Avoid reverse addressing, not sure if this is faster or slower but people

maybe are confused by it. The code needs to be optimized anyway.

Originally committed as revision 12854 to svn://svn.ffmpeg.org/ffmpeg/trunk

libavcodec/nellymoserdec.c

@@ -92,7 +92,7 @@ static const int16_t nelly_delta_table[32] = {
 typedef struct NellyMoserDecodeContext {
     AVCodecContext* avctx;
     DECLARE_ALIGNED_16(float,float_buf[NELLY_SAMPLES]);
-    float           state[64];
+    float           state[128];
     AVRandomState   random_state;
     GetBitContext   gb;
     int             add_bias;
@@ -121,12 +121,12 @@ static void overlap_and_window(NellyMoserDecodeContext *s, float *state, float *
 
     while (bot < NELLY_BUF_LEN/2) {
         audio[bot] = ( a_in[bot]*sine_window[bot]+state[bot]*sine_window[top])/s->scale_bias + s->add_bias;
-        audio[top] = ( a_in[top]*sine_window[top]+state[bot]*sine_window[bot])/s->scale_bias + s->add_bias;
-        state[bot] = a_in[bot + NELLY_BUF_LEN];
+        audio[top] = ( a_in[top]*sine_window[top]+state[top]*sine_window[bot])/s->scale_bias + s->add_bias;
 
         bot++;
         top--;
     }
+    memcpy(state, a_in + NELLY_BUF_LEN, sizeof(float)*NELLY_BUF_LEN);
 }
 
 static int sum_bits(short *buf, short shift, short off)