#include "libavutil/timer.h"
#include "avcodec.h"
-#include "dsputil.h"
+#include "blockdsp.h"
+#include "fdctdsp.h"
#include "internal.h"
#include "mpegvideo.h"
+#include "pixblockdsp.h"
#include "dnxhdenc.h"
-// The largest value that will not lead to overflow for 10bit samples.
+// The largest value that will not lead to overflow for 10-bit samples.
#define DNX10BIT_QMAT_SHIFT 18
#define RC_VARIANCE 1 // use variance or ssd for fast rc
#define LAMBDA_FRAC_BITS 10
static const AVOption options[] = {
{ "nitris_compat", "encode with Avid Nitris compatibility",
offsetof(DNXHDEncContext, nitris_compat), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, 1, VE },
+ { "ibias", "intra quant bias",
+ offsetof(DNXHDEncContext, intra_quant_bias), AV_OPT_TYPE_INT,
+ { .i64 = FF_DEFAULT_QUANT_BIAS }, INT_MIN, INT_MAX, VE },
{ NULL }
};
static const AVClass class = {
- "dnxhd",
- av_default_item_name,
- options,
- LIBAVUTIL_VERSION_INT
+ .class_name = "dnxhd",
+ .item_name = av_default_item_name,
+ .option = options,
+ .version = LIBAVUTIL_VERSION_INT,
};
static void dnxhd_8bit_get_pixels_8x4_sym(int16_t *restrict block,
const uint8_t *pixels,
- int line_size)
+ ptrdiff_t line_size)
{
int i;
for (i = 0; i < 4; i++) {
static av_always_inline
void dnxhd_10bit_get_pixels_8x4_sym(int16_t *restrict block,
const uint8_t *pixels,
- int line_size)
+ ptrdiff_t line_size)
{
int i;
int last_non_zero = 0;
int i;
- ctx->dsp.fdct(block);
+ ctx->fdsp.fdct(block);
// Divide by 4 with rounding, to compensate scaling of DCT coefficients
block[0] = (block[0] + 2) >> 2;
for (i = 1; i < 64; ++i) {
int j = scantable[i];
- int sign = block[j] >> 31;
+ int sign = FF_SIGNBIT(block[j]);
int level = (block[j] ^ sign) - sign;
level = level * qmat[j] >> DNX10BIT_QMAT_SHIFT;
block[j] = (level ^ sign) - sign;
if (ctx->cid_table->bit_depth == 8) {
for (i = 1; i < 64; i++) {
- int j = ctx->m.dsp.idct_permutation[ff_zigzag_direct[i]];
+ int j = ctx->m.idsp.idct_permutation[ff_zigzag_direct[i]];
weight_matrix[j] = ctx->cid_table->luma_weight[i];
}
- ff_convert_matrix(&ctx->m.dsp, ctx->qmatrix_l, ctx->qmatrix_l16,
- weight_matrix, ctx->m.intra_quant_bias, 1,
+ ff_convert_matrix(&ctx->m, ctx->qmatrix_l, ctx->qmatrix_l16,
+ weight_matrix, ctx->intra_quant_bias, 1,
ctx->m.avctx->qmax, 1);
for (i = 1; i < 64; i++) {
- int j = ctx->m.dsp.idct_permutation[ff_zigzag_direct[i]];
+ int j = ctx->m.idsp.idct_permutation[ff_zigzag_direct[i]];
weight_matrix[j] = ctx->cid_table->chroma_weight[i];
}
- ff_convert_matrix(&ctx->m.dsp, ctx->qmatrix_c, ctx->qmatrix_c16,
- weight_matrix, ctx->m.intra_quant_bias, 1,
+ ff_convert_matrix(&ctx->m, ctx->qmatrix_c, ctx->qmatrix_c16,
+ weight_matrix, ctx->intra_quant_bias, 1,
ctx->m.avctx->qmax, 1);
for (qscale = 1; qscale <= ctx->m.avctx->qmax; qscale++) {
// 10-bit
for (qscale = 1; qscale <= ctx->m.avctx->qmax; qscale++) {
for (i = 1; i < 64; i++) {
- int j = ctx->m.dsp.idct_permutation[ff_zigzag_direct[i]];
+ int j = ctx->m.idsp.idct_permutation[ff_zigzag_direct[i]];
/* The quantization formula from the VC-3 standard is:
* quantized = sign(block[i]) * floor(abs(block[i]/s) * p /
break;
default:
av_log(avctx, AV_LOG_ERROR,
- "pixel format is incompatible with DNxHD\n");
+ "Pixel format is incompatible with DNxHD, use yuv422p or yuv422p10.\n");
return AVERROR(EINVAL);
}
ctx->cid = ff_dnxhd_find_cid(avctx, bit_depth);
if (!ctx->cid) {
av_log(avctx, AV_LOG_ERROR,
- "video parameters incompatible with DNxHD\n");
+ "Video parameters incompatible with DNxHD, available CIDs:\n");
+ ff_dnxhd_list_cid(avctx);
return AVERROR(EINVAL);
}
av_log(avctx, AV_LOG_DEBUG, "cid %d\n", ctx->cid);
index = ff_dnxhd_get_cid_table(ctx->cid);
+ if (index < 0)
+ return index;
ctx->cid_table = &ff_dnxhd_cid_table[index];
ctx->m.avctx = avctx;
avctx->bits_per_raw_sample = ctx->cid_table->bit_depth;
- ff_dsputil_init(&ctx->m.dsp, avctx);
- ff_dct_common_init(&ctx->m);
+ ff_blockdsp_init(&ctx->bdsp);
+ ff_fdctdsp_init(&ctx->m.fdsp, avctx);
+ ff_mpv_idct_init(&ctx->m);
+ ff_mpegvideoencdsp_init(&ctx->m.mpvencdsp, avctx);
+ ff_pixblockdsp_init(&ctx->m.pdsp, avctx);
if (!ctx->m.dct_quantize)
ctx->m.dct_quantize = ff_dct_quantize_c;
ctx->m.mb_height = (avctx->height + 15) / 16;
ctx->m.mb_width = (avctx->width + 15) / 16;
- if (avctx->flags & CODEC_FLAG_INTERLACED_DCT) {
+ if (avctx->flags & AV_CODEC_FLAG_INTERLACED_DCT) {
ctx->interlaced = 1;
ctx->m.mb_height /= 2;
}
ctx->m.mb_num = ctx->m.mb_height * ctx->m.mb_width;
- if (avctx->intra_quant_bias != FF_DEFAULT_QUANT_BIAS)
- ctx->m.intra_quant_bias = avctx->intra_quant_bias;
// XXX tune lbias/cbias
- if ((ret = dnxhd_init_qmat(ctx, ctx->m.intra_quant_bias, 0)) < 0)
+ if ((ret = dnxhd_init_qmat(ctx, ctx->intra_quant_bias, 0)) < 0)
return ret;
/* Avid Nitris hardware decoder requires a minimum amount of padding
FF_ALLOCZ_OR_GOTO(ctx->m.avctx, ctx->mb_qscale,
ctx->m.mb_num * sizeof(uint8_t), fail);
- avctx->coded_frame = av_frame_alloc();
- if (!avctx->coded_frame)
- return AVERROR(ENOMEM);
-
+#if FF_API_CODED_FRAME
+FF_DISABLE_DEPRECATION_WARNINGS
avctx->coded_frame->key_frame = 1;
avctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
+FF_ENABLE_DEPRECATION_WARNINGS
+#endif
if (avctx->thread_count > MAX_THREADS) {
av_log(avctx, AV_LOG_ERROR, "too many threads\n");
static int dnxhd_write_header(AVCodecContext *avctx, uint8_t *buf)
{
DNXHDEncContext *ctx = avctx->priv_data;
- const uint8_t header_prefix[5] = { 0x00, 0x00, 0x02, 0x80, 0x01 };
+ static const uint8_t header_prefix[5] = { 0x00, 0x00, 0x02, 0x80, 0x01 };
memset(buf, 0, 640);
((mb_y << 4) * ctx->m.uvlinesize) + (mb_x << bs);
const uint8_t *ptr_v = ctx->thread[0]->src[2] +
((mb_y << 4) * ctx->m.uvlinesize) + (mb_x << bs);
- DSPContext *dsp = &ctx->m.dsp;
+ PixblockDSPContext *pdsp = &ctx->m.pdsp;
- dsp->get_pixels(ctx->blocks[0], ptr_y, ctx->m.linesize);
- dsp->get_pixels(ctx->blocks[1], ptr_y + bw, ctx->m.linesize);
- dsp->get_pixels(ctx->blocks[2], ptr_u, ctx->m.uvlinesize);
- dsp->get_pixels(ctx->blocks[3], ptr_v, ctx->m.uvlinesize);
+ pdsp->get_pixels(ctx->blocks[0], ptr_y, ctx->m.linesize);
+ pdsp->get_pixels(ctx->blocks[1], ptr_y + bw, ctx->m.linesize);
+ pdsp->get_pixels(ctx->blocks[2], ptr_u, ctx->m.uvlinesize);
+ pdsp->get_pixels(ctx->blocks[3], ptr_v, ctx->m.uvlinesize);
if (mb_y + 1 == ctx->m.mb_height && ctx->m.avctx->height == 1080) {
if (ctx->interlaced) {
ptr_v + ctx->dct_uv_offset,
ctx->m.uvlinesize);
} else {
- dsp->clear_block(ctx->blocks[4]);
- dsp->clear_block(ctx->blocks[5]);
- dsp->clear_block(ctx->blocks[6]);
- dsp->clear_block(ctx->blocks[7]);
+ ctx->bdsp.clear_block(ctx->blocks[4]);
+ ctx->bdsp.clear_block(ctx->blocks[5]);
+ ctx->bdsp.clear_block(ctx->blocks[6]);
+ ctx->bdsp.clear_block(ctx->blocks[7]);
}
} else {
- dsp->get_pixels(ctx->blocks[4],
- ptr_y + ctx->dct_y_offset, ctx->m.linesize);
- dsp->get_pixels(ctx->blocks[5],
- ptr_y + ctx->dct_y_offset + bw, ctx->m.linesize);
- dsp->get_pixels(ctx->blocks[6],
- ptr_u + ctx->dct_uv_offset, ctx->m.uvlinesize);
- dsp->get_pixels(ctx->blocks[7],
- ptr_v + ctx->dct_uv_offset, ctx->m.uvlinesize);
+ pdsp->get_pixels(ctx->blocks[4],
+ ptr_y + ctx->dct_y_offset, ctx->m.linesize);
+ pdsp->get_pixels(ctx->blocks[5],
+ ptr_y + ctx->dct_y_offset + bw, ctx->m.linesize);
+ pdsp->get_pixels(ctx->blocks[6],
+ ptr_u + ctx->dct_uv_offset, ctx->m.uvlinesize);
+ pdsp->get_pixels(ctx->blocks[7],
+ ptr_v + ctx->dct_uv_offset, ctx->m.uvlinesize);
}
}
if (avctx->mb_decision == FF_MB_DECISION_RD || !RC_VARIANCE) {
dnxhd_unquantize_c(ctx, block, i, qscale, last_index);
- ctx->m.dsp.idct(block);
+ ctx->m.idsp.idct(block);
ssd += dnxhd_ssd_block(block, src_block);
}
}
int varc;
if (!partial_last_row && mb_x * 16 <= avctx->width - 16) {
- sum = ctx->m.dsp.pix_sum(pix, ctx->m.linesize);
- varc = ctx->m.dsp.pix_norm1(pix, ctx->m.linesize);
+ sum = ctx->m.mpvencdsp.pix_sum(pix, ctx->m.linesize);
+ varc = ctx->m.mpvencdsp.pix_norm1(pix, ctx->m.linesize);
} else {
int bw = FFMIN(avctx->width - 16 * mb_x, 16);
int bh = FFMIN((avctx->height >> ctx->interlaced) - 16 * mb_y, 16);
ctx->mb_cmp[mb].mb = mb;
}
} else { // 10-bit
- int const linesize = ctx->m.linesize >> 1;
+ const int linesize = ctx->m.linesize >> 1;
for (mb_x = 0; mb_x < ctx->m.mb_width; ++mb_x) {
uint16_t *pix = (uint16_t *)ctx->thread[0]->src[0] +
((mb_y << 4) * linesize) + (mb_x << 4);
for (i = 0; i < 16; ++i) {
for (j = 0; j < 16; ++j) {
// Turn 16-bit pixels into 10-bit ones.
- int const sample = (unsigned) pix[j] >> 6;
+ const int sample = (unsigned) pix[j] >> 6;
sum += sample;
sqsum += sample * sample;
// 2^10 * 2^10 * 16 * 16 = 2^28, which is less than INT_MAX
if (bits > ctx->frame_bits)
break;
}
- // av_dlog(ctx->m.avctx,
- // "lambda %d, up %u, down %u, bits %d, frame %d\n",
- // lambda, last_higher, last_lower, bits, ctx->frame_bits);
if (end) {
if (bits > ctx->frame_bits)
return AVERROR(EINVAL);
down_step = 1<<LAMBDA_FRAC_BITS;
}
}
- //av_dlog(ctx->m.avctx, "out lambda %d\n", lambda);
ctx->lambda = lambda;
return 0;
}
if (bits > ctx->frame_bits)
break;
}
- // av_dlog(ctx->m.avctx,
- // "%d, qscale %d, bits %d, frame %d, higher %d, lower %d\n",
- // ctx->m.avctx->frame_number, qscale, bits, ctx->frame_bits,
- // last_higher, last_lower);
if (bits < ctx->frame_bits) {
if (qscale == 1)
return 1;
return AVERROR(EINVAL);
}
}
- //av_dlog(ctx->m.avctx, "out qscale %d\n", qscale);
ctx->qscale = qscale;
return 0;
}
ctx->thread[i]->dct_uv_offset = ctx->m.uvlinesize*8;
}
+#if FF_API_CODED_FRAME
+FF_DISABLE_DEPRECATION_WARNINGS
ctx->m.avctx->coded_frame->interlaced_frame = frame->interlaced_frame;
+FF_ENABLE_DEPRECATION_WARNINGS
+#endif
ctx->cur_field = frame->interlaced_frame && !frame->top_field_first;
}
DNXHDEncContext *ctx = avctx->priv_data;
int first_field = 1;
int offset, i, ret;
- uint8_t *buf;
+ uint8_t *buf, *sd;
if ((ret = ff_alloc_packet(pkt, ctx->cid_table->frame_size)) < 0) {
av_log(avctx, AV_LOG_ERROR,
goto encode_coding_unit;
}
+#if FF_API_CODED_FRAME
+FF_DISABLE_DEPRECATION_WARNINGS
avctx->coded_frame->quality = ctx->qscale * FF_QP2LAMBDA;
+FF_ENABLE_DEPRECATION_WARNINGS
+#endif
+
+ sd = av_packet_new_side_data(pkt, AV_PKT_DATA_QUALITY_FACTOR, sizeof(int));
+ if (!sd)
+ return AVERROR(ENOMEM);
+ *(int *)sd = ctx->qscale * FF_QP2LAMBDA;
pkt->flags |= AV_PKT_FLAG_KEY;
*got_packet = 1;
for (i = 1; i < avctx->thread_count; i++)
av_freep(&ctx->thread[i]);
- av_frame_free(&avctx->coded_frame);
-
return 0;
}
.init = dnxhd_encode_init,
.encode2 = dnxhd_encode_picture,
.close = dnxhd_encode_end,
- .capabilities = CODEC_CAP_SLICE_THREADS,
+ .capabilities = AV_CODEC_CAP_SLICE_THREADS,
.pix_fmts = (const enum AVPixelFormat[]) {
AV_PIX_FMT_YUV422P,
AV_PIX_FMT_YUV422P10,
projects / ffmpeg / blobdiff