[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1395072992.9668.59.camel@edumazet-glaptop2.roam.corp.google.com>
Date: Mon, 17 Mar 2014 09:16:32 -0700
From: Eric Dumazet <eric.dumazet@...il.com>
To: David Laight <David.Laight@...LAB.COM>
Cc: 'Thomas Graf' <tgraf@...g.ch>, David Miller <davem@...emloft.net>,
John Fastabend <john.fastabend@...il.com>,
"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: RE: [PATCH net-next] net: sched: use no more than one page in
struct fw_head
On Mon, 2014-03-17 at 16:00 +0000, David Laight wrote:
> Actually the -Os forces the divide.
Yep, this might be the case.
> With -O3 the generated code I get for amd64 is a multiply by reciprocal
> to get the quotient followed by an open coded multiply by 0x1ff and
> then a subtract. 13 instructions, only a few of which are register renames
> or are non-dependant.
I think its more 4 arithmetic instructions : 2 multiplies, one shift,
one sub.
Note that when John adds an rcu_head, the 0x1ff becomes 0x1fd
89 f8 mov %edi,%eax
ba b3 21 c1 80 mov $0x80c121b3,%edx
f7 e2 mul %edx
c1 ea 08 shr $0x8,%edx
69 d2 fd 01 00 00 imul $0x1fd,%edx,%edx
29 d7 sub %edx,%edi
89 f8 mov %edi,%eax
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists