https://issues.apache.org/ooo/show_bug.cgi?id=121421
--- Comment #27 from orcmid <[email protected]> --- Created attachment 79982 --> https://issues.apache.org/ooo/attachment.cgi?id=79982&action=edit Improved Generator for producing sets of results (In reply to comment #25) > For the initialization, you can try something like this: > > long SimpleSeeder(long prev) > { return (((69069L * prev) & 0x3FFFFFFF) | 1; > /* not all that random but nonzero and less than > the moduli, all up near 0x7FFFFFFF */ > } This was the right idea, but the bigger problem was having sufficient variation from starter seeds chosen only a few seconds apart. Here is a program using the 32-bit updated Whitmann-Hill pseudo-random generator, similar to the one Pedro has put in the form of a patch to Calc. This one can be run standalone to see how varied the sequences are, how there is useful variance in the initial result as well as the continuation of a long series, etc. This is not the same as strenuous testing. It's time for that now. -- You are receiving this mail because: You are on the CC list for the bug.
