Resampler modifications in preparation for arbitrary audioproc rates.

- Templatize PushResampler to support int16 and float.
- Add a helper method to PushSincResampler to compute the algorithmic
delay.

This is a prerequisite of:
http://review.webrtc.org/9919004/

BUG=2894
R=turaj@webrtc.org

Review URL: https://webrtc-codereview.appspot.com/12169004

git-svn-id: http://webrtc.googlecode.com/svn/trunk/webrtc@5943 4adac7df-926f-26a2-2b94-8c16560cd09d

12 files changed, 52 insertions, 48 deletions

diff --git a/common_audio/resampler/include/push_resampler.h b/common_audio/resampler/include/push_resampler.h
index 770a9926..f04dc0f3 100644
--- a/common_audio/resampler/include/push_resampler.h
+++ b/common_audio/resampler/include/push_resampler.h
@@ -20,6 +20,7 @@ class PushSincResampler;
 
 // Wraps PushSincResampler to provide stereo support.
 // TODO(ajm): add support for an arbitrary number of channels.
+template <typename T>
 class PushResampler {
  public:
   PushResampler();
@@ -32,22 +33,18 @@ class PushResampler {
 
   // Returns the total number of samples provided in destination (e.g. 32 kHz,
   // 2 channel audio gives 640 samples).
-  int Resample(const int16_t* src, int src_length, int16_t* dst,
-               int dst_capacity);
+  int Resample(const T* src, int src_length, T* dst, int dst_capacity);
 
  private:
-  int ResampleSinc(const int16_t* src, int src_length, int16_t* dst,
-                   int dst_capacity);
-
   scoped_ptr<PushSincResampler> sinc_resampler_;
   scoped_ptr<PushSincResampler> sinc_resampler_right_;
   int src_sample_rate_hz_;
   int dst_sample_rate_hz_;
   int num_channels_;
-  scoped_array<int16_t> src_left_;
-  scoped_array<int16_t> src_right_;
-  scoped_array<int16_t> dst_left_;
-  scoped_array<int16_t> dst_right_;
+  scoped_ptr<T[]> src_left_;
+  scoped_ptr<T[]> src_right_;
+  scoped_ptr<T[]> dst_left_;
+  scoped_ptr<T[]> dst_right_;
 };
 
 }  // namespace webrtc
diff --git a/common_audio/resampler/push_resampler.cc b/common_audio/resampler/push_resampler.cc
index 29944187..973c8f74 100644
--- a/common_audio/resampler/push_resampler.cc
+++ b/common_audio/resampler/push_resampler.cc
@@ -18,22 +18,21 @@
 
 namespace webrtc {
 
-PushResampler::PushResampler()
+template <typename T>
+PushResampler<T>::PushResampler()
     : src_sample_rate_hz_(0),
       dst_sample_rate_hz_(0),
-      num_channels_(0),
-      src_left_(NULL),
-      src_right_(NULL),
-      dst_left_(NULL),
-      dst_right_(NULL) {
+      num_channels_(0) {
 }
 
-PushResampler::~PushResampler() {
+template <typename T>
+PushResampler<T>::~PushResampler() {
 }
 
-int PushResampler::InitializeIfNeeded(int src_sample_rate_hz,
-                                      int dst_sample_rate_hz,
-                                      int num_channels) {
+template <typename T>
+int PushResampler<T>::InitializeIfNeeded(int src_sample_rate_hz,
+                                         int dst_sample_rate_hz,
+                                         int num_channels) {
   if (src_sample_rate_hz == src_sample_rate_hz_ &&
       dst_sample_rate_hz == dst_sample_rate_hz_ &&
       num_channels == num_channels_)
@@ -53,10 +52,10 @@ int PushResampler::InitializeIfNeeded(int src_sample_rate_hz,
   sinc_resampler_.reset(new PushSincResampler(src_size_10ms_mono,
                                               dst_size_10ms_mono));
   if (num_channels_ == 2) {
-    src_left_.reset(new int16_t[src_size_10ms_mono]);
-    src_right_.reset(new int16_t[src_size_10ms_mono]);
-    dst_left_.reset(new int16_t[dst_size_10ms_mono]);
-    dst_right_.reset(new int16_t[dst_size_10ms_mono]);
+    src_left_.reset(new T[src_size_10ms_mono]);
+    src_right_.reset(new T[src_size_10ms_mono]);
+    dst_left_.reset(new T[dst_size_10ms_mono]);
+    dst_right_.reset(new T[dst_size_10ms_mono]);
     sinc_resampler_right_.reset(new PushSincResampler(src_size_10ms_mono,
                                                       dst_size_10ms_mono));
   }
@@ -64,8 +63,9 @@ int PushResampler::InitializeIfNeeded(int src_sample_rate_hz,
   return 0;
 }
 
-int PushResampler::Resample(const int16_t* src, int src_length,
-                            int16_t* dst, int dst_capacity) {
+template <typename T>
+int PushResampler<T>::Resample(const T* src, int src_length, T* dst,
+                               int dst_capacity) {
   const int src_size_10ms = src_sample_rate_hz_ * num_channels_ / 100;
   const int dst_size_10ms = dst_sample_rate_hz_ * num_channels_ / 100;
   if (src_length != src_size_10ms || dst_capacity < dst_size_10ms)
@@ -74,13 +74,13 @@ int PushResampler::Resample(const int16_t* src, int src_length,
   if (src_sample_rate_hz_ == dst_sample_rate_hz_) {
     // The old resampler provides this memcpy facility in the case of matching
     // sample rates, so reproduce it here for the sinc resampler.
-    memcpy(dst, src, src_length * sizeof(int16_t));
+    memcpy(dst, src, src_length * sizeof(T));
     return src_length;
   }
   if (num_channels_ == 2) {
     const int src_length_mono = src_length / num_channels_;
     const int dst_capacity_mono = dst_capacity / num_channels_;
-    int16_t* deinterleaved[] = {src_left_.get(), src_right_.get()};
+    T* deinterleaved[] = {src_left_.get(), src_right_.get()};
     Deinterleave(src, src_length_mono, num_channels_, deinterleaved);
 
     int dst_length_mono =
@@ -98,4 +98,8 @@ int PushResampler::Resample(const int16_t* src, int src_length,
   }
 }
 
+// Explictly generate required instantiations.
+template class PushResampler<int16_t>;
+template class PushResampler<float>;
+
 }  // namespace webrtc
diff --git a/common_audio/resampler/push_resampler_unittest.cc b/common_audio/resampler/push_resampler_unittest.cc
index c40923bf..4449f4c6 100644
--- a/common_audio/resampler/push_resampler_unittest.cc
+++ b/common_audio/resampler/push_resampler_unittest.cc
@@ -16,7 +16,7 @@
 namespace webrtc {
 
 TEST(PushResamplerTest, VerifiesInputParameters) {
-  PushResampler resampler;
+  PushResampler<int16_t> resampler;
   EXPECT_EQ(-1, resampler.InitializeIfNeeded(-1, 16000, 1));
   EXPECT_EQ(-1, resampler.InitializeIfNeeded(16000, -1, 1));
   EXPECT_EQ(-1, resampler.InitializeIfNeeded(16000, 16000, 0));
diff --git a/common_audio/resampler/push_sinc_resampler.h b/common_audio/resampler/push_sinc_resampler.h
index fa1bb3e7..7c804fea 100644
--- a/common_audio/resampler/push_sinc_resampler.h
+++ b/common_audio/resampler/push_sinc_resampler.h
@@ -44,6 +44,9 @@ class PushSincResampler : public SincResamplerCallback {
   virtual void Run(int frames, float* destination) OVERRIDE;
 
   SincResampler* get_resampler_for_testing() { return resampler_.get(); }
+  static float AlgorithmicDelaySeconds(int source_rate_hz) {
+    return 1.f / source_rate_hz * SincResampler::kKernelSize / 2;
+  }
 
  private:
   scoped_ptr<SincResampler> resampler_;
diff --git a/modules/audio_coding/main/acm2/acm_resampler.h b/modules/audio_coding/main/acm2/acm_resampler.h
index 5d952e54..644a32f0 100644
--- a/modules/audio_coding/main/acm2/acm_resampler.h
+++ b/modules/audio_coding/main/acm2/acm_resampler.h
@@ -29,7 +29,7 @@ class ACMResampler {
                      int16_t* out_audio);
 
  private:
-  PushResampler resampler_;
+  PushResampler<int16_t> resampler_;
 };
 
 }  // namespace acm2
diff --git a/modules/audio_coding/main/source/acm_resampler.h b/modules/audio_coding/main/source/acm_resampler.h
index b50e722c..ceeae05f 100644
--- a/modules/audio_coding/main/source/acm_resampler.h
+++ b/modules/audio_coding/main/source/acm_resampler.h
@@ -15,7 +15,6 @@
 #include "webrtc/typedefs.h"
 
 namespace webrtc {
-
 namespace acm1 {
 
 class ACMResampler {
@@ -30,11 +29,10 @@ class ACMResampler {
                          uint8_t num_audio_channels);
 
  private:
-  PushResampler resampler_;
+  PushResampler<int16_t> resampler_;
 };
 
 }  // namespace acm1
-
 }  // namespace webrtc
 
 #endif  // WEBRTC_MODULES_AUDIO_CODING_MAIN_SOURCE_ACM_RESAMPLER_H_
diff --git a/voice_engine/channel.h b/voice_engine/channel.h
index 3377be0e..8f47b48c 100644
--- a/voice_engine/channel.h
+++ b/voice_engine/channel.h
@@ -519,8 +519,8 @@ private:
     bool _externalTransport;
     AudioFrame _audioFrame;
     scoped_ptr<int16_t[]> mono_recording_audio_;
-    // Resampler is used when input data is stereo while codec is mono.
-    PushResampler input_resampler_;
+    // Downsamples to the codec rate if necessary.
+    PushResampler<int16_t> input_resampler_;
     uint8_t _audioLevel_dBov;
     FilePlayer* _inputFilePlayerPtr;
     FilePlayer* _outputFilePlayerPtr;
diff --git a/voice_engine/output_mixer.h b/voice_engine/output_mixer.h
index bd17c3be..859ce64b 100644
--- a/voice_engine/output_mixer.h
+++ b/voice_engine/output_mixer.h
@@ -133,8 +133,10 @@ private:
     CriticalSectionWrapper& _fileCritSect;
     AudioConferenceMixer& _mixerModule;
     AudioFrame _audioFrame;
-    PushResampler resampler_;  // converts mixed audio to fit ADM format
-    PushResampler audioproc_resampler_;  // converts mixed audio to fit APM rate
+    // Converts mixed audio to the audio device output rate.
+    PushResampler<int16_t> resampler_;
+    // Converts mixed audio to the audio processing rate.
+    PushResampler<int16_t> audioproc_resampler_;
     AudioLevel _audioLevel;    // measures audio level for the combined signal
     DtmfInband _dtmfGenerator;
     int _instanceId;
diff --git a/voice_engine/transmit_mixer.h b/voice_engine/transmit_mixer.h
index b408614a..cc611308 100644
--- a/voice_engine/transmit_mixer.h
+++ b/voice_engine/transmit_mixer.h
@@ -200,7 +200,7 @@ private:
     // owns
     MonitorModule _monitorModule;
     AudioFrame _audioFrame;
-    PushResampler resampler_;  // ADM sample rate -> mixing rate
+    PushResampler<int16_t> resampler_;  // ADM sample rate -> mixing rate
     FilePlayer* _filePlayerPtr;
     FileRecorder* _fileRecorderPtr;
     FileRecorder* _fileCallRecorderPtr;
diff --git a/voice_engine/utility.cc b/voice_engine/utility.cc
index 5058aa32..b7eb885c 100644
--- a/voice_engine/utility.cc
+++ b/voice_engine/utility.cc
@@ -25,7 +25,7 @@ namespace voe {
 // ConvertToCodecFormat, but if we're to consolidate we should probably make a
 // real converter class.
 void RemixAndResample(const AudioFrame& src_frame,
-                      PushResampler* resampler,
+                      PushResampler<int16_t>* resampler,
                       AudioFrame* dst_frame) {
   const int16_t* audio_ptr = src_frame.data_;
   int audio_ptr_num_channels = src_frame.num_channels_;
@@ -76,7 +76,7 @@ void DownConvertToCodecFormat(const int16_t* src_data,
                               int codec_num_channels,
                               int codec_rate_hz,
                               int16_t* mono_buffer,
-                              PushResampler* resampler,
+                              PushResampler<int16_t>* resampler,
                               AudioFrame* dst_af) {
   assert(samples_per_channel <= kMaxMonoDataSizeSamples);
   assert(num_channels == 1 || num_channels == 2);
diff --git a/voice_engine/utility.h b/voice_engine/utility.h
index f6fa35bf..127bdba0 100644
--- a/voice_engine/utility.h
+++ b/voice_engine/utility.h
@@ -15,12 +15,12 @@
 #ifndef WEBRTC_VOICE_ENGINE_UTILITY_H_
 #define WEBRTC_VOICE_ENGINE_UTILITY_H_
 
+#include "webrtc/common_audio/resampler/include/push_resampler.h"
 #include "webrtc/typedefs.h"
 
 namespace webrtc {
 
 class AudioFrame;
-class PushResampler;
 
 namespace voe {
 
@@ -30,7 +30,7 @@ namespace voe {
 //
 // On failure, returns -1 and copies |src_frame| to |dst_frame|.
 void RemixAndResample(const AudioFrame& src_frame,
-                      PushResampler* resampler,
+                      PushResampler<int16_t>* resampler,
                       AudioFrame* dst_frame);
 
 // Downmix and downsample the audio in |src_data| to |dst_af| as necessary,
@@ -44,7 +44,7 @@ void DownConvertToCodecFormat(const int16_t* src_data,
                               int codec_num_channels,
                               int codec_rate_hz,
                               int16_t* mono_buffer,
-                              PushResampler* resampler,
+                              PushResampler<int16_t>* resampler,
                               AudioFrame* dst_af);
 
 void MixWithSat(int16_t target[],
diff --git a/voice_engine/utility_unittest.cc b/voice_engine/utility_unittest.cc
index a5d0bcd8..8f7efa87 100644
--- a/voice_engine/utility_unittest.cc
+++ b/voice_engine/utility_unittest.cc
@@ -39,7 +39,7 @@ class UtilityTest : public ::testing::Test {
                        int dst_channels, int dst_sample_rate_hz,
                        FunctionToTest function);
 
-  PushResampler resampler_;
+  PushResampler<int16_t> resampler_;
   AudioFrame src_frame_;
   AudioFrame dst_frame_;
   AudioFrame golden_frame_;
@@ -127,11 +127,11 @@ void VerifyFramesAreEqual(const AudioFrame& ref_frame,
 }
 
 void UtilityTest::RunResampleTest(int src_channels,
-                                      int src_sample_rate_hz,
-                                      int dst_channels,
-                                      int dst_sample_rate_hz,
-                                      FunctionToTest function) {
-  PushResampler resampler;  // Create a new one with every test.
+                                  int src_sample_rate_hz,
+                                  int dst_channels,
+                                  int dst_sample_rate_hz,
+                                  FunctionToTest function) {
+  PushResampler<int16_t> resampler;  // Create a new one with every test.
   const int16_t kSrcLeft = 30;  // Shouldn't overflow for any used sample rate.
   const int16_t kSrcRight = 15;
   const float resampling_factor = (1.0 * src_sample_rate_hz) /