[PATCH 4/5 v2] checksum: optimize loop and get rid of add16()

[Joakim Tjernlund]

Use some better vaiable names and get rid of add16 as
add32 will do just as well. Fold the 32 bit checkum
into 16 bits in the end.
---

 v2 - bug fix

 lib/checksum.c |   47 +++++++++++++++++++++++------------------------
 1 files changed, 23 insertions(+), 24 deletions(-)

diff --git a/lib/checksum.c b/lib/checksum.c
index 2e427a2..9e30bdc 100644
--- a/lib/checksum.c
+++ b/lib/checksum.c
@@ -15,12 +15,6 @@
 #include "nest/bird.h"
 #include "checksum.h"
 
-static u16				/* One-complement addition */
-add16(u16 sum, u16 x)
-{
-  u16 z = sum + x;
-  return z + (z < sum);
-}
 #ifdef __powerpc__
 static
 u32
@@ -42,10 +36,10 @@ add32(u32 sum, u32 x)
 #endif
 
 static u16
-ipsum_calc_block(u16 *x, unsigned len, u16 sum)
+ipsum_calc_block(u16 *buf, unsigned len, u16 sum)
 {
   int rest;
-  u32 tmp, *xx;
+  u32 acc_sum, *buf_u32;
 
   /*
    *  A few simple facts about the IP checksum (see RFC 1071 for detailed
@@ -62,23 +56,28 @@ ipsum_calc_block(u16 *x, unsigned len, u16 sum)
   ASSERT(!(len % 2));
   if (!len)
     return sum;
-  len >>= 1;
-  if ((unsigned long) x & 2)		/* Align to 32-bit boundary */
-    {
-      sum = add16(sum, *x++);
-      len--;
-    }
+
+  acc_sum = sum;
+  if ((u32) buf & 2) {		/* Align to 32-bit boundary */
+      acc_sum = add32(acc_sum, *buf++);
+      len =- 2;
+  }
   rest = len & 1;
-  len >>= 1;
-  tmp = 0;
-  xx = (u32 *) x;
-  for (xx--; len; --len);
-      tmp = add32(tmp, *++xx);
-  xx++;
-  sum = add16(sum, add16(tmp & 0xffff, tmp >> 16U));
-  if (rest)
-    sum = add16(sum, *(u16 *) xx);
-  return sum;
+  len >>= 2;
+  buf_u32 =  (u32 *) buf;
+  for(buf_u32--; len; --len)
+      acc_sum = add32(acc_sum, *++buf_u32);
+
+  if (rest) {
+      buf = (u16 *)(buf_u32++);
+      acc_sum = add32(acc_sum, *buf);
+  }
+
+  /* Add back carry outs from top 16 bits to low 16 bits.
+   */
+  acc_sum = (acc_sum >> 16) + (acc_sum & 0xffff);    /* add high-16 to low-16 */
+  acc_sum += (acc_sum >> 16); /* add carry */
+  return acc_sum;
 }
 
 static u16
-- 
1.6.4.4