From svetosch@gmx.net Thu Apr 30 03:23:13 2009
From: Sven Schreiber <svetosch@gmx.net>
To: gretl-devel@gretlml.univpm.it
Subject: [Gretl-devel] stopwatch problem
Date: Thu, 30 Apr 2009 09:22:26 +0200
Message-ID: <49F951B2.4030302@gmx.net>
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="===============2587305346445517071=="

--===============2587305346445517071==
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit

Hi,

I'm getting negative numbers from using $stopwatch differences, could
there be some sort of underflow problem?

What I do is a typical simulation exercise:

set stopwatch
t_sim1 = 0
loop for simrun=1..replications --quiet
	<create random numbers>
	loop for t=1..T	--quiet
		<do stuff with random numbers>
		start = $stopwatch
		<do more stuff>
		stop = $stopwatch
		<do more stuff>
		t_sim1 += stop-start
	end loop # t
end loop # simrun
print t_sim1

And t_sim1 is reproducibly negative.

However, in some other parts of the code, such cumulated running times
are not producing negative output, so it's more subtle. That's also the
reason why I'm not able to give you a minimal example so far, but I'll
keep trying.

Oh this is with a fairly new snapshot/cvs version, both on windows and
linux.

BTW and totally unrelated: Is it true that the user guide doesn't have a
section on plain VARs? (as opposed to the great VECM ch.)

thanks,
sven



--===============2587305346445517071==--


From cottrell@wfu.edu Thu Apr 30 20:23:57 2009
From: Allin Cottrell <cottrell@wfu.edu>
To: gretl-devel@gretlml.univpm.it
Subject: Re: [Gretl-devel] stopwatch problem
Date: Thu, 30 Apr 2009 20:23:56 -0400
Message-ID: <Pine.A41.4.58.0904302014360.393364@f1n11.sp2net.wfu.edu>
In-Reply-To: 49F951B2.4030302@gmx.net
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="===============2645902109039228804=="

--===============2645902109039228804==
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit

On Thu, 30 Apr 2009, Sven Schreiber wrote:

> I'm getting negative numbers from using $stopwatch differences,
> could there be some sort of underflow problem?
>
> What I do is a typical simulation exercise:
>
> set stopwatch
> t_sim1 = 0
> loop for simrun=1..replications --quiet
> 	<create random numbers>
> 	loop for t=1..T	--quiet
> 		<do stuff with random numbers>
> 		start = $stopwatch
> 		<do more stuff>
> 		stop = $stopwatch
> 		<do more stuff>
> 		t_sim1 += stop-start
> 	end loop # t
> end loop # simrun
> print t_sim1
>
> And t_sim1 is reproducibly negative.

Sven, please give a fully explicit test case.  I'm not sure I
understand the ellipses in your account.  I tried the following
complete example, broadly similar to what you stated:

<script>
set stopwatch
scalar t_sim1 = 0
scalar B = 100
scalar T = 20
loop for simrun=1..B --quiet
   matrix M1 = mnormal(100,100)
   matrix M2 = mnormal(100,100)
   start = $stopwatch
   loop for t=1..T --quiet
      matrix Mprod = M1 * M2
   end loop
   stop = $stopwatch
   t_sim1 += stop - start
end loop
print t_sim1
</script>

and it produced no negative results.

Allin.

--===============2645902109039228804==--


From cottrell@wfu.edu Sun May  3 16:10:30 2009
From: Allin Cottrell <cottrell@wfu.edu>
To: gretl-devel@gretlml.univpm.it
Subject: Re: [Gretl-devel] stopwatch problem
Date: Sun, 03 May 2009 16:10:30 -0400
Message-ID: <Pine.A41.4.58.0905031600170.745524@f1n11.sp2net.wfu.edu>
In-Reply-To: Pine.A41.4.58.0904302014360.393364@f1n11.sp2net.wfu.edu
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="===============5543848921013712350=="

--===============5543848921013712350==
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit

On Thu, 30 Apr 2009, Allin Cottrell wrote:

> On Thu, 30 Apr 2009, Sven Schreiber wrote:
>
> > I'm getting negative numbers from using $stopwatch differences,
> > could there be some sort of underflow problem?
> >
> > What I do is a typical simulation exercise:
> >
> > set stopwatch
> > t_sim1 = 0
> > loop for simrun=1..replications --quiet
> > 	<create random numbers>
> > 	loop for t=1..T	--quiet
> > 		<do stuff with random numbers>
> > 		start = $stopwatch
> > 		<do more stuff>
> > 		stop = $stopwatch
> > 		<do more stuff>
> > 		t_sim1 += stop-start
> > 	end loop # t
> > end loop # simrun
> > print t_sim1
> >
> > And t_sim1 is reproducibly negative.

Ah, I tried this again using an exact copy of your script with
<stuff> filled in, and I did indeed get negative values of t_sim1.
I was bothered at first, but then realized that's perfectly OK.
I'd failed to notice the problem in your script.

As per the manual, $stopwatch gives you the time _elapsed_ since
"set stopwatch", or since the previous call to $stopwatch.  Your
calls

start = $stopwatch
...
stop = $stopwatch
...
t_sim1 += stop-start

seem to be assuming that $stopwatch gives you the current time,
not the elapsed time: you want t_sim1 += $stopwatch.

Allin.

--===============5543848921013712350==--


From svetosch@gmx.net Mon May  4 05:01:32 2009
From: Sven Schreiber <svetosch@gmx.net>
To: gretl-devel@gretlml.univpm.it
Subject: Re: [Gretl-devel] stopwatch problem
Date: Mon, 04 May 2009 11:01:24 +0200
Message-ID: <49FEAEE4.2030707@gmx.net>
In-Reply-To: Pine.A41.4.58.0905031600170.745524@f1n11.sp2net.wfu.edu
MIME-Version: 1.0
Content-Type: multipart/mixed; boundary="===============1363850186467331386=="

--===============1363850186467331386==
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit

Am 03.05.2009 22:10, Allin Cottrell schrieb:
> On Thu, 30 Apr 2009, Allin Cottrell wrote:
> 
>> On Thu, 30 Apr 2009, Sven Schreiber wrote:
>>
>>> I'm getting negative numbers from using $stopwatch differences,
>>> could there be some sort of underflow problem?
>>>
>>> What I do is a typical simulation exercise:
>>>
>>> set stopwatch
>>> t_sim1 = 0
>>> loop for simrun=1..replications --quiet
>>> 	<create random numbers>
>>> 	loop for t=1..T	--quiet
>>> 		<do stuff with random numbers>
>>> 		start = $stopwatch
>>> 		<do more stuff>
>>> 		stop = $stopwatch
>>> 		<do more stuff>
>>> 		t_sim1 += stop-start
>>> 	end loop # t
>>> end loop # simrun
>>> print t_sim1
>>>
>>> And t_sim1 is reproducibly negative.
> 
> Ah, I tried this again using an exact copy of your script with
> <stuff> filled in, and I did indeed get negative values of t_sim1.
> I was bothered at first, but then realized that's perfectly OK.
> I'd failed to notice the problem in your script.
> 
> As per the manual, $stopwatch gives you the time _elapsed_ since
> "set stopwatch", or since the previous call to $stopwatch.  Your
> calls
> 
> start = $stopwatch
> ...
> stop = $stopwatch
> ...
> t_sim1 += stop-start
> 
> seem to be assuming that $stopwatch gives you the current time,
> not the elapsed time: you want t_sim1 += $stopwatch.
> 

Ah okay, my bad of course.

sorry and thanks,
sven

--===============1363850186467331386==--