n=390000 discussions and old reservations

[Historian]

Originally Posted by gribozavr

Yes, I'm working on 10G-150G, currently @ 46T. tpsieve needs 2.7 Gb of RAM. I started working on it when nobody talked about continuing previous search. Not much CPU time spent (a few days), so I can cancel it altogether if we decide that it is useless.

No, it's not useless; in fact I was thinking of picking up another range to LLR for this n. With that said, could you upload the sieve file for k<15M?

[gribozavr]

Originally Posted by Lennart

How does your tpconfig looks like ?

qmax=10e6 is not commented out. Everything else is the same.

Originally Posted by Historian

No, it's not useless; in fact I was thinking of picking up another range to LLR for this n. With that said, could you upload the sieve file for k<15M?

Too early for LLR. Currently sieving removes tens of factors per second.

[Lennart]

Originally Posted by gribozavr

qmax=10e6 is not commented out. Everything else is the same.

Ok. I run 2 tpsieve on my i7 920 both -t4 and I use ~85% cpu.

If I run -t8 it will go down to 70%.


Try raise nice.
priority=low to normal or 0

[TimSorbet]

Originally Posted by gribozavr

While we are talking about it, I want to ask a question. I've got a 4-core i5-750 and I start tpsieve with "-t 4". But top shows that it uses only 50% available time of each core. Why? RAM bandwidth is not enough?

Obvious question: Are you running any other math projects (or other programs) at the same priority that are taking the other half? Or is it actually idling?

# Sieve block size, in bytes. Should not exceed L2 cache size.
blocksize=512k

The L2 cache on my i5-750 is 4x256 KBytes (according to CPU-Z). Maybe your blocksize is too large to fit in your L2. Try setting blocksize to 256k or other values.
(I have blocksize set to 256k. When I run tpsieve with 4 threads it can use the equivalent of about 3.6 cores. I might be able to get this higher if I experimented more, but it's already certainly better than your performance of ~2.0 cores. blocksize is the most significant difference I have to your file that could be causing this. Another possibility is the priority, as Lennart suggested.)

[rajula]

I would like to work also on this project (at least for a while) and reserve 2.0M - 3.0M. I'll run it on one core so it should be done in about two days.

[Oddball]

Originally Posted by rajula

I would like to work also on this project (at least for a while) and reserve 2.0M - 3.0M. I'll run it on one core so it should be done in about two days.

I've added your reservation to the first post.

Welcome to the twin search, and happy prime hunting

[Lennart]

I shall do a test. I reserve 3M-10M

Lennart

[TimSorbet]

Originally Posted by Oddball

Stats:

Rank    Username    Amount sieved (in M)
1    Oddball        1.3
2    Mini-Geek    0.5
3    Historian    0.2

This isn't sieving. It's primality testing. You could replace "sieved" with "tested" or "LLRed" or "LLRd" or something like that.

[Oddball]

Originally Posted by Mini-Geek

This isn't sieving. It's primality testing.

Oops Fixed.

I'll reserve 10M-12M.

[Lennart]

Taking 12M-20M

Lennart