[FFmpeg-devel] [PATCH] libavcodec/fft_template: improve performance of the ff_fft_init in fft_template

[Steven Liu]

After patch:
init nbits = 17, get 10000 samples, duration: 15221
Before patch:
init nbits = 17, get 10000 samples, duration: 16105

test script:
DURATION=0
for((i=0;i<10000;i++)) do
./libavcodec/tests/fft -n 17 &>output
T_DURATION=`grep "duration" output | awk -F"=" '{ print $2}'`
DURATION=`expr $DURATION + $T_DURATION`
done
TOTAL=`expr $DURATION / 10000`
echo $TOTAL

Signed-off-by: Steven Liu <lq at chinaffmpeg.org>
---
 libavcodec/fft_template.c | 27 +++++++++++++++++----------
 1 file changed, 17 insertions(+), 10 deletions(-)

diff --git a/libavcodec/fft_template.c b/libavcodec/fft_template.c
index 762c014bc8..5f6f52275d 100644
--- a/libavcodec/fft_template.c
+++ b/libavcodec/fft_template.c
@@ -257,21 +257,28 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)
     }
 #endif /* FFT_FIXED_32 */
 
+#define SPLIT_RADIX_PERMUTATION(num) \
+    for(i=0; i<n; i++) { \
+        int k; \
+        j = i; \
+        if (s->fft_permutation == FF_FFT_PERM_SWAP_LSBS) \
+            j = (j&~3) | ((j>>1)&1) | ((j<<1)&2); \
+        k = -split_radix_permutation(i, n, s->inverse) & (n-1); \
+        j = (j&~3) | ((j>>1)&1) | ((j<<1)&2); \
+        k = -split_radix_permutation(i, n, s->inverse) & (n-1); \
+        s->revtab##num[k] = j; \
+    }
 
     if (s->fft_permutation == FF_FFT_PERM_AVX) {
         fft_perm_avx(s);
     } else {
-        for(i=0; i<n; i++) {
-            int k;
-            j = i;
-            if (s->fft_permutation == FF_FFT_PERM_SWAP_LSBS)
-                j = (j&~3) | ((j>>1)&1) | ((j<<1)&2);
-            k = -split_radix_permutation(i, n, s->inverse) & (n-1);
-            if (s->revtab)
-                s->revtab[k] = j;
-            if (s->revtab32)
-                s->revtab32[k] = j;
+        if (s->revtab) {
+            SPLIT_RADIX_PERMUTATION()
+        }
+        if (s->revtab32) {
+            SPLIT_RADIX_PERMUTATION(32)
         }
+#undef SPLIT_RADIX_PERMUTATION
     }
 
     return 0;
-- 
2.15.2 (Apple Git-101.1)