for loop taking too long to operate

I have a matrices (b1 and popx) with size 25000x50000, code is a vector which contains unique values in b1 (of length 300). For elements in popx which correspond to each unique value of b1, I need the sum of popx,

The below code takes very long (>2 hours) to operate, Any help to lower computation time is much appreciated

for i=1:300
  popx1=popx;
  popx1(b1~=code(i))=NaN;
  popn=nansum(nansum(popx1));
  popn1(i)=popn;
end;

Answer 1

[Jos (10584)]

Pre-allocation, and recoding using logical indexing and sum should speed things up

popn1 = nan(300,1) ; % pre-allocation
for i=1:300
  tf = b1 == code(i) ;
  popn1(i) = sum(popx(tf), 'all') ;
end;

[SChow]

much quicker , thanks