It's a pain keeping stuff synched on two sites, but these are the complete implementations, faster than FlashSort, which thrashes heavily for large N. It is no longer the champion it was. This code will be put in my library and submitted in its entirety to both net and org. even TJP, if it stops acting up. BTW, I have tried converting this for strings. It works but it is dismally slow, so I'll implement RADIX-256 for this process. The _CV() function seems to slow this process GREATLY, making it worse than string-specific code that does not require _CV().