKNOWN;
/* Interlaced encoding is not compatible with MFX_RATECONTROL_VCM,
* make sure profile */
mfx->RateControlMethod = MFX_RATECONTROL_CBR;
/* Too high (MFX_LEVEL_AVC_41) or low (MFX_LEVEL_AVC_21) level will not be
* accepted for interlaced encoding */
mfx->CodecLevel = MFX_LEVEL_UNKNOWN;
mfx->CodecProfile = iter;
mfx->FrameInfo.Width = GST_ROUND_UP_16 (320);
mfx->FrameInfo.Height = GST_ROUND_UP_32 (240);
mfx->FrameInfo.CropW = 320;
mfx->FrameInfo.CropH = 240;
mfx->FrameInfo.PicStruct = MFX_PICSTRUCT_FIELD_TFF;
status = MFXVideoENCODE_Query (session, &param, &param);
if (status == MFX_ERR_NONE) {
GST_INFO ("Interlaced encoding is supported");
supports_interlaced = true;
break;
}
}
}
/* *INDENT-ON* */
/* To cover both landscape and portrait,
* select max value (width in this case) */
guint resolution = MAX (max_resolution.width, max_resolution.height);
std::string sink_caps_str = "video/x-raw, format=(string) NV12";
sink_caps_str += ", width=(int) [ 16, " + std::to_string (resolution) + " ]";
sink_caps_str += ", height=(int) [ 16, " + std::to_string (resolution) + " ]";
if (!supports_interlaced)
sink_caps_str += ", interlace-mode = (string) progressive";
GstCaps *sink_caps = gst_caps_from_string (sink_caps_str.c_str ());
/* TODO: Add support for VA */
#ifdef G_OS_WIN32
GstCaps *d3d11_caps = gst_caps_copy (sink_caps);
GstCapsFeatures *caps_features =
gst_caps_features_new (GST_CAPS_FEATURE_MEMORY_D3D11_MEMORY, nullptr);
gst_caps_set_features_simple (d3d11_caps, caps_features);
gst_caps_append (d3d11_caps, sink_caps);
sink_caps = d3d11_caps;
qsvencoder: Add support for VA memory Use VA allocator and buffer pool implementation for zero-copy encoding with upstream VA elements Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/2030>
2022-03-25 13:04:05 +00:00
#else
GstCaps *va_caps = gst_caps_copy (sink_caps);
GstCapsFeatures *caps_features =
gst_caps_features_new (GST_CAPS_FEATURE_MEMORY_VA, nullptr);
gst_caps_set_features_simple (va_caps, caps_features);
gst_caps_append (va_caps, sink_caps);
sink_caps = va_caps;
qsv: Introduce H.264 Intel Quick Sync Video Encoder A new implementation of Intel Quick Sync Video plugin. This plugin supports both Windows and Linux but optimization for VA/DMABuf is not implemented yet. This new plugin has some notable differences compared with existing MSDK plugin. * Encoder will expose formats which can be natively supported without internal conversion. This will make encoder control/negotiation flow much simpler and cleaner than that of MSDK plugin. * This plugin includes QSV specific library loading helper, called dispatcher, with QSV SDK headers as a part of this plugin. So, there will be no more SDK version dependent #ifdef in the code and also there will be no more build-time MSDK/oneVPL SDK dependency. * Memory allocator interop between GStreamer and QSV is re-designed and decoupled. Instead of implementing QSV specific allocator/bufferpool, this plugin will make use of generic GStreamer memory allocator/bufferpool (e.g., GstD3D11Allocator and GstD3D11BufferPool). Specifically, GstQsvAllocator object will help interop between GstMemory and mfxFrameAllocator memory abstraction layers. Note that because of the design decision, VA/DMABuf support is not made as a part of this initial commit. We can add the optimization for Linux later once GstVA library exposes allocator/bufferpool implementation as an API like GstD3D11. * Initial encoder implementation supports interop with GstD3D11 infrastructure, including zero-copy encoding with upstream D3D11 element. Part-of: <https://gitlab.freedesktop.org/gstreamer/gstreamer/-/merge_requests/1408>
2022-02-06 12:34:43 +00:00
#endif
std::string src_caps_str = "video/x-h264";
src_caps_str += ", width=(int) [ 16, " + std::to_string (resolution) + " ]";
src_caps_str += ", height=(int) [ 16, " + std::to_string (resolution) + " ]";
src_caps_str += ", stream-format= (string) { avc, byte-stream }";
src_caps_str += ", alignment=(string) au";
/* *INDENT-OFF* */
if (supported_profiles.size () > 1) {
src_caps_str += ", profile=(string) { ";
bool first = true;
for (const auto &iter: supported_profiles) {
if (!first) {
src_caps_str += ", ";
}
src_caps_str += gst_qsv_h264_profile_to_string (iter);
first = false;
}
src_caps_str += " }";
} else {
src_caps_str += ", profile=(string) ";
src_caps_str += gst_qsv_h264_profile_to_string (supported_profiles[0]);
}
/* *INDENT-ON* */
GstCaps *src_caps = gst_caps_from_string (src_caps_str.c_str ());
GST_MINI_OBJECT_FLAG_SET (sink_caps, GST_MINI_OBJECT_FLAG_MAY_BE_LEAKED);
GST_MINI_OBJECT_FLAG_SET (src_caps, GST_MINI_OBJECT_FLAG_MAY_BE_LEAKED);
GstQsvH264EncClassData *cdata = g_new0 (GstQsvH264EncClassData, 1);
cdata->sink_caps = sink_caps;
cdata->src_caps = src_caps;
cdata->impl_index = impl_index;
#ifdef G_OS_WIN32
gint64 device_luid;
g_object_get (device, "adapter-luid", &device_luid, nullptr);
cdata->adapter_luid = device_luid;
#else
gchar *display_path;
g_object_get (device, "path", &display_path, nullptr);
cdata->display_path = display_path;
#endif
GType type;
gchar *type_name;
gchar *feature_name;
GTypeInfo type_info = {
sizeof (GstQsvH264EncClass),
nullptr,
nullptr,
(GClassInitFunc) gst_qsv_h264_enc_class_init,
nullptr,
cdata,
sizeof (GstQsvH264Enc),
0,
(GInstanceInitFunc) gst_qsv_h264_enc_init,
};
type_name = g_strdup ("GstQsvH264Enc");
feature_name = g_strdup ("qsvh264enc");
gint index = 0;
while (g_type_from_name (type_name)) {
index++;
g_free (type_name);
g_free (feature_name);
type_name = g_strdup_printf ("GstQsvH264Device%dEnc", index);
feature_name = g_strdup_printf ("qsvh264device%denc", index);
}
type = g_type_register_static (GST_TYPE_QSV_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);
}
Reference in a new issue Copy permalink