Initialize SVT-AV1 codec

This commit is contained in:
Atsushi Watanabe
2025-10-20 22:37:45 +09:00
parent a68a5ba4a6
commit 8840daf7ea
6 changed files with 437 additions and 0 deletions
+1
View File
@@ -27,6 +27,7 @@ jobs:
sudo apt-get update -qq \
&& sudo apt-get install --no-install-recommends -y \
libopus-dev \
libsvtav1enc-dev \
libva-dev \
libvpx-dev \
libx11-dev \
+74
View File
@@ -0,0 +1,74 @@
#include <EbSvtAv1Enc.h>
#include <EbSvtAv1ErrorCodes.h>
#include <stdint.h>
#define ERR_INIT_ENC_HANDLER 1
#define ERR_SET_ENC_PARAM 2
#define ERR_ENC_INIT 3
typedef struct Encoder {
EbSvtAv1EncConfiguration *param;
EbComponentType *handle;
} Encoder;
typedef struct Buffer {
unsigned char *data;
int len;
} Buffer;
int enc_new(Encoder **e) {
EbErrorType sret;
*e = malloc(sizeof(Encoder));
(*e)->param = malloc(sizeof(EbSvtAv1EncConfiguration));
sret = svt_av1_enc_init_handle(&(*e)->handle, *e, (*e)->param);
if (sret != EB_ErrorNone) {
free((*e)->param);
free(*e);
return ERR_INIT_ENC_HANDLER;
}
return 0;
}
int enc_init(Encoder *e) {
EbErrorType sret;
sret = svt_av1_enc_set_parameter(e->handle, e->param);
if (sret != EB_ErrorNone) {
return ERR_SET_ENC_PARAM;
}
sret = svt_av1_enc_init(e->handle);
if (sret != EB_ErrorNone) {
return ERR_ENC_INIT;
}
return 0;
}
int enc_apply_param(Encoder *e) {
EbErrorType sret = svt_av1_enc_set_parameter(e->handle, e->param);
if (sret != EB_ErrorNone) {
return ERR_SET_ENC_PARAM;
}
return 0;
}
unsigned char dummy[] = {0, 1, 2, 3};
int enc_encode(Encoder *e, Buffer *out, uint8_t *y, uint8_t *cb, uint8_t *cr) {
// TODO: implement
out->data = dummy;
out->len = 4;
return 0;
}
int enc_close(Encoder *e) {
free(e->param);
free(e);
return 0;
}
+12
View File
@@ -0,0 +1,12 @@
package svtav1
import (
"errors"
)
var (
ErrUnknownErrorCode = errors.New("unknown error code")
ErrInitEncHandler = errors.New("failed to initialize encoder handler")
ErrSetEncParam = errors.New("failed to set encoder parameters")
ErrEncInit = errors.New("failed to initialize encoder")
)
+38
View File
@@ -0,0 +1,38 @@
package svtav1
import (
"github.com/pion/mediadevices/pkg/codec"
"github.com/pion/mediadevices/pkg/io/video"
"github.com/pion/mediadevices/pkg/prop"
)
// Params stores libx264 specific encoding parameters.
type Params struct {
codec.BaseParams
// Preset configuration number of SVT-AV1
// 1-3: extremely high efficiency but heavy
// 4-6: a balance of efficiency and reasonable compute time
// 7-13: real-time encoding
Preset int
}
// NewParams returns default x264 codec specific parameters.
func NewParams() (Params, error) {
return Params{
BaseParams: codec.BaseParams{
KeyFrameInterval: 60,
},
Preset: 9,
}, nil
}
// RTPCodec represents the codec metadata
func (p *Params) RTPCodec() *codec.RTPCodec {
return codec.NewRTPAV1Codec(90000)
}
// BuildVideoEncoder builds x264 encoder with given params
func (p *Params) BuildVideoEncoder(r video.Reader, property prop.Media) (codec.ReadCloser, error) {
return newEncoder(r, property, *p)
}
+166
View File
@@ -0,0 +1,166 @@
// Package svtav1 implements AV1 encoder.
// This package requires libSvtAv1Enc headers and libraries to be built.
package svtav1
// #cgo pkg-config: SvtAv1Enc
// #include "bridge.h"
import "C"
import (
"image"
"io"
"sync"
"unsafe"
"github.com/pion/mediadevices/pkg/codec"
"github.com/pion/mediadevices/pkg/io/video"
"github.com/pion/mediadevices/pkg/prop"
)
type encoder struct {
engine *C.Encoder
r video.Reader
mu sync.Mutex
closed bool
}
func newEncoder(r video.Reader, p prop.Media, params Params) (codec.ReadCloser, error) {
var enc *C.Encoder
if params.KeyFrameInterval == 0 {
params.KeyFrameInterval = 60
}
if p.FrameRate == 0 {
p.FrameRate = 30
}
if err := errFromC(C.enc_new(&enc)); err != nil {
return nil, err
}
enc.param.source_width = C.uint32_t(p.Width)
enc.param.source_height = C.uint32_t(p.Height)
enc.param.encoder_bit_depth = 8
enc.param.encoder_color_format = C.EB_YUV420
enc.param.profile = C.MAIN_PROFILE
enc.param.level = 0 // auto
enc.param.hierarchical_levels = 0 // auto
enc.param.enc_mode = C.int8_t(params.Preset)
enc.param.tier = 0 // main
enc.param.rate_control_mode = C.SVT_AV1_RC_MODE_CBR
enc.param.pred_structure = 1 // LOW_DELAY
enc.param.qp = 50 // default
enc.param.target_bit_rate = C.uint32_t(params.BitRate)
enc.param.intra_period_length = -2 // auto
enc.param.frame_rate_numerator = C.uint32_t(p.FrameRate * 1000)
enc.param.frame_rate_denominator = 1000
enc.param.enable_tpl_la = 0 // LOW_DELAY requires no TPL
enc.param.max_qp_allowed = 63
enc.param.min_qp_allowed = 0
enc.param.intra_refresh_type = C.SVT_AV1_KF_REFRESH
if err := errFromC(C.enc_init(enc)); err != nil {
_ = C.enc_close(enc)
return nil, err
}
e := encoder{
engine: enc,
r: video.ToI420(r),
}
return &e, nil
}
func errFromC(ret C.int) error {
switch ret {
case 0:
return nil
case C.ERR_INIT_ENC_HANDLER:
return ErrInitEncHandler
case C.ERR_SET_ENC_PARAM:
return ErrSetEncParam
case C.ERR_ENC_INIT:
return ErrEncInit
default:
return ErrUnknownErrorCode
}
}
func (e *encoder) Read() ([]byte, func(), error) {
e.mu.Lock()
defer e.mu.Unlock()
if e.closed {
return nil, func() {}, io.EOF
}
img, release, err := e.r.Read()
if err != nil {
return nil, func() {}, err
}
defer release()
yuvImg := img.(*image.YCbCr)
var buf C.Buffer
ret := C.enc_encode(
e.engine,
&buf,
(*C.uchar)(&yuvImg.Y[0]),
(*C.uchar)(&yuvImg.Cb[0]),
(*C.uchar)(&yuvImg.Cr[0]),
)
if err := errFromC(ret); err != nil {
return nil, func() {}, err
}
encoded := C.GoBytes(unsafe.Pointer(buf.data), buf.len)
return encoded, func() {}, err
}
// TODO: Implement bit rate control
//var _ codec.BitRateController = (*encoder)(nil)
func (e *encoder) ForceKeyFrame() error {
e.mu.Lock()
defer e.mu.Unlock()
e.engine.param.force_key_frames = 1
if err := errFromC(C.enc_apply_param(e.engine)); err != nil {
return err
}
return nil
}
func (e *encoder) SetBitRate(bitrate int) error {
e.mu.Lock()
defer e.mu.Unlock()
e.engine.param.target_bit_rate = C.uint32_t(bitrate)
if err := errFromC(C.enc_apply_param(e.engine)); err != nil {
return err
}
return nil
}
func (e *encoder) Controller() codec.EncoderController {
return e
}
func (e *encoder) Close() error {
e.mu.Lock()
defer e.mu.Unlock()
if e.closed {
return nil
}
if err := errFromC(C.enc_close(e.engine)); err != nil {
return err
}
e.closed = true
return nil
}
+146
View File
@@ -0,0 +1,146 @@
package svtav1
import (
"image"
"testing"
"github.com/pion/mediadevices/pkg/codec"
"github.com/pion/mediadevices/pkg/codec/internal/codectest"
"github.com/pion/mediadevices/pkg/frame"
"github.com/pion/mediadevices/pkg/io/video"
"github.com/pion/mediadevices/pkg/prop"
)
func getTestVideoEncoder() (codec.ReadCloser, error) {
p, err := NewParams()
if err != nil {
return nil, err
}
p.BitRate = 200000
enc, err := p.BuildVideoEncoder(video.ReaderFunc(func() (image.Image, func(), error) {
return image.NewYCbCr(
image.Rect(0, 0, 256, 144),
image.YCbCrSubsampleRatio420,
), nil, nil
}), prop.Media{
Video: prop.Video{
Width: 256,
Height: 144,
FrameFormat: frame.FormatI420,
},
})
if err != nil {
return nil, err
}
return enc, nil
}
func TestEncoder(t *testing.T) {
t.Run("SimpleRead", func(t *testing.T) {
p, err := NewParams()
if err != nil {
t.Fatal(err)
}
p.BitRate = 200000
codectest.VideoEncoderSimpleReadTest(t, &p,
prop.Media{
Video: prop.Video{
Width: 256,
Height: 144,
FrameFormat: frame.FormatI420,
},
},
image.NewYCbCr(
image.Rect(0, 0, 256, 144),
image.YCbCrSubsampleRatio420,
),
)
})
t.Run("CloseTwice", func(t *testing.T) {
p, err := NewParams()
if err != nil {
t.Fatal(err)
}
p.BitRate = 200000
codectest.VideoEncoderCloseTwiceTest(t, &p, prop.Media{
Video: prop.Video{
Width: 640,
Height: 480,
FrameRate: 30,
FrameFormat: frame.FormatI420,
},
})
})
t.Run("ReadAfterClose", func(t *testing.T) {
p, err := NewParams()
if err != nil {
t.Fatal(err)
}
p.BitRate = 200000
codectest.VideoEncoderReadAfterCloseTest(t, &p,
prop.Media{
Video: prop.Video{
Width: 256,
Height: 144,
FrameFormat: frame.FormatI420,
},
},
image.NewYCbCr(
image.Rect(0, 0, 256, 144),
image.YCbCrSubsampleRatio420,
),
)
})
}
func TestShouldImplementKeyFrameControl(t *testing.T) {
e := &encoder{}
if _, ok := e.Controller().(codec.KeyFrameController); !ok {
t.Error()
}
}
func TestNoErrorOnForceKeyFrame(t *testing.T) {
enc, err := getTestVideoEncoder()
if err != nil {
t.Fatal(err)
}
kfc, ok := enc.Controller().(codec.KeyFrameController)
if !ok {
t.Fatal("Failed to get KeyFrameController")
}
if err := kfc.ForceKeyFrame(); err != nil {
t.Error(err)
}
_, rel, err := enc.Read() // try to read the encoded frame
rel()
if err != nil {
t.Fatal(err)
}
}
func TestShouldImplementBitRateControl(t *testing.T) {
e := &encoder{}
if _, ok := e.Controller().(codec.BitRateController); !ok {
t.Error()
}
}
func TestNoErrorOnSetBitRate(t *testing.T) {
enc, err := getTestVideoEncoder()
if err != nil {
t.Fatal(err)
}
brc, ok := enc.Controller().(codec.BitRateController)
if !ok {
t.Fatal("Failed to get BitRateController")
}
if err := brc.SetBitRate(1000); err != nil { // 1000 bit/second is ridiculously low, but this is a testcase.
t.Error(err)
}
_, rel, err := enc.Read() // try to read the encoded frame
rel()
if err != nil {
t.Fatal(err)
}
}