parisc: Do not hardcode registers in checksum functions

Do not hardcode processor registers r19 to r22 as scratch registers.
Instead let the compiler decide, which may give better optimization
results when the functions get inlined.
Signed-off-by: Helge Deller <deller@gmx.de>

arch/parisc/include/asm/checksum.h

@@ -42,31 +42,32 @@ extern __wsum csum_partial_copy_from_user(const void __user *src,
 static inline __sum16 ip_fast_csum(const void *iph, unsigned int ihl)
 {
 	unsigned int sum;
+	unsigned long t0, t1, t2;
 
 	__asm__ __volatile__ (
 "	ldws,ma		4(%1), %0\n"
 "	addib,<=	-4, %2, 2f\n"
 "\n"
-"	ldws		4(%1), %%r20\n"
-"	ldws		8(%1), %%r21\n"
-"	add		%0, %%r20, %0\n"
-"	ldws,ma		12(%1), %%r19\n"
-"	addc		%0, %%r21, %0\n"
-"	addc		%0, %%r19, %0\n"
-"1:	ldws,ma		4(%1), %%r19\n"
+"	ldws		4(%1), %4\n"
+"	ldws		8(%1), %5\n"
+"	add		%0, %4, %0\n"
+"	ldws,ma		12(%1), %3\n"
+"	addc		%0, %5, %0\n"
+"	addc		%0, %3, %0\n"
+"1:	ldws,ma		4(%1), %3\n"
 "	addib,<		0, %2, 1b\n"
-"	addc		%0, %%r19, %0\n"
+"	addc		%0, %3, %0\n"
 "\n"
-"	extru		%0, 31, 16, %%r20\n"
-"	extru		%0, 15, 16, %%r21\n"
-"	addc		%%r20, %%r21, %0\n"
-"	extru		%0, 15, 16, %%r21\n"
-"	add		%0, %%r21, %0\n"
+"	extru		%0, 31, 16, %4\n"
+"	extru		%0, 15, 16, %5\n"
+"	addc		%4, %5, %0\n"
+"	extru		%0, 15, 16, %5\n"
+"	add		%0, %5, %0\n"
 "	subi		-1, %0, %0\n"
 "2:\n"
-	: "=r" (sum), "=r" (iph), "=r" (ihl)
+	: "=r" (sum), "=r" (iph), "=r" (ihl), "=r" (t0), "=r" (t1), "=r" (t2)
 	: "1" (iph), "2" (ihl)
-	: "r19", "r20", "r21", "memory");
+	: "memory");
 
 	return (__force __sum16)sum;
 }
@@ -126,6 +127,10 @@ static __inline__ __sum16 csum_ipv6_magic(const struct in6_addr *saddr,
 					  __u32 len, __u8 proto,
 					  __wsum sum)
 {
+	unsigned long t0, t1, t2, t3;
+
+	len += proto;	/* add 16-bit proto + len */
+
 	__asm__ __volatile__ (
 
 #if BITS_PER_LONG > 32
@@ -136,20 +141,19 @@ static __inline__ __sum16 csum_ipv6_magic(const struct in6_addr *saddr,
 	** Try to keep 4 registers with "live" values ahead of the ALU.
 	*/
 
-"	ldd,ma		8(%1), %%r19\n"	/* get 1st saddr word */
-"	ldd,ma		8(%2), %%r20\n"	/* get 1st daddr word */
-"	add		%8, %3, %3\n"/* add 16-bit proto + len */
-"	add		%%r19, %0, %0\n"
-"	ldd,ma		8(%1), %%r21\n"	/* 2cd saddr */
-"	ldd,ma		8(%2), %%r22\n"	/* 2cd daddr */
-"	add,dc		%%r20, %0, %0\n"
-"	add,dc		%%r21, %0, %0\n"
-"	add,dc		%%r22, %0, %0\n"
+"	ldd,ma		8(%1), %4\n"	/* get 1st saddr word */
+"	ldd,ma		8(%2), %5\n"	/* get 1st daddr word */
+"	add		%4, %0, %0\n"
+"	ldd,ma		8(%1), %6\n"	/* 2nd saddr */
+"	ldd,ma		8(%2), %7\n"	/* 2nd daddr */
+"	add,dc		%5, %0, %0\n"
+"	add,dc		%6, %0, %0\n"
+"	add,dc		%7, %0, %0\n"
 "	add,dc		%3, %0, %0\n"  /* fold in proto+len | carry bit */
-"	extrd,u		%0, 31, 32, %%r19\n"	/* copy upper half down */
-"	depdi		0, 31, 32, %0\n"	/* clear upper half */
-"	add		%%r19, %0, %0\n"	/* fold into 32-bits */
-"	addc		0, %0, %0\n"		/* add carry */
+"	extrd,u		%0, 31, 32, %4\n"/* copy upper half down */
+"	depdi		0, 31, 32, %0\n"/* clear upper half */
+"	add		%4, %0, %0\n"	/* fold into 32-bits */
+"	addc		0, %0, %0\n"	/* add carry */
 
 #else
 
@@ -158,30 +162,29 @@ static __inline__ __sum16 csum_ipv6_magic(const struct in6_addr *saddr,
 	** Insn stream is serialized on the carry bit here too.
 	** result from the previous operation (eg r0 + x)
 	*/
-
-"	ldw,ma		4(%1), %%r19\n"	/* get 1st saddr word */
-"	ldw,ma		4(%2), %%r20\n"	/* get 1st daddr word */
-"	add		%8, %3, %3\n"	/* add 16-bit proto + len */
-"	add		%%r19, %0, %0\n"
-"	ldw,ma		4(%1), %%r21\n"	/* 2cd saddr */
-"	addc		%%r20, %0, %0\n"
-"	ldw,ma		4(%2), %%r22\n"	/* 2cd daddr */
-"	addc		%%r21, %0, %0\n"
-"	ldw,ma		4(%1), %%r19\n"	/* 3rd saddr */
-"	addc		%%r22, %0, %0\n"
-"	ldw,ma		4(%2), %%r20\n"	/* 3rd daddr */
-"	addc		%%r19, %0, %0\n"
-"	ldw,ma		4(%1), %%r21\n"	/* 4th saddr */
-"	addc		%%r20, %0, %0\n"
-"	ldw,ma		4(%2), %%r22\n"	/* 4th daddr */
-"	addc		%%r21, %0, %0\n"
-"	addc		%%r22, %0, %0\n"
+"	ldw,ma		4(%1), %4\n"	/* get 1st saddr word */
+"	ldw,ma		4(%2), %5\n"	/* get 1st daddr word */
+"	add		%4, %0, %0\n"
+"	ldw,ma		4(%1), %6\n"	/* 2nd saddr */
+"	addc		%5, %0, %0\n"
+"	ldw,ma		4(%2), %7\n"	/* 2nd daddr */
+"	addc		%6, %0, %0\n"
+"	ldw,ma		4(%1), %4\n"	/* 3rd saddr */
+"	addc		%7, %0, %0\n"
+"	ldw,ma		4(%2), %5\n"	/* 3rd daddr */
+"	addc		%4, %0, %0\n"
+"	ldw,ma		4(%1), %6\n"	/* 4th saddr */
+"	addc		%5, %0, %0\n"
+"	ldw,ma		4(%2), %7\n"	/* 4th daddr */
+"	addc		%6, %0, %0\n"
+"	addc		%7, %0, %0\n"
 "	addc		%3, %0, %0\n"	/* fold in proto+len, catch carry */
 
 #endif
-	: "=r" (sum), "=r" (saddr), "=r" (daddr), "=r" (len)
-	: "0" (sum), "1" (saddr), "2" (daddr), "3" (len), "r" (proto)
-	: "r19", "r20", "r21", "r22", "memory");
+	: "=r" (sum), "=r" (saddr), "=r" (daddr), "=r" (len),
+	  "=r" (t0), "=r" (t1), "=r" (t2), "=r" (t3)
+	: "0" (sum), "1" (saddr), "2" (daddr), "3" (len)
+	: "memory");
 	return csum_fold(sum);
 }