Factor not recorded by GPU72

[LaurV]

Might this be the cause why my "pinch" gets P-1 assignments every time I use the proxy, even if the preferences all the way are for first-time-LL?

You may remember I posted some time ago about the fact that I stopped using the proxy, because it gives me different work type than I requested. If I don't use the proxy, PrimeNet is giving me LL work. If I use the proxy, I get P-1 work which is not registered with PrimeNet! but it appears registered with GPU72.

This problem is not solved yet, and I check periodically (by enabling the proxy, requesting work, getting a different type of work, killing the proxy, manually unreserving all the work). I did this two days ago last time, you can check the GPU72 logs. This is pain in the ass.

I don't really mind who gives me the work, the only difference is that if I don't use the proxy, then I don't get GPU72 LL credit (which I did not get for more that 3 months, you can see my graphics, no purple bars!)

Can you do something about it? [edit: i.e. see why my "pinch" is getting wrong type of work? This does not apply to other computers, only to this one, other computers get the right assignment, it may be something related to the amount of the memory - this has a lot allocated for P95 - or some special treatment from GPU72?]

[bayanne]

Since mfaktc has stopped working (as it causes kernel panics) I have just had to run P-1 factoring.

I have now noticed that another factor has been missed on GPU to 72.

[Mon May 12 10:37:33 2014]
P-1 found a factor in stage #1, B1=675000.
UID: bayanne/MacPro, M68838181 has a factor: 180472425589257237885593

I wonder how many exponents which didn't find a factor have slipped through ...

[kracker]

Originally Posted by bayanne

Since mfaktc has stopped working (as it causes kernel panics) I have just had to run P-1 factoring.

I have now noticed that another factor has been missed on GPU to 72.

[Mon May 12 10:37:33 2014]
P-1 found a factor in stage #1, B1=675000.
UID: bayanne/MacPro, M68838181 has a factor: 180472425589257237885593

I wonder how many exponents which didn't find a factor have slipped through ...

No, it didn't slip though, that factor is 77 bits.

[chalsall]

Originally Posted by bayanne

I have now noticed that another factor has been missed on GPU to 72.

OK, thanks for reporting this. As I did previously, I can add an assignment record for this P-1 work, so the factor shows up in your records. As before, the database does not show that you were assigned this, but it's again probably because your Prime95 client asked for DC work, but you were given P-1 (I haven't drilled down to confirm this, nor have I had the time to add the code which handles such situations).

It would help me if you could tell me when this was assigned to your machine (UTC please).

Originally Posted by bayanne

I wonder how many exponents which didn't find a factor have slipped through ...

I am *really* busy with real work at the moment (something has to pay for the hosting behind GPU72.com), but if this is really important to you I could possibly, sometime in the future, write a script which runs through the logs picking up such assignments and creating assignment records so that such "no factor found" P-1 results are credited to your GPU72 account.

Please keep in mind, however, that GPU72 is simply a facilitator -- Primenet's statistics are canonical.

[chalsall]

Originally Posted by kracker

No, it didn't slip though, that factor is 77 bits.

It did slip through. This is a bug.

The API interface between clients and Primenet is rather complicated; when the client asks for work, it doesn't actually specify what work type it wants at that moment. The desired work type is defined in separate messages before hand, and is further complicated by the fact there can be a machine-wide preference, and per-CPU preferences.

[bayanne]

I hate to be seen to be nit picking, and I know that real life is way more important, however I just did a tally of the 57 P-1 factorizations that I have done since Apr 26 on one box, and of those 37 were caught by GPU to 72 and 20 were missed.


I did do a switch from DC to P-1, rather than switching back and forward ...

[chalsall]

Originally Posted by bayanne

I hate to be seen to be nit picking, and I know that real life is way more important, however I just did a tally of the 57 P-1 factorizations that I have done since Apr 26 on one box, and of those 37 were caught by GPU to 72 and 20 were missed.

Please be comfortable... "Real Programmers" want users to report issues.

As Isaac Asimov once said, "The most exciting phrase to hear in science, the one that heralds new discoveries, is not 'Eureka!' but 'That's funny...'".

Originally Posted by bayanne

I did do a switch from DC to P-1, rather than switching back and forward ...

Yeah. This is possibly the issue.

Please keep bringing issues forward.

[TheJudger]

bayanne: could you please say something like "GPUto72 didn't recognize a reported factor" instead of "missed factor"? When I read the latter (especially in combination with GPUto72) I'm always afraid that this a bug in my code.

Oliver

[kracker]

Originally Posted by TheJudger

bayanne: could you please say something like "GPUto72 didn't recognize a reported factor" instead of "missed factor"? When I read the latter (especially in combination with GPUto72) I'm always afraid that this a bug in my code.

Oliver

Oh, so that's what it's about...

[chalsall]

Originally Posted by TheJudger

bayanne: could you please say something like "GPUto72 didn't recognize a reported factor" instead of "missed factor"? When I read the latter (especially in combination with GPUto72) I'm always afraid that this a bug in my code.

A very important point.

There is a difference between a missed factor, and factor not recorded.

[bayanne]

I have been keeping a tally, and so far of 72 (including 8 factors) results recorded on Primenet only 46 (including 5 factors) results have been recorded on GPU to 72.