Dutch log

One of the things now that I’ve got some results is to prove the significance of them (significance in the statistical meaning). I’ve read a few things and it seems like I need runs of 30 to be able to compare. So I’ve been doing this. I’ve also been trying to reduce variations between runs so I’ve been running 60 seconds instead of 20 seconds per run. It now takes several days per test!While this has been going on I reworked my time difference server a little. Since three of the boxes were running the tests I used the old Compaq and the eMac. Needed to do a bit of writing of code to handle network byte order issues since hadn’t bothered before but with a little endian and big endian machine it becomes more of an issue… Spent a fair bit of time chasing around why that wasn’t working and it turned out to be my debug statements not working first off and then promotion orders of types. Interesting enough those two machines drift as much as 500 msecs per minute apart where the other machines I use it is < 5 msecs per minute.…more