Merge "Merge upstream SHA 04cb763"android-cts_7.1_r1android-cts-7.1_r9android-cts-7.1_r8android-cts-7.1_r7android-cts-7.1_r6android-cts-7.1_r5android-cts-7.1_r4android-cts-7.1_r3android-cts-7.1_r29android-cts-7.1_r28android-cts-7.1_r27android-cts-7.1_r26android-cts-7.1_r25android-cts-7.1_r24android-cts-7.1_r23android-cts-7.1_r22android-cts-7.1_r21android-cts-7.1_r20android-cts-7.1_r2android-cts-7.1_r19android-cts-7.1_r18android-cts-7.1_r17android-cts-7.1_r16android-cts-7.1_r15android-cts-7.1_r14android-cts-7.1_r13android-cts-7.1_r12android-cts-7.1_r11android-cts-7.1_r10android-cts-7.1_r1android-cts-7.0_r9android-cts-7.0_r8android-cts-7.0_r7android-cts-7.0_r6android-cts-7.0_r5android-cts-7.0_r4android-cts-7.0_r33android-cts-7.0_r32android-cts-7.0_r31android-cts-7.0_r30android-cts-7.0_r3android-cts-7.0_r29android-cts-7.0_r28android-cts-7.0_r27android-cts-7.0_r26android-cts-7.0_r25android-cts-7.0_r24android-cts-7.0_r23android-cts-7.0_r22android-cts-7.0_r21android-cts-7.0_r20android-cts-7.0_r2android-cts-7.0_r19android-cts-7.0_r18android-cts-7.0_r17android-cts-7.0_r16android-cts-7.0_r15android-cts-7.0_r14android-cts-7.0_r13android-cts-7.0_r12android-cts-7.0_r11android-cts-7.0_r10android-cts-7.0_r1android-7.1.2_r9android-7.1.2_r8android-7.1.2_r6android-7.1.2_r5android-7.1.2_r4android-7.1.2_r39android-7.1.2_r38android-7.1.2_r37android-7.1.2_r36android-7.1.2_r33android-7.1.2_r32android-7.1.2_r30android-7.1.2_r3android-7.1.2_r29android-7.1.2_r28android-7.1.2_r27android-7.1.2_r25android-7.1.2_r24android-7.1.2_r23android-7.1.2_r2android-7.1.2_r19android-7.1.2_r18android-7.1.2_r17android-7.1.2_r16android-7.1.2_r15android-7.1.2_r14android-7.1.2_r13android-7.1.2_r12android-7.1.2_r11android-7.1.2_r10android-7.1.2_r1android-7.1.1_r9android-7.1.1_r8android-7.1.1_r7android-7.1.1_r61android-7.1.1_r60android-7.1.1_r6android-7.1.1_r59android-7.1.1_r58android-7.1.1_r57android-7.1.1_r56android-7.1.1_r55android-7.1.1_r54android-7.1.1_r53android-7.1.1_r52android-7.1.1_r51android-7.1.1_r50android-7.1.1_r49android-7.1.1_r48android-7.1.1_r47android-7.1.1_r46android-7.1.1_r45android-7.1.1_r44android-7.1.1_r43android-7.1.1_r42android-7.1.1_r41android-7.1.1_r40android-7.1.1_r4android-7.1.1_r39android-7.1.1_r38android-7.1.1_r35android-7.1.1_r33android-7.1.1_r32android-7.1.1_r31android-7.1.1_r3android-7.1.1_r28android-7.1.1_r27android-7.1.1_r26android-7.1.1_r25android-7.1.1_r24android-7.1.1_r23android-7.1.1_r22android-7.1.1_r21android-7.1.1_r20android-7.1.1_r2android-7.1.1_r17android-7.1.1_r16android-7.1.1_r15android-7.1.1_r14android-7.1.1_r13android-7.1.1_r12android-7.1.1_r11android-7.1.1_r10android-7.1.1_r1android-7.1.0_r7android-7.1.0_r6android-7.1.0_r5android-7.1.0_r4android-7.1.0_r3android-7.1.0_r2android-7.1.0_r1android-7.0.0_r9android-7.0.0_r8android-7.0.0_r7android-7.0.0_r6android-7.0.0_r5android-7.0.0_r4android-7.0.0_r36android-7.0.0_r35android-7.0.0_r34android-7.0.0_r33android-7.0.0_r32android-7.0.0_r31android-7.0.0_r30android-7.0.0_r3android-7.0.0_r29android-7.0.0_r28android-7.0.0_r27android-7.0.0_r24android-7.0.0_r21android-7.0.0_r19android-7.0.0_r17android-7.0.0_r15android-7.0.0_r14android-7.0.0_r13android-7.0.0_r12android-7.0.0_r11android-7.0.0_r10android-7.0.0_r1nougat-releasenougat-mr2.3-releasenougat-mr2.2-releasenougat-mr2.1-releasenougat-mr2-security-releasenougat-mr2-releasenougat-mr2-pixel-releasenougat-mr2-devnougat-mr1.8-releasenougat-mr1.7-releasenougat-mr1.6-releasenougat-mr1.5-releasenougat-mr1.4-releasenougat-mr1.3-releasenougat-mr1.2-releasenougat-mr1.1-releasenougat-mr1-volantis-releasenougat-mr1-security-releasenougat-mr1-releasenougat-mr1-flounder-releasenougat-mr1-devnougat-mr1-cts-releasenougat-mr0.5-releasenougat-dr1-releasenougat-devnougat-cts-releasenougat-bugfix-release

am: 9a337512d9

* commit '9a337512d97e37afc142dee4fd50a41b741a87d2': (797 commits)
  Add tests for verifying transport feedback for audio and video.
  Eliminate defines in talk/
  Revert of Update with new default boringssl no-aes cipher suites. Re-enable tests. (patchset #3 id:40001 of https://codereview.webrtc.org/1550773002/ )
  Remove assert which was incorrectly added to TcpPort::OnSentPacket.
  Reland Connect TurnPort and TCPPort to AsyncPacketSocket::SignalSentPacket.
  Update with new default boringssl no-aes cipher suites. Re-enable tests.
  Revert of Connect TurnPort and TCPPort to AsyncPacketSocket::SignalSentPacket. (patchset #3 id:40001 of https://codereview.webrtc.org/1577873003/ )
  Re-land: "Use an explicit identifier in Config"
  Connect TurnPort and TCPPort to AsyncPacketSocket::SignalSentPacket.
  Revert of Delete remnants of non-square pixel support from cricket::VideoFrame. (patchset #1 id:1 of https://codereview.webrtc.org/1586613002/ )
  Remove libfuzzer trybot from default trybot set.
  Add ramp-up tests for transport sequence number with and w/o audio.
  Delete remnants of non-square pixel support from cricket::VideoFrame.
  Fix IPAddress::ToSensitiveString() to avoid dependency on inet_ntop().
  Revert of Storing raw audio sink for default audio track. (patchset #7 id:120001 of https://codereview.chromium.org/1551813002/ )
  Re-enable tests that failed under Linux_Msan.
  Revert of Use an explicit identifier in Config (patchset #4 id:60001 of https://codereview.webrtc.org/1538643004/ )
  Roll chromium_revision 346fea9..099be58 (369082:369139)
  Disable WebRtcVideoChannel2BaseTest.SendManyResizeOnce for TSan
  Add build_protobuf variable.
  ...

1 files changed, 95 insertions, 98 deletions

diff --git a/webrtc/modules/audio_processing/aec/aec_core_sse2.c b/webrtc/modules/audio_processing/aec/aec_core_sse2.c
index b1bffcbb9f..f897a4c0c7 100644
--- a/webrtc/modules/audio_processing/aec/aec_core_sse2.c
+++ b/webrtc/modules/audio_processing/aec/aec_core_sse2.c
@@ -29,67 +29,76 @@ __inline static float MulIm(float aRe, float aIm, float bRe, float bIm) {
   return aRe * bIm + aIm * bRe;
 }
 
-static void FilterFarSSE2(AecCore* aec, float yf[2][PART_LEN1]) {
+static void FilterFarSSE2(
+    int num_partitions,
+    int x_fft_buf_block_pos,
+    float x_fft_buf[2][kExtendedNumPartitions * PART_LEN1],
+    float h_fft_buf[2][kExtendedNumPartitions * PART_LEN1],
+    float y_fft[2][PART_LEN1]) {
+
   int i;
-  const int num_partitions = aec->num_partitions;
   for (i = 0; i < num_partitions; i++) {
     int j;
-    int xPos = (i + aec->xfBufBlockPos) * PART_LEN1;
+    int xPos = (i + x_fft_buf_block_pos) * PART_LEN1;
     int pos = i * PART_LEN1;
     // Check for wrap
-    if (i + aec->xfBufBlockPos >= num_partitions) {
+    if (i + x_fft_buf_block_pos >= num_partitions) {
       xPos -= num_partitions * (PART_LEN1);
     }
 
     // vectorized code (four at once)
     for (j = 0; j + 3 < PART_LEN1; j += 4) {
-      const __m128 xfBuf_re = _mm_loadu_ps(&aec->xfBuf[0][xPos + j]);
-      const __m128 xfBuf_im = _mm_loadu_ps(&aec->xfBuf[1][xPos + j]);
-      const __m128 wfBuf_re = _mm_loadu_ps(&aec->wfBuf[0][pos + j]);
-      const __m128 wfBuf_im = _mm_loadu_ps(&aec->wfBuf[1][pos + j]);
-      const __m128 yf_re = _mm_loadu_ps(&yf[0][j]);
-      const __m128 yf_im = _mm_loadu_ps(&yf[1][j]);
-      const __m128 a = _mm_mul_ps(xfBuf_re, wfBuf_re);
-      const __m128 b = _mm_mul_ps(xfBuf_im, wfBuf_im);
-      const __m128 c = _mm_mul_ps(xfBuf_re, wfBuf_im);
-      const __m128 d = _mm_mul_ps(xfBuf_im, wfBuf_re);
+      const __m128 x_fft_buf_re = _mm_loadu_ps(&x_fft_buf[0][xPos + j]);
+      const __m128 x_fft_buf_im = _mm_loadu_ps(&x_fft_buf[1][xPos + j]);
+      const __m128 h_fft_buf_re = _mm_loadu_ps(&h_fft_buf[0][pos + j]);
+      const __m128 h_fft_buf_im = _mm_loadu_ps(&h_fft_buf[1][pos + j]);
+      const __m128 y_fft_re = _mm_loadu_ps(&y_fft[0][j]);
+      const __m128 y_fft_im = _mm_loadu_ps(&y_fft[1][j]);
+      const __m128 a = _mm_mul_ps(x_fft_buf_re, h_fft_buf_re);
+      const __m128 b = _mm_mul_ps(x_fft_buf_im, h_fft_buf_im);
+      const __m128 c = _mm_mul_ps(x_fft_buf_re, h_fft_buf_im);
+      const __m128 d = _mm_mul_ps(x_fft_buf_im, h_fft_buf_re);
       const __m128 e = _mm_sub_ps(a, b);
       const __m128 f = _mm_add_ps(c, d);
-      const __m128 g = _mm_add_ps(yf_re, e);
-      const __m128 h = _mm_add_ps(yf_im, f);
-      _mm_storeu_ps(&yf[0][j], g);
-      _mm_storeu_ps(&yf[1][j], h);
+      const __m128 g = _mm_add_ps(y_fft_re, e);
+      const __m128 h = _mm_add_ps(y_fft_im, f);
+      _mm_storeu_ps(&y_fft[0][j], g);
+      _mm_storeu_ps(&y_fft[1][j], h);
     }
     // scalar code for the remaining items.
     for (; j < PART_LEN1; j++) {
-      yf[0][j] += MulRe(aec->xfBuf[0][xPos + j],
-                        aec->xfBuf[1][xPos + j],
-                        aec->wfBuf[0][pos + j],
-                        aec->wfBuf[1][pos + j]);
-      yf[1][j] += MulIm(aec->xfBuf[0][xPos + j],
-                        aec->xfBuf[1][xPos + j],
-                        aec->wfBuf[0][pos + j],
-                        aec->wfBuf[1][pos + j]);
+      y_fft[0][j] += MulRe(x_fft_buf[0][xPos + j],
+                           x_fft_buf[1][xPos + j],
+                           h_fft_buf[0][pos + j],
+                           h_fft_buf[1][pos + j]);
+      y_fft[1][j] += MulIm(x_fft_buf[0][xPos + j],
+                           x_fft_buf[1][xPos + j],
+                           h_fft_buf[0][pos + j],
+                           h_fft_buf[1][pos + j]);
     }
   }
 }
 
-static void ScaleErrorSignalSSE2(AecCore* aec, float ef[2][PART_LEN1]) {
+static void ScaleErrorSignalSSE2(int extended_filter_enabled,
+                                 float normal_mu,
+                                 float normal_error_threshold,
+                                 float x_pow[PART_LEN1],
+                                 float ef[2][PART_LEN1]) {
   const __m128 k1e_10f = _mm_set1_ps(1e-10f);
-  const __m128 kMu = aec->extended_filter_enabled ? _mm_set1_ps(kExtendedMu)
-                                                  : _mm_set1_ps(aec->normal_mu);
-  const __m128 kThresh = aec->extended_filter_enabled
+  const __m128 kMu = extended_filter_enabled ? _mm_set1_ps(kExtendedMu)
+      : _mm_set1_ps(normal_mu);
+  const __m128 kThresh = extended_filter_enabled
                              ? _mm_set1_ps(kExtendedErrorThreshold)
-                             : _mm_set1_ps(aec->normal_error_threshold);
+                             : _mm_set1_ps(normal_error_threshold);
 
   int i;
   // vectorized code (four at once)
   for (i = 0; i + 3 < PART_LEN1; i += 4) {
-    const __m128 xPow = _mm_loadu_ps(&aec->xPow[i]);
+    const __m128 x_pow_local = _mm_loadu_ps(&x_pow[i]);
     const __m128 ef_re_base = _mm_loadu_ps(&ef[0][i]);
     const __m128 ef_im_base = _mm_loadu_ps(&ef[1][i]);
 
-    const __m128 xPowPlus = _mm_add_ps(xPow, k1e_10f);
+    const __m128 xPowPlus = _mm_add_ps(x_pow_local, k1e_10f);
     __m128 ef_re = _mm_div_ps(ef_re_base, xPowPlus);
     __m128 ef_im = _mm_div_ps(ef_im_base, xPowPlus);
     const __m128 ef_re2 = _mm_mul_ps(ef_re, ef_re);
@@ -116,14 +125,14 @@ static void ScaleErrorSignalSSE2(AecCore* aec, float ef[2][PART_LEN1]) {
   // scalar code for the remaining items.
   {
     const float mu =
-        aec->extended_filter_enabled ? kExtendedMu : aec->normal_mu;
-    const float error_threshold = aec->extended_filter_enabled
+        extended_filter_enabled ? kExtendedMu : normal_mu;
+    const float error_threshold = extended_filter_enabled
                                       ? kExtendedErrorThreshold
-                                      : aec->normal_error_threshold;
+                                      : normal_error_threshold;
     for (; i < (PART_LEN1); i++) {
       float abs_ef;
-      ef[0][i] /= (aec->xPow[i] + 1e-10f);
-      ef[1][i] /= (aec->xPow[i] + 1e-10f);
+      ef[0][i] /= (x_pow[i] + 1e-10f);
+      ef[1][i] /= (x_pow[i] + 1e-10f);
       abs_ef = sqrtf(ef[0][i] * ef[0][i] + ef[1][i] * ef[1][i]);
 
       if (abs_ef > error_threshold) {
@@ -139,33 +148,36 @@ static void ScaleErrorSignalSSE2(AecCore* aec, float ef[2][PART_LEN1]) {
   }
 }
 
-static void FilterAdaptationSSE2(AecCore* aec,
-                                 float* fft,
-                                 float ef[2][PART_LEN1]) {
+static void FilterAdaptationSSE2(
+    int num_partitions,
+    int x_fft_buf_block_pos,
+    float x_fft_buf[2][kExtendedNumPartitions * PART_LEN1],
+    float e_fft[2][PART_LEN1],
+    float h_fft_buf[2][kExtendedNumPartitions * PART_LEN1]) {
+  float fft[PART_LEN2];
   int i, j;
-  const int num_partitions = aec->num_partitions;
   for (i = 0; i < num_partitions; i++) {
-    int xPos = (i + aec->xfBufBlockPos) * (PART_LEN1);
+    int xPos = (i + x_fft_buf_block_pos) * (PART_LEN1);
     int pos = i * PART_LEN1;
     // Check for wrap
-    if (i + aec->xfBufBlockPos >= num_partitions) {
+    if (i + x_fft_buf_block_pos >= num_partitions) {
       xPos -= num_partitions * PART_LEN1;
     }
 
     // Process the whole array...
     for (j = 0; j < PART_LEN; j += 4) {
-      // Load xfBuf and ef.
-      const __m128 xfBuf_re = _mm_loadu_ps(&aec->xfBuf[0][xPos + j]);
-      const __m128 xfBuf_im = _mm_loadu_ps(&aec->xfBuf[1][xPos + j]);
-      const __m128 ef_re = _mm_loadu_ps(&ef[0][j]);
-      const __m128 ef_im = _mm_loadu_ps(&ef[1][j]);
-      // Calculate the product of conjugate(xfBuf) by ef.
+      // Load x_fft_buf and e_fft.
+      const __m128 x_fft_buf_re = _mm_loadu_ps(&x_fft_buf[0][xPos + j]);
+      const __m128 x_fft_buf_im = _mm_loadu_ps(&x_fft_buf[1][xPos + j]);
+      const __m128 e_fft_re = _mm_loadu_ps(&e_fft[0][j]);
+      const __m128 e_fft_im = _mm_loadu_ps(&e_fft[1][j]);
+      // Calculate the product of conjugate(x_fft_buf) by e_fft.
       //   re(conjugate(a) * b) = aRe * bRe + aIm * bIm
       //   im(conjugate(a) * b)=  aRe * bIm - aIm * bRe
-      const __m128 a = _mm_mul_ps(xfBuf_re, ef_re);
-      const __m128 b = _mm_mul_ps(xfBuf_im, ef_im);
-      const __m128 c = _mm_mul_ps(xfBuf_re, ef_im);
-      const __m128 d = _mm_mul_ps(xfBuf_im, ef_re);
+      const __m128 a = _mm_mul_ps(x_fft_buf_re, e_fft_re);
+      const __m128 b = _mm_mul_ps(x_fft_buf_im, e_fft_im);
+      const __m128 c = _mm_mul_ps(x_fft_buf_re, e_fft_im);
+      const __m128 d = _mm_mul_ps(x_fft_buf_im, e_fft_re);
       const __m128 e = _mm_add_ps(a, b);
       const __m128 f = _mm_sub_ps(c, d);
       // Interleave real and imaginary parts.
@@ -176,10 +188,10 @@ static void FilterAdaptationSSE2(AecCore* aec,
       _mm_storeu_ps(&fft[2 * j + 4], h);
     }
     // ... and fixup the first imaginary entry.
-    fft[1] = MulRe(aec->xfBuf[0][xPos + PART_LEN],
-                   -aec->xfBuf[1][xPos + PART_LEN],
-                   ef[0][PART_LEN],
-                   ef[1][PART_LEN]);
+    fft[1] = MulRe(x_fft_buf[0][xPos + PART_LEN],
+                   -x_fft_buf[1][xPos + PART_LEN],
+                   e_fft[0][PART_LEN],
+                   e_fft[1][PART_LEN]);
 
     aec_rdft_inverse_128(fft);
     memset(fft + PART_LEN, 0, sizeof(float) * PART_LEN);
@@ -197,11 +209,11 @@ static void FilterAdaptationSSE2(AecCore* aec,
     aec_rdft_forward_128(fft);
 
     {
-      float wt1 = aec->wfBuf[1][pos];
-      aec->wfBuf[0][pos + PART_LEN] += fft[1];
+      float wt1 = h_fft_buf[1][pos];
+      h_fft_buf[0][pos + PART_LEN] += fft[1];
       for (j = 0; j < PART_LEN; j += 4) {
-        __m128 wtBuf_re = _mm_loadu_ps(&aec->wfBuf[0][pos + j]);
-        __m128 wtBuf_im = _mm_loadu_ps(&aec->wfBuf[1][pos + j]);
+        __m128 wtBuf_re = _mm_loadu_ps(&h_fft_buf[0][pos + j]);
+        __m128 wtBuf_im = _mm_loadu_ps(&h_fft_buf[1][pos + j]);
         const __m128 fft0 = _mm_loadu_ps(&fft[2 * j + 0]);
         const __m128 fft4 = _mm_loadu_ps(&fft[2 * j + 4]);
         const __m128 fft_re =
@@ -210,10 +222,10 @@ static void FilterAdaptationSSE2(AecCore* aec,
             _mm_shuffle_ps(fft0, fft4, _MM_SHUFFLE(3, 1, 3, 1));
         wtBuf_re = _mm_add_ps(wtBuf_re, fft_re);
         wtBuf_im = _mm_add_ps(wtBuf_im, fft_im);
-        _mm_storeu_ps(&aec->wfBuf[0][pos + j], wtBuf_re);
-        _mm_storeu_ps(&aec->wfBuf[1][pos + j], wtBuf_im);
+        _mm_storeu_ps(&h_fft_buf[0][pos + j], wtBuf_re);
+        _mm_storeu_ps(&h_fft_buf[1][pos + j], wtBuf_im);
       }
-      aec->wfBuf[1][pos] = wt1;
+      h_fft_buf[1][pos] = wt1;
     }
   }
 }
@@ -427,7 +439,8 @@ __inline static void _mm_add_ps_4x1(__m128 sum, float *dst) {
   sum = _mm_add_ps(sum, _mm_shuffle_ps(sum, sum, _MM_SHUFFLE(1, 1, 1, 1)));
   _mm_store_ss(dst, sum);
 }
-static int PartitionDelay(const AecCore* aec) {
+
+static int PartitionDelaySSE2(const AecCore* aec) {
   // Measures the energy in each filter partition and returns the partition with
   // highest energy.
   // TODO(bjornv): Spread computational cost by computing one partition per
@@ -476,7 +489,8 @@ static int PartitionDelay(const AecCore* aec) {
 static void SmoothedPSD(AecCore* aec,
                         float efw[2][PART_LEN1],
                         float dfw[2][PART_LEN1],
-                        float xfw[2][PART_LEN1]) {
+                        float xfw[2][PART_LEN1],
+                        int* extreme_filter_divergence) {
   // Power estimate smoothing coefficients.
   const float* ptrGCoh = aec->extended_filter_enabled
       ? WebRtcAec_kExtendedSmoothingCoefficients[aec->mult - 1]
@@ -595,19 +609,16 @@ static void SmoothedPSD(AecCore* aec,
     seSum += aec->se[i];
   }
 
-  // Divergent filter safeguard.
+  // Divergent filter safeguard update.
   aec->divergeState = (aec->divergeState ? 1.05f : 1.0f) * seSum > sdSum;
 
-  if (aec->divergeState)
-    memcpy(efw, dfw, sizeof(efw[0][0]) * 2 * PART_LEN1);
-
-  // Reset if error is significantly larger than nearend (13 dB).
-  if (!aec->extended_filter_enabled && seSum > (19.95f * sdSum))
-    memset(aec->wfBuf, 0, sizeof(aec->wfBuf));
+  // Signal extreme filter divergence if the error is significantly larger
+  // than the nearend (13 dB).
+  *extreme_filter_divergence = (seSum > (19.95f * sdSum));
 }
 
 // Window time domain data to be used by the fft.
-__inline static void WindowData(float* x_windowed, const float* x) {
+static void WindowDataSSE2(float* x_windowed, const float* x) {
   int i;
   for (i = 0; i < PART_LEN; i += 4) {
     const __m128 vec_Buf1 = _mm_loadu_ps(&x[i]);
@@ -627,8 +638,8 @@ __inline static void WindowData(float* x_windowed, const float* x) {
 }
 
 // Puts fft output data into a complex valued array.
-__inline static void StoreAsComplex(const float* data,
-                                    float data_complex[2][PART_LEN1]) {
+static void StoreAsComplexSSE2(const float* data,
+                               float data_complex[2][PART_LEN1]) {
   int i;
   for (i = 0; i < PART_LEN; i += 4) {
     const __m128 vec_fft0 = _mm_loadu_ps(&data[2 * i]);
@@ -649,32 +660,15 @@ __inline static void StoreAsComplex(const float* data,
 
 static void SubbandCoherenceSSE2(AecCore* aec,
                                  float efw[2][PART_LEN1],
+                                 float dfw[2][PART_LEN1],
                                  float xfw[2][PART_LEN1],
                                  float* fft,
                                  float* cohde,
-                                 float* cohxd) {
-  float dfw[2][PART_LEN1];
+                                 float* cohxd,
+                                 int* extreme_filter_divergence) {
   int i;
 
-  if (aec->delayEstCtr == 0)
-    aec->delayIdx = PartitionDelay(aec);
-
-  // Use delayed far.
-  memcpy(xfw,
-         aec->xfwBuf + aec->delayIdx * PART_LEN1,
-         sizeof(xfw[0][0]) * 2 * PART_LEN1);
-
-  // Windowed near fft
-  WindowData(fft, aec->dBuf);
-  aec_rdft_forward_128(fft);
-  StoreAsComplex(fft, dfw);
-
-  // Windowed error fft
-  WindowData(fft, aec->eBuf);
-  aec_rdft_forward_128(fft);
-  StoreAsComplex(fft, efw);
-
-  SmoothedPSD(aec, efw, dfw, xfw);
+  SmoothedPSD(aec, efw, dfw, xfw, extreme_filter_divergence);
 
   {
     const __m128 vec_1eminus10 =  _mm_set1_ps(1e-10f);
@@ -728,4 +722,7 @@ void WebRtcAec_InitAec_SSE2(void) {
   WebRtcAec_FilterAdaptation = FilterAdaptationSSE2;
   WebRtcAec_OverdriveAndSuppress = OverdriveAndSuppressSSE2;
   WebRtcAec_SubbandCoherence = SubbandCoherenceSSE2;
+  WebRtcAec_StoreAsComplex = StoreAsComplexSSE2;
+  WebRtcAec_PartitionDelay = PartitionDelaySSE2;
+  WebRtcAec_WindowData = WindowDataSSE2;
 }