upgrade to V25 on dual CPU - didn't get 2nd exp

[Jud McCranie]

I upgraded from version 24 to 25, and everything seemed to go well. I merged worktodo.ini and work0001.ini into worktodo.txt. But when the new version started, the second worker didn't have an exponent, it started on a new one. How can I restart the second exponent?

[mdettweiler]

Originally Posted by Jud McCranie

I upgraded from version 24 to 25, and everything seemed to go well. I merged worktodo.ini and work0001.ini into worktodo.txt. But when the new version started, the second worker didn't have an exponent, it started on a new one. How can I restart the second exponent?

Open worktodo.txt and you should see two sections--one labeled with "[Worker #1]" and the other labeled with "[Worker #2]". From what it sounds like, both of your original exponents are probably listed under Worker #1. Shut down Prime95, and cut/paste (note that I said cut, not copy) the respective line for your second original exponent into a new line directly after the line that says "[Worker #2]". This will put it at the front of Worker #2's testing queue, thus it will pause work on the new exponent that it grabbed, finish the original exponent, and then go back and pick up the new exponent at whatever progress it was at.

[Jud McCranie]

Thanks, that did it. I simply merged the two old ini files into the new txt file. If it said anything about "worker #1" I missed it.

[mdettweiler]

Originally Posted by Jud McCranie

Thanks, that did it. I simply merged the two old ini files into the new txt file. If it said anything about "worker #1" I missed it.

Probably that's because initially, the file was most likely blank (i.e. nothing about workers) and the "Worker #1" and "Worker #2" lines were added by Prime95 the next time you started it.

[Freightyard]

I am very thankful for the latest versions which support multiple workers, instead of configuring multiple copies of the software;

However, we're all going through the upgrade problems. I hope not too much work is lost or accounts abandoned due to the trouble.

[Jud McCranie]

Well, each of the ini files had one line in it and I made the txt file with those two lines. Didn't work though.

I was over 68% finished with the one that didn't show up - glad I didn't lose it.

I had the same problem as the person who started this thread.

I merged my worktodo files as instructed:

My original worktodo.ini file looked like:

Test=46928729,68,1

My original work0001.ini looked like:

Test=38009789,68,1

So, upon merging, as per the instructions, my new worktodo.txt file for version 25.8 was:

Test=46928729,68,1
Test=38009789,68,1

When I started running 25.8, my second exponent was lost, a new one assigned. My resulting worktodo.txt file became:

[Worker #1]
Test=F3AD1C66C9F4BA5B1AFB514795CFE82F,46928729,68,1

[Worker #2]
DoubleCheck=6EBC1590E69E0F917E33E289386AF5C5,21719573,67,1

Thanks to the responses to this thread, I edited my worktodo.txt file, to recover my second exponent, to look like this:

[Worker #1]
Test=F3AD1C66C9F4BA5B1AFB514795CFE82F,46928729,68,1

[Worker #2]
Test=38009789,68,1
DoubleCheck=6EBC1590E69E0F917E33E289386AF5C5,21719573,67,1

This kind of worked ... Worker #2 was now working on the correct exponent, but it was starting from 0.00%. I had lost the 50% completed work.

So, I copied the 2 save files for my second exponent from my old 2nd instance Prime95 folder into the current prime95 folder, allowing it to overwrite.

This let Worker #2 carry on from the correct point.


Now, since I was running the old Prime95 in the recommended way for a dual core Vista computer, using two prime95 folders with uniquely named exe's, set to different affinities, and I followed the software upgrade instructions as given, we must conclude that the developers have a serious lack of understanding in how to write instructions, they may not understand how their own software is being used, or their geek factor is just so high that they can't communicate effectively with the rest of us.

This must be fixed, as people upgrade the project will take a big hit in productivity from people not realizing they've lost weeks of work on each multi-core computer.

It also raises other concerns ... if this simple detail is screwed up, can we really trust that the actual prime number calculations are still working correctly?

I've been running Prime95 continuously on a few computers since 1997. At one time my modest contributions got me as high as 62nd on the top performers list.

Do you know how disturbed I was, when i went to the website and couldn't login, and couldnt see my account's results in the list!!!!!

Someone with real communication skills has got to re-write the website. At the very least it must state first thing that you won't see your past work until you create a V5 account and link to your old V4 account. The current bold headlines are pointless as they don't give the right message.

[cheesehead]

Originally Posted by Unregistered

I had the same problem as the person who started this thread.

< snip >

I generally sympathize with your complaints, but want to respond to a few points.

we must conclude that the developers have a serious lack of understanding in how to write instructions, they may not understand how their own software is being used, or their geek factor is just so high that they can't communicate effectively with the rest of us.

That's rather harsh. There are other possible conclusions.

(I presume you already appreciate some or all of the following factors, because you've been with the project a long time, and have given the developers some time to correct the flaws you mention. But I want to post this explanation for the sake of other readers, too.)

Before the hardware failure that took down the v4 server, v5/v25 alpha/beta testing had been proceeding systematically for over a year, with several intermediate goal achievements being reported to forum participants. There were active threads on this forum about the progress.

Because the changeover from v24/v4 to v25/v5 was planned to be several months later than it was, but had to be hurriedly done early because of the v4 server's hardware failure, not only the remaining alpha/beta testing of the v5 server and v25 client but also the writing and testing of transition instructions, with early feedback from a small group of dedicated testers before the new versions and instructions were released to the public, suddenly had to be compressed and performed simultaneously "in public", "warts and all".

Some aspects of the transition process could not have been finalized until developers had found solutions to certain difficulties encountered during testing, so public documentation such as transition instructions was, quite reasonably and understandably, not as advanced as software development and testing. There _was_ a solid plan for how the user transitions would be done easily, and if there had been time to finish and test the necessary software to accomplish that, user transitions would have been almost seamless! It was planned to have v24 client reports and assignments handled transparently via a v4/v5 interface to the v5 server. There would not have been any necessity for users to migrate their clients; there would not have been any necessity for much changeover instruction on the website -- only small notes about the availability of upgrades, similar to those posted in the past.

It has been argued to me that GIMPS administrators should have been able to bring up the v4 server software and data on the hardware system that was being used for v5 server development and testing, and that doing so would have minimized user trauma. I do not know enough to judge whether that was true.

Had the original schedule been able to be followed, the time and talents of the principal people could have been applied to those various areas in a more systematic and thorough way. Also, a larger number of part-time testers could have applied their talents to the portions of the projects where their special expertise lay.

It also raises other concerns ... if this simple detail is screwed up, can we really trust that the actual prime number calculations are still working correctly?

Oh, come on ... The basic computations were not being changed. It was the interface that was being revised.

I've been running Prime95 continuously on a few computers since 1997. At one time my modest contributions got me as high as 62nd on the top performers list.

... and I take your complaints seriously.

Do you know how disturbed I was, when i went to the website and couldn't login, and couldnt see my account's results in the list!!!!!

I and other long-time contributors all went through the same experience!

Someone with real communication skills has got to re-write the website. At the very least it must state first thing that you won't see your past work until you create a V5 account and link to your old V4 account. The current bold headlines are pointless as they don't give the right message.

I agree.