Merge "[svc] 1. Add two pass RC options in vp9_spatial_scalable_encoder. 2. Add read/write for RC stats file The two pass RC for svc does not work yet. This is just the first step. We need further development to make it working. Change-Id: I8ef0e177dff0b5ed3c97a916beea5123717cc6f2"

2014-03-07 14:02:35 -08:00 · 2014-03-07 14:02:35 -08:00 · 2740507142
--- a/examples.mk
+++ b/examples.mk
@ -62,6 +62,7 @@ vp9_spatial_scalable_encoder.SRCS += ivfenc.c ivfenc.h
 vp9_spatial_scalable_encoder.SRCS += tools_common.c tools_common.h
 vp9_spatial_scalable_encoder.SRCS += video_common.h
 vp9_spatial_scalable_encoder.SRCS += video_writer.h video_writer.c
+vp9_spatial_scalable_encoder.SRCS += vpxstats.c vpxstats.h
 vp9_spatial_scalable_encoder.GUID   = 4A38598D-627D-4505-9C7B-D4020C84100D
 vp9_spatial_scalable_encoder.DESCRIPTION = Spatial Scalable Encoder

--- a/examples/vp9_spatial_scalable_encoder.c
+++ b/examples/vp9_spatial_scalable_encoder.c
@ -26,6 +26,7 @@
 #include "vpx/svc_context.h"
 #include "vpx/vp8cx.h"
 #include "vpx/vpx_encoder.h"
+#include "./vpxstats.h"

 static const struct arg_enum_list encoding_mode_enum[] = {
  {"i", INTER_LAYER_PREDICTION_I},
@ -60,12 +61,19 @@ static const arg_def_t quantizers_arg =
 static const arg_def_t quantizers_keyframe_arg =
    ARG_DEF("qn", "quantizers-keyframe", 1, "quantizers for key frames (lowest "
        "to highest layer)");
+static const arg_def_t passes_arg =
+    ARG_DEF("p", "passes", 1, "Number of passes (1/2)");
+static const arg_def_t pass_arg =
+    ARG_DEF(NULL, "pass", 1, "Pass to execute (1/2)");
+static const arg_def_t fpf_name_arg =
+    ARG_DEF(NULL, "fpf", 1, "First pass statistics file name");

 static const arg_def_t *svc_args[] = {
  &encoding_mode_arg, &frames_arg,        &width_arg,       &height_arg,
  &timebase_arg,      &bitrate_arg,       &skip_frames_arg, &layers_arg,
  &kf_dist_arg,       &scale_factors_arg, &quantizers_arg,
-  &quantizers_keyframe_arg, NULL
+  &quantizers_keyframe_arg,               &passes_arg,       &pass_arg,
+  &fpf_name_arg,      NULL
 };

 static const SVC_ENCODING_MODE default_encoding_mode =
@ -85,6 +93,10 @@ typedef struct {
  const char *output_filename;
  uint32_t frames_to_code;
  uint32_t frames_to_skip;
+  struct VpxInputContext input_ctx;
+  stats_io_t rc_stats;
+  int passes;
+  int pass;
 } AppInput;

 static const char *exec_name;
@ -105,6 +117,9 @@ static void parse_command_line(int argc, const char **argv_,
  char **argi = NULL;
  char **argj = NULL;
  vpx_codec_err_t res;
+  int passes = 0;
+  int pass = 0;
+  const char *fpf_file_name = NULL;

  // initialize SvcContext with parameters that will be passed to vpx_svc_init
  svc_ctx->log_level = SVC_LOG_DEBUG;
@ -159,11 +174,53 @@ static void parse_command_line(int argc, const char **argv_,
      vpx_svc_set_quantizers(svc_ctx, arg.val, 0);
    } else if (arg_match(&arg, &quantizers_keyframe_arg, argi)) {
      vpx_svc_set_quantizers(svc_ctx, arg.val, 1);
+    } else if (arg_match(&arg, &passes_arg, argi)) {
+      passes = arg_parse_uint(&arg);
+      if (passes < 1 || passes > 2) {
+        die("Error: Invalid number of passes (%d)\n", passes);
+      }
+    } else if (arg_match(&arg, &pass_arg, argi)) {
+      pass = arg_parse_uint(&arg);
+      if (pass < 1 || pass > 2) {
+        die("Error: Invalid pass selected (%d)\n", pass);
+      }
+    } else if (arg_match(&arg, &fpf_name_arg, argi)) {
+      fpf_file_name = arg.val;
    } else {
      ++argj;
    }
  }

+  if (passes == 0 || passes == 1) {
+    if (pass) {
+      fprintf(stderr, "pass is ignored since there's only one pass\n");
+    }
+    enc_cfg->g_pass = VPX_RC_ONE_PASS;
+  } else {
+    if (pass == 0) {
+      die("pass must be specified when passes is 2\n");
+    }
+
+    if (fpf_file_name == NULL) {
+      die("fpf must be specified when passes is 2\n");
+    }
+
+    if (pass == 1) {
+      enc_cfg->g_pass = VPX_RC_FIRST_PASS;
+      if (!stats_open_file(&app_input->rc_stats, fpf_file_name, 0)) {
+        fatal("Failed to open statistics store");
+      }
+    } else {
+      enc_cfg->g_pass = VPX_RC_LAST_PASS;
+      if (!stats_open_file(&app_input->rc_stats, fpf_file_name, 1)) {
+        fatal("Failed to open statistics store");
+      }
+      enc_cfg->rc_twopass_stats_in = stats_get(&app_input->rc_stats);
+    }
+    app_input->passes = passes;
+    app_input->pass = pass;
+  }
+
  // Check for unrecognized options
  for (argi = argv; *argi; ++argi)
    if (argi[0][0] == '-' && strlen(argi[0]) > 1)
@ -234,10 +291,14 @@ int main(int argc, const char **argv) {
      VPX_CODEC_OK) {
    die("Failed to get output resolution");
  }
-  writer = vpx_video_writer_open(app_input.output_filename, kContainerIVF,
-                                 &info);
-  if (!writer)
-    die("Failed to open %s for writing\n", app_input.output_filename);
+
+  if (!(app_input.passes == 2 && app_input.pass == 1)) {
+    // We don't save the bitstream for the 1st pass on two pass rate control
+    writer = vpx_video_writer_open(app_input.output_filename, kContainerIVF,
+                                   &info);
+    if (!writer)
+      die("Failed to open %s for writing\n", app_input.output_filename);
+  }

  // skip initial frames
  for (i = 0; i < app_input.frames_to_skip; ++i)
@ -254,11 +315,18 @@ int main(int argc, const char **argv) {
    if (res != VPX_CODEC_OK) {
      die_codec(&codec, "Failed to encode frame");
    }
-    if (vpx_svc_get_frame_size(&svc_ctx) > 0) {
-      vpx_video_writer_write_frame(writer,
-                                   vpx_svc_get_buffer(&svc_ctx),
-                                   vpx_svc_get_frame_size(&svc_ctx),
-                                   pts);
+    if (!(app_input.passes == 2 && app_input.pass == 1)) {
+      if (vpx_svc_get_frame_size(&svc_ctx) > 0) {
+        vpx_video_writer_write_frame(writer,
+                                     vpx_svc_get_buffer(&svc_ctx),
+                                     vpx_svc_get_frame_size(&svc_ctx),
+                                     pts);
+      }
+    }
+    if (vpx_svc_get_rc_stats_buffer_size(&svc_ctx) > 0) {
+      stats_write(&app_input.rc_stats,
+                  vpx_svc_get_rc_stats_buffer(&svc_ctx),
+                  vpx_svc_get_rc_stats_buffer_size(&svc_ctx));
    }
    ++frame_cnt;
    pts += frame_duration;
@ -269,7 +337,12 @@ int main(int argc, const char **argv) {
  fclose(infile);
  if (vpx_codec_destroy(&codec)) die_codec(&codec, "Failed to destroy codec");

-  vpx_video_writer_close(writer);
+  if (app_input.passes == 2)
+    stats_close(&app_input.rc_stats, 1);
+
+  if (writer) {
+    vpx_video_writer_close(writer);
+  }

  vpx_img_free(&raw);

--- a/vpx/exports_enc
+++ b/vpx/exports_enc
@ -21,3 +21,5 @@ text vpx_svc_set_options
 text vpx_svc_set_quantizers
 text vpx_svc_set_scale_factors
 text vpx_svc_get_layer_resolution
+text vpx_svc_get_rc_stats_buffer_size
+text vpx_svc_get_rc_stats_buffer
--- a/vpx/src/svc_encodeframe.c
+++ b/vpx/src/svc_encodeframe.c
@ -81,6 +81,10 @@ typedef struct SvcInternal {
  size_t buffer_size;
  void *buffer;

+  char *rc_stats_buf;
+  size_t rc_stats_buf_size;
+  size_t rc_stats_buf_used;
+
  char message_buffer[2048];
  vpx_codec_ctx_t *codec_ctx;
 } SvcInternal;
@ -569,7 +573,6 @@ vpx_codec_err_t vpx_svc_init(SvcContext *svc_ctx, vpx_codec_ctx_t *codec_ctx,
  enc_cfg->ss_number_layers = si->layers;
  enc_cfg->ts_number_layers = 1;  // Temporal layers not used in this encoder.
  enc_cfg->kf_mode = VPX_KF_DISABLED;
-  enc_cfg->g_pass = VPX_RC_ONE_PASS;
  // Lag in frames not currently supported
  enc_cfg->g_lag_in_frames = 0;

@ -851,6 +854,7 @@ vpx_codec_err_t vpx_svc_encode(SvcContext *svc_ctx, vpx_codec_ctx_t *codec_ctx,

  memset(&superframe, 0, sizeof(superframe));
  svc_log_reset(svc_ctx);
+  si->rc_stats_buf_used = 0;

  si->layers = svc_ctx->spatial_layers;
  if (si->frame_within_gop >= si->kf_dist ||
@ -923,6 +927,25 @@ vpx_codec_err_t vpx_svc_encode(SvcContext *svc_ctx, vpx_codec_ctx_t *codec_ctx,
          }
          break;
        }
+        case VPX_CODEC_STATS_PKT: {
+          size_t new_size = si->rc_stats_buf_used +
+              cx_pkt->data.twopass_stats.sz;
+
+          if (new_size > si->rc_stats_buf_size) {
+            char *p = (char*)realloc(si->rc_stats_buf, new_size);
+            if (p == NULL) {
+              svc_log(svc_ctx, SVC_LOG_ERROR, "Error allocating stats buf\n");
+              break;
+            }
+            si->rc_stats_buf = p;
+            si->rc_stats_buf_size = new_size;
+          }
+
+          memcpy(si->rc_stats_buf + si->rc_stats_buf_used,
+                 cx_pkt->data.twopass_stats.buf, cx_pkt->data.twopass_stats.sz);
+          si->rc_stats_buf_used += cx_pkt->data.twopass_stats.sz;
+          break;
+        }
        default: {
          break;
        }
@ -1077,7 +1100,24 @@ void vpx_svc_release(SvcContext *svc_ctx) {
  si = (SvcInternal *)svc_ctx->internal;
  if (si != NULL) {
    free(si->buffer);
+    if (si->rc_stats_buf) {
+      free(si->rc_stats_buf);
+    }
    free(si);
    svc_ctx->internal = NULL;
  }
 }
+
+size_t vpx_svc_get_rc_stats_buffer_size(const SvcContext *svc_ctx) {
+  const SvcInternal *const si = get_const_svc_internal(svc_ctx);
+  if (svc_ctx == NULL || si == NULL) return 0;
+  return si->rc_stats_buf_used;
+}
+
+char *vpx_svc_get_rc_stats_buffer(const SvcContext *svc_ctx) {
+  const SvcInternal *const si = get_const_svc_internal(svc_ctx);
+  if (svc_ctx == NULL || si == NULL) return NULL;
+  return si->rc_stats_buf;
+}
+
+
--- a/vpx/svc_context.h
+++ b/vpx/svc_context.h
@ -113,6 +113,17 @@ size_t vpx_svc_get_frame_size(const SvcContext *svc_ctx);
 */
 void *vpx_svc_get_buffer(const SvcContext *svc_ctx);

+/**
+ * return size of two pass rate control stats data to be returned by
+ * vpx_svc_get_rc_stats_buffer
+ */
+size_t vpx_svc_get_rc_stats_buffer_size(const SvcContext *svc_ctx);
+
+/**
+ * return buffer two pass of rate control stats data
+ */
+char *vpx_svc_get_rc_stats_buffer(const SvcContext *svc_ctx);
+
 /**
 * return spatial resolution of the specified layer
 */