2024-06-26 20:47:26 +03:00
|
|
|
#include"node.h"
|
|
|
|
|
|
|
|
#include<vpx/vpx_encoder.h>
|
|
|
|
#include<vpx/vp8cx.h>
|
|
|
|
|
|
|
|
#include<new>
|
|
|
|
|
|
|
|
#include<mkvmuxer/mkvwriter.h>
|
|
|
|
|
|
|
|
#include<assert.h>
|
|
|
|
|
|
|
|
#include"mode.h"
|
|
|
|
|
|
|
|
#include"img.h"
|
|
|
|
#include<math.h>
|
|
|
|
|
|
|
|
#include<smmintrin.h>
|
|
|
|
|
|
|
|
#include<queue>
|
|
|
|
|
|
|
|
#include<string.h>
|
|
|
|
|
2024-06-30 14:43:13 +03:00
|
|
|
#include"minitrace.h"
|
|
|
|
|
2024-06-26 20:47:26 +03:00
|
|
|
#include"linearity.h"
|
|
|
|
|
|
|
|
struct CHiEncodeVP9Node {
|
|
|
|
vpx_codec_ctx_t codec;
|
|
|
|
vpx_codec_enc_cfg_t cfg;
|
|
|
|
|
|
|
|
enum {
|
|
|
|
WAITING, IN_PROGRESS
|
|
|
|
} state;
|
|
|
|
uint8_t *outY, *outU, *outV;
|
|
|
|
uint16_t strideY, strideU, strideV;
|
|
|
|
|
|
|
|
vpx_codec_iface_t *iface;
|
|
|
|
|
|
|
|
CHiPubNode pub;
|
|
|
|
};
|
|
|
|
|
|
|
|
static int encodevp9_perform(CHiPubNode *pub) {
|
|
|
|
CHiEncodeVP9Node *node = (CHiEncodeVP9Node*) ((uintptr_t) pub - offsetof(CHiEncodeVP9Node, pub));
|
|
|
|
|
2024-06-30 14:43:13 +03:00
|
|
|
MTR_BEGIN("CHi", "encodevp9_perform");
|
|
|
|
|
2024-06-26 20:47:26 +03:00
|
|
|
pub->sources[0].type = CUTIHI_VAL_VP9BS;
|
|
|
|
pub->sources[0].data.bitstream = NULL;
|
|
|
|
|
|
|
|
if(node->state == CHiEncodeVP9Node::WAITING) return 1;
|
|
|
|
|
|
|
|
CHiImage *rgbIn = (CHiImage*) CHi_Crawl(&pub->sinks[0])->data.sample;
|
|
|
|
|
|
|
|
#pragma omp parallel for simd
|
|
|
|
for(size_t y = 0; y < node->cfg.g_h; y += 2) {
|
|
|
|
for(size_t x = 0; x < node->cfg.g_w; x += 16) {
|
|
|
|
__m128i rgb, partY, partU, partV, dotY, dotU, dotV;
|
|
|
|
|
|
|
|
__m128i wipY0 = _mm_setzero_si128();
|
|
|
|
__m128i wipY1 = _mm_setzero_si128();
|
|
|
|
__m128i wipU = _mm_setzero_si128();
|
|
|
|
__m128i wipV = _mm_setzero_si128();
|
|
|
|
|
|
|
|
__m128i tempU = _mm_setzero_si128();
|
|
|
|
__m128i tempV = _mm_setzero_si128();
|
|
|
|
|
|
|
|
#define DO_DAH_DOO_DOO(LoOrHi, shufY, shufUV) \
|
|
|
|
/* Process top two */\
|
|
|
|
rgb = _mm_srli_epi16(apply_gamma_epi16(line0, _mm_set1_ps(1 / 2.2f)), 8); \
|
|
|
|
/* Start matrix multiplication (BT.709 + full->studio range) */\
|
|
|
|
partY = _mm_mullo_epi16(rgb, _mm_set_epi16(0, 47, 157, 16, 0, 47, 157, 16));\
|
|
|
|
partU = _mm_mullo_epi16(rgb, _mm_set_epi16(0, -25, -85, 110, 0, -25, -85, 110));\
|
|
|
|
partV = _mm_mullo_epi16(rgb, _mm_set_epi16(0, 110, -100, -10, 0, 110, -100, -10));\
|
|
|
|
/* Finish mat-mul with dot products */\
|
|
|
|
dotY = _mm_madd_epi16(partY, _mm_set1_epi16(1));\
|
|
|
|
dotY = _mm_hadd_epi32(dotY, _mm_setzero_si128());\
|
|
|
|
dotU = _mm_madd_epi16(partU, _mm_set1_epi16(1));\
|
|
|
|
dotU = _mm_hadd_epi32(dotU, _mm_setzero_si128());\
|
|
|
|
dotV = _mm_madd_epi16(partV, _mm_set1_epi16(1));\
|
|
|
|
dotV = _mm_hadd_epi32(dotV, _mm_setzero_si128());\
|
|
|
|
/* Insert Ys */\
|
|
|
|
wipY0 = _mm_or_si128(wipY0, _mm_shuffle_epi8(dotY, shufY));\
|
|
|
|
/* Save top UV */\
|
|
|
|
tempU = dotU;\
|
|
|
|
tempV = dotV;\
|
|
|
|
\
|
|
|
|
/* Process bottom two */\
|
|
|
|
rgb = _mm_srli_epi16(apply_gamma_epi16(line1, _mm_set1_ps(1 / 2.2f)), 8); \
|
|
|
|
/* Start matrix multiplication (BT.709 + full->studio range) */\
|
|
|
|
partY = _mm_mullo_epi16(rgb, _mm_set_epi16(0, 47, 157, 16, 0, 47, 157, 16));\
|
|
|
|
partU = _mm_mullo_epi16(rgb, _mm_set_epi16(0, -25, -85, 110, 0, -25, -85, 110));\
|
|
|
|
partV = _mm_mullo_epi16(rgb, _mm_set_epi16(0, 110, -100, -10, 0, 110, -100, -10));\
|
|
|
|
/* Finish mat-mul with dot products */\
|
|
|
|
dotY = _mm_madd_epi16(partY, _mm_set1_epi16(1));\
|
|
|
|
dotY = _mm_hadd_epi32(dotY, _mm_setzero_si128());\
|
|
|
|
dotU = _mm_madd_epi16(partU, _mm_set1_epi16(1));\
|
|
|
|
dotU = _mm_hadd_epi32(dotU, _mm_setzero_si128());\
|
|
|
|
dotV = _mm_madd_epi16(partV, _mm_set1_epi16(1));\
|
|
|
|
dotV = _mm_hadd_epi32(dotV, _mm_setzero_si128());\
|
|
|
|
/* Insert Ys */\
|
|
|
|
wipY1 = _mm_or_si128(wipY1, _mm_shuffle_epi8(dotY, shufY));\
|
|
|
|
/* Save bottom UVs */\
|
|
|
|
tempU = _mm_hadd_epi32(_mm_add_epi32(tempU, dotU), _mm_setzero_si128());\
|
|
|
|
tempV = _mm_hadd_epi32(_mm_add_epi32(tempV, dotV), _mm_setzero_si128());\
|
|
|
|
\
|
|
|
|
/* Insert UVs */\
|
|
|
|
wipU = _mm_or_si128(wipU, _mm_shuffle_epi8(_mm_srli_epi32(tempU, 2), shufUV));\
|
|
|
|
wipV = _mm_or_si128(wipV, _mm_shuffle_epi8(_mm_srli_epi32(tempV, 2), shufUV));
|
|
|
|
|
|
|
|
__m128i line0 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 0) * rgbIn->stride + (x + 0) * 8)); // Load two pixels
|
|
|
|
__m128i line1 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 1) * rgbIn->stride + (x + 0) * 8)); // Load two pixels
|
|
|
|
|
|
|
|
DO_DAH_DOO_DOO(_mm_unpacklo_epi8,
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, 5, 1),
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, 1));
|
|
|
|
|
|
|
|
line0 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 0) * rgbIn->stride + (x + 2) * 8)); // Load two pixels
|
|
|
|
line1 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 1) * rgbIn->stride + (x + 2) * 8)); // Load two pixels
|
|
|
|
|
|
|
|
DO_DAH_DOO_DOO(_mm_unpacklo_epi8,
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, 5, 1, -128, -128),
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, 1, -128));
|
|
|
|
|
|
|
|
line0 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 0) * rgbIn->stride + (x + 4) * 8)); // Load two pixels
|
|
|
|
line1 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 1) * rgbIn->stride + (x + 4) * 8)); // Load two pixels
|
|
|
|
|
|
|
|
DO_DAH_DOO_DOO(_mm_unpacklo_epi8,
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, 5, 1, -128, -128, -128, -128),
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, 1, -128, -128));
|
|
|
|
|
|
|
|
line0 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 0) * rgbIn->stride + (x + 6) * 8)); // Load two pixels
|
|
|
|
line1 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 1) * rgbIn->stride + (x + 6) * 8)); // Load two pixels
|
|
|
|
|
|
|
|
DO_DAH_DOO_DOO(_mm_unpacklo_epi8,
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, 5, 1, -128, -128, -128, -128, -128, -128),
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, 1, -128, -128, -128));
|
|
|
|
|
|
|
|
line0 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 0) * rgbIn->stride + (x + 8) * 8)); // Load two pixels
|
|
|
|
line1 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 1) * rgbIn->stride + (x + 8) * 8)); // Load two pixels
|
|
|
|
|
|
|
|
DO_DAH_DOO_DOO(_mm_unpacklo_epi8,
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, 5, 1, -128, -128, -128, -128, -128, -128, -128, -128),
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, 1, -128, -128, -128, -128));
|
|
|
|
|
|
|
|
line0 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 0) * rgbIn->stride + (x + 10) * 8)); // Load two pixels
|
|
|
|
line1 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 1) * rgbIn->stride + (x + 10) * 8)); // Load two pixels
|
|
|
|
|
|
|
|
DO_DAH_DOO_DOO(_mm_unpacklo_epi8,
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, 5, 1, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128),
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, -128, -128, 1, -128, -128, -128, -128, -128));
|
|
|
|
|
|
|
|
line0 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 0) * rgbIn->stride + (x + 12) * 8)); // Load two pixels
|
|
|
|
line1 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 1) * rgbIn->stride + (x + 12) * 8)); // Load two pixels
|
|
|
|
|
|
|
|
DO_DAH_DOO_DOO(_mm_unpacklo_epi8,
|
|
|
|
_mm_set_epi8(-128, -128, 5, 1, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128),
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, -128, 1, -128, -128, -128, -128, -128, -128));
|
|
|
|
|
|
|
|
line0 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 0) * rgbIn->stride + (x + 14) * 8)); // Load two pixels
|
|
|
|
line1 = _mm_load_si128((__m128i*) ((uintptr_t) rgbIn->data16 + (y + 1) * rgbIn->stride + (x + 14) * 8)); // Load two pixels
|
|
|
|
|
|
|
|
DO_DAH_DOO_DOO(_mm_unpacklo_epi8,
|
|
|
|
_mm_set_epi8( 5, 1, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128, -128),
|
|
|
|
_mm_set_epi8(-128, -128, -128, -128, -128, -128, -128, -128, 1, -128, -128, -128, -128, -128, -128, -128));
|
|
|
|
|
|
|
|
_mm_stream_si128((__m128i*) &node->outY[node->strideY * (y + 0) + x], _mm_add_epi8(_mm_set1_epi8(16), wipY0));
|
|
|
|
_mm_stream_si128((__m128i*) &node->outY[node->strideY * (y + 1) + x], _mm_add_epi8(_mm_set1_epi8(16), wipY1));
|
|
|
|
_mm_storeu_si128((__m128i*) &node->outU[node->strideU * (y / 2) + x / 2], _mm_add_epi8(wipU, _mm_set1_epi8(128)));
|
|
|
|
_mm_storeu_si128((__m128i*) &node->outV[node->strideV * (y / 2) + x / 2], _mm_add_epi8(wipV, _mm_set1_epi8(128)));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
vpx_image_t vpxraw;
|
|
|
|
vpxraw.fmt = VPX_IMG_FMT_I420;
|
|
|
|
vpxraw.cs = VPX_CS_BT_709;
|
|
|
|
vpxraw.range = VPX_CR_STUDIO_RANGE;
|
|
|
|
vpxraw.bit_depth = 8;
|
|
|
|
vpxraw.w = vpxraw.d_w = node->cfg.g_w;
|
|
|
|
vpxraw.h = vpxraw.d_h = node->cfg.g_h;
|
|
|
|
vpxraw.r_w = vpxraw.r_h = 0;
|
|
|
|
vpxraw.x_chroma_shift = vpxraw.y_chroma_shift = 1;
|
|
|
|
vpxraw.img_data_owner = 0;
|
|
|
|
vpxraw.self_allocd = 0;
|
|
|
|
vpxraw.bps = 12;
|
|
|
|
vpxraw.stride[VPX_PLANE_Y] = node->strideY;
|
|
|
|
vpxraw.planes[VPX_PLANE_Y] = node->outY;
|
|
|
|
vpxraw.stride[VPX_PLANE_U] = node->strideU;
|
|
|
|
vpxraw.planes[VPX_PLANE_U] = node->outU;
|
|
|
|
vpxraw.stride[VPX_PLANE_V] = node->strideV;
|
|
|
|
vpxraw.planes[VPX_PLANE_V] = node->outV;
|
|
|
|
|
|
|
|
vpx_codec_encode(&node->codec, &vpxraw, CHi_Time_Get(pub->ng) * 1000.f, 1, 0, VPX_DL_REALTIME);
|
|
|
|
|
|
|
|
auto ret = (CHiBSFrames*) malloc(sizeof(CHiBSFrames));
|
|
|
|
ret->count = 0;
|
|
|
|
|
|
|
|
vpx_codec_iter_t iter = NULL;
|
|
|
|
const vpx_codec_cx_pkt_t *pkt;
|
|
|
|
while((pkt = vpx_codec_get_cx_data(&node->codec, &iter)) != NULL) {
|
|
|
|
if(pkt->kind == VPX_CODEC_CX_FRAME_PKT) {
|
|
|
|
ret = (CHiBSFrames*) realloc(ret, sizeof(CHiBSFrames) + sizeof(CHiBSFrame) * (ret->count + 1));
|
|
|
|
ret->data[ret->count].timestamp = pkt->data.frame.pts;
|
|
|
|
ret->data[ret->count].sz = pkt->data.frame.sz;
|
|
|
|
ret->data[ret->count].flags = pkt->data.frame.flags & VPX_FRAME_IS_KEY;
|
|
|
|
ret->data[ret->count].ptr = malloc(ret->data[ret->count].sz);
|
|
|
|
memcpy(ret->data[ret->count].ptr, pkt->data.frame.buf, ret->data[ret->count].sz);
|
|
|
|
ret->count++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// if(pktRet) v->queueOut.enqueue(pktRet);
|
|
|
|
|
|
|
|
//memcpy(node->vpxraw.planes[VPX_PLANE_Y], VIPS_IMAGE_ADDR(y, 0, 0), node->vpxraw.stride[VPX_PLANE_Y] * node->vpxraw.d_h);
|
|
|
|
//memcpy(node->vpxraw.planes[VPX_PLANE_U], VIPS_IMAGE_ADDR(u, 0, 0), node->vpxraw.stride[VPX_PLANE_U] * (node->vpxraw.d_h >> node->vpxraw.y_chroma_shift));
|
|
|
|
//memcpy(node->vpxraw.planes[VPX_PLANE_V], VIPS_IMAGE_ADDR(v, 0, 0), node->vpxraw.stride[VPX_PLANE_V] * (node->vpxraw.d_h >> node->vpxraw.y_chroma_shift));
|
|
|
|
|
|
|
|
//const vpx_codec_cx_pkt_t *pkt;
|
|
|
|
//while(!node->queueOut.try_dequeue(pkt)) usleep(0);
|
|
|
|
|
|
|
|
pub->sources[0].data.bitstream = ret;
|
|
|
|
|
2024-06-30 14:43:13 +03:00
|
|
|
MTR_END("CHi", "encodevp9_perform");
|
|
|
|
|
2024-06-26 20:47:26 +03:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2024-06-30 14:43:13 +03:00
|
|
|
static void encodevp_destroy(CHiPubNode *pub) {
|
|
|
|
CHiEncodeVP9Node *node = (CHiEncodeVP9Node*) ((uintptr_t) pub - offsetof(CHiEncodeVP9Node, pub));
|
|
|
|
free(node);
|
|
|
|
}
|
|
|
|
|
2024-06-26 20:47:26 +03:00
|
|
|
CUTIVIS CHiPubNode *CHi_EncodeVP8() {
|
2024-06-30 14:43:13 +03:00
|
|
|
CHiEncodeVP9Node *n = (CHiEncodeVP9Node*) calloc(1, sizeof(*n));
|
2024-06-26 20:47:26 +03:00
|
|
|
new (n) CHiEncodeVP9Node();
|
|
|
|
n->pub.type = CUTIHI_T('CEnc','GVP8');
|
|
|
|
n->pub.Start = CHi_EncodeVP9_Start;
|
|
|
|
n->pub.Perform = encodevp9_perform;
|
|
|
|
n->pub.Stop = CHi_EncodeVP9_Stop;
|
2024-06-30 14:43:13 +03:00
|
|
|
n->pub.Destroy = encodevp_destroy;
|
2024-06-26 20:47:26 +03:00
|
|
|
n->pub.clean = 0;
|
|
|
|
n->pub.sinks = (CHiValue*) calloc(sizeof(*n->pub.sinks), n->pub.sinkCount = 1);
|
|
|
|
n->pub.sources = (CHiValue*) calloc(sizeof(*n->pub.sources), n->pub.sourceCount = 1);
|
|
|
|
n->state = CHiEncodeVP9Node::WAITING;
|
|
|
|
n->iface = vpx_codec_vp8_cx();
|
|
|
|
return &n->pub;
|
|
|
|
}
|
|
|
|
|
|
|
|
CUTIVIS CHiPubNode *CHi_EncodeVP9() {
|
2024-06-30 14:43:13 +03:00
|
|
|
CHiEncodeVP9Node *n = (CHiEncodeVP9Node*) calloc(1, sizeof(*n));
|
2024-06-26 20:47:26 +03:00
|
|
|
new (n) CHiEncodeVP9Node();
|
|
|
|
n->pub.type = CUTIHI_T('CEnc','GVP9');
|
|
|
|
n->pub.Start = CHi_EncodeVP9_Start;
|
|
|
|
n->pub.Perform = encodevp9_perform;
|
|
|
|
n->pub.Stop = CHi_EncodeVP9_Stop;
|
2024-06-30 14:43:13 +03:00
|
|
|
n->pub.Destroy = encodevp_destroy;
|
2024-06-26 20:47:26 +03:00
|
|
|
n->pub.clean = 0;
|
|
|
|
n->pub.sinks = (CHiValue*) calloc(sizeof(*n->pub.sinks), n->pub.sinkCount = 1);
|
|
|
|
n->pub.sources = (CHiValue*) calloc(sizeof(*n->pub.sources), n->pub.sourceCount = 1);
|
|
|
|
n->state = CHiEncodeVP9Node::WAITING;
|
|
|
|
n->iface = vpx_codec_vp9_cx();
|
|
|
|
return &n->pub;
|
|
|
|
}
|
|
|
|
|
|
|
|
CUTIVIS int CHi_EncodeVP9_Start(CHiPubNode *pubn) {
|
|
|
|
CHiEncodeVP9Node *node = (CHiEncodeVP9Node*) ((uintptr_t) pubn - offsetof(CHiEncodeVP9Node, pub));
|
|
|
|
|
|
|
|
node->state = CHiEncodeVP9Node::IN_PROGRESS;
|
|
|
|
|
|
|
|
CHiImage *firstFrame = (CHiImage*) CHi_Crawl(&pubn->sinks[0])->data.sample;
|
|
|
|
|
|
|
|
vpx_codec_enc_config_default(node->iface, &node->cfg, 0);
|
|
|
|
node->cfg.g_w = firstFrame->width;
|
|
|
|
node->cfg.g_h = firstFrame->height;
|
|
|
|
node->cfg.g_timebase.num = 1;
|
|
|
|
node->cfg.g_timebase.den = 30;
|
|
|
|
node->cfg.g_lag_in_frames = 0;
|
|
|
|
node->cfg.g_threads = 8;
|
|
|
|
node->cfg.kf_mode = VPX_KF_AUTO;
|
|
|
|
node->cfg.kf_max_dist = 300;
|
|
|
|
node->cfg.rc_end_usage = VPX_VBR;
|
|
|
|
node->cfg.rc_target_bitrate = 512;
|
|
|
|
node->cfg.rc_min_quantizer = 4;
|
|
|
|
node->cfg.rc_max_quantizer = 48;
|
|
|
|
|
|
|
|
vpx_codec_enc_init(&node->codec, node->iface, &node->cfg, 0);
|
|
|
|
vpx_codec_control(&node->codec, VP8E_SET_CPUUSED, 8);
|
|
|
|
vpx_codec_control(&node->codec, VP9E_SET_ROW_MT, 1);
|
|
|
|
vpx_codec_control(&node->codec, VP9E_SET_TILE_COLUMNS, 2);
|
|
|
|
vpx_codec_control(&node->codec, VP9E_SET_TILE_ROWS, 1);
|
|
|
|
vpx_codec_control(&node->codec, VP9E_SET_TUNE_CONTENT, VP9E_CONTENT_SCREEN);
|
|
|
|
|
|
|
|
node->strideY = (node->cfg.g_w + 64) & ~63;
|
|
|
|
node->strideU = (node->cfg.g_w / 2 + 64) & ~63;
|
|
|
|
node->strideV = (node->cfg.g_w / 2 + 64) & ~63;
|
|
|
|
|
|
|
|
node->outY = (uint8_t*) _mm_malloc(node->strideY * node->cfg.g_h, 16);
|
|
|
|
node->outU = (uint8_t*) _mm_malloc(node->strideU * node->cfg.g_h / 2, 16);
|
|
|
|
node->outV = (uint8_t*) _mm_malloc(node->strideV * node->cfg.g_h / 2, 16);
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
CUTIVIS int CHi_EncodeVP9_Stop(CHiPubNode *pubn) {
|
|
|
|
CHiEncodeVP9Node *node = (CHiEncodeVP9Node*) ((uintptr_t) pubn - offsetof(CHiEncodeVP9Node, pub));
|
|
|
|
|
|
|
|
node->state = CHiEncodeVP9Node::WAITING;
|
|
|
|
|
|
|
|
_mm_free(node->outY);
|
|
|
|
_mm_free(node->outU);
|
|
|
|
_mm_free(node->outV);
|
|
|
|
|
2024-06-30 14:43:13 +03:00
|
|
|
vpx_codec_destroy(&node->codec);
|
|
|
|
|
2024-06-26 20:47:26 +03:00
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
|
|
|
struct CHiMuxWebmNode {
|
|
|
|
CHiPubNode pub;
|
|
|
|
|
|
|
|
mkvmuxer::MkvWriter w;
|
|
|
|
mkvmuxer::Segment seg;
|
|
|
|
size_t videoTrack, audioTrack;
|
|
|
|
|
|
|
|
std::queue<CHiBSFrame> audioBacklog;
|
|
|
|
std::queue<CHiBSFrame> videoBacklog;
|
|
|
|
};
|
|
|
|
static int muxwebm_perform(CHiPubNode *pubn) {
|
|
|
|
using namespace mkvmuxer;
|
|
|
|
|
2024-06-30 14:43:13 +03:00
|
|
|
MTR_BEGIN("CHi", "muxwebm_perform");
|
|
|
|
|
2024-06-26 20:47:26 +03:00
|
|
|
CHiMuxWebmNode *alln = (CHiMuxWebmNode*) pubn;
|
|
|
|
|
|
|
|
if(pubn->sinks[1].data.linked.to) {
|
|
|
|
CHiBSFrames *opus = CHi_Crawl(&pubn->sinks[1])->data.bitstream;
|
|
|
|
for(size_t i = 0; i < opus->count; i++) {
|
|
|
|
alln->audioBacklog.push(opus->data[i]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
auto vp9bs = CHi_Crawl(&pubn->sinks[0])->data.bitstream;
|
|
|
|
if(vp9bs) {
|
|
|
|
for(size_t i = 0; i < vp9bs->count; i++) {
|
|
|
|
alln->videoBacklog.push(vp9bs->data[i]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
while(pubn->sinks[1].data.linked.to && alln->audioBacklog.size() > 0 && alln->videoBacklog.size() > 0 && alln->audioBacklog.front().timestamp <= alln->videoBacklog.front().timestamp) {
|
|
|
|
Frame frame;
|
|
|
|
frame.Init((const uint8_t*) alln->audioBacklog.front().ptr, alln->audioBacklog.front().sz);
|
|
|
|
frame.set_track_number(alln->audioTrack);
|
|
|
|
frame.set_timestamp(alln->audioBacklog.front().timestamp * 1000000L);
|
|
|
|
frame.set_is_key(true);
|
|
|
|
alln->seg.AddGenericFrame(&frame);
|
|
|
|
|
|
|
|
alln->audioBacklog.pop();
|
|
|
|
}
|
|
|
|
|
|
|
|
if(pubn->sinks[1].data.linked.to == NULL || (alln->audioBacklog.size() > 0 && alln->videoBacklog.size() > 0 && alln->audioBacklog.front().timestamp >= alln->videoBacklog.front().timestamp)) {
|
|
|
|
Frame frame;
|
|
|
|
if(!frame.Init((const uint8_t*) alln->videoBacklog.front().ptr, alln->videoBacklog.front().sz)) puts("INIT FAIL");
|
|
|
|
frame.set_track_number(alln->videoTrack);
|
|
|
|
frame.set_timestamp(alln->videoBacklog.front().timestamp * 1000000L);
|
|
|
|
frame.set_is_key(!!(alln->videoBacklog.front().flags & CUTIHI_BS_FLAG_KEY));
|
|
|
|
if(!alln->seg.AddGenericFrame(&frame)) puts("ADD FAIL");
|
|
|
|
|
|
|
|
alln->videoBacklog.pop();
|
|
|
|
}
|
|
|
|
|
2024-06-30 14:43:13 +03:00
|
|
|
MTR_END("CHi", "muxwebm_perform");
|
|
|
|
|
2024-06-26 20:47:26 +03:00
|
|
|
return 1;
|
|
|
|
}
|
2024-06-30 14:43:13 +03:00
|
|
|
static void muxwebm_destroy(CHiPubNode *pubn) {
|
|
|
|
CHiMuxWebmNode *n = (CHiMuxWebmNode*) pubn;
|
|
|
|
|
|
|
|
n->~CHiMuxWebmNode();
|
|
|
|
|
|
|
|
free(n);
|
|
|
|
}
|
2024-06-26 20:47:26 +03:00
|
|
|
CUTIVIS CHiPubNode *CHi_MuxWebm() {
|
2024-06-30 14:43:13 +03:00
|
|
|
CHiMuxWebmNode *n = (CHiMuxWebmNode*) calloc(1, sizeof(*n));
|
|
|
|
new (n) CHiMuxWebmNode;
|
2024-06-26 20:47:26 +03:00
|
|
|
n->pub.type = CUTIHI_T('CExp','Webm');
|
|
|
|
n->pub.Start = CHi_MuxWebm_Start;
|
|
|
|
n->pub.Perform = muxwebm_perform;
|
2024-06-30 14:43:13 +03:00
|
|
|
n->pub.Destroy = muxwebm_destroy;
|
2024-06-26 20:47:26 +03:00
|
|
|
n->pub.Stop = CHi_MuxWebm_Stop;
|
|
|
|
n->pub.clean = 0;
|
|
|
|
n->pub.sinks = (CHiValue*) calloc(sizeof(*n->pub.sinks), n->pub.sinkCount = 3);
|
|
|
|
n->pub.sourceCount = 0;
|
|
|
|
n->pub.sources = NULL;
|
|
|
|
return &n->pub;
|
|
|
|
}
|
|
|
|
CUTIVIS int CHi_MuxWebm_Start(CHiPubNode *pubn) {
|
|
|
|
using namespace mkvmuxer;
|
|
|
|
|
|
|
|
CHiMuxWebmNode *alln = (CHiMuxWebmNode*) pubn;
|
|
|
|
|
|
|
|
new (&alln->w) MkvWriter{};
|
|
|
|
alln->w.Open(CHi_Crawl(&pubn->sinks[CUTIHI_MUXWEBM_IN_FILENAME])->data.text);
|
|
|
|
|
|
|
|
new (&alln->seg) Segment{};
|
|
|
|
alln->seg.Init(&alln->w);
|
|
|
|
alln->seg.AccurateClusterDuration(true);
|
|
|
|
alln->seg.UseFixedSizeClusterTimecode(false);
|
|
|
|
alln->seg.set_mode(Segment::kFile);
|
|
|
|
alln->seg.OutputCues(true);
|
|
|
|
alln->seg.set_duration(pubn->ng->duration * 1000);
|
|
|
|
alln->seg.GetSegmentInfo()->set_timecode_scale(1000000);
|
|
|
|
alln->seg.GetSegmentInfo()->set_writing_app("Cuticle");
|
|
|
|
|
|
|
|
/* Hack into first frame to get resolution */
|
|
|
|
CHiPubNode *evp9 = pubn->sinks[0].data.linked.to;
|
|
|
|
CHiImage *firstFrame = (CHiImage*) CHi_Crawl(&evp9->sinks[0])->data.sample;
|
|
|
|
|
|
|
|
alln->videoTrack = alln->seg.AddVideoTrack(firstFrame->width, firstFrame->height, 0);
|
|
|
|
VideoTrack *track = (VideoTrack*) alln->seg.GetTrackByNumber(alln->videoTrack);
|
|
|
|
track->set_codec_id(CHi_Crawl(&pubn->sinks[0])->type == CUTIHI_VAL_VP9BS ? Tracks::kVp9CodecId : Tracks::kVp8CodecId);
|
|
|
|
track->set_frame_rate(30);
|
|
|
|
Colour colourspace;
|
|
|
|
colourspace.set_matrix_coefficients(Colour::MatrixCoefficients::kBt709);
|
|
|
|
colourspace.set_range(Colour::Range::kBroadcastRange);
|
|
|
|
colourspace.set_transfer_characteristics(Colour::TransferCharacteristics::kIturBt709Tc);
|
|
|
|
colourspace.set_primaries(Colour::Primaries::kIturBt709P);
|
|
|
|
track->SetColour(colourspace);
|
|
|
|
alln->seg.CuesTrack(alln->videoTrack);
|
|
|
|
|
|
|
|
if(pubn->sinks[1].data.linked.to) {
|
|
|
|
struct __attribute__((packed)) {
|
|
|
|
uint32_t magic1;
|
|
|
|
uint32_t magic2;
|
|
|
|
uint8_t version;
|
|
|
|
uint8_t channels; // NUMBER OF CHANNELS IS HARDCODED TO ONE
|
|
|
|
uint16_t preskip;
|
|
|
|
uint32_t sampleRate;
|
|
|
|
uint16_t gain;
|
|
|
|
uint8_t family;
|
|
|
|
} header = {0x7375704f, 0x64616548, 1, 1, 3840, 48000, 0, 0};
|
|
|
|
|
|
|
|
alln->audioTrack = alln->seg.AddAudioTrack(48000, 1, 0);
|
|
|
|
AudioTrack *atrack = (AudioTrack*) alln->seg.GetTrackByNumber(alln->audioTrack);
|
|
|
|
atrack->set_codec_id(Tracks::kOpusCodecId);
|
|
|
|
atrack->set_seek_pre_roll(80000000);
|
|
|
|
atrack->SetCodecPrivate((const uint8_t*) &header, sizeof(header));
|
|
|
|
}
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
CUTIVIS int CHi_MuxWebm_Stop(CHiPubNode *pubn) {
|
|
|
|
CHiMuxWebmNode *alln =(CHiMuxWebmNode*) pubn;
|
|
|
|
alln->seg.Finalize();
|
|
|
|
|
|
|
|
alln->w.Close();
|
|
|
|
|
|
|
|
return 1;
|
|
|
|
}
|