| Re: intarr speed in structure [message #10847] |
Mon, 16 February 1998 00:00 |
Kevin Ivory
Messages: 71
Registered: January 1997
|
Member |
|
|
c.c.mclean@ed.ac.uk wrote:
> When I allocate an array, as part of a structure, it
> takes far longer to perform the operation than normal.
>
> a) array=intarr(x,y,z)
> b) array={volume:intarr(x,y,z)}
>
> if I use x=2048,y=240,z=5 a) takes 0.1s, b) takes 0.4s
> if I use x=2048,y=240,z=15 a) takes 0.4s,b) takes 2mins
>
> The same thing happens with IDL 5.03 under win95 (P200, 32MB)
> and with IDL 4 on an HP workstation.
Just some wild guessing:
It looks like b) is making the array first, then the structure.
For the first example that would make 10 MB for a), 20 MB for b),
for the second example 30 MB for a), 60 MB for b)
Since you only have 32 MB RAM, IDL starts swapping to disk,
which is very slow.
For b) you might want to try: array={volume:intarr(x,y,z, /nozero)}
I didn't test it, but it might help.
Best regards
Kevin
--
Kevin Ivory Tel: +49 5556 979 434
Max-Planck-Institut fuer Aeronomie Fax: +49 5556 979 240
Max-Planck-Str. 2 mailto:Kevin.Ivory@linmpi.mpg.de
D-37191 Katlenburg-Lindau, GERMANY http://www.gwdg.de/~kivory2/
|
|
|
|
| Re: intarr speed in structure [message #10848 is a reply to message #10847] |
Mon, 16 February 1998 00:00  |
Dr. G. Scott Lett
Messages: 14
Registered: February 1998
|
Junior Member |
|
|
c.c.mclean@ed.ac.uk wrote:
> Hi,
>
> When I allocate an array, as part of a structure, it
> takes far longer to perform the operation than normal.
>
> ie. a) is much quicker than b) where:-
>
> a) array=intarr(x,y,z)
> b) array={volume:intarr(x,y,z)}
>
> if I use x=2048,y=240,z=5 a) takes 0.1s, b) takes 0.4s
> if I use x=2048,y=240,z=15 a) takes 0.4s,b) takes 2mins
>
> The same thing happens with IDL 5.03 under win95 (P200, 32MB)
> and with IDL 4 on an HP workstation.
>
> Can anybody explain why there is this difference!
>
> TIA
>
> Calum...
>
> -----== Posted via Deja News, The Leader in Internet Discussion ==-----
> http://www.dejanews.com/ Now offering spam-free web-based newsreading
In the case of the structure, IDL does the following:
Allocate enough memory for INTARR(x,y,z), and fill with zeroes.
Allocate enough memory for the structure, including enough to hold the
array.
Copy the (original) array into the structure.
Free the memory holding the (original) array.
I'm just guessing, but this looks like at least 3 times as much work as
simply
allocating memory for a single array, not counting for special hardware
and such.
Since 2 copies of the array exist temporarily, IDL uses (at least) twice
as much
memory to build the structure as the array. In your latter case,
intarr(2048,240,15)
requires about 14MB, so {volume:intarr(2048,240,15)} requires 28MB to
build.
Adding the memory for the IDL, the operating system, and any other memory
use,
you've probably exceeded the 32MB your system has, and you're using
virtual
memory, which is many times slower than RAM. This probably explains why
your second example is so much slower than the first.
--
========================
Dr. G. Scott Lett
slett@holisticmath.com
http://holisticmath.com/
========================
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
