Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 13 Sep 2006 21:42:41 -0700
From:      Gary Kline <kline@sage.thought.org>
To:        Chuck Swiger <cswiger@mac.com>
Cc:        Gary Kline <kline@sage.thought.org>, freebsd-stable <freebsd-stable@freebsd.org>
Subject:   Re: optimization levels for 6-STABLE build{kernel,world}
Message-ID:  <20060914044241.GA92358@thought.org>
In-Reply-To: <0B8BF03E-8F4A-4279-850B-2EA7FF5E1B89@mac.com>
References:  <200609130905.k8D95idk062789@lurza.secnetix.de> <4507CC9B.60704@sun-fish.com> <20060913234934.GA92067@thought.org> <0B8BF03E-8F4A-4279-850B-2EA7FF5E1B89@mac.com>

next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, Sep 13, 2006 at 05:25:35PM -0700, Chuck Swiger wrote:
> On Sep 13, 2006, at 4:49 PM, Gary Kline wrote:
> >	A couple of things.  Will having gcc unroll loops have any
> >	negative consequences?
> 
> Yes, it certainly can have negative consequences.  The primary intent  
> of using that option is to change a loop from executing the test or  
> control block for each iteration that the body is executed, to  
> executing the loop body several times and checking the test or  
> control block less often.  The issue is that there is often  
> additional setup or post-loop fixups to ensure that the loop body  
> actually is executed the right number of times, which makes the  
> generated binary code much larger.
> 
> This can mean that the loop no longer fits within the L1 instruction  
> cache, which will usually result in the program going slower, rather  
> than faster.  Using the option will always increase the size of  
> compiled executables.
> 
> >        (I can't imagine how:: but better
> >	informed than to have something crash inexplicability.)
> >	With 6.X safe at -O2 and with -funroll-loops, that should be
> >	a slight gain, right?
> 
> -funroll-loops is as likely to decrease performance for a particular  
> program as it is to help.


	Isn't the compiler intelligent enough to have a reasonable 
	limit, N, of the loops it will unroll to ensure a faster runtime?
	Something much less than 1000, say; possibly less than 100.  
	At least, if the initializiation and end-loop code *plus* the
	loop code itself were too large for the cache, my thought is that
	gcc would back out.  Imay be giving RMS too much credit; but 
	if memory serves, thed compiler was GNU's first project.  And 
	Stallman was into GOFAI, &c, for better/worse.[1]  Anyway, for now
	I'll comment out the unroll-loops arg.  

> 
> One particular caveat with using that option is that the increase in  
> program size apparently causes the initial bootloader code to no  
> longer fit within a single sector, making the system unbootable.
> 
> >	[Dumb] questions:: first, what does the compiler do with
> >	"-fno-strict-aliasing"?
> 
> It prevents the compiler from generating buggy output from source  
> code which uses type-punning.
> 
>    http://gcc.gnu.org/onlinedocs/gcc/Optimize-Options.html
> 
> A safe optimizer must assume that an arbitrary assignment via a  
> pointer dereference can change any value in memory, which means that  
> you have to spill and reload any data being cached in CPU registers  
> around the use of the pointer, except for const's, variables declared  
> as "register", and possibly function arguments being passed via  
> registers and not on the stack (cf "register windows" on the SPARC  
> hardware, or HP/PA's calling conventions).

	
	Well, I'd added the no-strict-aliasing flag to make.conf!
	Pointers give me indigestion ... even after all these years.
	Thanks for your insights.  And the URL.

	gary

[1]. Seems to me that "good old-fashioned AI" techniques would work in
     something like a compiler  where you probblyhave a good idea  of
     most heuristics.   -gk

> 
> -- 
> -Chuck
> 
> 
> 
> _______________________________________________
> freebsd-stable@freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-stable
> To unsubscribe, send any mail to "freebsd-stable-unsubscribe@freebsd.org"

-- 
   Gary Kline     kline@thought.org   www.thought.org     Public service Unix




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060914044241.GA92358>