Thank you for your help,

Sadly, it's dynamic data. Basically, I start with half a million 
integers, then I generate from this a bigger set of integers (using 
millions of calls to the include? method in the process), that I must 
add to the initial set while avoiding redundancy. Then rinse and repeat 
until around 20 million.

I tried using a long string encoded in UTF-32BE, but the include? method 
is much too slow. These numbers are out of range for NArray, and this 
has to run on Linux so I can't use RTensor either. A 512 MB Bitmap could 
work. I will start by trying Robert Klemme's BitSet. Thank you very 
much.

Regards,

George

-- 
Posted via http://www.ruby-forum.com/.