net: sctp: Optimise the way 'sctp_arg_t' values are initialised.
Even if memset() is inlined (as on x86) using it to zero the union
generates a memory word write of zero, followed by a write of the
smaller field, and then a read of the word.
As well as being a lot of instructions the sequence is unlikely to
be optimised by the store-load forward hardware so will be slow.
Instead allocate a field of the union that is the same size as the
entire union and write a zero value to it. The compiler will then
generate the required value in a register.
Zeroing the union shouldn't be necessary, but this patch series isn't
intended to have a behavioural change.
Signed-off-by: David Laight <david.laight@aculab.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
diff --git a/include/net/sctp/command.h b/include/net/sctp/command.h
index 0e91a42..589a191 100644
--- a/include/net/sctp/command.h
+++ b/include/net/sctp/command.h
@@ -118,6 +118,7 @@
#define SCTP_MAX_NUM_COMMANDS 14
typedef union {
+ void *zero_all; /* Set to NULL to clear the entire union */
__s32 i32;
__u32 u32;
__be32 be32;
@@ -154,7 +155,7 @@
static inline sctp_arg_t \
SCTP_## name (type arg) \
{ sctp_arg_t retval;\
- memset(&retval, 0, sizeof(sctp_arg_t));\
+ retval.zero_all = NULL;\
retval.elt = arg;\
return retval;\
}
@@ -191,7 +192,7 @@
static inline sctp_arg_t SCTP_NULL(void)
{
sctp_arg_t retval;
- memset(&retval, 0, sizeof(sctp_arg_t));
+ retval.zero_all = NULL;
return retval;
}
@@ -212,7 +213,8 @@
*/
static inline int sctp_init_cmd_seq(sctp_cmd_seq_t *seq)
{
- memset(seq, 0, sizeof(sctp_cmd_seq_t));
+ seq->next_free_slot = 0;
+ seq->next_cmd = 0;
return 1; /* We always succeed. */
}