Double Checking on GPU72

[kladner]

Originally Posted by davieddy

I feel a prime coming on

That would be quite a shock.

[davieddy]

Originally Posted by kladner

That would be quite a shock.

Probability per day = (tests per day)*(probability of error)/250,000

[retina]

Originally Posted by davieddy

Probability per day = (tests per day)*(probability of error)/250,000

Isn't there a factor missing: Namely the probability of a number being prime.

Probability per day = (tests per day)*(probability of error)/250,000 * probability of being prime

^{[sup][sup][sup][sup]Why is the colour tag missing the u?}[/sup][/sup][/sup][/sup]

[davieddy]

Originally Posted by retina

Isn't there a factor missing: Namely the probability of a number being prime.

Probability per day = (tests per day)*(probability of error)/250,000 * probability of being prime

^{[sup][sup][sup][sup]Why is the colour tag missing the u?}[/sup][/sup][/sup][/sup]

No.
What requires clarification is the LHS.
It is the probability of a shock, namely finding a prime.

As it stands it refers to Kieren's DCs.
For my LL it would be (1/50)/450,000.

Merry Christmas!

[axn]

Originally Posted by retina

Isn't there a factor missing: Namely the probability of a number being prime.

Probability per day = (tests per day)*(probability of error)/250,000 * probability of being prime

What is the role of the magic number "1/250000"?

Also, it is not "Probability per day", it is "Expected primes per day" (as in, with really large values, the result can easily be > 1)

[davieddy]

Originally Posted by axn

Also, it is not "Probability per day", it is "Expected primes per day" (as in, with really large values, the result can easily be > 1)

THX. An important distinction.
The expected primes per day for the whole project is 1/1500.
So the probability of one or more primes per day is 1 - e^=1/1500 = 1/1500.5

[davieddy]

Originally Posted by davieddy

The expected primes per day for the whole project is 1/1500.
So the probability of one or more primes per day is 1 - e^=1/1500 = 1/1500.5

@Phil: THX for allowing me to participate in the sort of mathsy fun which I enjoy.

Perhaps the 0.5 is worth looking into:

1 - e^-1/X ~ 1/X - 0.5*(1/X)²
= (1/X)*(1 - 0.5*(1/X))
~ 1/(X*(1 + 0.5*(1/X))
= 1/(X + 0.5)

Wish I had done it that way first time.
Damn those calculators

D