Merge remote branch 'qatar/master'

[ffmpeg] / libavcodec / dca.c

diff --git a/libavcodec/dca.c b/libavcodec/dca.c
index 14b0b2d10af68430f3f7ad0169f42a3c19ba2d9e..1e26eedc3cc486581b251b528e7ffb2b6b195c82 100644 (file)
--- a/libavcodec/dca.c
+++ b/libavcodec/dca.c
@@ -321,16 +321,16 @@ typedef struct {
 
     /* Subband samples history (for ADPCM) */
     float subband_samples_hist[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
-    DECLARE_ALIGNED(16, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
-    DECLARE_ALIGNED(16, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
+    DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
+    DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
     int hist_index[DCA_PRIM_CHANNELS_MAX];
-    DECLARE_ALIGNED(16, float, raXin)[32];
+    DECLARE_ALIGNED(32, float, raXin)[32];
 
     int output;                 ///< type of output
     float scale_bias;           ///< output scale
 
-    DECLARE_ALIGNED(16, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
-    DECLARE_ALIGNED(16, float, samples)[(DCA_PRIM_CHANNELS_MAX+1)*256];
+    DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
+    DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX+1)*256];
     const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX+1];
 
     uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
@@ -1626,6 +1626,7 @@ static int dca_decode_frame(AVCodecContext * avctx,
     int lfe_samples;
     int num_core_channels = 0;
     int i;
+    float *samples_flt = data;
     int16_t *samples = data;
     DCAContext *s = avctx->priv_data;
     int channels;
@@ -1812,9 +1813,10 @@ static int dca_decode_frame(AVCodecContext * avctx,
         return -1;
     }
 
-    if (*data_size < (s->sample_blocks / 8) * 256 * sizeof(int16_t) * channels)
+    /* ffdshow custom code */
+    if (*data_size < (s->sample_blocks / 8) * 256 * sizeof(samples[0]) * channels)
         return -1;
-    *data_size = 256 / 8 * s->sample_blocks * sizeof(int16_t) * channels;
+    *data_size = 256 / 8 * s->sample_blocks * sizeof(samples[0]) * channels;
 
     /* filter to get final output */
     for (i = 0; i < (s->sample_blocks / 8); i++) {
@@ -1833,8 +1835,14 @@ static int dca_decode_frame(AVCodecContext * avctx,
             }
         }
 
-        s->fmt_conv.float_to_int16_interleave(samples, s->samples_chanptr, 256, channels);
-        samples += 256 * channels;
+        /* interleave samples */
+        if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT) {
+            float_interleave(samples_flt, s->samples_chanptr, 256, channels);
+            samples_flt += 256 * channels;
+        } else {
+            s->fmt_conv.float_to_int16_interleave(samples, s->samples_chanptr, 256, channels);
+            samples += 256 * channels;
+        }
     }
 
     /* update lfe history */
@@ -1870,7 +1878,8 @@ static av_cold int dca_decode_init(AVCodecContext * avctx)
 
     for (i = 0; i < DCA_PRIM_CHANNELS_MAX+1; i++)
         s->samples_chanptr[i] = s->samples + i * 256;
-    avctx->sample_fmt = AV_SAMPLE_FMT_S16;
+    avctx->sample_fmt = avctx->request_sample_fmt == AV_SAMPLE_FMT_FLT ?
+                        AV_SAMPLE_FMT_FLT : AV_SAMPLE_FMT_S16;
 
     s->scale_bias = 1.0;