X-Git-Url: http://git.tuebingen.mpg.de/?p=paraslash.git;a=blobdiff_plain;f=wmadec_filter.c;h=c49489b70eea8d1d9ae271f318cefbb3a349e1e8;hp=15bb39253672dae4e4dd39e2d105c7ab41e026ba;hb=df6606e438a34c7bfc325e5f320843d1e97d3f7a;hpb=1899bae1378ad51dd8e43b3d426c174689b736fa

diff --git a/wmadec_filter.c b/wmadec_filter.c
index 15bb3925..c49489b7 100644
--- a/wmadec_filter.c
+++ b/wmadec_filter.c
@@ -131,6 +131,27 @@ struct private_wmadec_data {
 #define VLCBITS 9
 #define VLCMAX ((22+VLCBITS-1)/VLCBITS)
 
+DECLARE_ALIGNED(16, float, ff_sine_128[128]);
+DECLARE_ALIGNED(16, float, ff_sine_256[256]);
+DECLARE_ALIGNED(16, float, ff_sine_512[512]);
+DECLARE_ALIGNED(16, float, ff_sine_1024[1024]);
+DECLARE_ALIGNED(16, float, ff_sine_2048[2048]);
+DECLARE_ALIGNED(16, float, ff_sine_4096[4096]);
+
+static float *ff_sine_windows[6] = {
+	ff_sine_128, ff_sine_256, ff_sine_512, ff_sine_1024,
+	ff_sine_2048, ff_sine_4096
+};
+
+/* Generate a sine window. */
+static void sine_window_init(float *window, int n)
+{
+	int i;
+
+	for (i = 0; i < n; i++)
+		window[i] = sinf((i + 0.5) * (M_PI / (2.0 * n)));
+}
+
 static int wmadec_cleanup(struct private_wmadec_data *s)
 {
 	int i;
@@ -461,7 +482,7 @@ static int wma_decode_init(char *initial_buf, int len, struct private_wmadec_dat
 		return ret;
 	/* init MDCT */
 	for (i = 0; i < s->nb_block_sizes; i++) {
-		ret = imdct_init(s->frame_len_bits - i + 1, 1, &s->mdct_ctx[i]);
+		ret = imdct_init(s->frame_len_bits - i + 1, &s->mdct_ctx[i]);
 		if (ret < 0)
 			return ret;
 	}
@@ -608,18 +629,21 @@ static int decode_exp_vlc(struct private_wmadec_data *s, int ch)
 	return 0;
 }
 
-static void vector_fmul_add(float *dst, const float *src0, const float *src1,
-		const float *src2, int src3, int len, int step)
+/* compute src0 * src1 + src2 */
+static inline void vector_mult_add(float *dst, const float *src0, const float *src1,
+		const float *src2, int len)
 {
 	int i;
+
 	for (i = 0; i < len; i++)
-		dst[i * step] = src0[i] * src1[i] + src2[i] + src3;
+		dst[i] = src0[i] * src1[i] + src2[i];
 }
 
-static void vector_fmul_reverse_c(float *dst, const float *src0,
+static inline void vector_mult_reverse(float *dst, const float *src0,
 		const float *src1, int len)
 {
 	int i;
+
 	src1 += len - 1;
 	for (i = 0; i < len; i++)
 		dst[i] = src0[i] * src1[-i];
@@ -641,16 +665,15 @@ static void wma_window(struct private_wmadec_data *s, float *out)
 		block_len = s->block_len;
 		bsize = s->frame_len_bits - s->block_len_bits;
 
-		vector_fmul_add(out, in, s->windows[bsize],
-					 out, 0, block_len, 1);
+		vector_mult_add(out, in, s->windows[bsize], out, block_len);
 
 	} else {
 		block_len = 1 << s->prev_block_len_bits;
 		n = (s->block_len - block_len) / 2;
 		bsize = s->frame_len_bits - s->prev_block_len_bits;
 
-		vector_fmul_add(out + n, in + n, s->windows[bsize],
-					 out + n, 0, block_len, 1);
+		vector_mult_add(out + n, in + n, s->windows[bsize], out + n,
+			block_len);
 
 		memcpy(out + n + block_len, in + n + block_len,
 		       n * sizeof(float));
@@ -664,7 +687,7 @@ static void wma_window(struct private_wmadec_data *s, float *out)
 		block_len = s->block_len;
 		bsize = s->frame_len_bits - s->block_len_bits;
 
-		vector_fmul_reverse_c(out, in, s->windows[bsize], block_len);
+		vector_mult_reverse(out, in, s->windows[bsize], block_len);
 
 	} else {
 		block_len = 1 << s->next_block_len_bits;
@@ -673,7 +696,7 @@ static void wma_window(struct private_wmadec_data *s, float *out)
 
 		memcpy(out, in, n * sizeof(float));
 
-		vector_fmul_reverse_c(out + n, in + n, s->windows[bsize],
+		vector_mult_reverse(out + n, in + n, s->windows[bsize],
 				      block_len);
 
 		memset(out + n + block_len, 0, n * sizeof(float));