| Re: for loop is killing me [message #56648 is a reply to message #56590] |
Thu, 08 November 2007 07:26  |
Conor
Messages: 138
Registered: February 2007
|
Senior Member |
|
|
Okay, here's one more for you. Your m loop and phi loop can both be
made for loop-less. Here's your warning though: I haven't tested this
code in the slightest. It's hard to test stuff when it is part of a
much larger program, so I'll leave that to you. I'll just explain
what this chunk of code is doing. So this replaces your m and your
phi loops:
i_phi_max = 20.
phis = rebin(findgen(i_phi_max)+1,i_phi_max,i_phi_max)
i_ms = rebin(findgen(1,i_phi_max)+1,i_phi_max,i_phi_max)
gp = 2.0*(1./i_ms)*(r_minus/r_plus)^i_ms * cos(i_ms * phis*2*!PI/
i_phi_max )
gh = 2.0*(1./i_ms)*(r_minus*r_plus/a^2)^i_ms * cos(i_ms * phis*2*!PI/
i_phi_max )
G_in_phisum = total(gp,1)
G_out_phisum = total(gh,1)
The general idea here is to simply do all the calculations at once,
combined into one big array operation. So for instance from looking
at the phi loop you know that your phi's are going to go from 1-20 and
are going to be used 20 times by the m-loop. Imagine this as a 20x20
array where each row contains the numbers 1-20. The second line (phis
= rebin()) creates such a 20x20 array. Rebin is used to repeat the
findgen() 20 times vertically.
Next consider your i_m's. These will also have the values 1-20 and
will be used against the different values in the phis array. So you
can consider it to be a 20x20 array where each column has the values
1-20. Once again, this array is created using rebin and findgen in a
very similar way to how I created the phi array. Once this is
accomplished you have two 20x20 arrays with all the 400 possible
combinations of phi and m. Then, you perform your calculations
exactly as you did in the for loop and in one go calculate all the
possible results. Finally you use the dimension keyword to the total
function to sum up along rows. Viola, no more for loops! To make
sure things are clear, I want to explain this in a different way:
The idea is to do all your calculations at once. To make this happen
I created a 20x20 array for the phi loop and a 20x20 array for the m
loop. The idea is to pre-calculate all possible combinations of m and
phi and calculate all the possible values at once. That's what the
phis and ims arrays are for. Between them they contain all possible
values for i_m and phi. Traveling along a row is the equivelent of
looping over phi, and traveling along a column is the equivilent of
looping over i_m. So for instance if, with your original set up, you
were in the 3rd iteration of the m-loop and the 12th iteration of the
phi loop the i_m variable would have the value of 3 and the phi
variable would have the value of 12. With the new setup if you
printed out the values of i_ms[11,2] and phis[11,2] you would get the
values 3 and 12! Hopefully this all makes sense.
Also, please note that there is a potential source of confusion
between my code and yours because for your phi-loop you looped from
phi=0,19 but then added one later. In my code this meant that the
findgen() for phi had a +1 after it.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
