ac3dsp: Add some special-case handling for the C downmix function

This is about 200% faster for in-decoder downmixing of 5.0 and 5.1 content.
Signed-off-by: Diego Biurrun <diego@biurrun.de>

libavcodec/ac3dsp.c

@@ -174,9 +174,46 @@ static void ac3_extract_exponents_c(uint8_t *exp, int32_t *coef, int nb_coefs)
 static void ac3_downmix_c(float **samples, float **matrix,
                           int out_ch, int in_ch, int len)
 {
+    int **matrix_cmp = (int **)matrix;
     int i, j;
     float v0, v1;
-    if (out_ch == 2) {
+
+    if (in_ch == 5 && out_ch == 2 &&
+        !(matrix_cmp[1][0] | matrix_cmp[0][2]   |
+          matrix_cmp[1][3] | matrix_cmp[0][4]   |
+         (matrix_cmp[0][1] ^ matrix_cmp[1][1]) |
+         (matrix_cmp[0][0] ^ matrix_cmp[1][2]))) {
+        float front_mix    = matrix[0][0];
+        float center_mix   = matrix[0][1];
+        float surround_mix = matrix[0][3];
+
+        for (i = 0; i < len; i++) {
+            v0 = samples[0][i] * front_mix  +
+                 samples[1][i] * center_mix +
+                 samples[3][i] * surround_mix;
+
+            v1 = samples[1][i] * center_mix +
+                 samples[2][i] * front_mix  +
+                 samples[4][i] * surround_mix;
+
+            samples[0][i] = v0;
+            samples[1][i] = v1;
+        }
+    } else if (in_ch == 5 && out_ch == 1 &&
+               matrix_cmp[0][0] == matrix_cmp[0][2] &&
+               matrix_cmp[0][3] == matrix_cmp[0][4]) {
+        float front_mix    = matrix[0][0];
+        float center_mix   = matrix[0][1];
+        float surround_mix = matrix[0][3];
+
+        for (i = 0; i < len; i++) {
+            samples[0][i] = samples[0][i] * front_mix    +
+                            samples[1][i] * center_mix   +
+                            samples[2][i] * front_mix    +
+                            samples[3][i] * surround_mix +
+                            samples[4][i] * surround_mix;
+        }
+    } else if (out_ch == 2) {
         for (i = 0; i < len; i++) {
             v0 = v1 = 0.0f;
             for (j = 0; j < in_ch; j++) {