I've uploaded a new integer_sort release to http://sourceforge.net/projects/spreadsort/, and attached the source to this email. This uses randomaccessiterators just like std::sort, and should be just as flexible. I will see if using some generic algorithms for iterators improves the performance a little; I'm seeing a noticeable speed penalty with iterators instead of pointers. I get the same performance as the old code when I pass in pointers as the iterators, but a roughly 15% increase in runtime when I use .begin() and .end(). I tried the performance optimization I discussed in a prior email, and it didn't work. I'm going to leave it as is, because I've found the old way to give a safe, steady performance improvement across platforms, even if said improvement isn't huge. Suggestions are welcome. Steve