Fix compiler warnings

- celt/modes.c:430:14: warning: cast from 'const unsigned char *' to
  'opus_int16 *' increases required alignment from 1 to 2 [-Wcast-align]
- 'C[0][1]' may be used uninitialized [-Wmaybe-uninitialized]
- Unused variable/parameter
- Value stored is never read
- MSVC warnings about "possible loss of data" due to type conversions
- MSVC warning C4146: unary minus operator applied to unsigned type
- silk/NLSF_del_dec_quant.c:137:20: warning: array subscript is above
  array bounds [-Warray-bounds] (gcc -O3 false positive)
- src/mlp_train.h:39:20: warning: function declaration isn't a prototype
  [-Wstrict-prototypes]
- Remove SMALL_FOOTPRINT code from SSE 4.1 FIR implementation, matching
  the C implementation.

The clang -Wcast-align warnings with SSE intrinsics are a known
clang issue: https://llvm.org/bugs/show_bug.cgi?id=20670

7 files changed, 56 insertions, 51 deletions

diff --git a/src/analysis.c b/src/analysis.c
index 3e8a23bd..63d3b732 100644
--- a/src/analysis.c
+++ b/src/analysis.c
@@ -48,6 +48,8 @@
 #define M_PI 3.141592653
 #endif
 
+#ifndef DISABLE_FLOAT_API
+
 static const float dct_table[128] = {
         0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f,
         0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f, 0.250000f,
@@ -151,7 +153,7 @@ static opus_val32 silk_resampler_down2_hp(
     /* len2 can be up to 480, so we shift by 8 more to make it fit. */
     hp_ener = hp_ener >> (2*SIG_SHIFT + 8);
 #endif
-    return hp_ener;
+    return (opus_val32)hp_ener;
 }
 
 static opus_val32 downmix_and_resample(downmix_func downmix, const void *_x, opus_val32 *y, opus_val32 S[3], int subframe, int offset, int c1, int c2, int C, int Fs)
@@ -256,7 +258,7 @@ void tonality_get_info(TonalityAnalysisState *tonal, AnalysisInfo *info_out, int
          pos = 0;
       if (pos == tonal->write_pos)
          break;
-      info_out->tonality = MAX32(0, -.03 + MAX32(info_out->tonality, tonal->info[pos].tonality-.05));
+      info_out->tonality = MAX32(0, -.03f + MAX32(info_out->tonality, tonal->info[pos].tonality-.05f));
    }
    tonal->read_subframe += len/(tonal->Fs/400);
    while (tonal->read_subframe>=8)
@@ -284,8 +286,8 @@ void tonality_get_info(TonalityAnalysisState *tonal, AnalysisInfo *info_out, int
 }
 
 static const float std_feature_bias[9] = {
-      5.684947, 3.475288, 1.770634, 1.599784, 3.773215,
-      2.163313, 1.260756, 1.116868, 1.918795
+      5.684947f, 3.475288f, 1.770634f, 1.599784f, 3.773215f,
+      2.163313f, 1.260756f, 1.116868f, 1.918795f
 };
 
 static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt_mode, const void *x, int len, int offset, int c1, int c2, int C, int lsb_depth, downmix_func downmix)
@@ -346,7 +348,8 @@ static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt
     kfft = celt_mode->mdct.kfft[0];
     if (tonal->count==0)
        tonal->mem_fill = 240;
-    tonal->hp_ener_accum += downmix_and_resample(downmix, x, &tonal->inmem[tonal->mem_fill], tonal->downmix_state,
+    tonal->hp_ener_accum += (float)downmix_and_resample(downmix, x,
+          &tonal->inmem[tonal->mem_fill], tonal->downmix_state,
           IMIN(len, ANALYSIS_BUF_SIZE-tonal->mem_fill), offset, c1, c2, C, tonal->Fs);
     if (tonal->mem_fill+len < ANALYSIS_BUF_SIZE)
     {
@@ -374,8 +377,9 @@ static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt
     }
     OPUS_MOVE(tonal->inmem, tonal->inmem+ANALYSIS_BUF_SIZE-240, 240);
     remaining = len - (ANALYSIS_BUF_SIZE-tonal->mem_fill);
-    tonal->hp_ener_accum = downmix_and_resample(downmix, x, &tonal->inmem[240], tonal->downmix_state,
-          remaining, offset+ANALYSIS_BUF_SIZE-tonal->mem_fill, c1, c2, C, tonal->Fs);
+    tonal->hp_ener_accum = (float)downmix_and_resample(downmix, x,
+          &tonal->inmem[240], tonal->downmix_state, remaining,
+          offset+ANALYSIS_BUF_SIZE-tonal->mem_fill, c1, c2, C, tonal->Fs);
     tonal->mem_fill = 240 + remaining;
     opus_fft(kfft, in, out, tonal->arch);
 #ifndef FIXED_POINT
@@ -430,7 +434,7 @@ static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt
     for (i=2;i<N2-1;i++)
     {
        float tt = MIN32(tonality2[i], MAX32(tonality2[i-1], tonality2[i+1]));
-       tonality[i] = .9*MAX32(tonality[i], tt-.1);
+       tonality[i] = .9f*MAX32(tonality[i], tt-.1f);
     }
     frame_tonality = 0;
     max_frame_tonality = 0;
@@ -486,9 +490,9 @@ static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt
        if (tonal->highE[b] > tonal->lowE[b] + 7.5)
        {
           if (tonal->highE[b] - logE[b] > logE[b] - tonal->lowE[b])
-             tonal->highE[b] -= .01;
+             tonal->highE[b] -= .01f;
           else
-             tonal->lowE[b] += .01;
+             tonal->lowE[b] += .01f;
        }
        if (logE[b] > tonal->highE[b])
        {
@@ -534,7 +538,7 @@ static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt
     for (i=0;i<NB_FRAMES;i++)
     {
        int j;
-       float mindist = 1e15;
+       float mindist = 1e15f;
        for (j=0;j<NB_FRAMES;j++)
        {
           int k;
@@ -550,7 +554,7 @@ static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt
        }
        spec_variability += mindist;
     }
-    spec_variability = sqrt(spec_variability/NB_FRAMES/NB_TBANDS);
+    spec_variability = (float)sqrt(spec_variability/NB_FRAMES/NB_TBANDS);
     bandwidth_mask = 0;
     bandwidth = 0;
     maxE = 0;
@@ -590,7 +594,7 @@ static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt
     /* Special case for the last two bands, for which we don't have spectrum but only
        the energy above 12 kHz. */
     {
-       float E = hp_ener*(1./(240*240));
+       float E = hp_ener*(1.f/(240*240));
 #ifdef FIXED_POINT
        /* silk_resampler_down2_hp() shifted right by an extra 8 bits. */
        E *= ((opus_int32)1 << 2*SIG_SHIFT)*256.f;
@@ -622,7 +626,7 @@ static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt
     {
        float sum=0;
        for (b=0;b<16;b++)
-          sum += dct_table[i*16+b]*.5*(tonal->highE[b]+tonal->lowE[b]);
+          sum += dct_table[i*16+b]*.5f*(tonal->highE[b]+tonal->lowE[b]);
        midE[i] = sum;
     }
 
@@ -675,14 +679,13 @@ static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt
     }
     for (i=0;i<9;i++)
        features[11+i] = (float)sqrt(tonal->std[i]) - std_feature_bias[i];
-    features[18] = spec_variability-.78;;
-    features[20] = info->tonality - 0.154723;
-    features[21] = info->activity - 0.724643;
-    features[22] = frame_stationarity - 0.743717;
-    features[23] = info->tonality_slope + 0.069216;
-    features[24] = tonal->lowECount - 0.067930;
+    features[18] = spec_variability - 0.78f;
+    features[20] = info->tonality - 0.154723f;
+    features[21] = info->activity - 0.724643f;
+    features[22] = frame_stationarity - 0.743717f;
+    features[23] = info->tonality_slope + 0.069216f;
+    features[24] = tonal->lowECount - 0.067930f;
 
-#ifndef DISABLE_FLOAT_API
     mlp_process(&net, features, frame_probs);
     frame_probs[0] = .5f*(frame_probs[0]+1);
     /* Curve fitting between the MLP probability and the actual probability */
@@ -818,9 +821,6 @@ static void tonality_analysis(TonalityAnalysisState *tonal, const CELTMode *celt
        }
     }
     tonal->last_music = tonal->music_prob>.5f;
-#else
-    info->music_prob = 0;
-#endif
 #ifdef MLP_TRAINING
     for (i=0;i<25;i++)
        printf("%f ", features[i]);
@@ -862,3 +862,5 @@ void run_analysis(TonalityAnalysisState *analysis, const CELTMode *celt_mode, co
    analysis_info->valid = 0;
    tonality_get_info(analysis, analysis_info, frame_size);
 }
+
+#endif /* DISABLE_FLOAT_API */
diff --git a/src/mlp_data.c b/src/mlp_data.c
index 6e15dc5a..a819880b 100644
--- a/src/mlp_data.c
+++ b/src/mlp_data.c
@@ -95,18 +95,18 @@ static const float weights[450] = {
 -0.222056f, -0.508859f, -0.473369f, 0.484958f, -2.28411f,
 0.0139516f,
 /* output layer */
-3.90017, 1.71789, -1.43372, -2.70839, 1.77107,
-5.48006, 1.44661, 2.01134, -1.88383, -3.64958,
--1.26351, 0.779421, 2.11357, 3.10409, 1.68846,
--4.46197, -1.61455, 3.59832, 2.43531, -1.26458,
-0.417941, 1.47437, 2.16635, -1.909, -0.828869,
-1.38805, -2.67975, -0.110044, 1.95596, 0.697931,
--0.313226, -0.889315, 0.283236, 0.946102, };
+3.90017f, 1.71789f, -1.43372f, -2.70839f, 1.77107f,
+5.48006f, 1.44661f, 2.01134f, -1.88383f, -3.64958f,
+-1.26351f, 0.779421f, 2.11357f, 3.10409f, 1.68846f,
+-4.46197f, -1.61455f, 3.59832f, 2.43531f, -1.26458f,
+0.417941f, 1.47437f, 2.16635f, -1.909f, -0.828869f,
+1.38805f, -2.67975f, -0.110044f, 1.95596f, 0.697931f,
+-0.313226f, -0.889315f, 0.283236f, 0.946102f, };
 
 static const int topo[3] = {25, 16, 2};
 
 const MLP net = {
-	3,
-	topo,
-	weights
+    3,
+    topo,
+    weights
 };
diff --git a/src/mlp_train.c b/src/mlp_train.c
index a41f9271..8d9d127a 100644
--- a/src/mlp_train.c
+++ b/src/mlp_train.c
@@ -485,7 +485,7 @@ int main(int argc, char **argv)
     printf ("\n/* output layer */\n");
     for (i=0;i<(topo[1]+1)*topo[2];i++)
     {
-        printf ("%g,", net->weights[1][i]);
+        printf ("%gf,", net->weights[1][i]);
         if (i%5==4)
             printf("\n");
         else
@@ -494,8 +494,8 @@ int main(int argc, char **argv)
     printf ("};\n\n");
     printf ("static const int topo[3] = {%d, %d, %d};\n\n", topo[0], topo[1], topo[2]);
     printf ("const MLP net = {\n");
-    printf ("\t3,\n");
-    printf ("\ttopo,\n");
-    printf ("\tweights\n};\n");
+    printf ("    3,\n");
+    printf ("    topo,\n");
+    printf ("    weights\n};\n");
     return 0;
 }
diff --git a/src/mlp_train.h b/src/mlp_train.h
index 2786b40d..49404158 100644
--- a/src/mlp_train.h
+++ b/src/mlp_train.h
@@ -36,7 +36,7 @@ static inline double tansig_double(double x)
 {
     return 2./(1.+exp(-2.*x)) - 1.;
 }
-static inline void build_tansig_table()
+static inline void build_tansig_table(void)
 {
     int i;
     for (i=0;i<501;i++)
@@ -59,7 +59,7 @@ static inline double tansig_approx(double x)
     return y;
 }
 
-inline float randn(float sd)
+static inline float randn(float sd)
 {
    float U1, U2, S, x;
    do {
diff --git a/src/opus.c b/src/opus.c
index f76f125c..cdbd13a1 100644
--- a/src/opus.c
+++ b/src/opus.c
@@ -107,7 +107,7 @@ OPUS_EXPORT void opus_pcm_soft_clip(float *_x, int N, int C, float *declip_mem)
          /* Slightly boost "a" by 2^-22. This is just enough to ensure -ffast-math
             does not cause output values larger than +/-1, but small enough not
             to matter even for 24-bit output.  */
-         a += a*2.4e-7;
+         a += a*2.4e-7f;
          if (x[i*C]>0)
             a = -a;
          /* Apply soft clipping */
diff --git a/src/opus_demo.c b/src/opus_demo.c
index 9bd02947..50987c96 100644
--- a/src/opus_demo.c
+++ b/src/opus_demo.c
@@ -253,6 +253,7 @@ int main(int argc, char *argv[])
     opus_uint32 dec_final_range;
     int encode_only=0, decode_only=0;
     int max_frame_size = 48000*2;
+    size_t num_read;
     int curr_read=0;
     int sweep_bps = 0;
     int random_framesize=0, newsize=0, delayed_celt=0;
@@ -657,8 +658,8 @@ int main(int argc, char *argv[])
         if (decode_only)
         {
             unsigned char ch[4];
-            err = fread(ch, 1, 4, fin);
-            if (feof(fin))
+            num_read = fread(ch, 1, 4, fin);
+            if (num_read!=4)
                 break;
             len[toggle] = char_to_int(ch);
             if (len[toggle]>max_payload_bytes || len[toggle]<0)
@@ -666,14 +667,16 @@ int main(int argc, char *argv[])
                 fprintf(stderr, "Invalid payload length: %d\n",len[toggle]);
                 break;
             }
-            err = fread(ch, 1, 4, fin);
+            num_read = fread(ch, 1, 4, fin);
+            if (num_read!=4)
+                break;
             enc_final_range[toggle] = char_to_int(ch);
-            err = fread(data[toggle], 1, len[toggle], fin);
-            if (err<len[toggle])
+            num_read = fread(data[toggle], 1, len[toggle], fin);
+            if (num_read!=(size_t)len[toggle])
             {
                 fprintf(stderr, "Ran out of input, "
                                 "expecting %d bytes got %d\n",
-                                len[toggle],err);
+                                len[toggle],(int)num_read);
                 break;
             }
         } else {
@@ -685,8 +688,8 @@ int main(int argc, char *argv[])
                 opus_encoder_ctl(enc, OPUS_SET_FORCE_CHANNELS(mode_list[curr_mode][3]));
                 frame_size = mode_list[curr_mode][2];
             }
-            err = fread(fbytes, sizeof(short)*channels, frame_size-remaining, fin);
-            curr_read = err;
+            num_read = fread(fbytes, sizeof(short)*channels, frame_size-remaining, fin);
+            curr_read = (int)num_read;
             tot_in += curr_read;
             for(i=0;i<curr_read*channels;i++)
             {
@@ -795,7 +798,7 @@ int main(int argc, char *argv[])
                     if (!decode_only && tot_out + output_samples > tot_in)
                     {
                        stop=1;
-                       output_samples  = tot_in-tot_out;
+                       output_samples = (opus_int32)(tot_in - tot_out);
                     }
                     if (output_samples>skip) {
                        int i;
diff --git a/src/opus_encoder.c b/src/opus_encoder.c
index ddeaf05d..10bb62d5 100644
--- a/src/opus_encoder.c
+++ b/src/opus_encoder.c
@@ -1130,7 +1130,7 @@ opus_int32 opus_encode_native(OpusEncoder *st, const opus_val16 *pcm, int frame_
 
        /* Track the peak signal energy */
        if (!is_silence && analysis_info.activity_probability > DTX_ACTIVITY_THRESHOLD)
-          st->peak_signal_energy = MAX32(MULT16_32_Q15(QCONST16(0.999, 15), st->peak_signal_energy),
+          st->peak_signal_energy = MAX32(MULT16_32_Q15(QCONST16(0.999f, 15), st->peak_signal_energy),
                 compute_frame_energy(pcm, frame_size, st->channels, st->arch));
     }
 #else