]> git.sesse.net Git - ffmpeg/blobdiff - libavcodec/libx264.c
Deprecate avctx.coded_frame
[ffmpeg] / libavcodec / libx264.c
index 0db859419f0700b259aa4a604cb3cdd6e292cf2b..7d719d4e3ae4c1dfb61fa944d60d4d1fae89d853 100644 (file)
 #include "libavutil/opt.h"
 #include "libavutil/mem.h"
 #include "libavutil/pixdesc.h"
+#include "libavutil/stereo3d.h"
 #include "avcodec.h"
 #include "internal.h"
+
+#if defined(_MSC_VER)
+#define X264_API_IMPORTS 1
+#endif
+
 #include <x264.h>
 #include <float.h>
 #include <math.h>
@@ -39,7 +45,6 @@ typedef struct X264Context {
     x264_picture_t  pic;
     uint8_t        *sei;
     int             sei_size;
-    AVFrame         out_pic;
     char *preset;
     char *tune;
     char *profile;
@@ -56,6 +61,7 @@ typedef struct X264Context {
     int weightb;
     int ssim;
     int intra_refresh;
+    int bluray_compat;
     int b_bias;
     int b_pyramid;
     int mixed_refs;
@@ -70,6 +76,7 @@ typedef struct X264Context {
     int slice_max_size;
     char *stats;
     int nal_hrd;
+    char *x264_params;
 } X264Context;
 
 static void X264_log(void *p, int level, const char *fmt, va_list args)
@@ -121,6 +128,91 @@ static int encode_nals(AVCodecContext *ctx, AVPacket *pkt,
     return 1;
 }
 
+static void reconfig_encoder(AVCodecContext *ctx, const AVFrame *frame)
+{
+    X264Context *x4 = ctx->priv_data;
+    AVFrameSideData *side_data;
+
+
+    if (x4->params.b_tff != frame->top_field_first) {
+        x4->params.b_tff = frame->top_field_first;
+        x264_encoder_reconfig(x4->enc, &x4->params);
+    }
+    if (x4->params.vui.i_sar_height != ctx->sample_aspect_ratio.den ||
+        x4->params.vui.i_sar_width  != ctx->sample_aspect_ratio.num) {
+        x4->params.vui.i_sar_height = ctx->sample_aspect_ratio.den;
+        x4->params.vui.i_sar_width  = ctx->sample_aspect_ratio.num;
+        x264_encoder_reconfig(x4->enc, &x4->params);
+    }
+
+    if (x4->params.rc.i_vbv_buffer_size != ctx->rc_buffer_size / 1000 ||
+        x4->params.rc.i_vbv_max_bitrate != ctx->rc_max_rate    / 1000) {
+        x4->params.rc.i_vbv_buffer_size = ctx->rc_buffer_size / 1000;
+        x4->params.rc.i_vbv_max_bitrate = ctx->rc_max_rate    / 1000;
+        x264_encoder_reconfig(x4->enc, &x4->params);
+    }
+
+    if (x4->params.rc.i_rc_method == X264_RC_ABR &&
+        x4->params.rc.i_bitrate != ctx->bit_rate / 1000) {
+        x4->params.rc.i_bitrate = ctx->bit_rate / 1000;
+        x264_encoder_reconfig(x4->enc, &x4->params);
+    }
+
+    if (x4->crf >= 0 &&
+        x4->params.rc.i_rc_method == X264_RC_CRF &&
+        x4->params.rc.f_rf_constant != x4->crf) {
+        x4->params.rc.f_rf_constant = x4->crf;
+        x264_encoder_reconfig(x4->enc, &x4->params);
+    }
+
+    if (x4->params.rc.i_rc_method == X264_RC_CQP &&
+        x4->params.rc.i_qp_constant != x4->cqp) {
+        x4->params.rc.i_qp_constant = x4->cqp;
+        x264_encoder_reconfig(x4->enc, &x4->params);
+    }
+
+    if (x4->crf_max >= 0 &&
+        x4->params.rc.f_rf_constant_max != x4->crf_max) {
+        x4->params.rc.f_rf_constant_max = x4->crf_max;
+        x264_encoder_reconfig(x4->enc, &x4->params);
+    }
+
+    side_data = av_frame_get_side_data(frame, AV_FRAME_DATA_STEREO3D);
+    if (side_data) {
+        AVStereo3D *stereo = (AVStereo3D *)side_data->data;
+        int fpa_type;
+
+        switch (stereo->type) {
+        case AV_STEREO3D_CHECKERBOARD:
+            fpa_type = 0;
+            break;
+        case AV_STEREO3D_COLUMNS:
+            fpa_type = 1;
+            break;
+        case AV_STEREO3D_LINES:
+            fpa_type = 2;
+            break;
+        case AV_STEREO3D_SIDEBYSIDE:
+            fpa_type = 3;
+            break;
+        case AV_STEREO3D_TOPBOTTOM:
+            fpa_type = 4;
+            break;
+        case AV_STEREO3D_FRAMESEQUENCE:
+            fpa_type = 5;
+            break;
+        default:
+            fpa_type = -1;
+            break;
+        }
+
+        if (fpa_type != x4->params.i_frame_packing) {
+            x4->params.i_frame_packing = fpa_type;
+            x264_encoder_reconfig(x4->enc, &x4->params);
+        }
+    }
+}
+
 static int X264_frame(AVCodecContext *ctx, AVPacket *pkt, const AVFrame *frame,
                       int *got_packet)
 {
@@ -147,47 +239,52 @@ static int X264_frame(AVCodecContext *ctx, AVPacket *pkt, const AVFrame *frame,
             frame->pict_type == AV_PICTURE_TYPE_P ? X264_TYPE_P :
             frame->pict_type == AV_PICTURE_TYPE_B ? X264_TYPE_B :
                                             X264_TYPE_AUTO;
-        if (x4->params.b_tff != frame->top_field_first) {
-            x4->params.b_tff = frame->top_field_first;
-            x264_encoder_reconfig(x4->enc, &x4->params);
-        }
-        if (x4->params.vui.i_sar_height != ctx->sample_aspect_ratio.den ||
-            x4->params.vui.i_sar_width  != ctx->sample_aspect_ratio.num) {
-            x4->params.vui.i_sar_height = ctx->sample_aspect_ratio.den;
-            x4->params.vui.i_sar_width  = ctx->sample_aspect_ratio.num;
-            x264_encoder_reconfig(x4->enc, &x4->params);
-        }
+        reconfig_encoder(ctx, frame);
     }
-
     do {
         if (x264_encoder_encode(x4->enc, &nal, &nnal, frame? &x4->pic: NULL, &pic_out) < 0)
-            return -1;
+            return AVERROR_UNKNOWN;
 
         ret = encode_nals(ctx, pkt, nal, nnal);
         if (ret < 0)
-            return -1;
+            return ret;
     } while (!ret && !frame && x264_encoder_delayed_frames(x4->enc));
 
     pkt->pts = pic_out.i_pts;
     pkt->dts = pic_out.i_dts;
 
+#if FF_API_CODED_FRAME
+FF_DISABLE_DEPRECATION_WARNINGS
     switch (pic_out.i_type) {
     case X264_TYPE_IDR:
     case X264_TYPE_I:
-        x4->out_pic.pict_type = AV_PICTURE_TYPE_I;
+        ctx->coded_frame->pict_type = AV_PICTURE_TYPE_I;
         break;
     case X264_TYPE_P:
-        x4->out_pic.pict_type = AV_PICTURE_TYPE_P;
+        ctx->coded_frame->pict_type = AV_PICTURE_TYPE_P;
         break;
     case X264_TYPE_B:
     case X264_TYPE_BREF:
-        x4->out_pic.pict_type = AV_PICTURE_TYPE_B;
+        ctx->coded_frame->pict_type = AV_PICTURE_TYPE_B;
         break;
     }
+FF_ENABLE_DEPRECATION_WARNINGS
+#endif
 
     pkt->flags |= AV_PKT_FLAG_KEY*pic_out.b_keyframe;
-    if (ret)
-        x4->out_pic.quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
+    if (ret) {
+        uint8_t *sd = av_packet_new_side_data(pkt, AV_PKT_DATA_QUALITY_FACTOR,
+                                              sizeof(int));
+        if (!sd)
+            return AVERROR(ENOMEM);
+        *(int *)sd = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
+
+#if FF_API_CODED_FRAME
+FF_DISABLE_DEPRECATION_WARNINGS
+        ctx->coded_frame->quality = (pic_out.i_qpplus1 - 1) * FF_QP2LAMBDA;
+FF_ENABLE_DEPRECATION_WARNINGS
+#endif
+    }
 
     *got_packet = ret;
     return 0;
@@ -198,26 +295,33 @@ static av_cold int X264_close(AVCodecContext *avctx)
     X264Context *x4 = avctx->priv_data;
 
     av_freep(&avctx->extradata);
-    av_free(x4->sei);
+    av_freep(&x4->sei);
 
-    if (x4->enc)
+    if (x4->enc) {
         x264_encoder_close(x4->enc);
+        x4->enc = NULL;
+    }
 
     return 0;
 }
 
-static int convert_pix_fmt(enum PixelFormat pix_fmt)
+static int convert_pix_fmt(enum AVPixelFormat pix_fmt)
 {
     switch (pix_fmt) {
-    case PIX_FMT_YUV420P:
-    case PIX_FMT_YUVJ420P:
-    case PIX_FMT_YUV420P9:
-    case PIX_FMT_YUV420P10: return X264_CSP_I420;
-    case PIX_FMT_YUV422P:
-    case PIX_FMT_YUV422P10: return X264_CSP_I422;
-    case PIX_FMT_YUV444P:
-    case PIX_FMT_YUV444P9:
-    case PIX_FMT_YUV444P10: return X264_CSP_I444;
+    case AV_PIX_FMT_YUV420P:
+    case AV_PIX_FMT_YUVJ420P:
+    case AV_PIX_FMT_YUV420P9:
+    case AV_PIX_FMT_YUV420P10: return X264_CSP_I420;
+    case AV_PIX_FMT_YUV422P:
+    case AV_PIX_FMT_YUVJ422P:
+    case AV_PIX_FMT_YUV422P10: return X264_CSP_I422;
+    case AV_PIX_FMT_YUV444P:
+    case AV_PIX_FMT_YUVJ444P:
+    case AV_PIX_FMT_YUV444P9:
+    case AV_PIX_FMT_YUV444P10: return X264_CSP_I444;
+    case AV_PIX_FMT_NV12:      return X264_CSP_NV12;
+    case AV_PIX_FMT_NV16:
+    case AV_PIX_FMT_NV20:      return X264_CSP_NV16;
     };
     return 0;
 }
@@ -232,7 +336,14 @@ static av_cold int X264_init(AVCodecContext *avctx)
 {
     X264Context *x4 = avctx->priv_data;
 
+#if CONFIG_LIBX262_ENCODER
+    if (avctx->codec_id == AV_CODEC_ID_MPEG2VIDEO) {
+        x4->params.b_mpeg2 = 1;
+        x264_param_default_mpeg2(&x4->params);
+    } else
+#else
     x264_param_default(&x4->params);
+#endif
 
     x4->params.b_deblocking_filter         = avctx->flags & CODEC_FLAG_LOOP_FILTER;
 
@@ -272,13 +383,14 @@ static av_cold int X264_init(AVCodecContext *avctx)
             x4->params.rc.f_rf_constant_max = x4->crf_max;
     }
 
-    if (avctx->rc_buffer_size && avctx->rc_initial_buffer_occupancy &&
+    if (avctx->rc_buffer_size && avctx->rc_initial_buffer_occupancy > 0 &&
         (avctx->rc_initial_buffer_occupancy <= avctx->rc_buffer_size)) {
         x4->params.rc.f_vbv_buffer_init =
             (float)avctx->rc_initial_buffer_occupancy / avctx->rc_buffer_size;
     }
 
-    x4->params.rc.f_ip_factor             = 1 / fabs(avctx->i_quant_factor);
+    if (avctx->i_quant_factor > 0)
+        x4->params.rc.f_ip_factor         = 1 / fabs(avctx->i_quant_factor);
     x4->params.rc.f_pb_factor             = avctx->b_quant_factor;
     x4->params.analyse.i_chroma_qp_offset = avctx->chromaoffset;
 
@@ -351,6 +463,10 @@ static av_cold int X264_init(AVCodecContext *avctx)
         x4->params.analyse.b_ssim = x4->ssim;
     if (x4->intra_refresh >= 0)
         x4->params.b_intra_refresh = x4->intra_refresh;
+    if (x4->bluray_compat >= 0) {
+        x4->params.b_bluray_compat = x4->bluray_compat;
+        x4->params.b_vfr_input = 0;
+    }
     if (x4->b_bias != INT_MIN)
         x4->params.i_bframe_bias              = x4->b_bias;
     if (x4->b_pyramid >= 0)
@@ -402,11 +518,35 @@ static av_cold int X264_init(AVCodecContext *avctx)
 
     x4->params.i_slice_count  = avctx->slices;
 
-    x4->params.vui.b_fullrange = avctx->pix_fmt == PIX_FMT_YUVJ420P;
+    x4->params.vui.b_fullrange = avctx->pix_fmt == AV_PIX_FMT_YUVJ420P ||
+                                 avctx->pix_fmt == AV_PIX_FMT_YUVJ422P ||
+                                 avctx->pix_fmt == AV_PIX_FMT_YUVJ444P ||
+                                 avctx->color_range == AVCOL_RANGE_JPEG;
+
+    // x264 validates the values internally
+    x4->params.vui.i_colorprim = avctx->color_primaries;
+    x4->params.vui.i_transfer  = avctx->color_trc;
+    x4->params.vui.i_colmatrix = avctx->colorspace;
 
     if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER)
         x4->params.b_repeat_headers = 0;
 
+    if (x4->x264_params) {
+        AVDictionary *dict    = NULL;
+        AVDictionaryEntry *en = NULL;
+
+        if (!av_dict_parse_string(&dict, x4->x264_params, "=", ":", 0)) {
+            while ((en = av_dict_get(dict, "", en, AV_DICT_IGNORE_SUFFIX))) {
+                if (x264_param_parse(&x4->params, en->key, en->value) < 0)
+                    av_log(avctx, AV_LOG_WARNING,
+                           "Error parsing option '%s = %s'.\n",
+                            en->key, en->value);
+            }
+
+            av_dict_free(&dict);
+        }
+    }
+
     // update AVCodecContext with x264 parameters
     avctx->has_b_frames = x4->params.i_bframe ?
         x4->params.i_bframe_pyramid ? 2 : 1 : 0;
@@ -417,9 +557,7 @@ static av_cold int X264_init(AVCodecContext *avctx)
 
     x4->enc = x264_encoder_open(&x4->params);
     if (!x4->enc)
-        return -1;
-
-    avctx->coded_frame = &x4->out_pic;
+        return AVERROR_UNKNOWN;
 
     if (avctx->flags & CODEC_FLAG_GLOBAL_HEADER) {
         x264_nal_t *nal;
@@ -428,6 +566,8 @@ static av_cold int X264_init(AVCodecContext *avctx)
 
         s = x264_encoder_headers(x4->enc, &nal, &nnal);
         avctx->extradata = p = av_malloc(s);
+        if (!p)
+            return AVERROR(ENOMEM);
 
         for (i = 0; i < nnal; i++) {
             /* Don't put the SEI in extradata. */
@@ -435,6 +575,8 @@ static av_cold int X264_init(AVCodecContext *avctx)
                 av_log(avctx, AV_LOG_INFO, "%s\n", nal[i].p_payload+25);
                 x4->sei_size = nal[i].i_payload;
                 x4->sei      = av_malloc(x4->sei_size);
+                if (!x4->sei)
+                    return AVERROR(ENOMEM);
                 memcpy(x4->sei, nal[i].p_payload, nal[i].i_payload);
                 continue;
             }
@@ -447,23 +589,28 @@ static av_cold int X264_init(AVCodecContext *avctx)
     return 0;
 }
 
-static const enum PixelFormat pix_fmts_8bit[] = {
-    PIX_FMT_YUV420P,
-    PIX_FMT_YUVJ420P,
-    PIX_FMT_YUV422P,
-    PIX_FMT_YUV444P,
-    PIX_FMT_NONE
+static const enum AVPixelFormat pix_fmts_8bit[] = {
+    AV_PIX_FMT_YUV420P,
+    AV_PIX_FMT_YUVJ420P,
+    AV_PIX_FMT_YUV422P,
+    AV_PIX_FMT_YUVJ422P,
+    AV_PIX_FMT_YUV444P,
+    AV_PIX_FMT_YUVJ444P,
+    AV_PIX_FMT_NV12,
+    AV_PIX_FMT_NV16,
+    AV_PIX_FMT_NONE
 };
-static const enum PixelFormat pix_fmts_9bit[] = {
-    PIX_FMT_YUV420P9,
-    PIX_FMT_YUV444P9,
-    PIX_FMT_NONE
+static const enum AVPixelFormat pix_fmts_9bit[] = {
+    AV_PIX_FMT_YUV420P9,
+    AV_PIX_FMT_YUV444P9,
+    AV_PIX_FMT_NONE
 };
-static const enum PixelFormat pix_fmts_10bit[] = {
-    PIX_FMT_YUV420P10,
-    PIX_FMT_YUV422P10,
-    PIX_FMT_YUV444P10,
-    PIX_FMT_NONE
+static const enum AVPixelFormat pix_fmts_10bit[] = {
+    AV_PIX_FMT_YUV420P10,
+    AV_PIX_FMT_YUV422P10,
+    AV_PIX_FMT_YUV444P10,
+    AV_PIX_FMT_NV20,
+    AV_PIX_FMT_NONE
 };
 
 static av_cold void X264_init_static(AVCodec *codec)
@@ -501,6 +648,7 @@ static const AVOption options[] = {
     { "smart",         NULL, 0, AV_OPT_TYPE_CONST, {.i64 = X264_WEIGHTP_SMART},  INT_MIN, INT_MAX, VE, "weightp" },
     { "ssim",          "Calculate and print SSIM stats.",                 OFFSET(ssim),          AV_OPT_TYPE_INT,    { .i64 = -1 }, -1, 1, VE },
     { "intra-refresh", "Use Periodic Intra Refresh instead of IDR frames.",OFFSET(intra_refresh),AV_OPT_TYPE_INT,    { .i64 = -1 }, -1, 1, VE },
+    { "bluray-compat", "Bluray compatibility workarounds.",               OFFSET(bluray_compat) ,AV_OPT_TYPE_INT,    { .i64 = -1 }, -1, 1, VE },
     { "b-bias",        "Influences how often B-frames are used",          OFFSET(b_bias),        AV_OPT_TYPE_INT,    { .i64 = INT_MIN}, INT_MIN, INT_MAX, VE },
     { "b-pyramid",     "Keep some B-frames as references.",               OFFSET(b_pyramid),     AV_OPT_TYPE_INT,    { .i64 = -1 }, -1, INT_MAX, VE, "b_pyramid" },
     { "none",          NULL,                                  0, AV_OPT_TYPE_CONST, {.i64 = X264_B_PYRAMID_NONE},   INT_MIN, INT_MAX, VE, "b_pyramid" },
@@ -527,20 +675,15 @@ static const AVOption options[] = {
     { "none",          NULL, 0, AV_OPT_TYPE_CONST, {.i64 = X264_NAL_HRD_NONE}, INT_MIN, INT_MAX, VE, "nal-hrd" },
     { "vbr",           NULL, 0, AV_OPT_TYPE_CONST, {.i64 = X264_NAL_HRD_VBR},  INT_MIN, INT_MAX, VE, "nal-hrd" },
     { "cbr",           NULL, 0, AV_OPT_TYPE_CONST, {.i64 = X264_NAL_HRD_CBR},  INT_MIN, INT_MAX, VE, "nal-hrd" },
+    { "x264-params",  "Override the x264 configuration using a :-separated list of key=value parameters", OFFSET(x264_params), AV_OPT_TYPE_STRING, { 0 }, 0, 0, VE },
     { NULL },
 };
 
-static const AVClass class = {
-    .class_name = "libx264",
-    .item_name  = av_default_item_name,
-    .option     = options,
-    .version    = LIBAVUTIL_VERSION_INT,
-};
-
 static const AVCodecDefault x264_defaults[] = {
     { "b",                "0" },
     { "bf",               "-1" },
     { "g",                "-1" },
+    { "i_qfactor",        "-1" },
     { "qmin",             "-1" },
     { "qmax",             "-1" },
     { "qdiff",            "-1" },
@@ -559,11 +702,22 @@ static const AVCodecDefault x264_defaults[] = {
     { "cmp",              "-1" },
     { "threads",          AV_STRINGIFY(X264_THREADS_AUTO) },
     { "thread_type",      "0" },
+    { "flags",            "+cgop" },
+    { "rc_init_occupancy","-1" },
     { NULL },
 };
 
+#if CONFIG_LIBX264_ENCODER
+static const AVClass class = {
+    .class_name = "libx264",
+    .item_name  = av_default_item_name,
+    .option     = options,
+    .version    = LIBAVUTIL_VERSION_INT,
+};
+
 AVCodec ff_libx264_encoder = {
     .name             = "libx264",
+    .long_name        = NULL_IF_CONFIG_SMALL("libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
     .type             = AVMEDIA_TYPE_VIDEO,
     .id               = AV_CODEC_ID_H264,
     .priv_data_size   = sizeof(X264Context),
@@ -571,8 +725,36 @@ AVCodec ff_libx264_encoder = {
     .encode2          = X264_frame,
     .close            = X264_close,
     .capabilities     = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
-    .long_name        = NULL_IF_CONFIG_SMALL("libx264 H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
     .priv_class       = &class,
     .defaults         = x264_defaults,
     .init_static_data = X264_init_static,
+    .caps_internal    = FF_CODEC_CAP_INIT_THREADSAFE |
+                        FF_CODEC_CAP_INIT_CLEANUP,
+};
+#endif
+
+#if CONFIG_LIBX262_ENCODER
+static const AVClass X262_class = {
+    .class_name = "libx262",
+    .item_name  = av_default_item_name,
+    .option     = options,
+    .version    = LIBAVUTIL_VERSION_INT,
+};
+
+AVCodec ff_libx262_encoder = {
+    .name             = "libx262",
+    .long_name        = NULL_IF_CONFIG_SMALL("libx262 MPEG2VIDEO"),
+    .type             = AVMEDIA_TYPE_VIDEO,
+    .id               = AV_CODEC_ID_MPEG2VIDEO,
+    .priv_data_size   = sizeof(X264Context),
+    .init             = X264_init,
+    .encode2          = X264_frame,
+    .close            = X264_close,
+    .capabilities     = CODEC_CAP_DELAY | CODEC_CAP_AUTO_THREADS,
+    .priv_class       = &X262_class,
+    .defaults         = x264_defaults,
+    .pix_fmts         = pix_fmts_8bit,
+    .caps_internal    = FF_CODEC_CAP_INIT_THREADSAFE |
+                        FF_CODEC_CAP_INIT_CLEANUP,
 };
+#endif