[QUOTE=bchaffin;233908][URL="http://www.sendspace.com/file/uv1rcz"]50M-52M[/URL]

[/QUOTE]

Got it!

Under windows:

80M to 88M is running on a quad-core where the siever gives 0.46 sec/rel per core.
88M to 92M is running on a quad-core where the siever gives 0.6 sec/rel per core.

So the first range would finish in less than 11 days and the second in less than 7 days. Am I correct?

Second range should already be done, the first one is at ~51%. Started the runs on 13 Oct 2010.

[QUOTE=em99010pepe;234098]Under windows:

80M to 88M is running on a quad-core where the siever gives 0.46 sec/rel per core.
88M to 92M is running on a quad-core where the siever gives 0.6 sec/rel per core.

So the first range would finish in less than 11 days and the second in less than 7 days. Am I correct?

Second range should already be done, the first one is at ~51%. Started the runs on 13 Oct 2010.[/QUOTE]I think where you're calculating incorrectly is using the [B]seconds per relation[/B] time to figure how long to get through a [B]range of special-q[/B].

Here's the way I calculate your time required:[code].46 secs/rel means .115 secs/rel total speed on 4 cores (.46/4)

For each sp-q we get ~2 relations, so for a range of 8 million sp-q
we get 16 million relations.

16M * .115 seconds = 1.84M secs.

1.84M seconds = 511 hours.

511 hrs = ~21.3 days.[/code]So if you're at >50% on the 8M range after 8 days, you're actually doing better than estimated. AFAIK, the speed reported is just "total relations so far" divided by "total time taken so far", so any other activity on the sieve machine would skew that.

Calcualting 8M relations at .15 secs/rel, I get ~13.75 days, so almost twice your estimate on the other range.

I understand, I forgot that in 1M range we have more than 1M relations. On half of the range 80M to 88M I am getting 1.5M x 4 relations found so far.

[URL="http://www.sendspace.com/file/ltayzs"]60M-62M[/URL]

I'll take 62M-64M next.

[QUOTE=bchaffin;234115][URL="http://www.sendspace.com/file/ltayzs"]60M-62M[/URL]

I'll take 62M-64M next.[/QUOTE]

Downloaded.

Relations needed: [B]~97M unique[/B]

Relations received so far: [B]46.9M unique (~48.4%)[/B]

[QUOTE=em99010pepe;234102]I understand, I forgot that in 1M range we have more than 1M relations. [/QUOTE]Not to worry. I miscalculated my estimates all the time when I started.....

I've got about 1M left on my range, so I should be ready to start uploading by the middle of next week.

88M-92M done. 80M-88M still in progress.

Taking 64M-65M.

Relations needed: [B]~97M unique[/B]

Relations received so far: [B]51.4M unique (~52.9%)

[/B](found 7823322 duplicates and 51394117 unique relations)

I'm ignorant on the matter, here a few questions.

Why for this c170 we need to get 97M unique relations?

I ran a few post-processing jobs for XYYXF:

c188 we got 52964343 unique relations
c182 we got 55559545 unique relations
c237 we got 86491103 unique relations

and for OddPerfect project:

c226 we got 102244572 unique relations
c208 we got 50301290 unique relations

So what's the rule? Different poly, different weight, type of integer?

Edit: What's the estimated time for post-processing this c170? 4-5 days?

[QUOTE=em99010pepe;234214]I'm ignorant on the matter, here a few questions.

Why for this c170 we need to get 97M unique relations?

I ran a few post-processing jobs for XYYXF:

c188 we got 52964343 unique relations
c182 we got 55559545 unique relations
c237 we got 86491103 unique relations

and for OddPerfect project:

c226 we got 102244572 unique relations
c208 we got 50301290 unique relations

So what's the rule? Different poly, different weight, type of integer?

Edit: What's the estimated time for post-processing this c170? 4-5 days?[/QUOTE]
Depends mostly on the size of the large primes. Collisions among the large primes/ideals must occur (on both sides of the relation), so the bigger the large primes, the more relations needed. Here we have large primes up to 2^30 on both sides, so to get enough collisions we need something close to 2*pi(2^30) unique relations. A bit less than this, since not all ideals will be used. Generally the bare minimum to get a square matrix is about 0.8 * (pi(2^lpbr) + pi(2^lpba)) for current problems. The bare minimum will get you a huge matrix, and sieving some more will make the matrix smaller.