| Re: Cannot allocate memory! [message #47032] |
Tue, 17 January 2006 09:43 |
Liberum
Messages: 48
Registered: September 2005
|
Member |
|
|
Thanks for your help Craig,
I understand now.
The program worked with limiting the amount of arrays and without using
MEAN().
Cheers,
Sheldon
|
|
|
|
| Re: Cannot allocate memory! [message #47037 is a reply to message #47032] |
Tue, 17 January 2006 02:28  |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
"Sheldon" <shejo284@gmail.com> writes:
> Thanks David,
>
> IDL does have a very good manual.
> I have a side question that I hope is short enough for you to answer.
> When looking at the other solutions for averaging large 3D arrays, I
> noticed something, namely that some synthax for depicting a 3D array
> took the form: array[50,50,3] where the third dimension is placed at
> the end and not the beginning: array[3,50,50].
> Coming over from Python, this seems a bit strange. What is the
> difference between these two?
> Doesn't IDL interpret former as a 50 dimensional array of 50 col x 3
> rows and the latter as a 3 dimensional array of 50 col x 50 rows?
It doesn't really matter whether you call an array index a row or a
column. There is more discussion here:
http://www.dfanning.com/misc_tips/colrow_major.html
I say it doesn't matter because most often, you must refer to the
array indices explicitly, so you can define what you mean by a row and
column yourself. The only time this is not true that I'm aware of is
when doing matrix multiplication with the "#" and "##" operators.
It's best just to verify which one is right for you.
For implementation efficiency, it may be important to know which array
elements are adjacaent. For an array, the elements ARRAY(*,J,K) are
contiguous in memory.
Craig
--
------------------------------------------------------------ --------------
Craig B. Markwardt, Ph.D. EMAIL: craigmnet@REMOVEcow.physics.wisc.edu
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
------------------------------------------------------------ --------------
|
|
|
|
| Re: Cannot allocate memory! [message #47039 is a reply to message #47037] |
Tue, 17 January 2006 00:56 |
Liberum
Messages: 48
Registered: September 2005
|
Member |
|
|
Thanks David,
IDL does have a very good manual.
I have a side question that I hope is short enough for you to answer.
When looking at the other solutions for averaging large 3D arrays, I
noticed something, namely that some synthax for depicting a 3D array
took the form: array[50,50,3] where the third dimension is placed at
the end and not the beginning: array[3,50,50].
Coming over from Python, this seems a bit strange. What is the
difference between these two?
Doesn't IDL interpret former as a 50 dimensional array of 50 col x 3
rows and the latter as a 3 dimensional array of 50 col x 50 rows?
Cheers,
Sheldon
|
|
|
|
| Re: Cannot allocate memory! [message #47042 is a reply to message #47039] |
Mon, 16 January 2006 13:09 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Sheldon writes:
> I cannot average each array because I want the average at each pixel
> for the 119 or so arrays.
> What I am toying with now is this creating 2 1215*1215 arrays and use
> one to carry the new arrays and one to keep the running total:
> t = make_array(col,row)
> sum = make_array(col,row)
> for i=0, n-1 do begin
> t = read_hdf(f,str)
> sum = temporary(sum) + temporary(t)
> endfor
> avg = sum/float(n)
> Basicly I need to keep IDL from making many of these arrarys. I have
> not tested this yet.
> Now you will really laugh at this but it is best that I can come up
> with with what little experience I have using IDL. I am reading some
> older problems sent to the group but I am not done yet. Any comments?
I think this is more or less what Craig had in mind.
At least it is tending in the right direction. :-)
Cheers,
David
--
David Fanning, Ph.D.
Fanning Software Consulting, Inc.
Coyote's Guide to IDL Programming: http://www.dfanning.com/
|
|
|
|
| Re: Cannot allocate memory! [message #47043 is a reply to message #47042] |
Mon, 16 January 2006 12:57 |
Liberum
Messages: 48
Registered: September 2005
|
Member |
|
|
Hi Craig,
I cannot average each array because I want the average at each pixel
for the 119 or so arrays.
What I am toying with now is this creating 2 1215*1215 arrays and use
one to carry the new arrays and one to keep the running total:
t = make_array(col,row)
sum = make_array(col,row)
for i=0, n-1 do begin
t = read_hdf(f,str)
sum = temporary(sum) + temporary(t)
endfor
avg = sum/float(n)
Basicly I need to keep IDL from making many of these arrarys. I have
not tested this yet.
Now you will really laugh at this but it is best that I can come up
with with what little experience I have using IDL. I am reading some
older problems sent to the group but I am not done yet. Any comments?
Sheldon
|
|
|
|
|
|
| Re: Cannot allocate memory! [message #47045 is a reply to message #47044] |
Mon, 16 January 2006 12:26 |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
"Sheldon" <shejo284@gmail.com> writes:
> Hi everyone,
>
> I am working with very large arrays of varying quantities, i.e.,
> sometimes I have 89 arrays and other times I have 200 arrays. Now I
> would like to average these arrays and reduce them to one array.
> I use the MAKE_ARRAY() function but IDL screams that it cannot allocate
> the memory.
> Now these arrays are quite large (1215,1215) and I only need to hold
> them within a function and then they are averaged using MEAN().
> Is there away to do this so that IDL can allocate the memory?
> IDl complains about this:
> tmp = make_array([119,1215,1215], /float). I was thinking about
> changing to /integer but I think is not such a good fix.
If all you are doing is straight averaging, then you can average each
1215x1215 slice individually using a FOR loop, and then average those
averages. That allows you to keep only one image in memory at a time.
Good luck,
Craig
--
------------------------------------------------------------ --------------
Craig B. Markwardt, Ph.D. EMAIL: craigmnet@REMOVEcow.physics.wisc.edu
Astrophysics, IDL, Finance, Derivatives | Remove "net" for better response
------------------------------------------------------------ --------------
|
|
|
|
| Re: Cannot allocate memory! [message #47046 is a reply to message #47045] |
Mon, 16 January 2006 12:25 |
David Fanning
Messages: 11724
Registered: August 2001
|
Senior Member |
|
|
Sheldon writes:
> I am working with very large arrays of varying quantities, i.e.,
> sometimes I have 89 arrays and other times I have 200 arrays. Now I
> would like to average these arrays and reduce them to one array.
> I use the MAKE_ARRAY() function but IDL screams that it cannot allocate
> the memory.
> Now these arrays are quite large (1215,1215) and I only need to hold
> them within a function and then they are averaged using MEAN().
> Is there away to do this so that IDL can allocate the memory?
> IDl complains about this:
> tmp = make_array([119,1215,1215], /float). I was thinking about
> changing to /integer but I think is not such a good fix.
> Any takers?
Uh, have you been reading the recent discussion of MEAN()
in this newsgroup? There might be an idea or two in there. :-)
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
