| Re: about replicate_inplace [message #35901 is a reply to message #35899] |
Tue, 22 July 2003 14:46  |
Mark Hadfield
Messages: 783
Registered: May 1995
|
Senior Member |
|
|
Xiaoying Jin wrote:
> I am really puzzled by the IDL usage of memory.
Me too.
> Here is an example:
>
> T = systime(1)
> b = bytarr(nreg, nreg)
> print, systime(1) - T, ' seconds.'
>
> T = systime(1)
> replicate_inplace, b, 0
> print, systime(1) - T, ' seconds.'
>
> In the example, the first part is to create an array with values 0.
> The second part is to assign each element in the array the value 0.
> The running time for the first part is 0.12sec, while the running time
> for the second part is 0.37sec.
Interesting. On my system (Pentium 4 2.67 GHz with 1 GB RAM) I get
similar results, though the difference is not so marked. eg with nreg =
10000, my times are 0.22 s & 0.34 s respectively. (I have to run these
tests a couple of times to get stable results.)
However creating the array with REPLICATE takes 0.37 s, ie.
significantly longer than BYTARR and slightly less than REPLICATE_INPLACE.
> My question is: since the first part needs to allocate the memory and
> assign values to elements, it should take longer time than the second
> part. In IDL help, it says: "REPLICATE_INPLACE can be faster and use
> less memory than the IDL function REPLICATE or the IDL array notation
> for large arrays that already exist. " However, why the first part
> takes less time?
Well, based on my experience, REPLICATE_INPLACE *does* take (slightly)
less time than REPLICATE.
> Any suggestion?
For most purposes, just use the approach that gives the most
straightforward, readable code. Where performance or memory usage
becomes an issue, test different approaches.
--
Mark Hadfield "Ka puwaha te tai nei, Hoea tatou"
m.hadfield@niwa.co.nz
National Institute for Water and Atmospheric Research (NIWA)
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
