| Re: Shortcut needed [message #10366 is a reply to message #10365] |
Sun, 16 November 1997 00:00  |
sclaflin
Messages: 5
Registered: April 1996
|
Junior Member |
|
|
Daniel Williams (williams@idunn.srl.caltech.edu) wrote:
: Hi fellow IDL-users,
: I have a program which runs slower than I like, and wonder if someone
: could help me speed it up a bit.
: My data is an array of structures, call it 'a'. One element of the
: structure, call it 'n' is an integer, another is a quantity which I
: want to add up, call it 'x'.
: There are about 100000 elements in this array. The integer n, marks
: data as being of a certain type. I want to make a new array, call it
: 'y', which has as its nth element the sum of all the x's which have a
: n-value of n. I currently do this as follows,
: for n=0, n_max do begin
: y[n] = total( a[where(a.n eq n)].x )
: endfor
: The trouble is that the loop is slow, as for loops always are. Does
: anyone out there know a better way to do this kind of sorting?
Some weeks ago the virtues of the histogram function for sorting
problems were discussed in this news group. I haven't tested the
the following solution, but if you have a really slow process,
it might be worth a try. There is still a "for" loop, but the
"where" function does not have to be called each time through the loop.
h = histogram(a.n, Min=0, Max=n_max, Reverse_Indices=r)
h = 0b
y = intarr(n_max + 1)
for i=0,n_max do begin
if (r[i+1] gt r[i]) then begin
y[i] = total(a[r[r[i]:r[i+1]-1]].x)
endif
endfor
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
