src/media/base/audio_discard_helper_unittest.cc

   1 // Copyright 2014 The Chromium Authors. All rights reserved.
   2 // Use of this source code is governed by a BSD-style license that can be
   3 // found in the LICENSE file.
   4
   5 #include "base/memory/scoped_ptr.h"
   6 #include "media/base/audio_buffer.h"
   7 #include "media/base/audio_bus.h"
   8 #include "media/base/audio_discard_helper.h"
   9 #include "media/base/buffers.h"
  10 #include "media/base/decoder_buffer.h"
  11 #include "media/base/test_helpers.h"
  12 #include "testing/gtest/include/gtest/gtest.h"
  13
  14 namespace media {
  15
  16 static const float kDataStep = 0.01f;
  17 static const size_t kSampleRate = 48000;
  18
  19 static scoped_refptr<DecoderBuffer> CreateEncodedBuffer(
  20     base::TimeDelta timestamp,
  21     base::TimeDelta duration) {
  22   scoped_refptr<DecoderBuffer> result(new DecoderBuffer(1));
  23   result->set_timestamp(timestamp);
  24   result->set_duration(duration);
  25   return result;
  26 }
  27
  28 static scoped_refptr<AudioBuffer> CreateDecodedBuffer(int frames) {
  29   return MakeAudioBuffer(kSampleFormatPlanarF32,
  30                          CHANNEL_LAYOUT_MONO,
  31                          1,
  32                          kSampleRate,
  33                          0.0f,
  34                          kDataStep,
  35                          frames,
  36                          kNoTimestamp());
  37 }
  38
  39 static float ExtractDecodedData(const scoped_refptr<AudioBuffer>& buffer,
  40                                 int index) {
  41   // This is really inefficient, but we can't access the raw AudioBuffer if any
  42   // start trimming has been applied.
  43   scoped_ptr<AudioBus> temp_bus = AudioBus::Create(buffer->channel_count(), 1);
  44   buffer->ReadFrames(1, index, 0, temp_bus.get());
  45   return temp_bus->channel(0)[0];
  46 }
  47
  48 TEST(AudioDiscardHelperTest, TimeDeltaToFrames) {
  49   AudioDiscardHelper discard_helper(kSampleRate, 0);
  50
  51   EXPECT_EQ(0u, discard_helper.TimeDeltaToFrames(base::TimeDelta()));
  52   EXPECT_EQ(
  53       kSampleRate / 100,
  54       discard_helper.TimeDeltaToFrames(base::TimeDelta::FromMilliseconds(10)));
  55
  56   // Ensure partial frames are rounded down correctly.  The equation below
  57   // calculates a frame count with a fractional part < 0.5.
  58   const int small_remainder =
  59       base::Time::kMicrosecondsPerSecond * (kSampleRate - 0.9) / kSampleRate;
  60   EXPECT_EQ(kSampleRate - 1,
  61             discard_helper.TimeDeltaToFrames(
  62                 base::TimeDelta::FromMicroseconds(small_remainder)));
  63
  64   // Ditto, but rounded up using a fractional part > 0.5.
  65   const int large_remainder =
  66       base::Time::kMicrosecondsPerSecond * (kSampleRate - 0.4) / kSampleRate;
  67   EXPECT_EQ(kSampleRate,
  68             discard_helper.TimeDeltaToFrames(
  69                 base::TimeDelta::FromMicroseconds(large_remainder)));
  70 }
  71
  72 TEST(AudioDiscardHelperTest, BasicProcessBuffers) {
  73   AudioDiscardHelper discard_helper(kSampleRate, 0);
  74   ASSERT_FALSE(discard_helper.initialized());
  75
  76   const base::TimeDelta kTimestamp = base::TimeDelta();
  77
  78   // Use an estimated duration which doesn't match the number of decoded frames
  79   // to ensure the helper is correctly setting durations based on output frames.
  80   const base::TimeDelta kEstimatedDuration =
  81       base::TimeDelta::FromMilliseconds(9);
  82   const base::TimeDelta kActualDuration = base::TimeDelta::FromMilliseconds(10);
  83   const int kTestFrames = discard_helper.TimeDeltaToFrames(kActualDuration);
  84
  85   scoped_refptr<DecoderBuffer> encoded_buffer =
  86       CreateEncodedBuffer(kTimestamp, kEstimatedDuration);
  87   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
  88
  89   // Verify the basic case where nothing is discarded.
  90   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
  91   ASSERT_TRUE(discard_helper.initialized());
  92   EXPECT_EQ(kTimestamp, decoded_buffer->timestamp());
  93   EXPECT_EQ(kActualDuration, decoded_buffer->duration());
  94   EXPECT_EQ(kTestFrames, decoded_buffer->frame_count());
  95
  96   // Verify a Reset() takes us back to an uninitialized state.
  97   discard_helper.Reset(0);
  98   ASSERT_FALSE(discard_helper.initialized());
  99
 100   // Verify a NULL output buffer returns false.
 101   ASSERT_FALSE(discard_helper.ProcessBuffers(encoded_buffer, NULL));
 102 }
 103
 104 TEST(AudioDiscardHelperTest, NegativeTimestampClampsToZero) {
 105   AudioDiscardHelper discard_helper(kSampleRate, 0);
 106   ASSERT_FALSE(discard_helper.initialized());
 107
 108   const base::TimeDelta kTimestamp = -base::TimeDelta::FromSeconds(1);
 109   const base::TimeDelta kDuration = base::TimeDelta::FromMilliseconds(10);
 110   const int kTestFrames = discard_helper.TimeDeltaToFrames(kDuration);
 111
 112   scoped_refptr<DecoderBuffer> encoded_buffer =
 113       CreateEncodedBuffer(kTimestamp, kDuration);
 114   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
 115
 116   // Verify the basic case where nothing is discarded.
 117   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 118   ASSERT_TRUE(discard_helper.initialized());
 119   EXPECT_EQ(base::TimeDelta(), decoded_buffer->timestamp());
 120   EXPECT_EQ(kDuration, decoded_buffer->duration());
 121   EXPECT_EQ(kTestFrames, decoded_buffer->frame_count());
 122 }
 123
 124 TEST(AudioDiscardHelperTest, ProcessBuffersWithInitialDiscard) {
 125   AudioDiscardHelper discard_helper(kSampleRate, 0);
 126   ASSERT_FALSE(discard_helper.initialized());
 127
 128   const base::TimeDelta kTimestamp = base::TimeDelta();
 129   const base::TimeDelta kDuration = base::TimeDelta::FromMilliseconds(10);
 130   const int kTestFrames = discard_helper.TimeDeltaToFrames(kDuration);
 131
 132   // Tell the helper we want to discard half of the initial frames.
 133   const int kDiscardFrames = kTestFrames / 2;
 134   discard_helper.Reset(kDiscardFrames);
 135
 136   scoped_refptr<DecoderBuffer> encoded_buffer =
 137       CreateEncodedBuffer(kTimestamp, kDuration);
 138   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
 139
 140   // Verify half the frames end up discarded.
 141   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 142   ASSERT_TRUE(discard_helper.initialized());
 143   EXPECT_EQ(kTimestamp, decoded_buffer->timestamp());
 144   EXPECT_EQ(kDuration / 2, decoded_buffer->duration());
 145   EXPECT_EQ(kDiscardFrames, decoded_buffer->frame_count());
 146   ASSERT_FLOAT_EQ(kDiscardFrames * kDataStep,
 147                   ExtractDecodedData(decoded_buffer, 0));
 148 }
 149
 150 TEST(AudioDiscardHelperTest, ProcessBuffersWithLargeInitialDiscard) {
 151   AudioDiscardHelper discard_helper(kSampleRate, 0);
 152   ASSERT_FALSE(discard_helper.initialized());
 153
 154   const base::TimeDelta kTimestamp = base::TimeDelta();
 155   const base::TimeDelta kDuration = base::TimeDelta::FromMilliseconds(10);
 156   const int kTestFrames = discard_helper.TimeDeltaToFrames(kDuration);
 157
 158   // Tell the helper we want to discard 1.5 buffers worth of frames.
 159   discard_helper.Reset(kTestFrames * 1.5);
 160
 161   scoped_refptr<DecoderBuffer> encoded_buffer =
 162       CreateEncodedBuffer(kTimestamp, kDuration);
 163   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
 164
 165   // The first call should fail since no output buffer remains.
 166   ASSERT_FALSE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 167   ASSERT_TRUE(discard_helper.initialized());
 168
 169   // Generate another set of buffers and expect half the output frames.
 170   encoded_buffer = CreateEncodedBuffer(kTimestamp + kDuration, kDuration);
 171   decoded_buffer = CreateDecodedBuffer(kTestFrames);
 172   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 173
 174   // The timestamp should match that of the initial buffer.
 175   const int kDiscardFrames = kTestFrames / 2;
 176   EXPECT_EQ(kTimestamp, decoded_buffer->timestamp());
 177   EXPECT_EQ(kDuration / 2, decoded_buffer->duration());
 178   EXPECT_EQ(kDiscardFrames, decoded_buffer->frame_count());
 179   ASSERT_FLOAT_EQ(kDiscardFrames * kDataStep,
 180                   ExtractDecodedData(decoded_buffer, 0));
 181 }
 182
 183 TEST(AudioDiscardHelperTest, AllowNonMonotonicTimestamps) {
 184   AudioDiscardHelper discard_helper(kSampleRate, 0);
 185   ASSERT_FALSE(discard_helper.initialized());
 186
 187   const base::TimeDelta kTimestamp = base::TimeDelta();
 188   const base::TimeDelta kDuration = base::TimeDelta::FromMilliseconds(10);
 189   const int kTestFrames = discard_helper.TimeDeltaToFrames(kDuration);
 190
 191   scoped_refptr<DecoderBuffer> encoded_buffer =
 192       CreateEncodedBuffer(kTimestamp, kDuration);
 193   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
 194
 195   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 196   ASSERT_TRUE(discard_helper.initialized());
 197   EXPECT_EQ(kTimestamp, decoded_buffer->timestamp());
 198   EXPECT_EQ(kDuration, decoded_buffer->duration());
 199   EXPECT_EQ(kTestFrames, decoded_buffer->frame_count());
 200
 201   // Process the same input buffer again to ensure input timestamps which go
 202   // backwards in time are not errors.
 203   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 204   EXPECT_EQ(kTimestamp + kDuration, decoded_buffer->timestamp());
 205   EXPECT_EQ(kDuration, decoded_buffer->duration());
 206   EXPECT_EQ(kTestFrames, decoded_buffer->frame_count());
 207 }
 208
 209 TEST(AudioDiscardHelperTest, DiscardEndPadding) {
 210   AudioDiscardHelper discard_helper(kSampleRate, 0);
 211   ASSERT_FALSE(discard_helper.initialized());
 212
 213   const base::TimeDelta kTimestamp = base::TimeDelta();
 214   const base::TimeDelta kDuration = base::TimeDelta::FromMilliseconds(10);
 215   const int kTestFrames = discard_helper.TimeDeltaToFrames(kDuration);
 216
 217   scoped_refptr<DecoderBuffer> encoded_buffer =
 218       CreateEncodedBuffer(kTimestamp, kDuration);
 219   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
 220
 221   // Set a discard padding equivalent to half the buffer.
 222   encoded_buffer->set_discard_padding(
 223       std::make_pair(base::TimeDelta(), kDuration / 2));
 224
 225   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 226   ASSERT_TRUE(discard_helper.initialized());
 227   EXPECT_EQ(kTimestamp, decoded_buffer->timestamp());
 228   EXPECT_EQ(kDuration / 2, decoded_buffer->duration());
 229   EXPECT_EQ(kTestFrames / 2, decoded_buffer->frame_count());
 230 }
 231
 232 TEST(AudioDiscardHelperTest, BadDiscardEndPadding) {
 233   AudioDiscardHelper discard_helper(kSampleRate, 0);
 234   ASSERT_FALSE(discard_helper.initialized());
 235
 236   const base::TimeDelta kTimestamp = base::TimeDelta();
 237   const base::TimeDelta kDuration = base::TimeDelta::FromMilliseconds(10);
 238   const int kTestFrames = discard_helper.TimeDeltaToFrames(kDuration);
 239
 240   scoped_refptr<DecoderBuffer> encoded_buffer =
 241       CreateEncodedBuffer(kTimestamp, kDuration);
 242   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
 243
 244   // Set a discard padding equivalent to double the buffer size.
 245   encoded_buffer->set_discard_padding(
 246       std::make_pair(base::TimeDelta(), kDuration * 2));
 247
 248   // Verify the end discard padding is rejected.
 249   ASSERT_FALSE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 250   ASSERT_TRUE(discard_helper.initialized());
 251 }
 252
 253 TEST(AudioDiscardHelperTest, InitialDiscardAndDiscardEndPadding) {
 254   AudioDiscardHelper discard_helper(kSampleRate, 0);
 255   ASSERT_FALSE(discard_helper.initialized());
 256
 257   const base::TimeDelta kTimestamp = base::TimeDelta();
 258   const base::TimeDelta kDuration = base::TimeDelta::FromMilliseconds(10);
 259   const int kTestFrames = discard_helper.TimeDeltaToFrames(kDuration);
 260
 261   scoped_refptr<DecoderBuffer> encoded_buffer =
 262       CreateEncodedBuffer(kTimestamp, kDuration);
 263   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
 264
 265   // Set a discard padding equivalent to a quarter of the buffer.
 266   encoded_buffer->set_discard_padding(
 267       std::make_pair(base::TimeDelta(), kDuration / 4));
 268
 269   // Set an initial discard of a quarter of the buffer.
 270   const int kDiscardFrames = kTestFrames / 4;
 271   discard_helper.Reset(kDiscardFrames);
 272
 273   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 274   ASSERT_TRUE(discard_helper.initialized());
 275   EXPECT_EQ(kTimestamp, decoded_buffer->timestamp());
 276   EXPECT_EQ(kDuration / 2, decoded_buffer->duration());
 277   EXPECT_EQ(kTestFrames / 2, decoded_buffer->frame_count());
 278   ASSERT_FLOAT_EQ(kDiscardFrames * kDataStep,
 279                   ExtractDecodedData(decoded_buffer, 0));
 280 }
 281
 282 TEST(AudioDiscardHelperTest, InitialDiscardAndDiscardPadding) {
 283   AudioDiscardHelper discard_helper(kSampleRate, 0);
 284   ASSERT_FALSE(discard_helper.initialized());
 285
 286   const base::TimeDelta kTimestamp = base::TimeDelta();
 287   const base::TimeDelta kDuration = base::TimeDelta::FromMilliseconds(10);
 288   const int kTestFrames = discard_helper.TimeDeltaToFrames(kDuration);
 289
 290   scoped_refptr<DecoderBuffer> encoded_buffer =
 291       CreateEncodedBuffer(kTimestamp, kDuration);
 292   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
 293
 294   // Set all the discard values to be different to ensure each is properly used.
 295   const int kDiscardFrames = kTestFrames / 4;
 296   encoded_buffer->set_discard_padding(
 297       std::make_pair(kDuration / 8, kDuration / 16));
 298   discard_helper.Reset(kDiscardFrames);
 299
 300   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 301   ASSERT_TRUE(discard_helper.initialized());
 302   EXPECT_EQ(kTimestamp, decoded_buffer->timestamp());
 303   EXPECT_EQ(kDuration - kDuration / 4 - kDuration / 8 - kDuration / 16,
 304             decoded_buffer->duration());
 305   EXPECT_EQ(kTestFrames - kTestFrames / 4 - kTestFrames / 8 - kTestFrames / 16,
 306             decoded_buffer->frame_count());
 307 }
 308
 309 TEST(AudioDiscardHelperTest, InitialDiscardAndDiscardPaddingAndCodecDelay) {
 310   // Use a codec delay of 5ms.
 311   const int kCodecDelay = kSampleRate / 100 / 2;
 312   AudioDiscardHelper discard_helper(kSampleRate, kCodecDelay);
 313   ASSERT_FALSE(discard_helper.initialized());
 314   discard_helper.Reset(kCodecDelay);
 315
 316   const base::TimeDelta kTimestamp = base::TimeDelta();
 317   const base::TimeDelta kDuration = base::TimeDelta::FromMilliseconds(10);
 318   const int kTestFrames = discard_helper.TimeDeltaToFrames(kDuration);
 319
 320   scoped_refptr<DecoderBuffer> encoded_buffer =
 321       CreateEncodedBuffer(kTimestamp, kDuration);
 322   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
 323
 324   // Set a discard padding equivalent to half of the buffer.
 325   encoded_buffer->set_discard_padding(
 326       std::make_pair(kDuration / 2, base::TimeDelta()));
 327
 328   // All of the first buffer should be discarded.
 329   ASSERT_FALSE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 330   ASSERT_TRUE(discard_helper.initialized());
 331
 332   // Processing another buffer (with the same discard padding) should discard
 333   // the back half of the buffer since kCodecDelay is half a buffer.
 334   encoded_buffer->set_timestamp(kTimestamp + kDuration);
 335   decoded_buffer = CreateDecodedBuffer(kTestFrames);
 336   ASSERT_FLOAT_EQ(0.0f, ExtractDecodedData(decoded_buffer, 0));
 337   ASSERT_NEAR(kCodecDelay * kDataStep,
 338               ExtractDecodedData(decoded_buffer, kCodecDelay),
 339               kDataStep * 1000);
 340   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 341   EXPECT_EQ(kTimestamp, decoded_buffer->timestamp());
 342   EXPECT_EQ(kDuration / 2, decoded_buffer->duration());
 343   EXPECT_EQ(kTestFrames / 2, decoded_buffer->frame_count());
 344
 345   // Verify it was actually the latter half of the buffer that was removed.
 346   ASSERT_FLOAT_EQ(0.0f, ExtractDecodedData(decoded_buffer, 0));
 347 }
 348
 349 TEST(AudioDiscardHelperTest, DelayedDiscardInitialDiscardAndDiscardPadding) {
 350   AudioDiscardHelper discard_helper(kSampleRate, 0);
 351   ASSERT_FALSE(discard_helper.initialized());
 352
 353   const base::TimeDelta kTimestamp = base::TimeDelta();
 354   const base::TimeDelta kDuration = base::TimeDelta::FromMilliseconds(10);
 355   const int kTestFrames = discard_helper.TimeDeltaToFrames(kDuration);
 356
 357   scoped_refptr<DecoderBuffer> encoded_buffer =
 358       CreateEncodedBuffer(kTimestamp, kDuration);
 359
 360   // Set all the discard values to be different to ensure each is properly used.
 361   const int kDiscardFrames = kTestFrames / 4;
 362   encoded_buffer->set_discard_padding(
 363       std::make_pair(kDuration / 8, kDuration / 16));
 364   discard_helper.Reset(kDiscardFrames);
 365
 366   // Verify nothing is output for the first buffer, yet initialized is true.
 367   ASSERT_FALSE(discard_helper.ProcessBuffers(encoded_buffer, NULL));
 368   ASSERT_TRUE(discard_helper.initialized());
 369
 370   // Create an encoded buffer with no discard padding.
 371   encoded_buffer = CreateEncodedBuffer(kTimestamp + kDuration, kDuration);
 372   scoped_refptr<AudioBuffer> decoded_buffer = CreateDecodedBuffer(kTestFrames);
 373
 374   // Verify that when the decoded buffer is consumed, the discards from the
 375   // previous encoded buffer are applied.
 376   ASSERT_TRUE(discard_helper.ProcessBuffers(encoded_buffer, decoded_buffer));
 377   EXPECT_EQ(kTimestamp, decoded_buffer->timestamp());
 378   EXPECT_EQ(kDuration - kDuration / 4 - kDuration / 8 - kDuration / 16,
 379             decoded_buffer->duration());
 380   EXPECT_EQ(kTestFrames - kTestFrames / 4 - kTestFrames / 8 - kTestFrames / 16,
 381             decoded_buffer->frame_count());
 382 }
 383
 384 }  // namespace media