In JD's magnificent histogram tutorial, there is a description of how
to total data using a separate list of indices (with possible
repeats),
along with a vague hint that "For large histograms, there are even
more
efficient ways to do this with very short or no loops (e.g. using a
histogram of the histogram)." I have exactly that situation. Because
about 90% of my indices are not repeated, I have achieved a decent
speed-up using the following code for the "single" cases ("indices",
"data", and "hist" are the index list, data list, and final result
respectively, and hist is already pre-dimensioned).
indhist = histogram(indices, omin=om, reverse_indices=indri)
dupehist = histogram(indhist, min=1, reverse_indices=duperi)
; unique cases, so we can use them to index the LHS:
if dupehist[0] gt 0 then begin
just1 = duperi[duperi[0]:duperi[1]-1]
hist[just1+om] += data[indri[indri[just1]]]
endif
And going with the brute-force for loop for the rest:
; loop through the rest
if n_elements(dupehist) gt 1 then begin
multiples = duperi[duperi[1]:*]
for j=0l,n_elements(multiples)-1 do begin
elements = indri[indri[multiples[j]]:indri[multiples[j]+1]-1]
hist[multiples[j]+om] += total(data[elements])
endfor
endif
However, the loop is still going over hundreds of thousands of entries
and I can't help but suspect that another histogram and some fancy
footing with the i-vector would get rid of it. Does anyone have any
suggestions? Thanks.
|
