comp.lang.idl-pvwave archive: archive » Can I do this without using loops?

Home » Public Forums » archive » Can I do this without using loops?

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: Can I do this without using loops? [message #6404 is a reply to message #6323]

Sat, 15 June 1996 00:00

bowman
Messages: 121
Registered: September 1991

Senior Member

In article <31C17DE4.50C@zibmt.uni-ulm.de>, David Ritscher
<david.ritscher@zibmt.uni-ulm.de> wrote:

> The algorithms:
> 1. array = fltarr(rows, columns, /nozero)
> for j = 0L, columns-1 do array(*, j) = float(j)

> 3. array = FINDGEN(rows) # REPLICATE(1, columns)

> Here are the execution times, in seconds, under IDL and PV-Wave, for
> the specified array sizes, for each of the algorithms above. The left
> two columns show the times for inserting columns, the second two
> columns involve inserting rows.
>
> Tested using the following number of rows and columns: 3000, 3000
>
> Insert columns Insert rows
> IDL PV-Wave IDL PV-Wave
> 1. 22.84 17.58 30.48 21.31

> 3. 3.88 9.73 3.87 9.70
>
> Tested using the following number of rows and columns: 1000, 30000
>
> Insert columns Insert rows
> IDL PV-Wave IDL PV-Wave
> 1. 84.95 71.09 Inf Inf

> 3. 27.65 42.95 30.88 50.10

I confess to being surprised by how fast (3) is compared to (1). I always
thought these tricks using the # operator to avoid loops were 'too clever
by half'. At least on the face of it, (3) creates an unnecessary
temporary vector and does useless multiply operations. The multiplies can
probably be perfectly overlapped with the stores on a RISC machine (and so
lead to no additional cost), but even so it seems more complicated. There
may still be aspects of (1) that are interpreted, (the loop probably is),
but it would seem to be the simplest possible operation to write as
optimized code.

Thanks for the lesson.

This does point out how useful a good IDL profiling tool would be.

Ken Bowman

--
Kenneth P. Bowman, Assoc. Prof. 409-862-4060
Department of Meteorology 409-862-4132 fax
Texas A&M University bowman@csrp.tamu.edu
College Station, TX 77843-3150
Satellite ozone movies on CD-ROM --> http://www.lenticular.com/

Report message to a moderator

[Message index]

		Can I do this without using loops? By: Peter Clinch on Wed, 05 June 1996 00:00
		Re: Can I do this without using loops? By: steinhh on Sun, 16 June 1996 00:00
		Re: Can I do this without using loops? By: bowman on Sat, 15 June 1996 00:00
		Re: Can I do this without using loops? By: David Ritscher on Sat, 15 June 1996 00:00
		Re: Can I do this without using loops? By: steinhh on Fri, 14 June 1996 00:00
		Re: Can I do this without using loops? By: David Ritscher on Fri, 14 June 1996 00:00
		Re: Can I do this without using loops? By: steinhh on Tue, 11 June 1996 00:00
		Re: Can I do this without using loops? By: Trygve Sparr on Tue, 11 June 1996 00:00
		Re: Can I do this without using loops? By: David Ritscher on Mon, 10 June 1996 00:00

Previous Topic:	Searching for Constraint Optimization IDL Routine
Next Topic:	Re: polar surfaces

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Sat Nov 29 00:56:27 PST 2025

Total time taken to generate the page: 1.91698 seconds