Sieve:- Pick A Range

[opyrt]

OK, Thank you very much for your answer! :-)

(And ofcourse, I understand that time is always an issue. :-)

[Joe O]

In the attached graph, factors found for the combined sieving effort are color coded as follows:
PSP
SB
SB for April 2009

The lighter colors in these graphs are the more recent results.

As you can see, the leading edge has reached 39558T (39.6P).
Caution, the range 12.6P to 13.0P is not yet complete. The following reservations are outstanding:

Min (G) Max (G) User Comment Reserved Last Updated 
12694450 12699450 alpha ReAssigned from jasong April 11, 2009 April 11, 2009 
12732850 12732950 ohiomike Block 4 February 7, 2009 February 7, 2009 
12732950 12733050 ohiomike Block 5 February 7, 2009 February 7, 2009 
12770850 12773350 woolysheep Work February 19, 2009 March 22, 2009 
12773350 12775850 woolysheep Work February 19, 2009 March 22, 2009 
12790600 12795600 woolysheep Home February 21, 2009 March 22, 2009 
12795600 12800600 woolysheep Home February 21, 2009 March 22, 2009 
12811100 12821100 Freeze_XJ 12-3-09 - still taking long ;) February 25, 2009 March 12, 2009 
12825950 12826250 Frentik S1 February 27, 2009 March 9, 2009 
12826250 12826350 Elrond[NL]  February 28, 2009 February 28, 2009 
12826350 12827350 Elrond[NL]  February 28, 2009 February 28, 2009 
12837350 12837550 Golden_Frog test run March 1, 2009 March 1, 2009 
12867050 12868050 Elrond[NL]  March 7, 2009 March 7, 2009 
12878450 12878550 Frentik S2 March 9, 2009 March 9, 2009 
12904050 12905050 steinrar  March 15, 2009 March 15, 2009 
12905050 12906050 steinrar  March 15, 2009 March 15, 2009

[Joe O]

In the attached graph, factors found for the combined sieving effort are color coded as follows:
PSP
PSP for May 2009
SB
SB since May 19, 2009

As you can see, the leading edge has reached 44290T (44.20P).
Caution, the range 12.6P to 12.9P is not yet complete. The following reservations are outstanding:

Min (G) Max (G) User Comment Reserved Last Updated 
12694450 12699450 alpha ReAssigned from jasong April 11, 2009 April 11, 2009 
12732850 12732950 ohiomike Block 4 February 7, 2009 February 7, 2009 
12732950 12733050 ohiomike Block 5 February 7, 2009 February 7, 2009 
12811100 12821100 Freeze_XJ 12-3-09 - still taking long ;) February 25, 2009 March 12, 2009 
12825950 12826250 Frentik S1 February 27, 2009 March 9, 2009 
12837350 12837550 Golden_Frog test run March 1, 2009 March 1, 2009

[/QUOTE]

[Joe O]

In the attached graph, factors found for the combined sieving effort are color coded as follows:
PSP
PSP for June 2009
SB
SB since June 15, 2009

As you can see, the leading edge has reached 48297T (48.297P).
Caution, the range 12.6P to 12.9P is not yet complete. The following reservations are outstanding:

Min (G) -Max (G)   User Comment 					Reserved 	Last Updated 
12694450-12699450  alpha ReAssigned from jasong. 58%. ETA Mid July June 16, 2009 April 11, 2009 
12811100-12821100  Freeze_XJ 15-6-009 : still enjoying the range :D February 25, 2009 June 15, 2009

[Joe O]

In the attached graph, factors found for the combined sieving effort are color coded as follows:
PSP
PSP for July 2009
SB
SB for July 2009
SB since July 16, 2009

As you can see, the leading edge has reached 51500T (51.50P).

[opyrt]

For those of you running manual sieve.

From the releasenotes for sr2sieve version 1.8.11:

Reorganised setup64() so that Legendre symbols are computed earlier. This allows an earlier short-circuit in the case that no sequences pass the quadratic residue test. Faster when there are few sequences in the sieve, but a little slower when there are many.

I asked Geoff if he could specify, and this is his answer:

Version 1.8.11 should be faster when there are 4 or fewer sequences, although it might not be much faster unless there are just 1 or 2 sequences. This might vary a bit depending on machine type or other things, but you should probably use version 1.8.10 for 15 sequences.

So for the combined Sierpinski sieve, version 1.8.10 should be the fastest. Just wanted to let you know.

[Joe O]

In the attached graph, factors found for the combined sieving effort are color coded as follows:
PSP
PSP since June 26, 2010

SB
SB since June 26, 2010

As you can see, the leading edge has reached 79438T (79.438P).

[Joe O]

In the attached graph, factors found for the combined sieving effort are color coded as follows:
PSP
PSP since August 2, 2010

SB
SB since August 2, 2010

As you can see, the leading edge has reached (80.75P).

[mdettweiler]

Originally Posted by Joe O

In the attached graph, factors found for the combined sieving effort are color coded as follows:
PSP
PSP since August 2, 2010

SB
SB since August 2, 2010

As you can see, the leading edge has reached (80.75P).

Just curious, has the optimal depth been calculated for this sieve? I would guess that it's a ways away for the entire n<50M range, but it would be interesting to know whether n<10M is entirely optimal by now. Since PrimeGrid does its own doublechecks alongside the firstpass tests, I imagine that it won't be too long before all first-pass and doublecheck tests are completed below 10M; once that's done, it should be possible to get a big speed boost by removing all n<10M from the sieve file.

[Joe O]

Originally Posted by mdettweiler

Just curious, has the optimal depth been calculated for this sieve? I would guess that it's a ways away for the entire n<50M range, but it would be interesting to know whether n<10M is entirely optimal by now. Since PrimeGrid does its own doublechecks alongside the firstpass tests, I imagine that it won't be too long before all first-pass and doublecheck tests are completed below 10M; once that's done, it should be possible to get a big speed boost by removing all n<10M from the sieve file.

There is no real speed boost when you cut off the lower n this. The smaller file helps somewhat, but not as much as you might think. I did try to get jjsieve to be more efficient starting at higher n, but there was still a lot of computation necessary to get to the starting n. Another saving was in the hash table, but it still was not enough to warrant skipping the lower n in the sieve file. This comes up time and again, and time and again I have people actually try it and then admit that the saving is not as great as they thought.

[VJS]

And who knows we could eventually knock out that n=991