Remove unnecessary memory reallocation in fft (#2080)

fft_out needs to be twice the frame_size, not the frame_step.  It is resized in fft() anyway, but this change prevents an unnecessary reallocation.

n_fft must match the mel filter size, so it is best not to calculate it from the framesize.

We only need to get the magnitudes for half the spectrum since the other half is a mirror and not used in the mel filter loop later.

diff --git a/whisper.cpp b/whisper.cpp
index 1a6d889af285a25630ef4efded4e608ba5a688f6..f31309ed3b4b7fa532c32fb81a562376b4914288 100644 (file)
--- a/whisper.cpp
+++ b/whisper.cpp
@@ -2900,11 +2900,13 @@ static void log_mel_spectrogram_worker_thread(int ith, const std::vector<float>
                                               int n_samples, int frame_size, int frame_step, int n_threads,
                                               const whisper_filters & filters, whisper_mel & mel) {
     std::vector<float> fft_in(frame_size, 0.0);
-    std::vector<float> fft_out(2 * frame_step);
-    // make sure n_fft == 1 + (WHISPER_N_FFT / 2), bin_0 to bin_nyquist
-    int n_fft = 1 + (frame_size / 2);
+    std::vector<float> fft_out(2 * frame_size);
+    int n_fft = filters.n_fft;
     int i = ith;
 
+    // make sure n_fft == 1 + (WHISPER_N_FFT / 2), bin_0 to bin_nyquist
+    assert( n_fft == 1 + (frame_size / 2) );
+    
     // calculate FFT only when fft_in are not all zero
     for (; i < std::min(n_samples / frame_step + 1, mel.n_len); i += n_threads) {
         const int offset = i * frame_step;
@@ -2923,7 +2925,7 @@ static void log_mel_spectrogram_worker_thread(int ith, const std::vector<float>
 
         // Calculate modulus^2 of complex numbers
         // Use pow(fft_out[2 * j + 0], 2) + pow(fft_out[2 * j + 1], 2) causes inference quality problem? Interesting.
-        for (int j = 0; j < frame_size; j++) {
+        for (int j = 0; j < n_fft; j++) {
             fft_out[j] = (fft_out[2 * j + 0] * fft_out[2 * j + 0] + fft_out[2 * j + 1] * fft_out[2 * j + 1]);
         }