What determines which protein structure to be generated ?

[xman]

Since the DF is running without the need to communicate with the server, so what kind of protein structures will be generated depends on kind of random number ? If this is the case, then is that means starting the program with the same seed number would generate identical protein structures ? Thus, work will be redundant without benefit. :rolleyes:

[wirthi]

Hi,

copied from http://www.distributedfolding.org/details.html


Work depends only on the seed of the random number generator which is based on the time you start the program and the process ID (basically the two most random things on your machine). There are 4 billion possible random seeds and less than this number of users (so far...) so the chance is extremely small that work will be reproduced. Even though we plan to make 1 billion structures, assume that on average someone makes 1000 structures at a time, so that's only a million random seeds, far less than the 4 billion available

Greets,
Wirthi

[xman]

But my concern is that if I start many copies in a script, then the chances of using identical seed number would be increased.

[Jodie]

The increase would be miniscule unless the random number generator in your OS' libraries is broken. If you start two billion copies you'll likely have a single duplicate your self. And duplicate the work of a few other people. I think the statistical significance would be rather meaningless, though.

If you have two billion machines running offline, let Howard know - I'm positive knowing how attentive he is that he'd be willing to write a 64bit randomizer to get your two billion machines on the project. ;)

[MAD-ness]

LoL.

One of the funniest things I have read in a while.

I would not worry about the random seed, you get credit for the units even if they are duplicated by someone else, I believe, and the odds of duplicating them yourself are incredibly remote, from what I gather.

[Scoofy12]

I seem to remember Howard saying somewhere that your random seed was based on the PID of the process and the time - 2 of the most random things on your processor, and 2 things that will change even if you spawn a lot of copies.

[xman]

I see. That's better. I was worrying about it because I'm afraid that starting with the same time (same seed number) will generate duplicate copies, thus reduce the chance of getting good structures.

Jodie - you get what I meant ?

[Jodie]

Only thing I could think of would be kicking off a ton of copies and migrating them on a MOSIX (what I do) or PVM (rather tough) platform.

Unless they all fire in the same few milliseconds and with the same PID and probably using the randomize function on most systems, with the same hash'd content of memory - you're going to be pretty safe.

I mean, it's a perfectly valid (imho) concern. To date, I haven't seen any miss-credited units, although there may be some work going in that is invalid - and I'll never know. [shrug] Not one of those things that keeps me awake. ;)

:sleepy: