ivfenc

00001 /*
00002  *  Copyright (c) 2010 The VP8 project authors. All Rights Reserved.
00003  *
00004  *  Use of this source code is governed by a BSD-style license and patent
00005  *  grant that can be found in the LICENSE file in the root of the source
00006  *  tree. All contributing project authors may be found in the AUTHORS
00007  *  file in the root of the source tree.
00008  */
00009 
00010 
00011 /* This is a simple program that encodes YV12 files and generates ivf
00012  * files using the new interface.
00013  */
00014 #define USE_POSIX_MMAP HAVE_SYS_MMAN_H
00015 
00016 #include <stdio.h>
00017 #include <stdlib.h>
00018 #include <stdarg.h>
00019 #include <string.h>
00020 #include "vpx_encoder.h"
00021 #if USE_POSIX_MMAP
00022 #include <sys/types.h>
00023 #include <sys/stat.h>
00024 #include <sys/mman.h>
00025 #include <fcntl.h>
00026 #include <unistd.h>
00027 #endif
00028 #if CONFIG_VP8_ENCODER
00029 #include "vp8cx.h"
00030 #endif
00031 #include "vpx_ports/mem_ops.h"
00032 #include "vpx_ports/vpx_timer.h"
00033 
00034 static const char *exec_name;
00035 
00036 static const struct codec_item
00037 {
00038     char const              *name;
00039     const vpx_codec_iface_t *iface;
00040     unsigned int             fourcc;
00041 } codecs[] =
00042 {
00043 #if CONFIG_VP8_ENCODER
00044     {"vp8",  &vpx_codec_vp8_cx_algo, 0x30385056},
00045 #endif
00046 };
00047 
00048 static void usage_exit();
00049 
00050 void die(const char *fmt, ...)
00051 {
00052     va_list ap;
00053     va_start(ap, fmt);
00054     vprintf(fmt, ap);
00055     printf("\n");
00056     usage_exit();
00057 }
00058 
00059 static void ctx_exit_on_error(vpx_codec_ctx_t *ctx, const char *s)
00060 {
00061     if (ctx->err)
00062     {
00063         const char *detail = vpx_codec_error_detail(ctx);
00064 
00065         printf("%s: %s\n", s, vpx_codec_error(ctx));
00066 
00067         if (detail)
00068             printf("    %s\n", detail);
00069 
00070         exit(EXIT_FAILURE);
00071     }
00072 }
00073 
00074 /* This structure is used to abstract the different ways of handling
00075  * first pass statistics.
00076  */
00077 typedef struct
00078 {
00079     vpx_fixed_buf_t buf;
00080     int             pass;
00081     FILE           *file;
00082     char           *buf_ptr;
00083     size_t          buf_alloc_sz;
00084 } stats_io_t;
00085 
00086 int stats_open_file(stats_io_t *stats, const char *fpf, int pass)
00087 {
00088     int res;
00089 
00090     stats->pass = pass;
00091 
00092     if (pass == 0)
00093     {
00094         stats->file = fopen(fpf, "wb");
00095         stats->buf.sz = 0;
00096         stats->buf.buf = NULL,
00097                    res = (stats->file != NULL);
00098     }
00099     else
00100     {
00101 #if 0
00102 #elif USE_POSIX_MMAP
00103         struct stat stat_buf;
00104         int fd;
00105 
00106         fd = open(fpf, O_RDONLY);
00107         stats->file = fdopen(fd, "rb");
00108         fstat(fd, &stat_buf);
00109         stats->buf.sz = stat_buf.st_size;
00110         stats->buf.buf = mmap(NULL, stats->buf.sz, PROT_READ, MAP_PRIVATE,
00111                               fd, 0);
00112         res = (stats->buf.buf != NULL);
00113 #else
00114         size_t nbytes;
00115 
00116         stats->file = fopen(fpf, "rb");
00117 
00118         if (fseek(stats->file, 0, SEEK_END))
00119         {
00120             fprintf(stderr, "First-pass stats file must be seekable!\n");
00121             exit(EXIT_FAILURE);
00122         }
00123 
00124         stats->buf.sz = stats->buf_alloc_sz = ftell(stats->file);
00125         rewind(stats->file);
00126 
00127         stats->buf.buf = malloc(stats->buf_alloc_sz);
00128 
00129         if (!stats->buf.buf)
00130         {
00131             fprintf(stderr, "Failed to allocate first-pass stats buffer (%d bytes)\n",
00132                     stats->buf_alloc_sz);
00133             exit(EXIT_FAILURE);
00134         }
00135 
00136         nbytes = fread(stats->buf.buf, 1, stats->buf.sz, stats->file);
00137         res = (nbytes == stats->buf.sz);
00138 #endif
00139     }
00140 
00141     return res;
00142 }
00143 
00144 int stats_open_mem(stats_io_t *stats, int pass)
00145 {
00146     int res;
00147     stats->pass = pass;
00148 
00149     if (!pass)
00150     {
00151         stats->buf.sz = 0;
00152         stats->buf_alloc_sz = 64 * 1024;
00153         stats->buf.buf = malloc(stats->buf_alloc_sz);
00154     }
00155 
00156     stats->buf_ptr = stats->buf.buf;
00157     res = (stats->buf.buf != NULL);
00158     return res;
00159 }
00160 
00161 
00162 void stats_close(stats_io_t *stats)
00163 {
00164     if (stats->file)
00165     {
00166         if (stats->pass == 1)
00167         {
00168 #if 0
00169 #elif USE_POSIX_MMAP
00170             munmap(stats->buf.buf, stats->buf.sz);
00171 #else
00172             free(stats->buf.buf);
00173 #endif
00174         }
00175 
00176         fclose(stats->file);
00177         stats->file = NULL;
00178     }
00179     else
00180     {
00181         if (stats->pass == 1)
00182             free(stats->buf.buf);
00183     }
00184 }
00185 
00186 void stats_write(stats_io_t *stats, const void *pkt, size_t len)
00187 {
00188     if (stats->file)
00189     {
00190         fwrite(pkt, 1, len, stats->file);
00191     }
00192     else
00193     {
00194         if (stats->buf.sz + len > stats->buf_alloc_sz)
00195         {
00196             size_t  new_sz = stats->buf_alloc_sz + 64 * 1024;
00197             char   *new_ptr = realloc(stats->buf.buf, new_sz);
00198 
00199             if (new_ptr)
00200             {
00201                 stats->buf_ptr = new_ptr + (stats->buf_ptr - (char *)stats->buf.buf);
00202                 stats->buf.buf = new_ptr;
00203                 stats->buf_alloc_sz = new_sz;
00204             } /* else ... */
00205         }
00206 
00207         memcpy(stats->buf_ptr, pkt, len);
00208         stats->buf.sz += len;
00209         stats->buf_ptr += len;
00210     }
00211 }
00212 
00213 vpx_fixed_buf_t stats_get(stats_io_t *stats)
00214 {
00215     return stats->buf;
00216 }
00217 
00218 #define IVF_FRAME_HDR_SZ (4+8) /* 4 byte size + 8 byte timestamp */
00219 static int read_frame(FILE *f, vpx_image_t *img, unsigned int is_ivf)
00220 {
00221     int plane = 0;
00222 
00223     if (is_ivf)
00224     {
00225         char junk[IVF_FRAME_HDR_SZ];
00226 
00227         /* Skip the frame header. We know how big the frame should be. See
00228          * write_ivf_frame_header() for documentation on the frame header
00229          * layout.
00230          */
00231         fread(junk, 1, IVF_FRAME_HDR_SZ, f);
00232     }
00233 
00234     for (plane = 0; plane < 3; plane++)
00235     {
00236         unsigned char *ptr;
00237         int w = (plane ? (1 + img->d_w) / 2 : img->d_w);
00238         int h = (plane ? (1 + img->d_h) / 2 : img->d_h);
00239         int r;
00240 
00241         /* Determine the correct plane based on the image format. The for-loop
00242          * always counts in Y,U,V order, but this may not match the order of
00243          * the data on disk.
00244          */
00245         switch (plane)
00246         {
00247         case 1:
00248             ptr = img->planes[img->fmt==IMG_FMT_YV12? PLANE_V : PLANE_U];
00249             break;
00250         case 2:
00251             ptr = img->planes[img->fmt==IMG_FMT_YV12?PLANE_U : PLANE_V];
00252             break;
00253         default:
00254             ptr = img->planes[plane];
00255         }
00256 
00257         for (r = 0; r < h; r++)
00258         {
00259             fread(ptr, 1, w, f);
00260             ptr += img->stride[plane];
00261         }
00262     }
00263 
00264     return !feof(f);
00265 }
00266 
00267 
00268 #define IVF_FILE_HDR_SZ (32)
00269 unsigned int file_is_ivf(FILE *infile,
00270                          unsigned int *fourcc,
00271                          unsigned int *width,
00272                          unsigned int *height)
00273 {
00274     char raw_hdr[IVF_FILE_HDR_SZ];
00275     int is_ivf = 0;
00276 
00277     /* See write_ivf_file_header() for more documentation on the file header
00278      * layout.
00279      */
00280     if (fread(raw_hdr, 1, IVF_FILE_HDR_SZ, infile) == IVF_FILE_HDR_SZ)
00281     {
00282         if (raw_hdr[0] == 'D' && raw_hdr[1] == 'K'
00283             && raw_hdr[2] == 'I' && raw_hdr[3] == 'F')
00284         {
00285             is_ivf = 1;
00286 
00287             if (mem_get_le16(raw_hdr + 4) != 0)
00288                 fprintf(stderr, "Error: Unrecognized IVF version! This file may not"
00289                         " decode properly.");
00290 
00291             *fourcc = mem_get_le32(raw_hdr + 8);
00292         }
00293     }
00294 
00295     if (is_ivf)
00296     {
00297         *width = mem_get_le16(raw_hdr + 12);
00298         *height = mem_get_le16(raw_hdr + 14);
00299     }
00300     else
00301         rewind(infile);
00302 
00303     return is_ivf;
00304 }
00305 
00306 
00307 static void write_ivf_file_header(FILE *outfile,
00308                                   const vpx_codec_enc_cfg_t *cfg,
00309                                   unsigned int fourcc,
00310                                   int frame_cnt)
00311 {
00312     char header[32];
00313 
00314     if (cfg->g_pass != VPX_RC_ONE_PASS && cfg->g_pass != VPX_RC_LAST_PASS)
00315         return;
00316 
00317     header[0] = 'D';
00318     header[1] = 'K';
00319     header[2] = 'I';
00320     header[3] = 'F';
00321     mem_put_le16(header + 4,  0);                 /* version */
00322     mem_put_le16(header + 6,  32);                /* headersize */
00323     mem_put_le32(header + 8,  fourcc);            /* headersize */
00324     mem_put_le16(header + 12, cfg->g_w);          /* width */
00325     mem_put_le16(header + 14, cfg->g_h);          /* height */
00326     mem_put_le32(header + 16, cfg->g_timebase.den); /* rate */
00327     mem_put_le32(header + 20, cfg->g_timebase.num); /* scale */
00328     mem_put_le32(header + 24, frame_cnt);         /* length */
00329     mem_put_le32(header + 28, 0);                 /* unused */
00330 
00331     fwrite(header, 1, 32, outfile);
00332 }
00333 
00334 
00335 static void write_ivf_frame_header(FILE *outfile,
00336                                    const vpx_codec_cx_pkt_t *pkt)
00337 {
00338     char             header[12];
00339     vpx_codec_pts_t  pts;
00340 
00341     if (pkt->kind != VPX_CODEC_CX_FRAME_PKT)
00342         return;
00343 
00344     pts = pkt->data.frame.pts;
00345     mem_put_le32(header, pkt->data.frame.sz);
00346     mem_put_le32(header + 4, pts & 0xFFFFFFFF);
00347     mem_put_le32(header + 8, pts >> 32);
00348 
00349     fwrite(header, 1, 12, outfile);
00350 }
00351 
00352 #include "args.h"
00353 
00354 static const arg_def_t use_yv12 = ARG_DEF(NULL, "yv12", 0,
00355                                   "Input file is YV12 ");
00356 static const arg_def_t use_i420 = ARG_DEF(NULL, "i420", 0,
00357                                   "Input file is I420 (default)");
00358 static const arg_def_t codecarg = ARG_DEF(NULL, "codec", 1,
00359                                   "Codec to use");
00360 static const arg_def_t passes           = ARG_DEF("p", "passes", 1,
00361         "Number of passes (1/2)");
00362 static const arg_def_t pass_arg         = ARG_DEF(NULL, "pass", 1,
00363         "Pass to execute (1/2)");
00364 static const arg_def_t fpf_name         = ARG_DEF(NULL, "fpf", 1,
00365         "First pass statistics file name");
00366 static const arg_def_t limit = ARG_DEF(NULL, "limit", 1,
00367                                        "Stop encoding after n input frames");
00368 static const arg_def_t deadline         = ARG_DEF("d", "deadline", 1,
00369         "Deadline per frame (usec)");
00370 static const arg_def_t best_dl          = ARG_DEF(NULL, "best", 0,
00371         "Use Best Quality Deadline");
00372 static const arg_def_t good_dl          = ARG_DEF(NULL, "good", 0,
00373         "Use Good Quality Deadline");
00374 static const arg_def_t rt_dl            = ARG_DEF(NULL, "rt", 0,
00375         "Use Realtime Quality Deadline");
00376 static const arg_def_t verbosearg       = ARG_DEF("v", "verbose", 0,
00377         "Show encoder parameters");
00378 static const arg_def_t psnrarg          = ARG_DEF(NULL, "psnr", 0,
00379         "Show PSNR in status line");
00380 static const arg_def_t *main_args[] =
00381 {
00382     &codecarg, &passes, &pass_arg, &fpf_name, &limit, &deadline, &best_dl, &good_dl, &rt_dl,
00383     &verbosearg, &psnrarg,
00384     NULL
00385 };
00386 
00387 static const arg_def_t usage            = ARG_DEF("u", "usage", 1,
00388         "Usage profile number to use");
00389 static const arg_def_t threads          = ARG_DEF("t", "threads", 1,
00390         "Max number of threads to use");
00391 static const arg_def_t profile          = ARG_DEF(NULL, "profile", 1,
00392         "Bitstream profile number to use");
00393 static const arg_def_t width            = ARG_DEF("w", "width", 1,
00394         "Frame width");
00395 static const arg_def_t height           = ARG_DEF("h", "height", 1,
00396         "Frame height");
00397 static const arg_def_t timebase         = ARG_DEF(NULL, "timebase", 1,
00398         "Stream timebase (frame duration)");
00399 static const arg_def_t error_resilient  = ARG_DEF(NULL, "error-resilient", 1,
00400         "Enable error resiliency features");
00401 static const arg_def_t lag_in_frames    = ARG_DEF(NULL, "lag-in-frames", 1,
00402         "Max number of frames to lag");
00403 
00404 static const arg_def_t *global_args[] =
00405 {
00406     &use_yv12, &use_i420, &usage, &threads, &profile,
00407     &width, &height, &timebase, &error_resilient,
00408     &lag_in_frames, NULL
00409 };
00410 
00411 static const arg_def_t dropframe_thresh   = ARG_DEF(NULL, "drop-frame", 1,
00412         "Temporal resampling threshold (buf %)");
00413 static const arg_def_t resize_allowed     = ARG_DEF(NULL, "resize-allowed", 1,
00414         "Spatial resampling enabled (bool)");
00415 static const arg_def_t resize_up_thresh   = ARG_DEF(NULL, "resize-up", 1,
00416         "Upscale threshold (buf %)");
00417 static const arg_def_t resize_down_thresh = ARG_DEF(NULL, "resize-down", 1,
00418         "Downscale threshold (buf %)");
00419 static const arg_def_t end_usage          = ARG_DEF(NULL, "end-usage", 1,
00420         "VBR=0 | CBR=1");
00421 static const arg_def_t target_bitrate     = ARG_DEF(NULL, "target-bitrate", 1,
00422         "Bitrate (kbps)");
00423 static const arg_def_t min_quantizer      = ARG_DEF(NULL, "min-q", 1,
00424         "Minimum (best) quantizer");
00425 static const arg_def_t max_quantizer      = ARG_DEF(NULL, "max-q", 1,
00426         "Maximum (worst) quantizer");
00427 static const arg_def_t undershoot_pct     = ARG_DEF(NULL, "undershoot-pct", 1,
00428         "Datarate undershoot (min) target (%)");
00429 static const arg_def_t overshoot_pct      = ARG_DEF(NULL, "overshoot-pct", 1,
00430         "Datarate overshoot (max) target (%)");
00431 static const arg_def_t buf_sz             = ARG_DEF(NULL, "buf-sz", 1,
00432         "Client buffer size (ms)");
00433 static const arg_def_t buf_initial_sz     = ARG_DEF(NULL, "buf-initial-sz", 1,
00434         "Client initial buffer size (ms)");
00435 static const arg_def_t buf_optimal_sz     = ARG_DEF(NULL, "buf-optimal-sz", 1,
00436         "Client optimal buffer size (ms)");
00437 static const arg_def_t *rc_args[] =
00438 {
00439     &dropframe_thresh, &resize_allowed, &resize_up_thresh, &resize_down_thresh,
00440     &end_usage, &target_bitrate, &min_quantizer, &max_quantizer,
00441     &undershoot_pct, &overshoot_pct, &buf_sz, &buf_initial_sz, &buf_optimal_sz,
00442     NULL
00443 };
00444 
00445 
00446 static const arg_def_t bias_pct = ARG_DEF(NULL, "bias-pct", 1,
00447                                   "CBR/VBR bias (0=CBR, 100=VBR)");
00448 static const arg_def_t minsection_pct = ARG_DEF(NULL, "minsection-pct", 1,
00449                                         "GOP min bitrate (% of target)");
00450 static const arg_def_t maxsection_pct = ARG_DEF(NULL, "maxsection-pct", 1,
00451                                         "GOP max bitrate (% of target)");
00452 static const arg_def_t *rc_twopass_args[] =
00453 {
00454     &bias_pct, &minsection_pct, &maxsection_pct, NULL
00455 };
00456 
00457 
00458 static const arg_def_t kf_min_dist = ARG_DEF(NULL, "kf-min-dist", 1,
00459                                      "Minimum keyframe interval (frames)");
00460 static const arg_def_t kf_max_dist = ARG_DEF(NULL, "kf-max-dist", 1,
00461                                      "Maximum keyframe interval (frames)");
00462 static const arg_def_t *kf_args[] =
00463 {
00464     &kf_min_dist, &kf_max_dist, NULL
00465 };
00466 
00467 
00468 #if CONFIG_VP8_ENCODER
00469 static const arg_def_t noise_sens = ARG_DEF(NULL, "noise-sensitivity", 1,
00470                                     "Noise sensitivity (frames to blur)");
00471 static const arg_def_t sharpness = ARG_DEF(NULL, "sharpness", 1,
00472                                    "Filter sharpness (0-7)");
00473 static const arg_def_t static_thresh = ARG_DEF(NULL, "static-thresh", 1,
00474                                        "Motion detection threshold");
00475 #endif
00476 
00477 #if CONFIG_VP8_ENCODER
00478 static const arg_def_t cpu_used = ARG_DEF(NULL, "cpu-used", 1,
00479                                   "CPU Used (-16..16)");
00480 #endif
00481 
00482 
00483 #if CONFIG_VP8_ENCODER
00484 static const arg_def_t token_parts = ARG_DEF(NULL, "token-parts", 1,
00485                                      "Number of token partitions to use, log2");
00486 static const arg_def_t auto_altref = ARG_DEF(NULL, "auto-alt-ref", 1,
00487                                      "Enable automatic alt reference frames");
00488 static const arg_def_t arnr_maxframes = ARG_DEF(NULL, "arnr-maxframes", 1,
00489                                         "alt_ref Max Frames");
00490 static const arg_def_t arnr_strength = ARG_DEF(NULL, "arnr-strength", 1,
00491                                        "alt_ref Strength");
00492 static const arg_def_t arnr_type = ARG_DEF(NULL, "arnr-type", 1,
00493                                    "alt_ref Type");
00494 
00495 static const arg_def_t *vp8_args[] =
00496 {
00497     &cpu_used, &auto_altref, &noise_sens, &sharpness, &static_thresh,
00498     &token_parts, &arnr_maxframes, &arnr_strength, &arnr_type, NULL
00499 };
00500 static const int vp8_arg_ctrl_map[] =
00501 {
00502     VP8E_SET_CPUUSED, VP8E_SET_ENABLEAUTOALTREF,
00503     VP8E_SET_NOISE_SENSITIVITY, VP8E_SET_SHARPNESS, VP8E_SET_STATIC_THRESHOLD,
00504     VP8E_SET_TOKEN_PARTITIONS,
00505     VP8E_SET_ARNR_MAXFRAMES, VP8E_SET_ARNR_STRENGTH , VP8E_SET_ARNR_TYPE, 0
00506 };
00507 #endif
00508 
00509 static const arg_def_t *no_args[] = { NULL };
00510 
00511 static void usage_exit()
00512 {
00513     int i;
00514 
00515     printf("Usage: %s <options> src_filename dst_filename\n", exec_name);
00516 
00517     printf("\n_options:\n");
00518     arg_show_usage(stdout, main_args);
00519     printf("\n_encoder Global Options:\n");
00520     arg_show_usage(stdout, global_args);
00521     printf("\n_rate Control Options:\n");
00522     arg_show_usage(stdout, rc_args);
00523     printf("\n_twopass Rate Control Options:\n");
00524     arg_show_usage(stdout, rc_twopass_args);
00525     printf("\n_keyframe Placement Options:\n");
00526     arg_show_usage(stdout, kf_args);
00527 #if CONFIG_VP8_ENCODER
00528     printf("\n_vp8 Specific Options:\n");
00529     arg_show_usage(stdout, vp8_args);
00530 #endif
00531     printf("\n"
00532            "Included encoders:\n"
00533            "\n");
00534 
00535     for (i = 0; i < sizeof(codecs) / sizeof(codecs[0]); i++)
00536         printf("    %-6s - %s\n",
00537                codecs[i].name,
00538                vpx_codec_iface_name(codecs[i].iface));
00539 
00540     exit(EXIT_FAILURE);
00541 }
00542 
00543 #define ARG_CTRL_CNT_MAX 10
00544 
00545 
00546 int main(int argc, const char **argv_)
00547 {
00548     vpx_codec_ctx_t        encoder;
00549     const char                  *in_fn = NULL, *out_fn = NULL, *stats_fn = NULL;
00550     int                    i;
00551     FILE                  *infile, *outfile;
00552     vpx_codec_enc_cfg_t    cfg;
00553     vpx_codec_err_t        res;
00554     int                    pass, one_pass_only = 0;
00555     stats_io_t             stats;
00556     vpx_image_t            raw;
00557     const struct codec_item  *codec = codecs;
00558     int                    frame_avail, got_data;
00559 
00560     struct arg               arg;
00561     char                   **argv, **argi, **argj;
00562     int                      arg_usage = 0, arg_passes = 1, arg_deadline = 0;
00563     int                      arg_ctrls[ARG_CTRL_CNT_MAX][2], arg_ctrl_cnt = 0;
00564     int                      arg_limit = 0;
00565     static const arg_def_t **ctrl_args = no_args;
00566     static const int        *ctrl_args_map = NULL;
00567     int                      verbose = 0, show_psnr = 0;
00568     int                      arg_use_i420 = 1;
00569     unsigned long            cx_time = 0;
00570     unsigned int             is_ivf, fourcc;
00571 
00572     exec_name = argv_[0];
00573 
00574     if (argc < 3)
00575         usage_exit();
00576 
00577 
00578     /* First parse the codec and usage values, because we want to apply other
00579      * parameters on top of the default configuration provided by the codec.
00580      */
00581     argv = argv_dup(argc - 1, argv_ + 1);
00582 
00583     for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step)
00584     {
00585         arg.argv_step = 1;
00586 
00587         if (arg_match(&arg, &codecarg, argi))
00588         {
00589             int j, k = -1;
00590 
00591             for (j = 0; j < sizeof(codecs) / sizeof(codecs[0]); j++)
00592                 if (!strcmp(codecs[j].name, arg.val))
00593                     k = j;
00594 
00595             if (k >= 0)
00596                 codec = codecs + k;
00597             else
00598                 die("Error: Unrecognized argument (%s) to --codec\n",
00599                     arg.val);
00600 
00601         }
00602         else if (arg_match(&arg, &passes, argi))
00603         {
00604             arg_passes = arg_parse_uint(&arg);
00605 
00606             if (arg_passes < 1 || arg_passes > 2)
00607                 die("Error: Invalid number of passes (%d)\n", arg_passes);
00608         }
00609         else if (arg_match(&arg, &pass_arg, argi))
00610         {
00611             one_pass_only = arg_parse_uint(&arg);
00612 
00613             if (one_pass_only < 1 || one_pass_only > 2)
00614                 die("Error: Invalid pass selected (%d)\n", one_pass_only);
00615         }
00616         else if (arg_match(&arg, &fpf_name, argi))
00617             stats_fn = arg.val;
00618         else if (arg_match(&arg, &usage, argi))
00619             arg_usage = arg_parse_uint(&arg);
00620         else if (arg_match(&arg, &deadline, argi))
00621             arg_deadline = arg_parse_uint(&arg);
00622         else if (arg_match(&arg, &best_dl, argi))
00623             arg_deadline = VPX_DL_BEST_QUALITY;
00624         else if (arg_match(&arg, &good_dl, argi))
00625             arg_deadline = VPX_DL_GOOD_QUALITY;
00626         else if (arg_match(&arg, &rt_dl, argi))
00627             arg_deadline = VPX_DL_REALTIME;
00628         else if (arg_match(&arg, &use_yv12, argi))
00629         {
00630             arg_use_i420 = 0;
00631         }
00632         else if (arg_match(&arg, &use_i420, argi))
00633         {
00634             arg_use_i420 = 1;
00635         }
00636         else if (arg_match(&arg, &verbosearg, argi))
00637             verbose = 1;
00638         else if (arg_match(&arg, &limit, argi))
00639             arg_limit = arg_parse_uint(&arg);
00640         else if (arg_match(&arg, &psnrarg, argi))
00641             show_psnr = 1;
00642         else
00643             argj++;
00644     }
00645 
00646     /* Ensure that --passes and --pass are consistent. If --pass is set and --passes=2,
00647      * ensure --fpf was set.
00648      */
00649     if (one_pass_only)
00650     {
00651         /* DWIM: Assume the user meant passes=2 if pass=2 is specified */
00652         if (one_pass_only > arg_passes)
00653         {
00654             printf("Warning: Assuming --pass=%d implies --passes=%d\n",
00655                    one_pass_only, one_pass_only);
00656             arg_passes = one_pass_only;
00657         }
00658 
00659         if (arg_passes == 2 && !stats_fn)
00660             die("Must specify --fpf when --pass=%d and --passes=2\n", one_pass_only);
00661     }
00662 
00663     /* Populate encoder configuration */
00664     res = vpx_codec_enc_config_default(codec->iface, &cfg, arg_usage);
00665 
00666     if (res)
00667     {
00668         printf("Failed to get config: %s\n", vpx_codec_err_to_string(res));
00669         return EXIT_FAILURE;
00670     }
00671 
00672     /* Now parse the remainder of the parameters. */
00673     for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step)
00674     {
00675         arg.argv_step = 1;
00676 
00677         if (0);
00678         else if (arg_match(&arg, &threads, argi))
00679             cfg.g_threads = arg_parse_uint(&arg);
00680         else if (arg_match(&arg, &profile, argi))
00681             cfg.g_profile = arg_parse_uint(&arg);
00682         else if (arg_match(&arg, &width, argi))
00683             cfg.g_w = arg_parse_uint(&arg);
00684         else if (arg_match(&arg, &height, argi))
00685             cfg.g_h = arg_parse_uint(&arg);
00686         else if (arg_match(&arg, &timebase, argi))
00687             cfg.g_timebase = arg_parse_rational(&arg);
00688         else if (arg_match(&arg, &error_resilient, argi))
00689             cfg.g_error_resilient = arg_parse_uint(&arg);
00690         else if (arg_match(&arg, &lag_in_frames, argi))
00691             cfg.g_lag_in_frames = arg_parse_uint(&arg);
00692         else if (arg_match(&arg, &dropframe_thresh, argi))
00693             cfg.rc_dropframe_thresh = arg_parse_uint(&arg);
00694         else if (arg_match(&arg, &resize_allowed, argi))
00695             cfg.rc_resize_allowed = arg_parse_uint(&arg);
00696         else if (arg_match(&arg, &resize_up_thresh, argi))
00697             cfg.rc_resize_up_thresh = arg_parse_uint(&arg);
00698         else if (arg_match(&arg, &resize_down_thresh, argi))
00699             cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
00700         else if (arg_match(&arg, &resize_down_thresh, argi))
00701             cfg.rc_resize_down_thresh = arg_parse_uint(&arg);
00702         else if (arg_match(&arg, &end_usage, argi))
00703             cfg.rc_end_usage = arg_parse_uint(&arg);
00704         else if (arg_match(&arg, &target_bitrate, argi))
00705             cfg.rc_target_bitrate = arg_parse_uint(&arg);
00706         else if (arg_match(&arg, &min_quantizer, argi))
00707             cfg.rc_min_quantizer = arg_parse_uint(&arg);
00708         else if (arg_match(&arg, &max_quantizer, argi))
00709             cfg.rc_max_quantizer = arg_parse_uint(&arg);
00710         else if (arg_match(&arg, &undershoot_pct, argi))
00711             cfg.rc_undershoot_pct = arg_parse_uint(&arg);
00712         else if (arg_match(&arg, &overshoot_pct, argi))
00713             cfg.rc_overshoot_pct = arg_parse_uint(&arg);
00714         else if (arg_match(&arg, &buf_sz, argi))
00715             cfg.rc_buf_sz = arg_parse_uint(&arg);
00716         else if (arg_match(&arg, &buf_initial_sz, argi))
00717             cfg.rc_buf_initial_sz = arg_parse_uint(&arg);
00718         else if (arg_match(&arg, &buf_optimal_sz, argi))
00719             cfg.rc_buf_optimal_sz = arg_parse_uint(&arg);
00720         else if (arg_match(&arg, &bias_pct, argi))
00721         {
00722             cfg.rc_2pass_vbr_bias_pct = arg_parse_uint(&arg);
00723 
00724             if (arg_passes < 2)
00725                 printf("Warning: option %s ignored in one-pass mode.\n",
00726                        arg.name);
00727         }
00728         else if (arg_match(&arg, &minsection_pct, argi))
00729         {
00730             cfg.rc_2pass_vbr_minsection_pct = arg_parse_uint(&arg);
00731 
00732             if (arg_passes < 2)
00733                 printf("Warning: option %s ignored in one-pass mode.\n",
00734                        arg.name);
00735         }
00736         else if (arg_match(&arg, &maxsection_pct, argi))
00737         {
00738             cfg.rc_2pass_vbr_maxsection_pct = arg_parse_uint(&arg);
00739 
00740             if (arg_passes < 2)
00741                 printf("Warning: option %s ignored in one-pass mode.\n",
00742                        arg.name);
00743         }
00744         else if (arg_match(&arg, &kf_min_dist, argi))
00745             cfg.kf_min_dist = arg_parse_uint(&arg);
00746         else if (arg_match(&arg, &kf_max_dist, argi))
00747             cfg.kf_max_dist = arg_parse_uint(&arg);
00748         else
00749             argj++;
00750     }
00751 
00752     /* Handle codec specific options */
00753 #if CONFIG_VP8_ENCODER
00754 
00755     if (codec->iface == &vpx_codec_vp8_cx_algo)
00756     {
00757         ctrl_args = vp8_args;
00758         ctrl_args_map = vp8_arg_ctrl_map;
00759     }
00760 
00761 #endif
00762 
00763     for (argi = argj = argv; (*argj = *argi); argi += arg.argv_step)
00764     {
00765         int match = 0;
00766 
00767         arg.argv_step = 1;
00768 
00769         for (i = 0; ctrl_args[i]; i++)
00770         {
00771             if (arg_match(&arg, ctrl_args[i], argi))
00772             {
00773                 match = 1;
00774 
00775                 if (arg_ctrl_cnt < ARG_CTRL_CNT_MAX)
00776                 {
00777                     arg_ctrls[arg_ctrl_cnt][0] = ctrl_args_map[i];
00778                     arg_ctrls[arg_ctrl_cnt][1] = arg_parse_int(&arg);
00779                     arg_ctrl_cnt++;
00780                 }
00781             }
00782         }
00783 
00784         if (!match)
00785             argj++;
00786     }
00787 
00788     /* Check for unrecognized options */
00789     for (argi = argv; *argi; argi++)
00790         if (argi[0][0] == '-')
00791             die("Error: Unrecognized option %s\n", *argi);
00792 
00793     /* Handle non-option arguments */
00794     in_fn = argv[0];
00795     out_fn = argv[1];
00796 
00797     if (!in_fn || !out_fn)
00798         usage_exit();
00799 
00800     /* Parse certain options from the input file, if possible */
00801     infile = fopen(in_fn, "rb");
00802 
00803     if (!infile)
00804     {
00805         printf("Failed to open input file");
00806         return EXIT_FAILURE;
00807     }
00808 
00809     is_ivf = file_is_ivf(infile, &fourcc, &cfg.g_w, &cfg.g_h);
00810 
00811     if (is_ivf)
00812     {
00813         switch (fourcc)
00814         {
00815         case 0x32315659:
00816             arg_use_i420 = 0;
00817             break;
00818         case 0x30323449:
00819             arg_use_i420 = 1;
00820             break;
00821         default:
00822             printf("Unsupported fourcc (%08x) in IVF\n", fourcc);
00823             return EXIT_FAILURE;
00824         }
00825     }
00826 
00827     fclose(infile);
00828 
00829 
00830 #define SHOW(field) printf("    %-28s = %d\n", #field, cfg.field)
00831 
00832     if (verbose)
00833     {
00834         printf("Codec: %s\n", vpx_codec_iface_name(codec->iface));
00835         printf("Source file: %s Format: %s\n", in_fn, arg_use_i420 ? "I420" : "YV12");
00836         printf("Destination file: %s\n", out_fn);
00837         printf("Encoder parameters:\n");
00838 
00839         SHOW(g_usage);
00840         SHOW(g_threads);
00841         SHOW(g_profile);
00842         SHOW(g_w);
00843         SHOW(g_h);
00844         SHOW(g_timebase.num);
00845         SHOW(g_timebase.den);
00846         SHOW(g_error_resilient);
00847         SHOW(g_pass);
00848         SHOW(g_lag_in_frames);
00849         SHOW(rc_dropframe_thresh);
00850         SHOW(rc_resize_allowed);
00851         SHOW(rc_resize_up_thresh);
00852         SHOW(rc_resize_down_thresh);
00853         SHOW(rc_end_usage);
00854         SHOW(rc_target_bitrate);
00855         SHOW(rc_min_quantizer);
00856         SHOW(rc_max_quantizer);
00857         SHOW(rc_undershoot_pct);
00858         SHOW(rc_overshoot_pct);
00859         SHOW(rc_buf_sz);
00860         SHOW(rc_buf_initial_sz);
00861         SHOW(rc_buf_optimal_sz);
00862         SHOW(rc_2pass_vbr_bias_pct);
00863         SHOW(rc_2pass_vbr_minsection_pct);
00864         SHOW(rc_2pass_vbr_maxsection_pct);
00865         SHOW(kf_mode);
00866         SHOW(kf_min_dist);
00867         SHOW(kf_max_dist);
00868     }
00869 
00870     vpx_img_alloc(&raw, arg_use_i420 ? IMG_FMT_I420 : IMG_FMT_YV12,
00871                   cfg.g_w, cfg.g_h, 1);
00872 
00873     // This was added so that ivfenc will create monotically increasing
00874     // timestamps.  Since we create new timestamps for alt-reference frames
00875     // we need to make room in the series of timestamps.  Since there can
00876     // only be 1 alt-ref frame ( current bitstream) multiplying by 2
00877     // gives us enough room.
00878     cfg.g_timebase.den *= 2;
00879 
00880     memset(&stats, 0, sizeof(stats));
00881 
00882     for (pass = one_pass_only ? one_pass_only - 1 : 0; pass < arg_passes; pass++)
00883     {
00884         int frames_in = 0, frames_out = 0;
00885         unsigned long nbytes = 0;
00886 
00887         infile = fopen(in_fn, "rb");
00888 
00889         if (!infile)
00890         {
00891             printf("Failed to open input file");
00892             return EXIT_FAILURE;
00893         }
00894 
00895         outfile = fopen(out_fn, "wb");
00896 
00897         if (!outfile)
00898         {
00899             printf("Failed to open output file");
00900             return EXIT_FAILURE;
00901         }
00902 
00903         if (stats_fn)
00904         {
00905             if (!stats_open_file(&stats, stats_fn, pass))
00906             {
00907                 printf("Failed to open statistics store\n");
00908                 return EXIT_FAILURE;
00909             }
00910         }
00911         else
00912         {
00913             if (!stats_open_mem(&stats, pass))
00914             {
00915                 printf("Failed to open statistics store\n");
00916                 return EXIT_FAILURE;
00917             }
00918         }
00919 
00920         cfg.g_pass = arg_passes == 2
00921                      ? pass ? VPX_RC_LAST_PASS : VPX_RC_FIRST_PASS
00922                  : VPX_RC_ONE_PASS;
00923 #if VPX_ENCODER_ABI_VERSION > (1 + VPX_CODEC_ABI_VERSION)
00924 
00925         if (pass)
00926         {
00927             cfg.rc_twopass_stats_in = stats_get(&stats);
00928         }
00929 
00930 #endif
00931 
00932         write_ivf_file_header(outfile, &cfg, codec->fourcc, 0);
00933 
00934 
00935         /* Construct Encoder Context */
00936         if (cfg.kf_min_dist == cfg.kf_max_dist)
00937             cfg.kf_mode = VPX_KF_FIXED;
00938 
00939         vpx_codec_enc_init(&encoder, codec->iface, &cfg,
00940                            show_psnr ? VPX_CODEC_USE_PSNR : 0);
00941         ctx_exit_on_error(&encoder, "Failed to initialize encoder");
00942 
00943         /* Note that we bypass the vpx_codec_control wrapper macro because
00944          * we're being clever to store the control IDs in an array. Real
00945          * applications will want to make use of the enumerations directly
00946          */
00947         for (i = 0; i < arg_ctrl_cnt; i++)
00948         {
00949             if (vpx_codec_control_(&encoder, arg_ctrls[i][0], arg_ctrls[i][1]))
00950                 printf("Error: Tried to set control %d = %d\n",
00951                        arg_ctrls[i][0], arg_ctrls[i][1]);
00952 
00953             ctx_exit_on_error(&encoder, "Failed to control codec");
00954         }
00955 
00956         frame_avail = 1;
00957         got_data = 0;
00958 
00959         while (frame_avail || got_data)
00960         {
00961             vpx_codec_iter_t iter = NULL;
00962             const vpx_codec_cx_pkt_t *pkt;
00963             struct vpx_usec_timer timer;
00964 
00965             if (!arg_limit || frames_in < arg_limit)
00966             {
00967                 frame_avail = read_frame(infile, &raw, is_ivf);
00968 
00969                 if (frame_avail)
00970                     frames_in++;
00971 
00972                 printf("\rPass %d/%d frame %4d/%-4d %7ldB \033[K", pass + 1,
00973                        arg_passes, frames_in, frames_out, nbytes);
00974             }
00975             else
00976                 frame_avail = 0;
00977 
00978             vpx_usec_timer_start(&timer);
00979 
00980             // since we halved our timebase we need to double the timestamps
00981             // and duration we pass in.
00982             vpx_codec_encode(&encoder, frame_avail ? &raw : NULL, (frames_in - 1) * 2,
00983                              2, 0, arg_deadline);
00984             vpx_usec_timer_mark(&timer);
00985             cx_time += vpx_usec_timer_elapsed(&timer);
00986             ctx_exit_on_error(&encoder, "Failed to encode frame");
00987             got_data = 0;
00988 
00989             while ((pkt = vpx_codec_get_cx_data(&encoder, &iter)))
00990             {
00991                 got_data = 1;
00992                 nbytes += pkt->data.raw.sz;
00993 
00994                 switch (pkt->kind)
00995                 {
00996                 case VPX_CODEC_CX_FRAME_PKT:
00997                     frames_out++;
00998                     printf(" %6luF",
00999                            (unsigned long)pkt->data.frame.sz);
01000                     write_ivf_frame_header(outfile, pkt);
01001                     fwrite(pkt->data.frame.buf, 1, pkt->data.frame.sz, outfile);
01002                     break;
01003                 case VPX_CODEC_STATS_PKT:
01004                     frames_out++;
01005                     printf(" %6luS",
01006                            (unsigned long)pkt->data.twopass_stats.sz);
01007                     stats_write(&stats,
01008                                 pkt->data.twopass_stats.buf,
01009                                 pkt->data.twopass_stats.sz);
01010                     break;
01011                 case VPX_CODEC_PSNR_PKT:
01012 
01013                     if (show_psnr)
01014                     {
01015                         int i;
01016 
01017                         for (i = 0; i < 4; i++)
01018                             printf("%.3lf ", pkt->data.psnr.psnr[i]);
01019                     }
01020 
01021                     break;
01022                 default:
01023                     break;
01024                 }
01025             }
01026 
01027             fflush(stdout);
01028         }
01029 
01030         /* this bitrate calc is simplified and relies on the fact that this
01031          * application uses 1/timebase for framerate.
01032          */
01033         printf("\rPass %d/%d frame %4d/%-4d %7ldB %7ldb/f %7"PRId64"b/s"
01034                " %7lu %s (%.2f fps)\033[K", pass + 1,
01035                arg_passes, frames_in, frames_out, nbytes, nbytes * 8 / frames_in,
01036                nbytes * 8 *(int64_t)cfg.g_timebase.den / cfg.g_timebase.num / frames_in,
01037                cx_time > 9999999 ? cx_time / 1000 : cx_time,
01038                cx_time > 9999999 ? "ms" : "us",
01039                (float)frames_in * 1000000.0 / (float)cx_time);
01040 
01041         vpx_codec_destroy(&encoder);
01042 
01043         fclose(infile);
01044 
01045         if (!fseek(outfile, 0, SEEK_SET))
01046             write_ivf_file_header(outfile, &cfg, codec->fourcc, frames_out);
01047 
01048         fclose(outfile);
01049         stats_close(&stats);
01050         printf("\n");
01051 
01052         if (one_pass_only)
01053             break;
01054     }
01055 
01056     vpx_img_free(&raw);
01057     free(argv);
01058     return EXIT_SUCCESS;
01059 }

Generated on 25 Jun 2011 for WebM VP8 Codec SDK by  doxygen 1.6.1