Compute the tail time for an IIRFilter from its coefficients

third_party/WebKit/Source/modules/webaudio/IIRDSPKernel.cpp

Index: third_party/WebKit/Source/modules/webaudio/IIRDSPKernel.cpp
diff --git a/third_party/WebKit/Source/modules/webaudio/IIRDSPKernel.cpp b/third_party/WebKit/Source/modules/webaudio/IIRDSPKernel.cpp
index e3e38525c1180bde57f2924d1d8a47e2a44ee4cd..1a86b092464bdc402539a792804fda5a1390af41 100644
--- a/third_party/WebKit/Source/modules/webaudio/IIRDSPKernel.cpp
+++ b/third_party/WebKit/Source/modules/webaudio/IIRDSPKernel.cpp
@@ -8,6 +8,12 @@
 
 namespace blink {
 
+IIRDSPKernel::IIRDSPKernel(IIRProcessor* processor)
+    : AudioDSPKernel(processor),
+      iir_(processor->Feedforward(), processor->Feedback()) {
+  tail_time_ = iir_.TailTime(processor->SampleRate());

[hongchan, 2017/05/12 16:24:04]

Is this the only moment that the tail time is computed?

[Raymond Toy, 2017/05/12 17:33:18]

Since the filter can't change, we only need to compute the tail time once during
construction of the filter.

+}
+
 void IIRDSPKernel::Process(const float* source,
                            float* destination,
                            size_t frames_to_process) {
@@ -41,14 +47,7 @@ void IIRDSPKernel::GetFrequencyResponse(int n_frequencies,
 }
 
 double IIRDSPKernel::TailTime() const {
-  // TODO(rtoy): This is true mathematically (infinite impulse response), but
-  // perhaps it should be limited to a smaller value, possibly based on the
-  // actual filter coefficients.  To do that, we would probably need to find the
-  // pole, r, with largest magnitude and select some threshold, eps, such that
-  // |r|^n < eps for all n >= N.  N is then the tailTime for the filter.  If the
-  // the magnitude of r is greater than or equal to 1, the infinity is the right
-  // answer. (There is no constraint on the IIR filter that it be stable.)
-  return std::numeric_limits<double>::infinity();
+  return tail_time_;
 }
 
 double IIRDSPKernel::LatencyTime() const {