From tradke at aei.mpg.de  Mon Sep 15 10:29:27 2008
From: tradke at aei.mpg.de (Thomas Radke)
Date: Mon, 15 Sep 2008 17:29:27 +0200
Subject: [Developers] Correct error in parameter checking
In-Reply-To: <9535CBF4-C589-4A30-AD8A-3834DCF5DB41@cct.lsu.edu>
References: <9535CBF4-C589-4A30-AD8A-3834DCF5DB41@cct.lsu.edu>
Message-ID: <48CE7F57.4060103@aei.mpg.de>

Erik Schnetter wrote:
> Cactus has currently an error in its parameter checking.  It  
> interprets ranges consisting of single numbers as allowing this number  
> and all larger numbers.  For example, the specification
> 
> INT spatial_order "Order of spatial differencing"
> {
>    2 :: "second order"
>    4 :: "fourth order"
> } 2
> 
> allows all numbers not less than 2.
> 
> The enclosed patch correct this.  Since this involves regular  
> expressions which are often difficult to read, I have broken up the  
> regular expressions into macros containing sub-expressions.
> 
> I have also replaced calls to atoi and atof with calls to strtol and  
> strtod, since these allow error checking which also was not present  
> before.

Hi Erik,

I only noticed today that the (already applied) patch is incomplete. It 
doesn't cover the case where the specification of a parameter value is 
given just as '*' and no explicit ranges, as in:

INT bla "some description"
{
   * :: "any integer"
} 0

Parameter values are then checked against an interval [INT_MIN, INT_MIN] 
whereas it should be [INT_MIN, INT_MAX]; similarly for real-type parameters.

Attached is a patch correcting the assumed upper range. Ok to apply ?

-- 
Cheers, Thomas.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: parameter-checking.patch
Type: text/x-patch
Size: 912 bytes
Desc: not available
Url : http://www.cactuscode.org/pipermail/developers/attachments/20080915/9d1be37a/attachment.bin 

From schnetter at cct.lsu.edu  Mon Sep 15 11:09:23 2008
From: schnetter at cct.lsu.edu (Erik Schnetter)
Date: Mon, 15 Sep 2008 11:09:23 -0500
Subject: [Developers] Correct error in parameter checking
In-Reply-To: <48CE7F57.4060103@aei.mpg.de>
References: <9535CBF4-C589-4A30-AD8A-3834DCF5DB41@cct.lsu.edu>
	<48CE7F57.4060103@aei.mpg.de>
Message-ID: <86B02B97-6B3E-427A-905B-2CA84C9F4379@cct.lsu.edu>

On Sep 15, 2008, at 10:29:27, Thomas Radke wrote:

> Erik Schnetter wrote:
>> Cactus has currently an error in its parameter checking.  It   
>> interprets ranges consisting of single numbers as allowing this  
>> number  and all larger numbers.  For example, the specification
>> INT spatial_order "Order of spatial differencing"
>> {
>>   2 :: "second order"
>>   4 :: "fourth order"
>> } 2
>> allows all numbers not less than 2.
>> The enclosed patch correct this.  Since this involves regular   
>> expressions which are often difficult to read, I have broken up  
>> the  regular expressions into macros containing sub-expressions.
>> I have also replaced calls to atoi and atof with calls to strtol  
>> and  strtod, since these allow error checking which also was not  
>> present  before.
>
> Hi Erik,
>
> I only noticed today that the (already applied) patch is incomplete.  
> It doesn't cover the case where the specification of a parameter  
> value is given just as '*' and no explicit ranges, as in:
>
> INT bla "some description"
> {
>  * :: "any integer"
> } 0
>
> Parameter values are then checked against an interval [INT_MIN,  
> INT_MIN] whereas it should be [INT_MIN, INT_MAX]; similarly for real- 
> type parameters.

I mis-thought that this case was not allowed by the specification.

> Attached is a patch correcting the assumed upper range. Ok to apply ?


Yes.

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/

My email is as private as my paper mail.  I therefore support encrypting
and signing email messages.  Get my PGP key from www.keyserver.net.



-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 194 bytes
Desc: This is a digitally signed message part
Url : http://www.cactuscode.org/pipermail/developers/attachments/20080915/4c4f140e/attachment.bin 

From gkhanna at umassd.edu  Mon Sep 22 07:56:56 2008
From: gkhanna at umassd.edu (Gaurav Khanna)
Date: Mon, 22 Sep 2008 08:56:56 -0400
Subject: [Developers] Cell port & optimizations
Message-ID: <793176D0-EA95-4AF7-B78D-0C66F2DCB2DD@umassd.edu>


Hi

Is there any current interest/effort to develop Cactus with Cell  
processor optimizations? The Cell is quite mature now, with several  
stable development tools and initial issues (like double-precision  
performance) fully addressed. And with several large compute clusters  
being built using Cell architecture, there is likely to be a "market"  
for such development.

I'm quite interested in this myself. Please let me know if there are  
others ..

Thanks,
Gaurav


-----------------------------------------------

GAURAV KHANNA
UMass Dartmouth, Physics
(508) 910 6605
http://gravity.phy.umassd.edu/

"Black holes are where God divided by zero." - Steven Wright




-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www.cactuscode.org/pipermail/developers/attachments/20080922/29a222a6/attachment-0001.html 

From schnetter at cct.lsu.edu  Mon Sep 22 19:34:04 2008
From: schnetter at cct.lsu.edu (Erik Schnetter)
Date: Mon, 22 Sep 2008 19:34:04 -0500
Subject: [Developers] Adding OpenMP directives to CactusBase/MoL
Message-ID: <6E3C75EA-0A3B-4E41-8C67-9F2EA5509B5A@cct.lsu.edu>

The enclosed patch adds OpenMP parallel for directives to MoL.

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/

My email is as private as my paper mail.  I therefore support encrypting
and signing email messages.  Get my PGP key from www.keyserver.net.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: openmp.diff
Type: application/octet-stream
Size: 20271 bytes
Desc: not available
Url : http://www.cactuscode.org/pipermail/developers/attachments/20080922/d344ea74/attachment.obj 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 194 bytes
Desc: This is a digitally signed message part
Url : http://www.cactuscode.org/pipermail/developers/attachments/20080922/d344ea74/attachment.bin 

From tradke at aei.mpg.de  Tue Sep 23 04:31:37 2008
From: tradke at aei.mpg.de (Thomas Radke)
Date: Tue, 23 Sep 2008 11:31:37 +0200
Subject: [Developers] Adding OpenMP directives to CactusBase/MoL
In-Reply-To: <6E3C75EA-0A3B-4E41-8C67-9F2EA5509B5A@cct.lsu.edu>
References: <6E3C75EA-0A3B-4E41-8C67-9F2EA5509B5A@cct.lsu.edu>
Message-ID: <48D8B779.9030205@aei.mpg.de>

Erik Schnetter wrote:
> The enclosed patch adds OpenMP parallel for directives to MoL.

Very good, please apply !

One could also replace the memcpy calls in InitialCopy.c and SandR.c by 
explicit for-loops with surrounding OpenMP directives.

-- 
Cheers, Thomas.

From i.hawke at soton.ac.uk  Tue Sep 23 04:49:23 2008
From: i.hawke at soton.ac.uk (I.Hawke)
Date: Tue, 23 Sep 2008 10:49:23 +0100
Subject: [Developers] Adding OpenMP directives to CactusBase/MoL
In-Reply-To: <6E3C75EA-0A3B-4E41-8C67-9F2EA5509B5A@cct.lsu.edu>
References: <6E3C75EA-0A3B-4E41-8C67-9F2EA5509B5A@cct.lsu.edu>
Message-ID: <48D8BBA3.3020009@soton.ac.uk>

Erik Schnetter wrote:
> The enclosed patch adds OpenMP parallel for directives to MoL.
>   
In the patch for RK3.c, around line 150, a CCTK_DELTA_TIME is changed to 
a cctk_delta_time. Is this intentional? It looks like a bug to me (with 
mesh refinement...).

Ian



From schnetter at cct.lsu.edu  Tue Sep 23 10:46:36 2008
From: schnetter at cct.lsu.edu (Erik Schnetter)
Date: Tue, 23 Sep 2008 10:46:36 -0500
Subject: [Developers] Adding OpenMP directives to CactusBase/MoL
In-Reply-To: <48D8BBA3.3020009@soton.ac.uk>
References: <6E3C75EA-0A3B-4E41-8C67-9F2EA5509B5A@cct.lsu.edu>
	<48D8BBA3.3020009@soton.ac.uk>
Message-ID: <50CC6CB7-9940-4697-BEAA-EDA60E39C65D@cct.lsu.edu>

On Sep 23, 2008, at 04:49:23, I.Hawke wrote:

> Erik Schnetter wrote:
>> The enclosed patch adds OpenMP parallel for directives to MoL.
>>
> In the patch for RK3.c, around line 150, a CCTK_DELTA_TIME is  
> changed to
> a cctk_delta_time. Is this intentional? It looks like a bug to me  
> (with
> mesh refinement...).


Thanks; I corrected this.

I applied the patch.

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/

My email is as private as my paper mail.  I therefore support encrypting
and signing email messages.  Get my PGP key from www.keyserver.net.



-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 194 bytes
Desc: This is a digitally signed message part
Url : http://www.cactuscode.org/pipermail/developers/attachments/20080923/44e95262/attachment-0001.bin