Convert IntelligibilityEnhancer to multi-threaded mode

webrtc/modules/audio_processing/noise_suppression_impl.cc

 /*
  *  Copyright (c) 2012 The WebRTC project authors. All Rights Reserved.
  *
  *  Use of this source code is governed by a BSD-style license
  *  that can be found in the LICENSE file in the root of the source
  *  tree. An additional intellectual property rights grant can be found
  *  in the file PATENTS.  All contributing project authors may
  *  be found in the AUTHORS file in the root of the source tree.
  */
 
@@ skipping 182 matching lines @@
     const uint32_t* noise = WebRtcNsx_noise_estimate(suppressor->state());
     for (size_t i = 0; i < noise_estimate.size(); ++i) {
       noise_estimate[i] += kNormalizationFactor *
           static_cast<float>(noise[i]) / suppressors_.size();
     }
   }
 #endif
   return noise_estimate;
 }
 
+size_t NoiseSuppressionImpl::num_noise_bins() const {
+  rtc::CritScope cs(crit_);
+  size_t num_freq = 0u;

[hlundin-webrtc, 2016/03/08 10:28:48]

Exactly one of WEBRTC_NS_FLOAT and _FIXED must be defined, right? If so,
simplify this function:

{
  rtc::CritScope cs(crit_);
  #if defined(WEBRTC_NS_FLOAT)
  return WebRtcNs_num_freq();
  #elif defined(WEBRTC_NS_FIXED)
  return WebRtcNsx_num_freq();
  #else
  static_assert(false, "You are doing it wrong.")
  #endif
}

(If you don't want to verify that exactly one of the defines are set, skip the
static assert and make the #elif a simple #else.)

[kwiberg-webrtc, 2016/03/08 10:40:22]

This still doesn't fail if both are defined.

Also, it's customary to use #error and not static_assert(false) in situations
like this, both for historical reasons and because it's less complicated and
works in more places.

[aluebs-webrtc, 2016/03/08 10:53:00]

In general I think the functions with only one return statement are less bug
prone, but I can understand the argument for simplicity, so I accepted your
suggestion. I am not adding the verification, because, as kwiberg pointed out,
it doesn't fail if both are defined and because the whole file assumes one of
them to be defined (and that is how they are defined in audio_processing.gypi).
But if you think I should add this additional verification, I am happy to do so.

[hlundin-webrtc, 2016/03/08 13:35:19]

This is good. Don't add any verification. This is not the place for it. (And I
think you will get implicit verification that at least one is defined the way
you wrote it now.)

+#if defined(WEBRTC_NS_FLOAT)
+  num_freq = WebRtcNs_num_freq();
+#elif defined(WEBRTC_NS_FIXED)
+  num_freq = WebRtcNsx_num_freq();
+#endif
+  return num_freq;
+}
+
 }  // namespace webrtc