| Re: Random convolution ? [message #40081] |
Fri, 09 July 2004 08:25 |
R.G. Stockwell
Messages: 363
Registered: July 1999
|
Senior Member |
|
|
"Chris Torrence" <chris_torrence@NO-SPAM.yahoo.com> wrote in message news:10erieidmldig88@corp.supernews.com...
> Hi Emmanuel and Bob,
>
> Can you tell me your IDL version and platform? And also whether you have
> multiple CPUs on your machine? The following IDL commands should give this
> information:
...
> -Chris Torrence
> Research Systems, Inc.
Hi Chris,
I have not come across any other cases that have failed.
Here is the info from my system (Dell Inspirion 8100 - Laptop):
IDL> print,!version
{ x86 Win32 Windows Microsoft Windows 6.0 Jun 27 2003 32 64}
IDL> help, !cpu, /struct
** Structure !CPU, 6 tags, length=24, data length=24:
HW_VECTOR LONG 0
VECTOR_ENABLE LONG 0
HW_NCPU LONG 1
TPOOL_NTHREADS LONG 1
TPOOL_MIN_ELTS LONG 100000
TPOOL_MAX_ELTS LONG 0
-bob stockwell
|
|
|
|
| Re: Random convolution ? [message #40083 is a reply to message #40081] |
Thu, 08 July 2004 23:03  |
Emmanuel Christophe
Messages: 11
Registered: March 2004
|
Junior Member |
|
|
Hi
Here are the information about the system where convol bug:
IDL> print, !version
{ sparc sunos unix Solaris 6.0.3 Feb 26 2004 32 64}
IDL> help, !cpu, /struct
** Structure !CPU, 6 tags, length=24, data length=24:
HW_VECTOR LONG 0
VECTOR_ENABLE LONG 0
HW_NCPU LONG 3
TPOOL_NTHREADS LONG 3
TPOOL_MIN_ELTS LONG 100000
TPOOL_MAX_ELTS LONG 0
IDL>
On another system, same version, same operating system, but more memory
and only 2 CPUs, everything seems to be fine.
Chris Torrence wrote:
> Hi Emmanuel and Bob,
>
> Can you tell me your IDL version and platform? And also whether you have
> multiple CPUs on your machine? The following IDL commands should give this
> information:
>
> IDL> print,!version
> { x86 Win32 Windows Microsoft Windows 6.0 Jun 27 2003 32 64}
>
> IDL> help, !cpu, /struct
> ** Structure !CPU, 6 tags, length=24, data length=24:
> HW_VECTOR LONG 0
> VECTOR_ENABLE LONG 0
> HW_NCPU LONG 2
> TPOOL_NTHREADS LONG 2
> TPOOL_MIN_ELTS LONG 100000
> TPOOL_MAX_ELTS LONG 0
>
> I can't seem to reproduce your bug on my machine using IDL6.0. Do you have
> any other cases that fail?
>
> However, there was a known bug with CONVOL in IDL5.6 and IDL6.0, where it
> could produce incorrect results using multiple threads. This has been fixed
> for IDL6.1.
>
> -Chris Torrence
> Research Systems, Inc.
>
>
|
|
|
|
| Re: Random convolution ? [message #40084 is a reply to message #40083] |
Thu, 08 July 2004 15:30 |
Chris[2]
Messages: 39
Registered: August 2003
|
Member |
|
|
Hi Emmanuel and Bob,
Can you tell me your IDL version and platform? And also whether you have
multiple CPUs on your machine? The following IDL commands should give this
information:
IDL> print,!version
{ x86 Win32 Windows Microsoft Windows 6.0 Jun 27 2003 32 64}
IDL> help, !cpu, /struct
** Structure !CPU, 6 tags, length=24, data length=24:
HW_VECTOR LONG 0
VECTOR_ENABLE LONG 0
HW_NCPU LONG 2
TPOOL_NTHREADS LONG 2
TPOOL_MIN_ELTS LONG 100000
TPOOL_MAX_ELTS LONG 0
I can't seem to reproduce your bug on my machine using IDL6.0. Do you have
any other cases that fail?
However, there was a known bug with CONVOL in IDL5.6 and IDL6.0, where it
could produce incorrect results using multiple threads. This has been fixed
for IDL6.1.
-Chris Torrence
Research Systems, Inc.
|
|
|
|
| Re: Random convolution ? [message #40085 is a reply to message #40084] |
Thu, 08 July 2004 13:28 |
K. Bowman
Messages: 330
Registered: May 2000
|
Senior Member |
|
|
In article <MPG.1b57087d5ec808c9897cb@news.frii.com>,
David Fanning <davidf@dfanning.com> wrote:
>> The first time, I got
>> IDL> .GO
>> 85.961538
>> 793.00000
>> 768.00000
>>
>> and the second time:
>> IDL> .GO
>> 793.00000
>> 793.00000
>> 0.0000000
>>
>> And all results should be zero... I suspect a memory effect but how ?
>
> Don't know. I get all zeros every time I run this.
Hmm, me too.
IDL Version 6.0, Mac OS X (darwin ppc m32). (c) 2003, Research Systems,
Inc.
IDL> @convol_test
0.0000000
0.0000000
0.0000000
IDL> @convol_test
0.0000000
0.0000000
0.0000000
Ken Bowman
|
|
|
|
| Re: Random convolution ? [message #40086 is a reply to message #40085] |
Thu, 08 July 2004 10:21 |
R.G. Stockwell
Messages: 363
Registered: July 1999
|
Senior Member |
|
|
"Emmanuel Christophe" <melaneum555@yahoo.fr> wrote in message news:ccjkta$o61$1@news.cict.fr...
...
> The first time, I got
> IDL> .GO
> 85.961538
> 793.00000
> 768.00000
>
> and the second time:
> IDL> .GO
> 793.00000
> 793.00000
> 0.0000000
Hi,
here are my results, (win2000 idl 6.0, 512mb ram, 1gb swap,
usage peaked at ~750 mb)
Weird!
% Compiled module: $MAIN$.
IDL> .GO
112.00000
112.00000
0.00000000
IDL> .GO
0.00000000
0.00000000
0.00000000
I haven't even looked into it, i might have some time tomorrow.
Cheers,
bob
|
|
|
|
| Re: Random convolution ? [message #40089 is a reply to message #40086] |
Thu, 08 July 2004 08:20 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Emmanuel Christophe writes:
> Do you reach the memory limits of your systems ? Or after several tries
> ? I also have the problem on only one of two system...
No, I didn't run into memory limits. I have 1GByte of RAM
and 2 GByte of virtual memory on this machine.
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
| Re: Random convolution ? [message #40090 is a reply to message #40089] |
Thu, 08 July 2004 07:50 |
Emmanuel Christophe
Messages: 11
Registered: March 2004
|
Junior Member |
|
|
David Fanning wrote:
> Emmanuel Christophe writes:
>
>
>> I got a random effect using convolution:
>>
>> ------------------------------------------------------
>> imagetest=indgen(256,256,224)
>> ker=indgen(1,3,5)
>> r1=convol(double(imagetest),double(ker),/edge_truncate)
>> r2=convol(double(imagetest),double(ker),/edge_truncate)
>> r3=convol(double(imagetest),double(ker),/edge_truncate)
>> print, max(abs(r1-r2))
>> print, max(abs(r1-r3))
>> print, max(abs(r2-r3))
>> end
>> ------------------------------------------------------
>>
>> The first time, I got
>> IDL> .GO
>> 85.961538
>> 793.00000
>> 768.00000
>>
>> and the second time:
>> IDL> .GO
>> 793.00000
>> 793.00000
>> 0.0000000
>>
>> And all results should be zero... I suspect a memory effect but how ?
>
>
> Don't know. I get all zeros every time I run this.
>
> Cheers,
>
> David
>
Do you reach the memory limits of your systems ? Or after several tries
? I also have the problem on only one of two system...
|
|
|
|
| Re: Random convolution ? [message #40092 is a reply to message #40090] |
Thu, 08 July 2004 07:30 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Emmanuel Christophe writes:
> I got a random effect using convolution:
>
> ------------------------------------------------------
> imagetest=indgen(256,256,224)
> ker=indgen(1,3,5)
> r1=convol(double(imagetest),double(ker),/edge_truncate)
> r2=convol(double(imagetest),double(ker),/edge_truncate)
> r3=convol(double(imagetest),double(ker),/edge_truncate)
> print, max(abs(r1-r2))
> print, max(abs(r1-r3))
> print, max(abs(r2-r3))
> end
> ------------------------------------------------------
>
> The first time, I got
> IDL> .GO
> 85.961538
> 793.00000
> 768.00000
>
> and the second time:
> IDL> .GO
> 793.00000
> 793.00000
> 0.0000000
>
> And all results should be zero... I suspect a memory effect but how ?
Don't know. I get all zeros every time I run this.
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
