| Programming style [message #3375] |
Wed, 18 January 1995 07:33  |
lars
Messages: 4
Registered: January 1995
|
Junior Member |
|
|
Does anybody out there know if IDL passes arguments to functions by value
or by refrence? I have developed a programming-style where I tend to pass huge
datastructures as parameters and not through common blocks. This is
obviously inefficient in the pass-by-value-case, so I might have to change
my style.
Greetings,
Lars
|
|
|
|
| Re: Programming style [message #11732 is a reply to message #3375] |
Sat, 16 May 1998 00:00  |
steinhh
Messages: 260
Registered: June 1994
|
Senior Member |
|
|
J.D. Smith wrote:
> z=findgen(nx*ny*nz)
> x2=((z mod (nx*ny)) mod nx-nx/2)^2
> y2=((z mod (nx*ny))/nx-ny/2)^2
> z2=((temporary(z)/(nx*ny)) - nz/2)^2
> array=bytarr(nx,ny,nz)
> array[where(z2+x2/e11^2+y2/e21^2 le c1^2 AND z2+x2/e12^2+y2/e22^2 ge
> c2^2)]=1b
>
> Note the two ellipses are centered on the midpoint of the array and are
> concentric. This can be modified by changing the subtracted value in
> each of x2,y2,z2. Definitely faster than loops. Elegance is in the eye
> of the beholder, though.
My $0.02 modification (speed/elegance?)
x2 = rebin((findgen(nx,1,1)-nx/2)^2,nx,ny,nz)
y2 = rebin((findgen(1,ny,1)-ny/2)^2,nx,ny,nz)
z2 = rebin((findgen(1,1,nz)-nz/2)^2,nx,ny,nz)
array = z2 + x2/e11^2 + y2/e21^2 LE c1^2 AND $
z2 + x2/e12^2 + y2/e22^2 GE c2^2
In case nx/ny/nz are large, this should save considerable time..
Since we're not using such huge indices any longer, we could
also substitute FINDGEN with INDGEN (if you don't need fractional
pixel positioning of the zero point (center) of the x2,y2,z2
arrays).
Stein Vidar
|
|
|
|
| Re: Programming style [message #11736 is a reply to message #3375] |
Sat, 16 May 1998 00:00 |
pford
Messages: 33
Registered: September 1996
|
Member |
|
|
In article <355CAD61.6BF75748@astrosun.tn.cornell.edu>, "J.D. Smith"
<jdsmith@astrosun.tn.cornell.edu> wrote:
> Patrick V. Ford wrote:
>>
>> I have a general style/algorithm question.
>>
>> I want to plot in a 3-D array an ellipsoid within an in ellipsoid where the
>> voxels between the boundaries are non-zero and else where zero.
>>
>> The general function is (x/a)^2 + (y/b)^2 + (z/c)^2 = 1.0
>>
>> I have already done this using for loops and conditional statements but it
>> occurred to me that there may be some IDL matrix-boolean logic combination
>> that could accomplish this in a faster and more 'elegant' fashion.
>>
>> I am now open for suggestions?
>>
>> Thanks in advance.
>>
>> Patrick Ford, MD
>> Department of Radiology
>> Baylor College of Medicine
>> pford@bcm.tmc.edu
>
> How about:
>
> recast as:
>
> Eq. 1: z^2+(x/e11)^2+(y/e21)^2 z^2=c1^2 (outer ellipse)
> Eq. 2: z^2+(x/e12)^2+(y/e22)^2 z^2=c2^2 (inner ellipse)
>
> let your array be nx by ny by nz.
>
> Then:
>
> z=findgen(nx*ny*nz)
> x2=((z mod (nx*ny)) mod nx-nx/2)^2
> y2=((z mod (nx*ny))/nx-ny/2)^2
> z2=((temporary(z)/(nx*ny)) - nz/2)^2
> array=bytarr(nx,ny,nz)
> array[where(z2+x2/e11^2+y2/e21^2 le c1^2 AND z2+x2/e12^2+y2/e22^2 ge
> c2^2)]=1b
>
> Note the two ellipses are centered on the midpoint of the array and are
> concentric. This can be modified by changing the subtracted value in
> each of x2,y2,z2. Definitely faster than loops. Elegance is in the eye
> of the beholder, though.
>
> NB: The x,y, and z index vectors must be floats, since for 3-d
> data,indices get large pretty quick. E.g. 100x100x100 would choke with
> longs (since 100^3^2=10^12=2^39.86!). This introduces some "fuzziness"
> at the boundaries due to roundoff. You can throw in a floor() statement
> to eliminate this if you really want.
>
> JD
>
> --
> J.D. Smith |*| WORK: (607) 255-5842
> Cornell University Dept. of Astronomy |*| (607) 255-4083
> 206 Space Sciences Bldg. |*| FAX: (607) 255-5875
> Ithaca, NY 14853 |*|
Interesting. I am going to have to think about this since it is not
immediately obvious to me. The maximum size of the cube I am working with
is 64X64X64 ( x8 since it is a cyclical dynamic objec) with the actual
object a lot smaller. I am actually using a hemi ellipsoid to grossly
represent the heart muscle of the left ventrical of the heart in a
radionuclide myocardial perfusion image. Fuzziness for this 'model' is
actually an advantage.
--
Patrick Ford
pford@bcm.tmc.edu
|
|
|
|
| Re: Programming style [message #11740 is a reply to message #3375] |
Fri, 15 May 1998 00:00 |
J.D. Smith
Messages: 214
Registered: August 1996
|
Senior Member |
|
|
Patrick V. Ford wrote:
>
> I have a general style/algorithm question.
>
> I want to plot in a 3-D array an ellipsoid within an in ellipsoid where the
> voxels between the boundaries are non-zero and else where zero.
>
> The general function is (x/a)^2 + (y/b)^2 + (z/c)^2 = 1.0
>
> I have already done this using for loops and conditional statements but it
> occurred to me that there may be some IDL matrix-boolean logic combination
> that could accomplish this in a faster and more 'elegant' fashion.
>
> I am now open for suggestions?
>
> Thanks in advance.
>
> Patrick Ford, MD
> Department of Radiology
> Baylor College of Medicine
> pford@bcm.tmc.edu
How about:
recast as:
Eq. 1: z^2+(x/e11)^2+(y/e21)^2 z^2=c1^2 (outer ellipse)
Eq. 2: z^2+(x/e12)^2+(y/e22)^2 z^2=c2^2 (inner ellipse)
let your array be nx by ny by nz.
Then:
z=findgen(nx*ny*nz)
x2=((z mod (nx*ny)) mod nx-nx/2)^2
y2=((z mod (nx*ny))/nx-ny/2)^2
z2=((temporary(z)/(nx*ny)) - nz/2)^2
array=bytarr(nx,ny,nz)
array[where(z2+x2/e11^2+y2/e21^2 le c1^2 AND z2+x2/e12^2+y2/e22^2 ge
c2^2)]=1b
Note the two ellipses are centered on the midpoint of the array and are
concentric. This can be modified by changing the subtracted value in
each of x2,y2,z2. Definitely faster than loops. Elegance is in the eye
of the beholder, though.
NB: The x,y, and z index vectors must be floats, since for 3-d
data,indices get large pretty quick. E.g. 100x100x100 would choke with
longs (since 100^3^2=10^12=2^39.86!). This introduces some "fuzziness"
at the boundaries due to roundoff. You can throw in a floor() statement
to eliminate this if you really want.
JD
--
J.D. Smith |*| WORK: (607) 255-5842
Cornell University Dept. of Astronomy |*| (607) 255-4083
206 Space Sciences Bldg. |*| FAX: (607) 255-5875
Ithaca, NY 14853 |*|
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
