Fwd: Re: [Shootout-list] Re: Integer overflow

[Brent Fulgham]

--0-2143457917-1096305505=:1934
Content-Type: text/plain; charset=us-ascii
Content-Id: 
Content-Disposition: inline

I meant to send this to the list, but ended up just
sending to bengt...
--0-2143457917-1096305505=:1934
Content-Type: message/rfc822

Received: from [66.80.54.58] by web81201.mail.yahoo.com via HTTP; Mon, 27 Sep 2004 10:17:36 PDT
Date: Mon, 27 Sep 2004 10:17:36 -0700 (PDT)
From: Brent Fulgham <bfulg@pacbell.net>
Subject: Re: [Shootout-list] Re: Integer overflow
To: Bengt Kleberg <bengt.kleberg@ericsson.com>
In-Reply-To: <415844D8.3020906@ericsson.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Length: 878

--- Bengt Kleberg wrote:

> sorry, but that is the message sending test. in it i
> do not want to create lots of threads dependent 
> upon n. instead i am advocating keeping 
> the number of threads constant (say 10000, or
> similar) and let n decided the number of times the 
> message goes around the ring. however, sofar i 
> have not been able to persuade mr fulgham.

Please!  Call me Brent!

No further pursuasion is necessary.  I will implement
this change (rather, I think you have already done so
for me in a separate e-mail) and update the site
tonight or tomorrow.  I understand the issue better
now, and I concede your point.

I don't think it's unfair for a language to allocate a
static array of messengers, but I may very well be
wrong.  We could also do something like have N be the
number of processes/threads involved in the ring, and
require that the messages be sent around the right
N*1000 times or something.  Would this provide any
benefit to the test?  It seems like we would be trying
to measure two separate things by doing so.

I think Bengt's idea is better for answering the
question "what does it cost to communicate between
processes/threads?"

> no, this thread creation test i am writing about is
> another test i have submitted to the shotout 
> (called procinst). here n is directly related 
> to the number of threads. it has not been
> published/made official in the shootout.

I agree we should add this test, and will do so ASAP. 
In case I lost it in my mailer mishap of a few weeks
ago, can you please forward it to me again?

Thanks,

-Brent

--0-2143457917-1096305505=:1934--