Very Prime Riesel and Sierpinski k

[Thomas11]

Originally Posted by pinhodecarlos

Sorry Robert but I won't.

I'm currently quite busy with the Sierpinskis, but at least I could help to do some sieving (I'm running ksieve with up to 100 Ks in parallel). Just let me know...

[pinhodecarlos]

Originally Posted by Thomas11

I'm currently quite busy with the Sierpinskis, but at least I could help to do some sieving (I'm running ksieve with up to 100 Ks in parallel). Just let me know...

Right now I remotely access my working machine where payam client is running due to the fact that it is a forget and run client and because I am not near the machine. For me it's more efficient to run payam rather than sieving and PRP. It's a lot of manual work that until late July I can't do it.

Carlos

[pinhodecarlos]

For M66 Riesel side I have something like 3 weeks to finish it from iteration 20 to 100. My question, is should I keep running it until iteration 200 or can I pick up M28 or M36 and take them too up to iteration 100? By looking at Robert's table M28 and M36 (Riesel side) are free.

[Thomas11]

Originally Posted by pinhodecarlos

For M66 Riesel side I have something like 3 weeks to finish it from iteration 20 to 100. My question, is should I keep running it until iteration 200 or can I pick up M28 or M36 and take them too up to iteration 100? By looking at Robert's table M28 and M36 (Riesel side) are free.

The program can only handle E>=52. The M28 and M36 Riesel VPS are found by an other (older) program and a lot of manual work (sieving + PFGW testing).

So, you should continue with E=66 (or perhaps E=60). But wait for Robert's comment...

BTW.: I could provide a modified version of the program which automatically stops at a given iteration/subiteration (I). I'm already using this on my Linux machines. It's quite useful when running multiple instances for the same E level. Just let me know if you're interested.

[pinhodecarlos]

Originally Posted by Thomas11

BTW.: I could provide a modified version of the program which automatically stops at a given iteration/subiteration (I). I'm already using this on my Linux machines. It's quite useful when running multiple instances for the same E level. Just let me know if you're interested.

Yes, I am interested. Just post it here, 64-bit windows version. Thank you.

[robert44444uk]

Originally Posted by pinhodecarlos

For M66 Riesel side I have something like 3 weeks to finish it from iteration 20 to 100. My question, is should I keep running it until iteration 200 or can I pick up M28 or M36 and take them too up to iteration 100? By looking at Robert's table M28 and M36 (Riesel side) are free.

Carlos - E66 is totally yours until you wish to change it, so no problems there.

I understand that this is a part time project for you, so don't worry about taking these values higher, someone will pick up the slack for sure, especially with the juicy results you have produced to date.

BTW taking M28 to 100 iterations would be a lifetime's work, there are far too many Payams at that level, and that level (and E36) proved not terribly productive for VPS although at low p counts they are superior. 99%+ of the really interesting candidates for us are E52 through E100.

[robert44444uk]

Originally Posted by robert44444uk

BTW taking M28 to 100 iterations would be a lifetime's work, there are far too many Payams at that level, and that level (and E36) proved not terribly productive for VPS although at low p counts they are superior. 99%+ of the really interesting candidates for us are E52 through E100.

Here is a table of payam number frequencies at each level, taken from my paper on VPS

E series	Payam # frequency per CRM calc on y in y*M(E+1)
10	1 in 9
12	1 in 20
18	1 in 35
28	1 in 235
36	1 in 895
52	1 in 26,802
58	1 in 30,339
60	1 in 31,123
66	1 in 33,886
82	1 in 278,551
100	1 in 20.4 million
106	1 in 28.6 million
130	1 in 130 million
138	1 in 537 million
148	1 in 981 million
162	1 in 10.6 billion

As each iteration covers 2.8*10^12 or so y, you can see that the E28 candidates needing checking for 1 iteration are approx 10^10 or approximatey 1 year on a lap top. Multiply by 100 iterations, and, yes, a lifetime give or take.

There are huge differences between 36 and 52 and again between 82 and 100

[Thomas11]

Here comes the 64bit Windows version with automatic stop feature.
Included is the source code as well as the two input files.

Note the two additional lines in the progress.txt file:

c -1
E 66
iteration 100
I 0
maxiter 100
maxI 200

If the two lines are omitted, the program behaves like before, e.g. runs until iteration=401776.

Perhaps you should first check it for a small interval (as given by the example file) before starting a long run.

[gd_barnes]

Originally Posted by Thomas11

The program can only handle E>=52. The M28 and M36 Riesel VPS are found by an other (older) program and a lot of manual work (sieving + PFGW testing).

So, you should continue with E=66 (or perhaps E=60). But wait for Robert's comment...

BTW.: I could provide a modified version of the program which automatically stops at a given iteration/subiteration (I). I'm already using this on my Linux machines. It's quite useful when running multiple instances for the same E level. Just let me know if you're interested.

Can the program be modified for E<52? It seems like it would be a fairly simple modification. At some point I would like to work on E28 or E36 and break some smaller P records.

[gd_barnes]

Originally Posted by pinhodecarlos

Sorry Robert but I won't.

I will test some of these higher over the next few weeks using the files that you have posted already.

[gd_barnes]

Originally Posted by robert44444uk

BTW taking M28 to 100 iterations would be a lifetime's work, there are far too many Payams at that level, and that level (and E36) proved not terribly productive for VPS although at low p counts they are superior. 99%+ of the really interesting candidates for us are E52 through E100.

I was wondering...would it make sense to test E28 and E36 to something less like 5 or 10 iterations? IMHO, just because there are a lot of y's to test in 1 iteration doesn't mean we shouldn't test it extensively. Even though unlikely to break some of the larger P recoreds, we could break some of the smaller ones. I believe that is where we'll find some of the 13/13 or 14/14's that we've been looking for.

I would look forward to seeing this great program modified to handle all E>=10.