From f74471e043c6788a59bf6e8143f71d3968353152 Mon Sep 17 00:00:00 2001
From: Loren Merritt <lorenm@u.washington.edu>
Date: Tue, 25 Sep 2007 23:30:29 +0000
Subject: [PATCH] optimize encode_residual_lpc() 37%/45%/90% faster on
 core2/k8/p4, making flac encoding overall 15%/17%/40% faster at
 compression_level>=8 (less at low levels).

Originally committed as revision 10585 to svn://svn.ffmpeg.org/ffmpeg/trunk
---
 libavcodec/flacenc.c | 17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/libavcodec/flacenc.c b/libavcodec/flacenc.c
index 9dd6c7eb87..8dd55eabe0 100644
--- a/libavcodec/flacenc.c
+++ b/libavcodec/flacenc.c
@@ -838,17 +838,22 @@ static void encode_residual_lpc(int32_t *res, const int32_t *smp, int n,
                                 int order, const int32_t *coefs, int shift)
 {
     int i, j;
-    int32_t pred;
 
     for(i=0; i<order; i++) {
         res[i] = smp[i];
     }
-    for(i=order; i<n; i++) {
-        pred = 0;
-        for(j=0; j<order; j++) {
-            pred += coefs[j] * smp[i-j-1];
+    for(i=order; i<n; i+=2) {
+        int32_t c = coefs[0];
+        int32_t p0 = 0, p1 = c*smp[i];
+        for(j=1; j<order; j++) {
+            int32_t s = smp[i-j];
+            p0 += c*s;
+            c = coefs[j];
+            p1 += c*s;
         }
-        res[i] = smp[i] - (pred >> shift);
+        p0 += c*smp[i-order];
+        res[i+0] = smp[i+0] - (p0 >> shift);
+        res[i+1] = smp[i+1] - (p1 >> shift);
     }
 }