Advice on a few bottlenecks in my code

According to the profiler, I have four lines of code taking up 99% of my CPU time.

if(statment)  
    pCond = (bold_Phi(k,2) <= data(thisIdx : idx, 7)); 
else
    pCond = (data(thisIdx : idx, 7) <= bold_Phi(k,2)); 
end
  IDY = find( (data(thisIdx : idx, 1) == -1) & (pCond) & (data(thisIdx : idx, 11) == bold_Phi(k,p + 10))  ); 
  IDZ = find( (dV(thisIdx : idx, :) <= -bold_Phi(k,3)) & (pCond) & (sCond)  );

where bold_Phi and data are 2D matrices and dV is a vector.

I dont think the time penalty is coming from the indexing as the following statement is very quick

~isnan(data(thisIdx : idx, 3));

anything obvious I have missed that will help me speed up these few statements?

thank you

[Matt Kindig]

How large are bold_Phi and dV?

[Matlab2010]

dv = 1x500K. bold_Phi = 1E5 x 5E2 data = 12 x 500K.

[Sean de Wolski]

What is statement?

[Matlab2010]

statement is a bool.

Answer 1

[Kye Taylor]

If you don't need the actual indices, you can omit the find command to save time. In particular, notice that the following code blocks are equivalent but the first does not use find:

data = rand(1000,1);
isBig = data>0.5;
bigData = data(isBig);

and

data = rand(1000,1);
isBigIdx = find(data>0.5);
bigData2 = data(isBigIdx);

[Matlab2010]

this has helped, but only a little bit. Sped up by c. 10%.

Answer 2

[Mark Whirdy]

(Almost-) Never use the find() function!

I'd need to see the contents of the variables, but at first glance you can maybe just remove it altogether with same functionality?