[QUOTE=Kevin;119763]The range is twice as large, but since the primes become more spread out the further you go, you don't necessarily have twice as many factors to test.[/QUOTE]

While technically what you said is correct, in practical implementations, it is highly inefficient to try to find the primes in such a high range.

Instead Prime95 will generate all the numbers in that range without small factors and use them to do the TF. Naturally, this means that P95 not only uses the primes, but also some composites to do the TF. It is faster this way.

The net effect is that the density of candidates used in TF remains pretty much constant, independent of the bit level being tested.

Sorry, I do pure math, what's this "practical implementation" you speak of? :whistle:

[QUOTE=Prime95;119768]Was this due to the fact that finding a factor saved you from testing all the 59-bit factors?[/QUOTE]

Forgive me as I still trying to understand how all of this works. I do agree that finding a factor did save me from testing all 59-bit factors. Guess what I was trying to say is if I would have quit testing this exponent to 2^58 bits, the server would have given me 0.0002 GHz-days credit but since I went into the 59-bit range and found a factor - the server only gave me 0.0001 GHz-days credit. Just seems for time credit I would have been better not finding a factor. To me this is sort of the same problem with credit that James Heinrich reported with F-PM1 and NF-PM1, so that is why I reported it.

[QUOTE=harlee;119785]...if I would have quit testing this exponent to 2^58 bits, the server would have given me 0.0002 GHz-days credit but since I went into the 59-bit range and found a factor - the server only gave me 0.0001 GHz-days credit.[/QUOTE]That should be true if you started from the same point. However, if you factored from 2^0 to 2^58 you should get more credit than if you factored partially from 2^58 to 2^59 and found a factor therein.

[QUOTE=harlee;119785]Guess what I was trying to say is if I would have quit testing this exponent to 2^58 bits, the server would have given me 0.0002 GHz-days credit but since I went into the 59-bit range and found a factor - the server only gave me 0.0001 GHz-days credit. Just seems for time credit I would have been better not finding a factor.[/QUOTE]

Unlike the last version of prime95, every time you complete a bit level it is reported to the server. Thus, you should get credit for factoring to 2^58 plus credit for finding the 59 bit factor.

However, TF-LMH blows through bit levels so quickly that I changed the client to only report every bit level after a point (I think its 60 or 61 bits). When I made this change I probably did not update the server's cpu credit formula. I will investigate.

Hi George,
It might be an idea to revert to the old reporting protocol for TF-LMH and the new one for TF. Since these are different worktypes, the server will probably be able to handle the difference, no?

[QUOTE=garo;119808]It might be an idea to revert to the old reporting protocol for TF-LMH and the new one for TF. Since these are different worktypes, the server will probably be able to handle the difference, no?[/QUOTE]

In fact, these are not different work types. There are 2 different concepts going on. One is work type - such as trial factoring, LL testing, ECM, etc. The other is work preference - such as LL world record, LL first time, TF, TF-LMH, ECM on Mersennes, ECM on Fermats. Different work preferences can return the same work type.

In theory, we can add more work preferences to the server without changing the client because it only handles a few known work types.

[QUOTE=Prime95;119807]However, TF-LMH blows through bit levels so quickly that I changed the client to only report every bit level after a point (I think its 60 or 61 bits). When I made this change I probably did not update the server's cpu credit formula. I will investigate.[/QUOTE]

Thanks. I'll take the blame (fault) for stopping prime95, changing the test_to_bit_factor without coordinating with you. I was wondering on what the credit values were for the different bit levels.

FYI, the point of reporting the bit level starts at 61 bits, which I noticed I factored some exponents to ^62 and ^63 bit levels. One exponent that I started to factor to ^63 bits, went to ^61 bits without a factor and the results were reported to the server and I got credit for it. The same thing for for ^62 bits. Then it started to ^63 bits then found a factor, reported it to the server and I got some credit for it. Sorry, didn't save of the credit value but seem to recall it was 0.0011 GHz-days.

I just pulled some ECM on Mersennes to factor and noticed I got some duplicate exponents. Just wondering why?

[CODE]ECM2=3FFFB7B027B7B46E6AA6473BF54F4A5C,1,2,194609,-1,50000,5000000,1
ECM2=D6FE058AB49069D4B0B441D4D42CB84C,1,2,194609,-1,50000,5000000,1
ECM2=98C57FEB601237BA1AF6BBF5E64A45A0,1,2,194609,-1,50000,5000000,1

ECM2=9A8A7E6DA19FF08DF898CA16FE2E1E34,1,2,209581,-1,50000,5000000,1
ECM2=7ADCF57AB961101A413A9E0A650C1ED0,1,2,209581,-1,50000,5000000,1
ECM2=78F004244F51367C7CB3A011275AC284,1,2,209581,-1,50000,5000000,1

ECM2=C6647358CDFA8BBBEA00F3EA044D09AB,1,2,248533,-1,50000,5000000,1
ECM2=82DD689FDC837A6515AB5B9193017962,1,2,248533,-1,50000,5000000,1

ECM2=805EE5066D0AE9DEB3EA07B343D135F5,1,2,305717,-1,50000,5000000,1
ECM2=B8F079435D3D226A1B61F62670C3F202,1,2,305717,-1,50000,5000000,1

ECM2=02C57834E983F74CE6A4E8B3D788D85D,1,2,305719,-1,50000,5000000,1
ECM2=B09DF9416F887B19D415020EC4FE559F,1,2,305719,-1,50000,5000000,1

ECM2=69EE3D62B21BBCA57301663A89969CB3,1,2,399523,-1,50000,5000000,1
ECM2=D2A5201E5866A627D4E84D2C52493FD6,1,2,399523,-1,50000,5000000,1
[/CODE]

[QUOTE=harlee;119854]I'll take the blame (fault) for stopping prime95, changing the test_to_bit_factor without coordinating with you. I was wondering on what the credit values were for the different bit levels.[/QUOTE]

No problem - that's what beta testers are supposed to do. I corrected the CPU credit formula on the server.

[QUOTE=harlee;119857]I just pulled some ECM on Mersennes to factor and noticed I got some duplicate exponents. Just wondering why?[/QUOTE]

A side effect of some changes I made on the server. This would have worked itself out eventually, but I've made another change which should reduce but not eliminate this in the near future.