Blackfin arch: Faster Implementation of csum_tcpudp_nofold()

Avoid conditional branch instructions during carry bit additions.
Special thanks to Bernd.
Simplify: Use ((len + proto) << 8) like every other __LITTLE_ENDIAN__ machine

Cc: Bernd Schmidt <bernds_cb1@t-online.de>
Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Bryan Wu <cooloney@kernel.org>

arch/blackfin/include/asm/checksum.h

@@ -63,23 +63,23 @@ static inline __wsum
 csum_tcpudp_nofold(__be32 saddr, __be32 daddr, unsigned short len,
 		   unsigned short proto, __wsum sum)
 {
-
-	__asm__ ("%0 = %0 + %1;\n\t"
-		 "CC = AC0;\n\t"
-		 "if !CC jump 4;\n\t"
-		 "%0 = %0 + %4;\n\t"
-		 "%0 = %0 + %2;\n\t"
-		 "CC = AC0;\n\t"
-                 "if !CC jump 4;\n\t"
-                 "%0 = %0 + %4;\n\t"
- 		 "%0 = %0 + %3;\n\t"
-		 "CC = AC0;\n\t"
-                 "if !CC jump 4;\n\t"
-                 "%0 = %0 + %4;\n\t"
-                 "NOP;\n\t"
- 		 : "=d" (sum)
-		 : "d" (daddr), "d" (saddr), "d" ((ntohs(len)<<16)+proto*256), "d" (1), "0"(sum)
-		 : "CC");
+	unsigned int carry;
+
+	__asm__ ("%0 = %0 + %2;\n\t"
+		"CC = AC0;\n\t"
+		"%1 = CC;\n\t"
+		"%0 = %0 + %1;\n\t"
+		"%0 = %0 + %3;\n\t"
+		"CC = AC0;\n\t"
+		"%1 = CC;\n\t"
+		"%0 = %0 + %1;\n\t"
+		"%0 = %0 + %4;\n\t"
+		"CC = AC0;\n\t"
+		"%1 = CC;\n\t"
+		"%0 = %0 + %1;\n\t"
+		: "=d" (sum), "=&d" (carry)
+		: "d" (daddr), "d" (saddr), "d" ((len + proto) << 8), "0"(sum)
+		: "CC");
 
 	return (sum);
 }