Primenet web design

[Madpoo]

Originally Posted by Madpoo

Hmm... that's unfortunate. I'll have to look at that scenario. It's hopefully just a question of me adjusting the z-index on that header. There is an option for that somewhere I think and it should be less than what the drop downs use.

Yes, that was it... zindex option is adjusted and it should be good now.

[Madpoo]

I'm working on a way to export an XML version of the exponent report, such as this one:
http://www.mersenne.org/report_expon...exp_hi=&full=1

Here's the format I've come up with (for this example I truncated some of the history items just for clarity).

Comments or suggestions?

<exponents expStart="70000141" expEnd="70000141">
  <data exponent="70000141">
    <factoredToBits>72</factoredToBits>
    <PM1_Bound1>820000</PM1_Bound1>
    <PM1_Bound2>23780000</PM1_Bound2>
    <llResults>
      <llResult result="Unverified" residue="28A1D8DFCE5B35__" dateReceived="2011-10-27 19:35" userName="Lorenzo"/>
    </llResults>
    <history>
      <result dateReceived="2008-12-31 01:02" userName="starrynte" resultType="NF" resultText="no factors by trial factoring">
        <resultMessage>no factor from 2^64 to 2^65</resultMessage>
      </result>
      <result dateReceived="2010-09-07 08:50" userName="timbit" resultType="NF-PM1" resultText="no factors by P-1">
        <resultMessage>B1=820000, B2=23780000</resultMessage>
      </result>
      <result dateReceived="2011-10-27 19:35" userName="Lorenzo" resultType="C" resultText="LL test composite">
        <resultMessage>28a1d8dfce5b35__</resultMessage>
      </result>
    </history>
  </data>
</exponents>

[garo]

Hi Madpoo, thanks for all your excellent work. This has probably been mentioned before but is it possible to show the Recent Results report without TF-LMH results. They really clog it up.It would make sense to shunt the TF-LMH results off to a separate report. Thanks.

[Madpoo]

Originally Posted by garo

Hi Madpoo, thanks for all your excellent work. This has probably been mentioned before but is it possible to show the Recent Results report without TF-LMH results. They really clog it up.It would make sense to shunt the TF-LMH results off to a separate report. Thanks.

Right now that report is a "canned" one that runs on an hourly basis, just because of the amount of data included. I think displaying it ad-hoc with customized features for people would impact performance.

I guess it depends on what people use that report for... right now it's just the last 3000 things checked in, plain and simple.

If you want to see something more interesting, the "recent cleared" might be more up your alley, showing the 3000 most recently checked in things without the "no factor found" stuff clogging it up.

I don't know if that's quite what you meant? When you mention excluding the LMH results, did you just mean the "no factors found" check ins from them (there are a lot), or even when a factor *was* found, which I think most people would find more useful?

[kladner]

I think the issue may be that a high-results-volume range can effectively push a lot of lower-volume work off the display very quickly.

[snme2pm1]

Originally Posted by Madpoo

When you mention excluding the LMH results, did you just mean...

I remember once looking but failing to find a definition of that term.
I have several vague notions as to what it might have been intended to mean, but I suspect there are different classes of LMH workers.
Some explore shallow depth across a broad range, or intensive and becoming deeper in a narrow range, and variously within sight of benefit for near term LL testing or way out in the future.
It might be handy if there really were a definition in a glossary somewhere!
p.s. I recognise the literal acronym of Lone Mersenne Hunter, but that does not really correspond well to the circumstances that the LMH term has been used.

[TheMawn]

LMH is Lone Mersenne Hunters. It boils down to people looking for Mersennes outside the "conventional" zone, most of which are 100M digits (although I think there's some minor interest in 1B digits). There are a few LL tests done / being done but the vast majority is trial factoring.

If finding factors is your thing, then doing a bit-level first TF run through 100M - 1000M will get you a lot, fast. Current work is trial factoring tens of millions of exponents to 66 from 65. I did a few runs myself in the 400M range and I ended up going from 65 to 68 because 65 to 66 was too fast.

The amount of work in the LMH area might be small in terms of GHz but in terms of # of results it is massive, so the last-1000 reported results are often mostly from the upper ranges that most of us don't actually care about.


One possible solution is putting on filters for the results, one of which could be the exponent range.

[James Heinrich]

Originally Posted by TheMawn

The amount of work in the LMH area might be small in terms of GHz but in terms of # of results it is massive

Fortunately the work on exponents above 1000M is outside the purview of GIMPS; I'm currently doing TF on 2⁵⁰-2⁶⁴ and finding ~40000 factors per day.

[garo]

Originally Posted by Madpoo

Right now that report is a "canned" one that runs on an hourly basis, just because of the amount of data included. I think displaying it ad-hoc with customized features for people would impact performance.

I understand that which is why I proposed splitting the report in two. As others have pointed out the issue is the large volume of results in the TF-LMH range that means one usually cannot see more than a couple of hours of recent results. Recent Cleared is only partially useful as you miss out on NF results on the wavefront (70M ish range).

Hence I propose creating two "canned" reports. One that shows "Recent Results" for exponents < 100M and another for exponents > 100M. This should resolve most of the issues and give us more visibility on ranges where "actual" work is being done.

Any factors found or LL/DC results in > 100M range should continue to appear in the Recent Cleared report as currently.

[snme2pm1]

Originally Posted by James Heinrich

Fortunately the work on exponents above 1000M is outside the purview of GIMPS; I'm currently doing TF on 2⁵⁰-2⁶⁴ and finding ~40000 factors per day.

Yeah, from some prior comments I thought you would probably be well into that.
So when are you planning to open 10G for general craziness?

[Prime95]

Originally Posted by garo

I understand that which is why I proposed splitting the report in two. As others have pointed out the issue is the large volume of results in the TF-LMH range that means one usually cannot see more than a couple of hours of recent results. Recent Cleared is only partially useful as you miss out on NF results on the wavefront (70M ish range).

Hence I propose creating two "canned" reports. One that shows "Recent Results" for exponents < 100M and another for exponents > 100M. This should resolve most of the issues and give us more visibility on ranges where "actual" work is being done.

Any factors found or LL/DC results in > 100M range should continue to appear in the Recent Cleared report as currently.

I can see value in this idea. It is trivial to create two (or more) canned hourly reports. We could also output the hourly data to a temporary table rather than an ugly text report. From the temporary table, we could then produce pretty HTML tables.

Also, I don't see why a dynamic (instead of hourly) recent results report should create a terrible server load. There is an index on date-received, so SQLServer should be able to execute any recent results query efficiently (even a non-LMH report that tosses 90% of result rows because they are LMH results) -- we should run some sample queries to test that out.

I can't do the tests as I'm on a cruise ship and cannot access the server. Yes, you folks have failed to find a new Mersenne Prime while I'm on vacation -- most disappointing. You have a week left to rectify the situation