Replace call of celt_inner_prod_c() (step 1)

Should call celt_inner_prod().

This change is bit exact as original, except for x86 floating-point.
In x86 floating-point, it calls celt_inner_prod_sse() which may have
different output with the change of floating-point operations' orders.

Change-Id: Ia2381e2e198a84296ac28305183f15be842b3454

Signed-off-by: Jean-Marc Valin <jmvalin@jmvalin.ca>

5 files changed, 10 insertions, 9 deletions

diff --git a/celt/bands.c b/celt/bands.c
index d4348182..3b1f5cfb 100644
--- a/celt/bands.c
+++ b/celt/bands.c
@@ -92,10 +92,11 @@ int bitexact_log2tan(int isin,int icos)
 
 #ifdef FIXED_POINT
 /* Compute the amplitude (sqrt energy) in each of the bands */
-void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM)
+void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM, int arch)
 {
    int i, c, N;
    const opus_int16 *eBands = m->eBands;
+   (void)arch;
    N = m->shortMdctSize<<LM;
    c=0; do {
       for (i=0;i<end;i++)
@@ -155,7 +156,7 @@ void normalise_bands(const CELTMode *m, const celt_sig * OPUS_RESTRICT freq, cel
 
 #else /* FIXED_POINT */
 /* Compute the amplitude (sqrt energy) in each of the bands */
-void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM)
+void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM, int arch)
 {
    int i, c, N;
    const opus_int16 *eBands = m->eBands;
@@ -164,7 +165,7 @@ void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *band
       for (i=0;i<end;i++)
       {
          opus_val32 sum;
-         sum = 1e-27f + celt_inner_prod_c(&X[c*N+(eBands[i]<<LM)], &X[c*N+(eBands[i]<<LM)], (eBands[i+1]-eBands[i])<<LM);
+         sum = 1e-27f + celt_inner_prod(&X[c*N+(eBands[i]<<LM)], &X[c*N+(eBands[i]<<LM)], (eBands[i+1]-eBands[i])<<LM, arch);
          bandE[i+c*m->nbEBands] = celt_sqrt(sum);
          /*printf ("%f ", bandE[i+c*m->nbEBands]);*/
       }
diff --git a/celt/bands.h b/celt/bands.h
index 8cb9b6ea..2488c185 100644
--- a/celt/bands.h
+++ b/celt/bands.h
@@ -44,7 +44,7 @@ int bitexact_log2tan(int isin,int icos);
  * @param X Spectrum
  * @param bandE Square root of the energy for each band (returned)
  */
-void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM);
+void compute_band_energies(const CELTMode *m, const celt_sig *X, celt_ener *bandE, int end, int C, int LM, int arch);
 
 /*void compute_noise_energies(const CELTMode *m, const celt_sig *X, const opus_val16 *tonality, celt_ener *bandE);*/
 
diff --git a/celt/celt_encoder.c b/celt/celt_encoder.c
index 6c5e9e1c..053e5a3b 100644
--- a/celt/celt_encoder.c
+++ b/celt/celt_encoder.c
@@ -1655,7 +1655,7 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
    if (secondMdct)
    {
       compute_mdcts(mode, 0, in, freq, C, CC, LM, st->upsample, st->arch);
-      compute_band_energies(mode, freq, bandE, effEnd, C, LM);
+      compute_band_energies(mode, freq, bandE, effEnd, C, LM, st->arch);
       amp2Log2(mode, effEnd, end, bandE, bandLogE2, C);
       for (i=0;i<C*nbEBands;i++)
          bandLogE2[i] += HALF16(SHL16(LM, DB_SHIFT));
@@ -1664,7 +1664,7 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
    compute_mdcts(mode, shortBlocks, in, freq, C, CC, LM, st->upsample, st->arch);
    if (CC==2&&C==1)
       tf_chan = 0;
-   compute_band_energies(mode, freq, bandE, effEnd, C, LM);
+   compute_band_energies(mode, freq, bandE, effEnd, C, LM, st->arch);
 
    if (st->lfe)
    {
@@ -1788,7 +1788,7 @@ int celt_encode_with_ec(CELTEncoder * OPUS_RESTRICT st, const opus_val16 * pcm,
          isTransient = 1;
          shortBlocks = M;
          compute_mdcts(mode, shortBlocks, in, freq, C, CC, LM, st->upsample, st->arch);
-         compute_band_energies(mode, freq, bandE, effEnd, C, LM);
+         compute_band_energies(mode, freq, bandE, effEnd, C, LM, st->arch);
          amp2Log2(mode, effEnd, end, bandE, bandLogE, C);
          /* Compensate for the scaling of short vs long mdcts */
          for (i=0;i<C*nbEBands;i++)
diff --git a/celt/pitch.c b/celt/pitch.c
index 42f526bf..1ca68bf6 100644
--- a/celt/pitch.c
+++ b/celt/pitch.c
@@ -378,7 +378,7 @@ void pitch_search(const opus_val16 * OPUS_RESTRICT x_lp, opus_val16 * OPUS_RESTR
       for (j=0;j<len>>1;j++)
          sum += SHR32(MULT16_16(x_lp[j],y[i+j]), shift);
 #else
-      sum = celt_inner_prod_c(x_lp, y+i, len>>1);
+      sum = celt_inner_prod(x_lp, y+i, len>>1, arch);
 #endif
       xcorr[i] = MAX32(-1, sum);
 #ifdef FIXED_POINT
diff --git a/src/opus_multistream_encoder.c b/src/opus_multistream_encoder.c
index e56ad7fb..032fc003 100644
--- a/src/opus_multistream_encoder.c
+++ b/src/opus_multistream_encoder.c
@@ -326,7 +326,7 @@ void surround_analysis(const CELTMode *celt_mode, const void *pcm, opus_val16 *b
                freq[i] = 0;
          }
 
-         compute_band_energies(celt_mode, freq, tmpE, 21, 1, LM);
+         compute_band_energies(celt_mode, freq, tmpE, 21, 1, LM, arch);
          /* If we have multiple frames, take the max energy. */
          for (i=0;i<21;i++)
             bandE[i] = MAX32(bandE[i], tmpE[i]);