Beta 3 available

[Brian the Fist]

I have completed making the changes to the client from our phase 2 beta test, and the 3rd beta is ready to download. It is the same place as the last (see end of this message). Please see the readme.txt enclosed to see what has been changed, and try to test these features specifically now.

It was also necessary to wipe out the beta 2 database and start clean again for testing purposes, so I hope you didn't get too attached to your structures from the first 2 betas. The first 2 betas will no longer work now, sorry.

For the moment all uploads will be rejected but this will be fixed momentarily. Also, the screensaver is now available too so please test this if you can. Post all bugs in THIS thread now please. This will hopefully be the final beta. Thanks!

To install, download the following package for your OS and overwrite files with the same name in a 'standard' distributed folding client. Download it fresh from the main website or use a recent copy you have. For the windows screensaver, be sure to put Foldtraj.scr wherever it currently is - c:\winnt\system32 or something like that usually.

Files are at:

ftp://ftp.mshri.on.ca/pub/distribfold/download/distribfold-beta-linux-i386.tar.gz

for Linux

ftp://ftp.mshri.on.ca/pub/distribfold/download/distribfold-beta-win9x.zip

for Windows text client

ftp://ftp.mshri.on.ca/pub/distribfold/download/distribfoldss-beta-win9x.zip

for Windows screensaver

[Brian the Fist]

An important note is that we are now using our energy function to choose which structure to follow for the next generation. This is more of a 'blind test' since we are pretending we do NOT know the native structure of the protein. Thus it is likely we will not see as low an RMSD as we did for the first betas. However, if the energy function works well (as we hope) we may get structure approaching as good as before. The only way we are still 'cheating' now is in choosing which of the generation 50 structures is best. But how we get to generation 50, unlike before, is now done completely without knowledge of the true structure. (Hope thats clear to everyone).

[Pascal]

Originally posted by Brian the Fist
An important note is that we are now using our energy function to choose which structure to follow for the next generation. This is more of a 'blind test' since we are pretending we do NOT know the native structure of the protein. Thus it is likely we will not see as low an RMSD as we did for the first betas. However, if the energy function works well (as we hope) we may get structure approaching as good as before. The only way we are still 'cheating' now is in choosing which of the generation 50 structures is best. But how we get to generation 50, unlike before, is now done completely without knowledge of the true structure. (Hope thats clear to everyone).

Looking at the project's plan, we're still in phase Ib - ain't we? so, there won't be any true blind test at all.
I can't tell you anything about, wether the stucutes of the 50th generation will be the nearest to the nature's given or better than them - I don't know.
But for the betatest would it perhaps be possible to decrease the time for actualising the stats? You may have an advantage of this. I've already seen, that the structures of generation 0 are not counted - but that's no problem at all. ;)
Shall we again check every parameter set?
Actually the client has

-it -df -qF -rt

under SuSE Linux version 8.1 - it does work properly.

[Brian the Fist]

Like I said, if possible please look through the changes in the readme, and especially if you were the one you suggested it or pointed out a bug, check if it's fixed now. It should not be necessary to test this 'from scratch' again, just please check the things that have changed and the bugs that were supposedly fixed.

and hopefully someone will try EVERYTHING on the screensaver...

[mighty]

Running Windows 2000
Switches: -rt

When I started the new beta3 for the first time, it made a few structures and then changed from "0 gen. buffered" to "1 gen. buferred". How can it buffer a generation when it's doing its very first generation ever?

[Aegion]

Originally posted by mighty
Running Windows 2000
Switches: -rt

When I started the new beta3 for the first time, it made a few structures and then changed from "0 gen. buffered" to "1 gen. buferred". How can it buffer a generation when it's doing its very first generation ever?
I've noticed the same thing. This 1 group of structures continues to be buffered even when I've completed and uploaded multiple other generations.

Update: After processing some additional structures it apparrently finally got uploaded.

[Guff®]

Have the text-client BETA running on WinXP and WinME systems, currently all switched as -qf -it -rt.
Nothing to report other than the quick 1 gen. buffer previously noted.
Does the "Data uploaded to server - continuing" dialogue still remain on the screen longer than it needs to, or is it actually taking that long to get all the bits uploaded?

[Brian the Fist]

Originally posted by Guff®
Have the text-client BETA running on WinXP and WinME systems, currently all switched as -qf -it -rt.
Nothing to report other than the quick 1 gen. buffer previously noted.
Does the "Data uploaded to server - continuing" dialogue still remain on the screen longer than it needs to, or is it actually taking that long to get all the bits uploaded?

The '1 generation buffered' is normal (though evidently confusing to you.. And the message says 'data uploaded', as in past tense. When you see that message, the data is accepted already by our server. It just stays on so you don't blink and miss it...

[AMD_is_logical]

Originally posted by Brian the Fist
Like I said, if possible please look through the changes in the readme, and especially if you were the one you suggested it or pointed out a bug, check if it's fixed now. The problem of excessive network trafic to my NFS server has been fixed. :thumbs:

So far, no problems. :thumbs:

[m0ti]

I see that it also resumes folds exactly where it left off for gens 1-50. :thumbs:

[Michael H.W. Weber]

I just launched the 3rd DF beta client. First of all I'd like to congratulate Howard for the obvious improvements made (evident in terms of rapidly progressing decrease of RMSD) which also reflects that all our computation efforts were worth the investment of time & money. :)

Here comes the problem:

1. For the first generation, 500 structures have to be calculated - which I did. After uploading, however, the stats display only 40. How is this possible? I see that this happened for others, too. I am aware that there was a discussion concerning the crediting of the first generation. If it is not credited at all, then why giving a score of 40 structures? Or is there some score weighing formula?
2. dfGUI (in its latest Windos-based version) reports the user's handle as an "invalid integer number". :rotfl:
3. I am a little annoyed that the previous beta stats were lost. I think these should be kept, too, since they indicate which people are really ready to help develop the client and not only run the stable stuff. Of course I won't cry over it the whole day now... :D

Michael.

P.S.: Hope this is news, since I had not taken the time to read all the stuff above. If not - just forget what I have written. :D

[edit]: The second generation was scored with EXACTLY 28 structures (20 were calculated as indicated by DF client) - why ever...

[Brian the Roman]

Is it just my imagination or is there evidence that the luck factor seems to have been largely removed by the betas?
Before, there seemded to me that there were more cases of users who's total # of structures was more out of line with their best result. If luck were completely removed then one would expect that one's rank in # of structures would always match one's rank in terms of RMSD. It looks to me like we're a little closer to that now.

ms

[Pascal]

At first, I wanna say a big Thank you very much!
Great work is done up to now, Howard :thumbs:

The client didn't make any probs on my system up to now (SuSE Linux 8.1)

I have another idea, which would make the client's behaviour a bit more logical - there is one point, that's not so lgical at all.

In our team, we discussed about the structures, that get stuck and restarted after three minutes.
But - in comparisation of different CPUs and client options - there are huge differencies in the computing speeds, that result in using a barrier of 3 min as a break.
So - if someone has a slower CPU, the client's behaviour maks him a loser.
Also someone is a loser, if his CPU works on another client.

Don't get too shocked - we have thought about a solution:
The client should count the CPU cycles during running the client - or makes a little benchmark, that defines a timestep y.

Now the client stops the stuck structures after x CPU cycles or - after a timestep y.
So - in this way - less CPU cycles are wasted. If you have a faster CPU - you may win with this in comparison with a slower machine a little bit - but in gerenal - it's an enhancement of the client's speed. ;)

[Digital Parasite]

Originally posted by Michael H.W. Weber
2. dfGUI (in its latest Windos-based version) reports the user's handle as an "invalid integer number". :rotfl:

Michael, I have stated several times that dfGUI doesn't understand the new beta and will produce strange results. dfGUI uses the progress.txt and filelist.txt files to gather information and those have been drastically changed for the beta client.

Software is sometimes backwards compatible but very few can predict the future. ;)

I am working on a new version of dfGUI that will support the beta client.

Jeff.

[Brian the Fist]

Originally posted by Pascal
In our team, we discussed about the structures, that get stuck and restarted after three minutes.
But - in comparisation of different CPUs and client options - there are huge differencies in the computing speeds, that result in using a barrier of 3 min as a break.
So - if someone has a slower CPU, the client's behaviour maks him a loser.
Also someone is a loser, if his CPU works on another client.


We're a step ahead of you - the 3 minutes is not the limit anymore - it now 'times out' after 3 minutes or about 250000 tries (to place a residue), whichever comes first. On a 2GHz machine, these turn out to be about the same. The 250000 tries was carefully chosen based on previous observations and is not arbitrary. It will not be lowered for any reason as soing so will damage potential results.

Anyways, sounds like it's mostly bug free now (knock on wood) - anyone tried playing with the screensaver yet?? Pretty please?

[IronBits]

Where are the stats pages for this Beta?

[IronBits]

Originally posted by Brian the Fist
Anyways, sounds like it's mostly bug free now (knock on wood) - anyone tried playing with the screensaver yet?? Pretty please?
I hate it when you do that. :p
I started it up and it went right out and got an update. I hope it's still the BETA... maybe update the TEXT on the screen to indicate it's a BETA. So far so good otherwise.
I had to install the regular screensaver, then overwrite it with the BETA client *.scr and the other two files that were in the archive.

[AMD_is_logical]

Originally posted by IronBits
Where are the stats pages for this Beta? Go to: http://beta.distributedfolding.org/index.html
Then click on "teams" (to find your team) or the "Stats" button (to get the stats summary).

[IronBits]

When I double click on the .scr file, it complains about someone messing with the filelist.txt file.
I don't see an easy way to change my options... I hope you don't expect folks to mess with the registry just to make changes. :confused:
When I press Q it doesn't upload the results...
next time I start it, it uploads the work and gets more.

[Pascal]

Originally posted by Brian the Fist
We're a step ahead of you - the 3 minutes is not the limit anymore - it now 'times out' after 3 minutes or about 250000 tries (to place a residue), whichever comes first. On a 2GHz machine, these turn out to be about the same. The 250000 tries was carefully chosen based on previous observations and is not arbitrary. It will not be lowered for any reason as soing so will damage potential results.
..

Hm. Not very pretty, another question:

How big is the part of people with a 2 GHz or faster computer? Throw away these three minutes - how much can you do on a non-optimated 600 MC/s Pentium III within 3 minutes? I think, that's rather a little bit rubbish, if there is no time left.
Or not? :confused:

[PinHead]

2 Part request ---------

1.) What is the current points system? I am having a tough time telling if the dual processor system is duplicating work.


2.) Would it be possible to get a -da switch?

-da -1 would "die after" ( program exit ) the current complete set of folds were finished.

-da -some positive number would "die after" (program exit) the positive number of complete sets of folds were finished.


By complete sets I mean all of the sets involved before starting from scratch to pick new values to drill down into.

p.s. I know it is a personality flaw, but I just hate leaving half completed work when I have to stop the client.

[Mikus]

There may be a user perception problem once the beta goes live.

With the current software, the user sees the count of structures going up steadily. Stalls are rare and usually last less than three minutes. The user gets a sense of accomplishment from being recognized for all the structures he builds.

With the beta software, the user keeps seeing stalls all the time. Most take the full three minutes. There can even be stall after stall on the same structure. If structures are still counted as the measure of his contribution, the user can get a sense of frustration seeing all those stalls, since the count of structures now increases more slowly.

One possible way to restore the user's sense of steady contribution might be to recognize the user's INPUT to the project, rather than counting the user's OUTPUT (structures built). Suppose the count (suitably scaled) of ASCII characters drawn to the screen were used as a measure of participation. Now, even when looking at the screen during a stall, the user could SEE the contribution he was making.

mikus

[Pascal]

I think I have found a bug, perhaps it is already found and does not seem to be too serious.
I'm still on my machine listed at the bottom of this posting with SuSE linux 8.1 (KDE 3.1) as OS.

For several hours three or four clients did really well work at the same time (each one started in a bash shell). But several times (perhaps 5-6 times) I got this error message:

...
pascal@athlon:~/Projekte/df/distribfold> ./foldit
[foldtrajlite] FATAL ERROR: [000.000] Unable to find file rod9saoi_0_rod9saoi_protein_1_0000012.val; cannot continue - replace file and start again, or manually delete filelist.txt
Hit Return
...

I had a look at a kind of system monitor - there was no swapping at all. CPU load was on an acceptable level - also the temperatures if die and chipset. So - is there perhaps anyone with a linux system that can reconstruct this bug?

EDIT: This only happened with the -qt parameter.

[Brian the Roman]

I installed the ss and ran it for a little while. I'll keep going back to it once in a while. seemed to work fine except for a message about a corrupt filelist.txt. I went in and deleted it and everything was fine.

My suggestion is for the screen saver to delete the file itself and restart automatically. I am assuming the SS is intended for completely uninformed users. For example, my wife would have had no idea how to respond to the message. In other words, the SS should be idiot proof.:D

ms

[Pascal]

./foldtrajlite -f protein -n native -it -rt


hecking for newer versions...
[foldtrajlite] FATAL ERROR: [000.000] Unable to find file rod9saoi_0_rod9saoi_protein_1_0000012.val; cannot continue - replace file and start again, or manually delete filelist.txt


Client crached. Same situation as above - no swapping.
Temperatures are okay - indeed.

At that time - three clients were running - two with -it -rt, one with -it -rt -qt

[IronBits]

Originally posted by Brian the Roman
I installed the ss and ran it for a little while. I'll keep going back to it once in a while. seemed to work fine except for a message about a corrupt filelist.txt. I went in and deleted it and everything was fine.
ms The SS does not upload the work when it quits. By deleting the filelist.txt file, you didn't get credit...
It does upload it the next time you start it tho...
I had the same error message if I double clicked on the .scr file. ;(

[Insidious]

So I said... OK, let's give the SS version a try.

The install went fine (overwrote the regular files as instructed)

my play time consisted of using 'preview' in display ss settings to see how it worked.

I set it up to buffer more data, use more memory, and connect to the internet at will.

Display settings were the 75% screen size, 1024x768x32 with high detail.

I tried the various renderings and with each, the display was very
"twitchey" There is NO WAY I could have that jumpy thing in the corner of my eye when I'm trying to get some work done. Screensavers need to be very smooth and unobtrusive to be tolerated by me. That thing in the background has about the same effect as one too many cups of coffee.... :rolleyes:

next, as I used 'p' to pause it and played with the rotate, drag, zoom etc. I eventually came to the time for it to prepare for the next generation. Well The machine decided I no longer needed to be in charge and put up it's little "minimizing....." note and let me sit and wait until IT decided I could use my machine once again. That would NEVER do! You need to make this thing so that when I move my mouse or hit a key it IMMEDIATELY lets me get to work.

"Uh, yes sir... let me pull that information up for you.... yes, I realize you are my boss and in an extreme hurry, but I need to wait until my screen saver decides I can use the computer again"

UH, I DON'T THINK SO :swear:

Oh yeah, no lockups or error messages!

-Sid

[PinHead]

OK I must have done something wrong on the screen saver install.

Downloaded and installed the normal screensaver. Configured it and it kicked off and started folding. Unzipped the beta and let it overwrite the files.

Went to the display properties and set the time down to 1 minute. BTW any change to the ss using display properties looks like it's hung, but you just have to wait a while.

Waited for 1 minute and when it kicked off it downloaded an update. Now I have a black screen with what looks like a DOS box saying that the screensaver was updated successfully. It now filcker about every minute (screensaver timing) blanking the screen and coming back to the same black screen with a dos box.

Claims to have updated protein.trj and native.val.


What did I do wrong? The beta should not have picked up an update and the original was just downloaded so it should not have had an update. I say that even though I am positive I overwrote the original files.

[PinHead]

Screen Saver

The error log is showing a checksum failure on df_logo.bmp.

The machine above is a PIII-500 with 98SE on it.

[Insidious]

PinHead,

Be sure you unzipped the .scr file to the correct directory. It goes to a different place than the rest of them.

for my machine:

all files EXCEPT the .scr file go to the distributed folding directory under program files. (sorry already uninstalled and don't remember the exact name...)

the .scr file goes to windows/system32

-Sid

[PinHead]

Thanks Insidious!

I knew I did something wrong.

It was c:\windows\system for win98se.

Now I have the right engine on the right protein.:rotfl:

[IronBits]

DOH !!! Thanks for letting me know where to put that .scr file .
\winnt\system32
Now it shows I'm using protein 96aa, and it's faster :)
It is also in fullscreen, entire screen is black except for the TEXT and the picture of the protein as it moves around...

[IronBits]

Minimizing Energy of best structure pops up, and so does the task bar at the bottom. (I'm in preview mode)
And the mouse pointer is staying on the screen.

After I drag it around the screen, rotate it, zoom in/out etc. (which causes the mouse to look like a large + and disappears) until that "minimizing energy" box pops up again, then the tray shows up and the mouse pointer shows up again, for awhile before disappearing again.
I can see where you might want to have an option to put it back the to the default view. ;)

[Killerrabbit]

I am a newbie at Beta and I downloaded the screesaver since it would be best to see how it runs and see if I could break it. I have Windows 98 and my computer and a AMD chip but I can't remeber what it is I think its like a K-6 or something like that.

First off the downlaoding went smooth and that was good. I then unzipped the files wrote down my handle and double clicked on the screen saver icon to start it up. A weird thing with the screen going blank and then seeing my Windows in a small square in the right hand corner and I could still click on things and such. Note please make it possible to go straight to the screen saver and avoid that part, for people not knowing what it is might think the program didn't work.

I then got a thing to put my handle number in and I did thinking it would go striaght on. It almost did until I got an error message and now it won't let me back on. Here is the error message that it says it was: "
========================[ Feb 28, 2003 11:04 PM ]========================
ERROR: [001.008] {trajtools.c, line 95} CheckMD5: File not found .\df_logo.bmp
ERROR: [001.007] {saver.c, line 1452} File failed checksum .\df_logo.bmp

========================[ Feb 28, 2003 11:05 PM ]========================
ERROR: [001.008] {trajtools.c, line 95} CheckMD5: File not found .\df_logo.bmp
ERROR: [001.007] {saver.c, line 1452} File failed checksum .\df_logo.bmp

========================[ Feb 28, 2003 11:06 PM ]========================
ERROR: [001.008] {trajtools.c, line 95} CheckMD5: File not found .\df_logo.bmp
ERROR: [001.007] {saver.c, line 1452} File failed checksum .\df_logo.bmp
ERROR: [001.002] {rotlib.c, line 86} Unable to open input file .\rotlib.bin.bz2
FATAL ERROR: [001.008] {foldtrajlite2.c, line 2160} Cannot open rotamer library"

The bottom one is the one I got and now I can't get into the screen saver it says that I need to go and downlaod it again or seomthing like that. Could someone explain to me using layman terms what to do and what went wrong. Remeber I represent the common user that won't know a lot of computer jargon.

Thanks,
Wabbit

[PinHead]

I am not sure if I understand the screen saver features!

Drag to rotate:
Well movining the mouse of course turns of the screen saver.
Hold click and drag does nothing.

Rotate:
Shift and move the mouse does nothing.
Shift, hold click and move the mouse causes video memory corruption and a hard lock.

Zoom:
Ctrl and move the mouse does nothing.
Ctrl, hold click and move the mouse causes video memory corruption and a hard lock.

OS is win98se

Oddity that I observed. When attempting the above functions, the cusror shows as an arrow on some parts of the screen and as a text input I-beam on other parts of the screen.

-----------------------------------------------------

As a side note:
Maybe when the screen saver calculates energy and the trajectory, you could do away with the graphic progress bar and just use a text block progress bar. The constant flashing lets you barely make out what the progress % is , but you can't read what it is doing. I guess it is only doing 1 of 2 things.

[Insidious]

you have to hit 'p' first to pause the screensaver, then you can
do the drag, rotate and zoom stuff.

-Sid

[IronBits]

I didn't have to pause it before moving around/zooming...
using the BETA... and it blanks out the whole screen (windows 2000)

[Insidious]

Originally posted by IronBits
I didn't have to pause it before moving around/zooming...
using the BETA... and it blanks out the whole screen (windows 2000)

everytime I hit a key or moved my mouse, it stopped the SS and went to the desktop. How does your computer know if you are moving the mouse to play with the screen saver or moving your mouse to get to the desktop?

(I suspected I was missing something.... :crazy: )

[PinHead]

Pausing did the trick in win98se!

Thanks again Insidious:cool:

[PinHead]

Killerrabbit!

You may want to look at Insidious post that is just a few posts back.

The errors you are seeing are similar, all though not exact, to the errors I was receiving when trying to click on the foldtraj.scr file when it was located in c:\ program files\Distributed Folding Project.

Do a search for foldtraj.scr . The one under c:\windows\... or c:\winnt\... should be the corrrect location.

Then copy or unzip just the foldtraj.scr to that location.
(since the other beta files are already installed in the correct location)

You may or may not have to reboot for it to work correctly. On win98se, I always had to reboot after copying over the file.

[Killerrabbit]

Thanks Sir Pinhead but I just did a file search and I didn't fidn any foldtraj.scr that hadn't already been unzipped. When I try to restart the screensaver it tells me to re-download the client and overwrite all existing files, I will try this and see what happens.

Ni

[Killerrabbit]

Originally posted by Insidious
PinHead,

Be sure you unzipped the .scr file to the correct directory. It goes to a different place than the rest of them.

for my machine:

all files EXCEPT the .scr file go to the distributed folding directory under program files. (sorry already uninstalled and don't remember the exact name...)

the .scr file goes to windows/system32

-Sid



Thing with it seems that the .scr file went with the rest. I am again downloading the client and hopefully this time ti will work properly. I have to agree the screensaver is going to have to be made idoit proof. A question, should I download the latest copy of the screensaver, I don't mean the Beta version??

Ni

[Killerrabbit]

Still get the same Setup error saying that because of an earlier error it won't let me run the program and that I should download it again and start over. New plan of attack I will first download the latest client and then the Beta and overwrite those files and try it that way.

Ni

[Guff®]

Originally posted by Killerrabbit
New plan of attack I will first download the latest client and then the Beta and overwrite those files and try it that way. That's exactly how I did it and it worked fine. Good luck, Killerrabbit!

[Brian the Fist]

Clarification on screensaver

I seems some are having trouble with my instructions so lets try again on how to install the beta screensaver.

1. Download and install screensaver from normal website - if you already have it, run it while connected to the internet to let it auto-update to the latest version, if it is not already. If you have been using the screensaver before the beta, you will need to erase filelist.txt in the screensaver directory. If in doubt or if you had problems, uninstall it (control panel-> add/remove programs -> Distributed Folding Project), re-download a clean copy and re-install it.

2. Download the beta package above and unzip to a temporary directory.

3. move protein.trj and native.val and readme.txt to c:\program files\distributed fodling project, or wherver you installed the screensaver

4. Move Foldtraj.scr to the following directory: C:\winnt\system32 (NT 4.0 and Win 2000) or c:\windows\system32 (Windows XP) or c:\windows\system (Win98, WinME). There should ALREADY be a foldtraj.scr there, and you must overwrite it. If there's not foldtraj.scr you are either putting it in the wrong place or didn't install the initial screensaver.

Ok, now as for the bug reports, please keep the following in mind:

1. The screensaver display may vary from video card to video card. Because something looks weird, or flickers, or is messed up on your video card, it does not mean it is like that on other makes/models.

2. Thus it is important that you report ANY visual anomalies along with your video card make and manufacturer (and driver version if possible), as well as a screenshot if possible.

3. To help me help you, in all bug reports please remember to state version of Windows, video card model and driver version (if problem is display related), and the exact steps you took to receive the error. Remember I can't fix anything unless I can reproduce the problem myself (in most cases).

As for intended behaviour for those who haven't played with the screensaver before:

1. You can exit ANY time (even energy minimization) by moving the mouse with the buttons not pressed (unless you do it really slowly)

2. You can rotate the molecule by clicking and dragging

3. You can zoom / drag the molecule by holding shift/ctrl while click-dragging with the mouse

4. While paused, moving the mouse will not exit

5. The taskbar does show up during energy minimization (this is a known 'bug')

6. Anything that looks strange / flickery / unusual is probably a bug so see steps 1-3 for bug reports above.

For those who've already posted problems, first make sure you've installed the beta correctly (as per above) and then try to make your bug reports clearer for me as outlined in this message. Thanks in advance.

[PinHead]

Howard

Do I understand the above post correctly?

I should be able to rotate / Move / Zoom without pausing first.

[Akkermans]

Howard,

The issue which I reported in beta2, with the -qt option, getting a '[]======[]=====' as screen output has disappeared.
I have tested it on both WinNT and Win2k.

Cheers

[mighty]

Running Windows 2000 with -rt switch
Running on an Athlon XP 2100+ (1733MHz)


Originally posted by Brian the Fist
We're a step ahead of you - the 3 minutes is not the limit anymore - it now 'times out' after 3 minutes or about 250000 tries (to place a residue), whichever comes first. On a 2GHz machine, these turn out to be about the same. The 250000 tries was carefully chosen based on previous observations and is not arbitrary. It will not be lowered for any reason as soing so will damage potential results.

Sofar I have done about 32 generations on beta 3. The average time is about 1,5-2 hours pr. generation. On beta 2 it was about 20 minutes. Is this increase due to the new 250000 or 3 minutes threshold? I have timed a couple of "stuck" structures, and it seems to start over after about 3 minutes and 10 seconds every time.

Have anybody experienced a similar rise in time pr. generation compared to beta2?

Also a suggestion:
With the new non-crazy ASCII art, could you maybe put in some sort of counter, so you can see how many tries it has reached on the current structure? For the untrained eye it looks very stuck most of the time (although I know it isn't ;) )

[Killerrabbit]

Thanks for the clarification. I deleted all my files and am now staring from scartch will know in a little while.

About the Foldtraj.scr going to that Systems place in Windows. I checked last itme and there was no foldtraj.scr. The foldtraj.scr went along with the rest of the files of the Beta to its own folder in unzipped files. Am I doing something wrong?

I run Windows 98 don't know if it is SE though it came already installed on the computer.

Ni

[Killerrabbit]

It worked perfectly I am now up and running.

One complaint at this moment when runngin the screensaver all my options at the bottom are like paues, rotate, move, and drag. Where is the save/quit and the upload options? If this has already been answered then ignore this question and just say so because I think the above posts were mainly about the TEXT client.

Ni

[Killerrabbit]

In the middle of a protein it just seems to stop for a moment and then sometimes the screen blinks and that particluar protein starts over at 0 to 96 and then goes on to the next. Most of the time it will freeze for a few seconds and then continue.

Not all bad though. It seems to be running faster than the older screensaver version and its not trying to do 5000 WU only 500 or I might have mistaken this project for another if I have my apoligies. Sorry about so many posts but I post when I find out something and try not wait to long after the fact.

It should tell you when you start the screensaver that it uploading work if you are already connected to the internet if you have a modem. Also if the program doesn't know you are connected to the internet then it should or there should be a seperate icon that says you can upload results so you don't have to start the screensaver.

Ni

[AMD_is_logical]

The client seems to be stable now, so the time has come ...

... to CHEAT! :crazy:

After all, if I can figure out a way, anyone can. :p

So I stopped the client at structure 20 of generation 9 and made a backup of the directory. I then let the client crunch through generation 19.

Using the struct 20 gen 9 backup, I tried completing the last structure and uploading. After several attempts, I had gotten no points for doing this. :thumbs:

Then I let it crunch and upload gen 10. I got standard gen 10 credit (63 points). :eek:
There was a warning in the error.log file, but that didn't keep me from getting points.

I let it crunch a couple more generations, but I got no more points. There was a structure compression error in the log for these upload attempts.

Going back to the backup, I again got 63 points for gen 10, (and none for 11).

A third time, and I yet again got 63 points for gen 10. This time there was both a warning and an error in the log:

ERROR: [000.000] {foldtrajlite2.c, line 3641} Warning during upload: STATUS 910 MISSING PREVIOUS OR ILLEGAL GENERATION

ERROR: [000.000] {foldtrajlite2.c, line 3718} Error during upload: STATUS 906 STRUCTURE COMPRESSION ERROR

The server seems to have seen that something was amiss, yet it still gave me points. :confused:

Getting credit for generation 10 again and again is a bug. If the same is true for a higher generation, then someone could crunch only that generation over and over, getting lots of points but not contributing to the science.

[Killerrabbit]

I believe I have a bug to report about the screensaver. I was planning to do an edurance run through the entire night. I started at about 9:50pm and I came into check on it at 10:38pm PST. I saw a little box that said there was an error and that the program was going to close; or as I call it the little blue box of death. I then clicked on details and copied it and then click closed. My screen went black and nothing came up and so I waited for a reasonable amount of time (4 minutes) and then rebooted. Sorry but for some reason I lost the copy of the details I will try to get it next time I am sorry. It happened during the first generation at about protein 132 or so for me. Should I try to run the screensaver on an edurance run or wait and just do little things? I will probably try it again one more time tonight. If I get it again I will write it down hard copy. Forgot to mention I run Windows 98 and like a AMD K-6 processor or soemthing like that, sorry I don't know my video card or driver I will try to find it by the end of the week.

Ni

[AMD_is_logical]

I think I've found a rather nasty bug.

I had a node no-netting and it had a lot of work spanning sets of generations. When I tried to upload that work, it had fatal errors.

The errors seemed related to a file that started with my handle and ended with ".bz2". There was only one such file in the directory, and its name showed it had been created during the earlier set of generations.

My guess is that when the client started a new set of generations, if failed to make a new <handle>...bz2 file, and instead kept using the old one. It then found a better structure, and put it in the old <handle>...bz2 file. That file then had a name from an earlier set of generations, and a structure from a later set of generations, which made it appear corrupt.

[Killerrabbit]

I tried the endurance run over the night abd I give it and F overall. Only reasson it got an F is that sometime over the night the program completely froze. I mean completely froze there was nothign that said what was wrong with it and it wouldn't even exit so I had to reboot. I will not try another endurance run until I read the board tonight. Again I run Windows 98, I have a AMD K-6 processor or something like that I still don't know what my video card or driver is. Could someone please let me know where I might find that information on my computer that would be most appricated.

Thanks

Ni

[grobinette]

Ok I ran this on a Win98se machine 650MHZ processor with 256Mg of ram using the -rt switch.

The lag in opening and closing applications during energy minimization was still present when the client was forced into high priority using TaskInfo.. But not nearly as bad as the last time.

When I set the priority back to normal during the same minimzation process it was much better - as in barely noticable. This is an improvement over the last beta, because with that one I could not even open the TaskInfo application to reset the priority.

So forcing a machine into using a high CPU priority is not such a good idea if you are also going to use the machine for work.

[Killerrabbit]

It looks like some more error messages but I don't really see how it is affecting the program. Here is the list it is from foldtraj_screensaver-Notepad:

"========================[ Mar 2, 2003 6:49 PM ]========================
ERROR: [001.007] {saver.c, line 1452} File failed checksum C:\PROGRA~1\DISTRI~1\df_logo.bmp
ERROR: [001.001] {foldtrajlite2.c, line 5307} Sun Mar 02 18:49:42 2003
Unable to check for new version, will try again later...

========================[ Mar 2, 2003 6:54 PM ]========================
ERROR: [001.007] {saver.c, line 1452} File failed checksum C:\PROGRA~1\DISTRI~1\df_logo.bmp
ERROR: [001.001] {foldtrajlite2.c, line 5307} Sun Mar 02 18:55:19 2003
Unable to check for new version, will try again later...

========================[ Mar 2, 2003 7:15 PM ]========================
ERROR: [001.007] {saver.c, line 1452} File failed checksum C:\PROGRA~1\DISTRI~1\df_logo.bmp

========================[ Mar 2, 2003 7:24 PM ]========================
ERROR: [001.007] {saver.c, line 1452} File failed checksum C:\PROGRA~1\DISTRI~1\df_logo.bmp
ERROR: [001.001] {foldtrajlite2.c, line 5307} Sun Mar 02 19:25:06 2003
Unable to check for new version, will try again later...

========================[ Mar 2, 2003 7:30 PM ]========================

========================[ Mar 2, 2003 7:47 PM ]========================
ERROR: [001.007] {saver.c, line 1452} File failed checksum C:\PROGRA~1\DISTRI~1\df_logo.bmp
ERROR: [001.001] {foldtrajlite2.c, line 5307} Sun Mar 02 19:47:34 2003
Unable to check for new version, will try again later...

========================[ Mar 2, 2003 7:52 PM ]========================

========================[ Mar 2, 2003 7:58 PM ]========================

========================[ Mar 2, 2003 9:50 PM ]========================
ERROR: [001.007] {saver.c, line 1452} File failed checksum C:\PROGRA~1\DISTRI~1\df_logo.bmp
ERROR: [001.001] {foldtrajlite2.c, line 5307} Sun Mar 02 21:50:47 2003
Unable to check for new version, will try again later...

========================[ Mar 2, 2003 9:55 PM ]========================

========================[ Mar 2, 2003 10:31 PM ]========================

========================[ Mar 2, 2003 11:03 PM ]========================
ERROR: [001.007] {saver.c, line 1452} File failed checksum C:\PROGRA~1\DISTRI~1\df_logo.bmp
ERROR: [001.001] {foldtrajlite2.c, line 5307} Sun Mar 02 23:03:10 2003
Unable to check for new version, will try again later...

========================[ Mar 2, 2003 11:08 PM ]========================

========================[ Mar 2, 2003 11:13 PM ]========================

========================[ Mar 2, 2003 11:18 PM ]========================

========================[ Mar 2, 2003 11:24 PM ]========================

========================[ Mar 2, 2003 11:29 PM ]========================

========================[ Mar 2, 2003 11:34 PM ]========================

========================[ Mar 3, 2003 9:43 PM ]========================
ERROR: [001.007] {saver.c, line 1452} File failed checksum C:\PROGRA~1\DISTRI~1\df_logo.bmp
ERROR: [001.001] {foldtrajlite2.c, line 5307} Mon Mar 03 21:43:25 2003
Unable to check for new version, will try again later...

========================[ Mar 3, 2003 9:48 PM ]========================

========================[ Mar 3, 2003 9:53 PM ]========================
ERROR: [001.007] {saver.c, line 1452} File failed checksum C:\PROGRA~1\DISTRI~1\df_logo.bmp
ERROR: [001.001] {foldtrajlite2.c, line 5307} Mon Mar 03 21:53:40 2003
Unable to check for new version, will try again later...
"

There it is. Also I am having problems trying to upload my data it seems I already have like 9 generations buffered I am just confused on how to upload them? Could someone answer that

Ni

[m0ti]

WinXP Pro
CLI version
switches: -g 2 -rt -df

My computer is still becoming quite unresponsive during energy minimization / trajectory distribution.

Perhaps you could place some yields in your code for this section (i.e. give up the processor if another application wants it)?

[Killerrabbit]

Windows 98
Screensaver version

It just froze again last night. When I mean froze I mean froze. Nothing was moving and it wouldn't let me exit so I just had to hit the power button on my computer. It hadn't even gone 30 straight miniutes before it froze. It was in generation 9 in the middle of structure 12 and was going mine for the first few minutes before it completely froze. I won't be running it much if it is going to be doing this.

Ni

[AMD_is_logical]

I saw an anomaly in the memory usage of the previous beta. I didn't report it at the time, but maybe I should have.

I was running several clients on a Win2k machine using the -rt switch. They were running online.

About the time the non-beta client was updated, the beta server went down for a while. My windows machine gave a warning message that it was running out of memory and was increasing the swap file size. I used task manager and saw that the beta clients were using twice as much memory as they usually do. Once the server came back up the clients returned to normal.

[KWSN_CutSack]

Ni!

Just FYI...running beta3, cmd-line, -rt -df, for 2 days now.

System: PIII 560 MHz, W2K, SP3, 256 MB SDRAM.

No problems what so ever :cool:

cheers,
KWSN_CutSack

[PinHead]

Screen Saver bug.

System has been running the screen saver without interruption for about 2 days straight.

Move the mouse and the screen saver stops folding but the screen does not go away and reveal the desktop. The screensaver is set for 1 minute and starts folding again before the previous screen goes away. Moved the mouse again and the same thing happens. Hit a key on the keyboard and the previously stoped screen saver screen goes away and reveals the desktop.

Also, moving the mouse to stop the screen saver and then right clicking causes video corruption. However, after the 1 minute, the screen saver kicks in again like nothing was wrong and folds away.

This is the same system I mentioned earlier when you were asking for video card info. So here it is:

P3 500
Win98SE
ATI All-in-Wonder 128 AGP
Video Driver Package 4.12.6269-WEB

This is a somewhat older video card with drivers from a time period when ATI was having difficulty with OPENGL drivers.

Does the screen saver use or require OPENGL drivers?

Also, since DirectX, affects ATI drivers:
DirectX 8.0 and DirectDraw 1.0 are in use on this system.

[Brian the Fist]

The screensaver is 100% OpenGL so buggy OpenGL drivers are often to blame for strange visual anomalies but we will look into all reports shortly. Thanks and please keep playing with it if you think there are bugs you still may not have found..

[Killerrabbit]

Same things happen with me when running screensaver. I move the mouse and it never goes back to the desktop. Haven't tried just pressing a key. Had to use the power button when it never made it to the desktop.

Actually I have found it frozen before I try to exit the program and I probably have the same video driver as PinHead. So how can I run it when it is already frozen after running it for 20 minutes and I don't even try to exit the program??

Ni

[Digital Parasite]

Would anyone be interested in helping me beta test a new version of dfGUI that works with the new beta client?

All: The new DF beta client acts much differently than the current one. Originally I could just count the number of structures produced and use the time it took to generate those to calculate the benchmark data (structs/sec,min,hour,day). The beta client now works in generations and at each generation it resets the structure count.

Right now all it can do is benchmark your current generation, the benmark getting reset when a new generation is started. Is that ok for people? Would you like to have Generation benchmarking information ie: time it took to complete the last generation? Average time per generation, that sort of thing?

Jeff.

[pointwood]

Suggestion: What about adding a benchmark WU to the client?

Running "foldtrajlite -bench" or something would run a benchmark with a special WU used only for that and never submitted to the server.

Howard?

[Tawcan]

I'm running the text client in Window XP. I just let it run and so far there doesn't seem to be any bugs.

Sometimes though, the client seems to be stuck on a certain structure for like hours. I would then hit "q", restart the client and the structure would get finished very quickly. I wonder why. :confused:

[Welnic]

I'm running the beta client on Debian Linux with just the -rt switch.

I start it up and foldtrajlite gets PID 602. It grabs about 81MBs of of memory and starts crunching. At some point this goes up to about 88. When it gets through with a generation, it spawns a couple of daughter processes, also called foldtrajlite, PID 607 and 608. All three grab about 24 MBs of memory, then 602 and 607 go to sleep. 608 crunches full time minimizing the energy. When that is over 608 quits, 609 starts and calculates the next generation trajectory. Then 609 quits, and 602 grabs about about 88 MBs and starts crunching the next generation. 607 also grabs 88Mbs and just stays asleep.

For the next end of a generation 2 more processes come alive and do the minimizing and trajectory, and 607 always just grabs the same memory as everyone else and does nothing.

[Brian the Fist]

Originally posted by Welnic
I'm running the beta client on Debian Linux with just the -rt switch.

I start it up and foldtrajlite gets PID 602. It grabs about 81MBs of of memory and starts crunching. At some point this goes up to about 88. When it gets through with a generation, it spawns a couple of daughter processes, also called foldtrajlite, PID 607 and 608. All three grab about 24 MBs of memory, then 602 and 607 go to sleep. 608 crunches full time minimizing the energy. When that is over 608 quits, 609 starts and calculates the next generation trajectory. Then 609 quits, and 602 grabs about about 88 MBs and starts crunching the next generation. 607 also grabs 88Mbs and just stays asleep.

For the next end of a generation 2 more processes come alive and do the minimizing and trajectory, and 607 always just grabs the same memory as everyone else and does nothing.

Is this a bug report or just an observation...
When threads are spawned, a complete copy of all allocated memory is made in some cases, regardless of what the thread is doing. Is this what you are seeing?

[pfb]

Something I've noticed with the Beta 3 is that the Minimising energy part seems to run at higher priority...if I'm watching a DVD or TV, playing CD or music or gaming during the Minimising part of the process I get brief judders as if DF is taking more CPU time. I can reproduce it by running the normal DF at priority 0 - which is why I am thinking it is running at a higher priority than the rest of DF.

As I say, only noticed it with this Beta - the others didn't seem to cause this behaviour...

System:

Windows XP Home SP1, Pentium 4 @ 2.7, 512MB DDR

[Welnic]

Originally posted by Brian the Fist
Is this a bug report or just an observation...
When threads are spawned, a complete copy of all allocated memory is made in some cases, regardless of what the thread is doing. Is this what you are seeing?

I consider it a bug. I don't have a problem with all of the threads having the same memory, especially since if you are running with the -rt switch when you have another thread minimizing or calculating the trajectory the total is less than the one thread that calculates the structures.

But there is a thread (607, in this case) that doesn't seem to do anything, it never has any cumulative time. If you kill it doesn't actually die, but the other threads don't get called. But while you are actually folding a generation it means that you are using twice the memory that you should be.

When you first start folding it is not there until the end of the generation. It seems to me that it should go away during the calculation of each generation and then come back to call the other two threads (if that is what it actually does).

[AMD_is_logical]

I have been trying to quantify a rare but serious BUG which I've now seen several times.

I am using the switches -rt -if.

A script periodically wakes up, shuts down the client by removing the .lock file, waits for it to terminate, runs the client with -ut, then restarts the client with -rt -if and goes back to sleep.

Here an example of what the bug causes in the error.log file when running with -ut:

ERROR: [000.000] {foldtrajlite2.c, line 3648} Error during upload, cannot open file: fold_6_<handle>_1_<handle>_protein_15.log.bz2
ERROR: [000.000] {foldtrajlite2.c, line 3641} Warning during upload: STATUS 910 MISSING PREVIOUS OR ILLEGAL GENERATION

Later runs with the -ut switch always fail with this error:

ERROR: [000.000] {foldtrajlite2.c, line 3718} Error during upload: STATUS 906 STRUCTURE COMPRESSION ERROR

Note that the middle number of the fold_*.bz2 file (that it says it can't open) is a '1'. So far that has been the case each time the bug has hit.

There was no fold_*_15.log.bz2 or fold_*_16.log.bz2 in the directory. There was a file fold_6_<handle>_7_<handle>_protein_17.log.bz2 .
Also, the only <handle>*.bz2 file was <handle>_6_<handle>_protein_16_0000013_min.val.bz2 .
The same was true each time the bug hit (with a different number instead of 15/16/17, of course).

The client will keep crunching (with -rt -if), but it never recovers consistency. All work after the bug hits appears to be unrecoverable. The only thing I've been able to do is to completely delete current work files and start over. That is why I consider the bug serious.

[Digital Parasite]

I am releasing the first beta copy of dfGUI v2.2 which supports the new BETA DF client. It will NOT work with the current regular release DF client.

I am hoping that those of you trying out the DF beta client will download and try this version of dfGUI to help me work out any bugs. It will also give you a chance to suggest features and or changes to the current version.

You can download the beta version here:
http://gilchrist.ca/jeff/dfGUI/dfGUIv22beta.zip

Jeff.

Here is a screen shot:

[mighty]

dfGUI beta seems to be working fine. I have a suggestion though:
Show the average time pr. generation.

Regards
Ole

[Brian the Roman]

Digital Parasite;
dfGUI beta seems to work OK. I'll play with it a bit tomorrow.

ms

[bwkaz]

Well, then, time to get working on upgrading the Linux port to 2.2beta. ;)

[Brian the Fist]

Originally posted by AMD_is_logical
I think I've found a rather nasty bug.

I had a node no-netting and it had a lot of work spanning sets of generations. When I tried to upload that work, it had fatal errors.

The errors seemed related to a file that started with my handle and ended with ".bz2". There was only one such file in the directory, and its name showed it had been created during the earlier set of generations.

My guess is that when the client started a new set of generations, if failed to make a new <handle>...bz2 file, and instead kept using the old one. It then found a better structure, and put it in the old <handle>...bz2 file. That file then had a name from an earlier set of generations, and a structure from a later set of generations, which made it appear corrupt.

Has this occurred again? Can you possibly give me a more detailed description - its not too useful as is. Give me an exact error message please, and what you did to receive it.

[Brian the Fist]

Originally posted by m0ti
WinXP Pro
CLI version
switches: -g 2 -rt -df

My computer is still becoming quite unresponsive during energy minimization / trajectory distribution.

Perhaps you could place some yields in your code for this section (i.e. give up the processor if another application wants it)?

Do me a favor and next time it is minimizing, open task manager and sort by CPU usage, then open some big application like outlook or excel or something and watch the task manager - or better yet, get a snapshot of it and post it. Frankly, I think you are just plain wrong here and I won't be convinced unless I see otherwise.

[Brian the Fist]

Originally posted by AMD_is_logical
I saw an anomaly in the memory usage of the previous beta. I didn't report it at the time, but maybe I should have.

I was running several clients on a Win2k machine using the -rt switch. They were running online.

About the time the non-beta client was updated, the beta server went down for a while. My windows machine gave a warning message that it was running out of memory and was increasing the swap file size. I used task manager and saw that the beta clients were using twice as much memory as they usually do. Once the server came back up the clients returned to normal.

Again please give numbers - what is 'twice as much as usual'? I cannot test this unless I know exactly what you mean.

[Brian the Fist]

I think I have enough info now to make the beta 4 client (but please answer those above 3 questions if you can). Thanks again for your thorough testing. Besides fixing the one or two bugs left (not counting the screensaver which Im still not too sure about - Ill try it more on Win98 I think), we will also be changing the algorithm a bit based on our observations. We will switch back to using the RMSD as the scoring functions, change the generation size to 10000 for gen. 1 and 50 for gen 2+, and will go for a total of 300 or so generations. This will allow us to see how far we can go before no improvement occurs any longer. It is clear form the graphs we've been making that 50 generations is not enough and we should go longer.

I'll start a new thread again later this week when the next release is ready and we hope you will continue to help us tweak the algorithm in this way before we release it to the teeming masses.

[AMD_is_logical]

Originally posted by Brian the Fist
Has this occurred again? Can you possibly give me a more detailed description - its not too useful as is. Give me an exact error message please, and what you did to receive it. I apologize for the earlier confused bug report. At that time I had only seen it once. I've now seen it a total of four times since the start of this beta. I described it better in my post above that starts "I have been trying to quantify a rare but serious BUG which I've now seen several times." Please let me know if you need more details than are in that post. I have some backups of the DF directory made after some of the bug hits.

[AMD_is_logical]

Originally posted by Brian the Fist
Again please give numbers - what is 'twice as much as usual'? I cannot test this unless I know exactly what you mean. I didn't write down the exact numbers, but when I go into task manager (This is on a Win2k system, so I ctrl-alt-del, then click on "Task Manager", use the "Processes" tab, and look at "Mem usage") I usually see around 80MB or so for the client. It went to around 160MB for a while when the server was down.

[Brian the Roman]

Howard;
you might want to make the # of generations dynamic, based on improvement over the previous ones. Clearly, not every generation is an improvement over its predecessor but if there's been no improvement for, say, 10 or 15 generations then you could give up. This would allow you to optimize the entire, how many gens do I go issue.

ms

[Georgina]

Originally posted by Brian the Fist
I think I have enough info now to make the beta 4 client (snip)

I'll start a new thread again later this week when the next release is ready and we hope you will continue to help us tweak the algorithm in this way before we release it to the teeming masses.

Will we have to download again or will the beta autoupdate?

G

[m0ti]

Originally posted by Brian the Roman
Howard;
you might want to make the # of generations dynamic, based on improvement over the previous ones. Clearly, not every generation is an improvement over its predecessor but if there's been no improvement for, say, 10 or 15 generations then you could give up. This would allow you to optimize the entire, how many gens do I go issue.

ms

I agree with that. It makes more sense to test for convergence than to assign an arbitrary cut-off. For example: start over at gen 0 if the improvement made over the last X generations is less than Y (where Y should be kept close to 0).

[AMD_is_logical]

Originally posted by Brian the Roman
Howard;
you might want to make the # of generations dynamic, based on improvement over the previous ones. Clearly, not every generation is an improvement over its predecessor but if there's been no improvement for, say, 10 or 15 generations ... You are making a wild guess at where the point of diminishing returns is. The primary purpose of the next beta is to find out where the point of diminishing returns is, and to do that it must go well beyond that point.

[Brian the Roman]

Yes, I believe that's why it's going out to 300 instead of, say 100. But, it also seems likely to me that the # of generations required may vary from protein to protein. So designing a fixed # of generations based on the requirements of this protein won't help long term. Doing it dynamically will help. The only thing you'd need to do is figure out (from a probabilistic perspective) the # of generations of no improvement to allow before you give up.

ms

[Brian the Fist]

As AMD says, we first wish to go way beyond, and then when we inspect the results at that point, we can make a better judgment about how far to actually go in generations with no improvement before giving up and starting over. Thus the 300 thingy is just a test for the next beta and by no means permanent.

P.S. I believe I fixed the cheating bug posted earlier allowing the same data to be uploaded multiple times for credit, please try to cheat again and let me know :)

[Brian the Fist]

Originally posted by AMD_is_logical
I have been trying to quantify a rare but serious BUG which I've now seen several times.

I am using the switches -rt -if.

A script periodically wakes up, shuts down the client by removing the .lock file, waits for it to terminate, runs the client with -ut, then restarts the client with -rt -if and goes back to sleep.
.

Try as I might, I cannot reproduce this. Are you sure there's nothing else you're doing that you forgot to mention. IS there a set of steps you can take to make the error happen, or is it random - i.e. run it with -if for X generations, the run it with -ut and it fails, or whatever? I need to be able to reproduce the error here on my machine if I am to fix it so I need exact step by step instructions on what to do - trying what you described above has not produced any error whatsoever for me.

[AMD_is_logical]

Originally posted by Brian the Fist
Try as I might, I cannot reproduce this. Are you sure there's nothing else you're doing that you forgot to mention. IS there a set of steps you can take to make the error happen, or is it random - i.e. run it with -if for X generations, the run it with -ut and it fails, or whatever? I need to be able to reproduce the error here on my machine if I am to fix it so I need exact step by step instructions on what to do - trying what you described above has not produced any error whatsoever for me. :(

It seems to be random. It can hit after a day, or after many days of crunching.

There are a couple of possible clues that may or may not be due to chance.

The first time the bug hit I was doing the -ut upload manually and infrequently. When I started using a script to do the -ut uploads, the far more frequent uploads did not seem to make the bug more frequent. From this I get the impression that the -ut upload is not causing the problem. That's just when the problem is noticed.

I have a node that has never been hit by the bug. This node has the -g0 switch in addition to the other switches. (That turns off checkpointing, right?)

I think I'll add a -g1 switch and see if that causes the bug to hit more often.

I'll keep trying to find out more about this bug.

[Brian the Fist]

Originally posted by AMD_is_logical
:(

I think I'll add a -g1 switch and see if that causes the bug to hit more often.

I'll keep trying to find out more about this bug.
Ok thanks. Its appreciated not only by us, but future users :D
If it helps at all, of those three numbers inthe filename, in case you don't know, the last number is the generation, the middle number is the structure number of the generation (for the log file, it is the number of the first structure counted in that particular log). The first number is an increasing unique integer, so that if you buffer more than 50 generations, the filenames are still guaranteed to be unique. It will increase forever or until you delete filelist.txt effectively restarting from scratch.

[Brian the Fist]

Beta 3 has now been superceded by Beta 4. Please post any future comments in the new Beta 4 thread. (This one was getting rather long anyhow..)