From: Miles Bader <miles@lsi.nec.co.jp>
To: Muli Ben-Yehuda <mulix@mulix.org>
Cc: Jamie Lokier <jamie@shareable.org>, Andrew Morton <akpm@osdl.org>,
Linux-Kernel <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] document optimizing macro for translating PROT_ to VM_ bits
Date: 30 Sep 2003 16:34:31 +0900 [thread overview]
Message-ID: <buohe2u3f20.fsf@mcspd15.ucom.lsi.nec.co.jp> (raw)
In-Reply-To: <20030930071005.GY729@actcom.co.il>
Muli Ben-Yehuda <mulix@mulix.org> writes:
> Ah, much nicer, thank you. I'll submit it, but first, what do you
> think about this version? it keeps the optimization and enforces the
> "bit1 and bit2 must be single bits only" rule.
On my arch it results in huge bloated code, including mul and div insns!
Here's what gcc 3.3.x does for both the prefix macro versions:
#define _calc_vm_trans_2(x,bit1,bit2) \
((bit1) <= (bit2) ? ((x) & (bit1)) * ((bit2) / (bit1)) \
: ((x) & (bit1)) / ((bit1) / (bit2)))
int test2_0 (unsigned val)
{
return _calc_vm_trans_2 (val,0x20,0x80);
}
==>
_test2_0:
andi 32,r6,r10 #, val, tmp43
shl 2,r10 #, <result>
jmp [r31]
Here's what gcc 3.3.x does for your inlined function version:
static inline void assert_single_bit(unsigned long bit)
{
extern __bug() __attribute__ ((noreturn));
if (bit & (bit - 1))
__bug ();
}
static inline unsigned long
_calc_vm_trans_3(unsigned long x, unsigned long bit1, unsigned long bit2)
{
assert_single_bit(bit1);
assert_single_bit(bit2);
return ((bit1) <= (bit2) ? ((x) & (bit1)) * ((bit2) / (bit1))
: ((x) & (bit1)) / ((bit1) / (bit2)));
}
int test3_0 (unsigned val)
{
return _calc_vm_trans_3 (val,0x20,0x80);
}
==>
_test3_0:
movea lo(32),r0,r7 #, bit1
movea lo(128),r0,r5 #, bit2
divu r7,r5,r10 # bit1, bit2, tmp55
mov r6,r10 # val, tmp52
and r7,r10 # bit1, tmp52
mul r5,r10,r0 # bit2, tmp52
jmp [r31]
[I've no idea what's up with that, I don't even think the generated code
makes sense, but it seems like the presence of inline function is
confusing something.]
-Miles
--
In New York, most people don't have cars, so if you want to kill a person, you
have to take the subway to their house. And sometimes on the way, the train
is delayed and you get impatient, so you have to kill someone on the subway.
[George Carlin]
next prev parent reply other threads:[~2003-09-30 7:35 UTC|newest]
Thread overview: 16+ messages / expand[flat|nested] mbox.gz Atom feed top
2003-09-29 9:06 Muli Ben-Yehuda
2003-09-29 15:34 ` Jamie Lokier
2003-09-29 15:51 ` Valdis.Kletnieks
2003-09-29 17:11 ` Jamie Lokier
2003-09-29 17:25 ` Maciej Zenczykowski
2003-09-30 2:15 ` Miles Bader
2003-09-30 7:10 ` Muli Ben-Yehuda
2003-09-30 7:34 ` Miles Bader [this message]
2003-09-30 7:41 ` Muli Ben-Yehuda
2003-09-30 7:59 ` Miles Bader
2003-09-30 9:24 ` Muli Ben-Yehuda
2003-09-30 10:00 ` Miles Bader
2003-09-30 10:06 ` Miles Bader
2003-10-02 11:37 ` Muli Ben-Yehuda
2003-09-30 10:35 ` Thomas Schlichter
2003-09-30 8:02 ` Frédéric St-Martin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=buohe2u3f20.fsf@mcspd15.ucom.lsi.nec.co.jp \
--to=miles@lsi.nec.co.jp \
--cc=akpm@osdl.org \
--cc=jamie@shareable.org \
--cc=linux-kernel@vger.kernel.org \
--cc=miles@gnu.org \
--cc=mulix@mulix.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®