| Re: help on optimization [message #76685 is a reply to message #76684] |
Thu, 23 June 2011 05:44  |
Wout De Nolf
Messages: 194
Registered: October 2008
|
Senior Member |
|
|
On Thu, 23 Jun 2011 02:38:00 -0700 (PDT), stefania
<stefania.giodini@gmail.com> wrote:
> Any help/idea/feedback will be very useful!
1. I don't think the preselection speeds things up, so skip it.
2. Avoid using WHERE by specifying a MAX in histogram (use 100^ to
avoid SQRT).
3. Remove some redundancy in total / cumulative total.
xgaspos=file_gaspos[0,*]
ygaspos=file_gaspos[1,*]
for i=0,nhalos do begin
;Distance of particles to the halo
dis2=(xrand[i]-xgaspos)^2+(yrand[i]-ygaspos)^2
;histogram distances within 10Mpc
hist =histogram(dis2[ind_gas],binsize=bin_mpc,$
max=10000,reverse_indices=ri)
gasmass=file_gasmass[ind_gas]
for ll=0,n_elements(hist)-1 do $
if ri[ll] eq ri[ll+1] then gasm[ll]=0 $
else gasm[ll]=total(gasmass[ ri[ ri[ll]:ri[ll+1]-1 ] ])
ymin_r_gas+=total(gasm,/cumul,/pres) ; Sigma(<r)
yplot_gas+=gasm ; Sigma(r)
endfor
4. You can think of calculating "dis" outside the loop but it depends
on how many columns file_gaspos has (memory issues). Assume xrand en
yrand are row vectors:
dis=sqrt((rebin(xrand,nhalos,np,/sample)-rebin(file_gaspos[0 ,*],nhalos,np,/sample))^2+$
(rebin(yrand,nhalos,np,/sample)-rebin(file_gaspos[1,*],nhalo s,np,/sample))^2)
You can also devide nhalos in smaller chunks. However you still need
to do the histogram and mass summation for each column separatly.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
