From bfdd5bc1f15c0c632dd8ad683fa03a63b4dec9f6 Mon Sep 17 00:00:00 2001
From: Loren Merritt <lorenm@u.washington.edu>
Date: Sat, 29 Sep 2007 23:44:34 +0000
Subject: unroll finite differences to avoid swapping registers. 1.5x faster
 encode_residual_fixed(). overall flac encoding: 3% faster at
 compression_levels 0-2.

Originally committed as revision 10625 to svn://svn.ffmpeg.org/ffmpeg/trunk
---
 libavcodec/flacenc.c | 21 ++++++++++++---------
 1 file changed, 12 insertions(+), 9 deletions(-)

(limited to 'libavcodec/flacenc.c')

diff --git a/libavcodec/flacenc.c b/libavcodec/flacenc.c
index e2627a8f9e..fb69561010 100644
--- a/libavcodec/flacenc.c
+++ b/libavcodec/flacenc.c
@@ -841,33 +841,36 @@ static void encode_residual_fixed(int32_t *res, const int32_t *smp, int n,
             res[i]= smp[i] - smp[i-1];
     }else if(order==2){
         int a = smp[order-1] - smp[order-2];
-        for(i=order; i<n; i++) {
+        for(i=order; i<n; i+=2) {
             int b = smp[i] - smp[i-1];
             res[i]= b - a;
-            a = b;
+            a = smp[i+1] - smp[i];
+            res[i+1]= a - b;
         }
     }else if(order==3){
         int a = smp[order-1] - smp[order-2];
         int c = smp[order-1] - 2*smp[order-2] + smp[order-3];
-        for(i=order; i<n; i++) {
+        for(i=order; i<n; i+=2) {
             int b = smp[i] - smp[i-1];
             int d = b - a;
             res[i]= d - c;
-            a = b;
-            c = d;
+            a = smp[i+1] - smp[i];
+            c = a - b;
+            res[i+1]= c - d;
         }
     }else{
         int a = smp[order-1] - smp[order-2];
         int c = smp[order-1] - 2*smp[order-2] + smp[order-3];
         int e = smp[order-1] - 3*smp[order-2] + 3*smp[order-3] - smp[order-4];
-        for(i=order; i<n; i++) {
+        for(i=order; i<n; i+=2) {
             int b = smp[i] - smp[i-1];
             int d = b - a;
             int f = d - c;
             res[i]= f - e;
-            a = b;
-            c = d;
-            e = f;
+            a = smp[i+1] - smp[i];
+            c = a - b;
+            e = c - d;
+            res[i+1]= e - f;
         }
     }
 }
-- 
cgit v1.2.3