Commit 21371f76 authored by Eric Dumazet's avatar Eric Dumazet Committed by David S. Miller

[SOCK] Avoid divides in sk_stream_pages() and __sk_stream_mem_reclaim()

sk_forward_alloc being signed, we should take care of divides by
SK_STREAM_MEM_QUANTUM we do in sk_stream_pages() and
__sk_stream_mem_reclaim()

This patchs introduces SK_STREAM_MEM_QUANTUM_SHIFT, defined
as ilog2(SK_STREAM_MEM_QUANTUM), to be able to use right
shifts instead of plain divides.

This should help compiler to choose right shifts instead of
expensive divides (as seen with CONFIG_CC_OPTIMIZE_FOR_SIZE=y on x86)
Signed-off-by: default avatarEric Dumazet <dada1@cosmosbay.com>
Signed-off-by: default avatarDavid S. Miller <davem@davemloft.net>
parent b15c4bcd
......@@ -716,10 +716,11 @@ extern void __sk_stream_mem_reclaim(struct sock *sk);
extern int sk_stream_mem_schedule(struct sock *sk, int size, int kind);
#define SK_STREAM_MEM_QUANTUM ((int)PAGE_SIZE)
#define SK_STREAM_MEM_QUANTUM_SHIFT ilog2(SK_STREAM_MEM_QUANTUM)
static inline int sk_stream_pages(int amt)
{
return DIV_ROUND_UP(amt, SK_STREAM_MEM_QUANTUM);
return (amt + SK_STREAM_MEM_QUANTUM - 1) >> SK_STREAM_MEM_QUANTUM_SHIFT;
}
static inline void sk_stream_mem_reclaim(struct sock *sk)
......
......@@ -196,7 +196,7 @@ EXPORT_SYMBOL(sk_stream_error);
void __sk_stream_mem_reclaim(struct sock *sk)
{
atomic_sub(sk->sk_forward_alloc / SK_STREAM_MEM_QUANTUM,
atomic_sub(sk->sk_forward_alloc >> SK_STREAM_MEM_QUANTUM_SHIFT,
sk->sk_prot->memory_allocated);
sk->sk_forward_alloc &= SK_STREAM_MEM_QUANTUM - 1;
if (*sk->sk_prot->memory_pressure &&
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment