#include "libavutil/opt.h"
#include "avcodec.h"
#include "dsputil.h"
+#include "internal.h"
#include "mpegvideo.h"
-#include "mpegvideo_common.h"
#include "dnxhdenc.h"
#define VE AV_OPT_FLAG_VIDEO_PARAM | AV_OPT_FLAG_ENCODING_PARAM
#define DNX10BIT_QMAT_SHIFT 18 // The largest value that will not lead to overflow for 10bit samples.
static const AVOption options[]={
- {"nitris_compat", "encode with Avid Nitris compatibility", offsetof(DNXHDEncContext, nitris_compat), FF_OPT_TYPE_INT, {.dbl = 0}, 0, 1, VE},
+ {"nitris_compat", "encode with Avid Nitris compatibility", offsetof(DNXHDEncContext, nitris_compat), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, VE},
{NULL}
};
static const AVClass class = { "dnxhd", av_default_item_name, options, LIBAVUTIL_VERSION_INT };
const uint8_t *scantable= ctx->intra_scantable.scantable;
const int *qmat = ctx->q_intra_matrix[qscale];
int last_non_zero = 0;
+ int i;
ctx->dsp.fdct(block);
// Divide by 4 with rounding, to compensate scaling of DCT coefficients
block[0] = (block[0] + 2) >> 2;
- for (int i = 1; i < 64; ++i) {
+ for (i = 1; i < 64; ++i) {
int j = scantable[i];
int sign = block[j] >> 31;
int level = (block[j] ^ sign) - sign;
int i, index, bit_depth;
switch (avctx->pix_fmt) {
- case PIX_FMT_YUV422P:
+ case AV_PIX_FMT_YUV422P:
bit_depth = 8;
break;
- case PIX_FMT_YUV422P10:
+ case AV_PIX_FMT_YUV422P10:
bit_depth = 10;
break;
default:
avctx->bits_per_raw_sample = ctx->cid_table->bit_depth;
- dsputil_init(&ctx->m.dsp, avctx);
+ ff_dsputil_init(&ctx->m.dsp, avctx);
ff_dct_common_init(&ctx->m);
if (!ctx->m.dct_quantize)
- ctx->m.dct_quantize = dct_quantize_c;
+ ctx->m.dct_quantize = ff_dct_quantize_c;
if (ctx->cid_table->bit_depth == 10) {
ctx->m.dct_quantize = dnxhd_10bit_dct_quantize;
ctx->block_width_l2 = 3;
}
-#if HAVE_MMX
- ff_dnxhd_init_mmx(ctx);
-#endif
+ if (ARCH_X86)
+ ff_dnxhdenc_init_x86(ctx);
ctx->m.mb_height = (avctx->height + 15) / 16;
ctx->m.mb_width = (avctx->width + 15) / 16;
for (i = 0; i < 8; i++) {
DCTELEM *block = ctx->blocks[i];
- int last_index, overflow;
- int n = dnxhd_switch_matrix(ctx, i);
- last_index = ctx->m.dct_quantize(&ctx->m, block, i, qscale, &overflow);
+ int overflow, n = dnxhd_switch_matrix(ctx, i);
+ int last_index = ctx->m.dct_quantize(&ctx->m, block, i,
+ qscale, &overflow);
//START_TIMER;
dnxhd_encode_block(ctx, block, last_index, n);
//STOP_TIMER("encode_block");
for (mb_x = 0; mb_x < ctx->m.mb_width; ++mb_x, pix += 16) {
unsigned mb = mb_y * ctx->m.mb_width + mb_x;
int sum = ctx->m.dsp.pix_sum(pix, ctx->m.linesize);
- int varc = (ctx->m.dsp.pix_norm1(pix, ctx->m.linesize) - (((unsigned)(sum*sum))>>8)+128)>>8;
+ int varc = (ctx->m.dsp.pix_norm1(pix, ctx->m.linesize) - (((unsigned)sum*sum)>>8)+128)>>8;
ctx->mb_cmp[mb].value = varc;
ctx->mb_cmp[mb].mb = mb;
}
int sum = 0;
int sqsum = 0;
int mean, sqmean;
+ int i, j;
// Macroblocks are 16x16 pixels, unlike DCT blocks which are 8x8.
- for (int i = 0; i < 16; ++i) {
- for (int j = 0; j < 16; ++j) {
+ 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;
sum += sample;
int qscale = 1;
int mb = y*ctx->m.mb_width+x;
for (q = 1; q < avctx->qmax; q++) {
- unsigned score = ctx->mb_rc[q][mb].bits*lambda+(ctx->mb_rc[q][mb].ssd<<LAMBDA_FRAC_BITS);
+ unsigned score = ctx->mb_rc[q][mb].bits*lambda+
+ ((unsigned)ctx->mb_rc[q][mb].ssd<<LAMBDA_FRAC_BITS);
if (score < min) {
min = score;
qscale = q;
lambda = (lambda+last_higher)>>1;
else
lambda -= down_step;
- down_step *= 5; // XXX tune ?
+ down_step = FFMIN((int64_t)down_step*5, INT_MAX);
up_step = 1<<LAMBDA_FRAC_BITS;
lambda = FFMAX(1, lambda);
if (lambda == last_lower)
ctx->cur_field = frame->interlaced_frame && !frame->top_field_first;
}
-static int dnxhd_encode_picture(AVCodecContext *avctx, unsigned char *buf, int buf_size, void *data)
+static int dnxhd_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
+ const AVFrame *frame, int *got_packet)
{
DNXHDEncContext *ctx = avctx->priv_data;
int first_field = 1;
int offset, i, ret;
+ uint8_t *buf;
- if (buf_size < ctx->cid_table->frame_size) {
+ if ((ret = ff_alloc_packet(pkt, ctx->cid_table->frame_size)) < 0) {
av_log(avctx, AV_LOG_ERROR, "output buffer is too small to compress picture\n");
- return -1;
+ return ret;
}
+ buf = pkt->data;
- dnxhd_load_picture(ctx, data);
+ dnxhd_load_picture(ctx, frame);
encode_coding_unit:
for (i = 0; i < 3; i++) {
first_field = 0;
ctx->cur_field ^= 1;
buf += ctx->cid_table->coding_unit_size;
- buf_size -= ctx->cid_table->coding_unit_size;
goto encode_coding_unit;
}
ctx->frame.quality = ctx->qscale*FF_QP2LAMBDA;
- return ctx->cid_table->frame_size;
+ pkt->flags |= AV_PKT_FLAG_KEY;
+ *got_packet = 1;
+ return 0;
}
static int dnxhd_encode_end(AVCodecContext *avctx)
}
AVCodec ff_dnxhd_encoder = {
- "dnxhd",
- AVMEDIA_TYPE_VIDEO,
- CODEC_ID_DNXHD,
- sizeof(DNXHDEncContext),
- dnxhd_encode_init,
- dnxhd_encode_picture,
- dnxhd_encode_end,
- .capabilities = CODEC_CAP_SLICE_THREADS,
- .pix_fmts = (const enum PixelFormat[]){PIX_FMT_YUV422P, PIX_FMT_YUV422P10, PIX_FMT_NONE},
- .long_name = NULL_IF_CONFIG_SMALL("VC3/DNxHD"),
- .priv_class = &class,
+ .name = "dnxhd",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .id = AV_CODEC_ID_DNXHD,
+ .priv_data_size = sizeof(DNXHDEncContext),
+ .init = dnxhd_encode_init,
+ .encode2 = dnxhd_encode_picture,
+ .close = dnxhd_encode_end,
+ .capabilities = CODEC_CAP_SLICE_THREADS,
+ .pix_fmts = (const enum AVPixelFormat[]){ AV_PIX_FMT_YUV422P,
+ AV_PIX_FMT_YUV422P10,
+ AV_PIX_FMT_NONE },
+ .long_name = NULL_IF_CONFIG_SMALL("VC3/DNxHD"),
+ .priv_class = &class,
};
projects / ffmpeg / blobdiff