Sender: chris AT mindspring DOT com
Message-ID: <38936999.1C044730@ix.netcom.com>
Date: Sat, 29 Jan 2000 14:28:41 -0800
From: Chris Sears <cbsears AT ix DOT netcom DOT com>
X-Mailer: Mozilla 4.7 [en] (X11; I; Linux 2.2.13-7mdk i686)
X-Accept-Language: en
MIME-Version: 1.0
To: pgcc AT delorie DOT com
Subject: Re: pgcc and egcs alignment -- function, basic block and string
References: <38921CD6 DOT 2A725779 AT ix DOT netcom DOT com> <20000129032101 DOT A25630 AT atrey DOT karlin DOT mff DOT cuni DOT cz> <38927310 DOT 2033EED4 AT ix DOT netcom DOT com> <20000129143544 DOT G1024 AT cerebro DOT laendle>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: pgcc AT delorie DOT com


Marc,

thanks.

> While intel recommends relatively large  alignments, "common knowledge"
> (linus for example ;) recommends no alignments at all.

> In _my_ tests large alignment is a very very slight win, but in the real
> world, the increased code size might not be worth it (cache effect, long
> nops, AGI because of lea-nops).

> It's a must on 486, though, and a bit better on ppro and later.

This is about loop body alignment, right, not function entry alignment?
I didn't get a clear idea of what you thought about function alignment.

> Yes, -malign-loops=xxx, -malign-functions=xxx etc.. Just read the
> manual(tm).

Oops.  I read the man page, really I did.  -malign-xxx isn't there,
but it is in the manual (2.8).  Thanks, I'll experiment with it.
But there isn't an alignment switch for strings.

> > Admittedly, a cache line is word aligned as well,
> > but wouldn't .align 4 suffice to align to a word boundary?
>
> It depends. The only way data alginment hurts is by wasting space, which
> is rarely a problem (despite cache problems). aligning to a cache line
> helps in practise, although one would think that natural alignment would
> suffice.

It isn't a huge issue, true.  But in the kernel, there is no paging.
String pages are pinned in memory.  A similar situation arises
in embedded applications.

> Maybe a aligning on longwords is a good trade-off, but that needs
> extensive benchmark data.

Or a switch.

But the .align 4 for egcs seems like a real gaffer to me.
Any idea who I should it to?  I've seen mention of it on a newsgroup
when I was researching the problem, so it may be known.


Chris Sears
cbsears AT ix DOT netcom DOT com