FAHClient V7.4.2 (Open-Beta)

Moderators: Site Moderators, PandeGroup

FAHClient V7.4.2 (Open-Beta)

Postby jcoffland » Fri Jan 24, 2014 11:36 pm

This release fixes the most significant bugs found testing v7.4.0.

Beta Downloads
Image

Documentation
Installation and user guides can be found here:
Pictorial Installation Guide (Windows) -> A detailed pictorial guide of the V7 installation.
Pictorial Installation Guide (Linux) -> A detailed pictorial guide of the V7 installation.
Pictorial Installation Guide (OSX) -> A detailed pictorial guide of the V7 installation.
FAHControl -> The Graphical User Interface (GUI) what controls the Slots.
FAHViewer -> It shows the protein being folded, if applicable.
Client Remote Interface -> Documentation for 3rd party developers.
Main Page -> Main page of the V7.

Getting Help
Aside from the documentation the best place to get help is in this forum. If you do have a problem post a message. There are many knowledgeable people ready and willing to help. Keep in mind, we greatly appreciate thorough reports delivered by patient people who can keep a cool head even when things go wrong.

Bugs/Tickets
Open Tickets Ordered by Milestone and Priority
Active Tickets by Change Time
Note: Some tickets may be closed because they are fixed in an upcoming internal release but are not yet fixed in the beta release.


Change Logs:
FAHClient:
v7.4.1:
  • Save config.xml on any changes.
  • Fixed QRB calculation, again. #1044
  • Report progress down to 0.01%.
  • Correct for cores which report incorrect frame/step. #888
  • OSX: workaround for slow communication on 10.9 as service. #1103
  • Removed trigger-save command, no longer necessary.
  • Only allow one visualization type per WU.
  • Update GPUs.txt when gpu=true, log message on update.

FAHControl:
v7.4.1:
  • Removed trigger-save command, no longer necessary.

FAHViewer:
No change

FAHScreensaver:
No change

Notes
This release addresses a number of related issues:

  1. The client will now save config.xml when it detects any change but no more often than once per minute and at application exit. This fixes several problems with remembering state between restarts.
  2. In v7.3.11, I attempted to fix QRB. In fact, I made things much worse. This has now been corrected and triple checked so hopefully the calculation is correct now.
  3. With some help from Calxalot I finally figured out why TFP and PPD oscillate on some WUs. It turns out that most of the cores have a problem where they divide the current and total steps by 1,000 and round down before reporting it to FAHClient. HFM.net is not susceptible to this problem because the core prints the correct values to the log. This causes oscillations for any project who's total number of steps is not evenly divisible by 100,000. For example, if the total steps is 250,000 then at %1 the current step will be 2,500. But the core divides this by 1,000 and rounds down so the client sees 2. Note, that the cores only report once every 1%. When this happens the client sees the following progression of reports: 0/250, 2/250, 5/250, 7/250, 10/250... So it computes the current percent progress as: 0%, 0.8%, 2%, 2.8%, 4%... I've worked around this by adding code to the client which attempts to detect this situation and correct for it.
  4. The failure to download at %99 was caused by the previous error. The client would see 98.8% and then at the end 100%.
  5. The FAHClient will now attempt to update GPUs.txt every 5 minutes whenever GPUs.txt is more than 30 days old. It will also print to the log when it does update. It is still necessary to restart FAHClient in order to configure any new GPU slots. This should help with the problem in Linux where the network is not always available when FAHClient first starts.

Notes
If you find any significant bugs don't panic. We'll fix them and make a new release soon.

You might be wondering why there are no ChangeLog entries for v7.4.2. There were only minor changes and some things went wrong with the v7.4.1 release that were most easily fixed by a new release.

Our immediate goal is to go public with the v7.4.x release ASAP. v7.4.x has a number of improvements over the current public release so the criteria for going public is very simple. Once v7.4.x has no major bugs that are not already in v7.3.x we will go public. So concentrate on bugs and think about features for v7.5.x.

Thanks again for your help. Beta testing Folding@home is another great way to contribute to disease research.

Joseph Coffland
Folding@home Developer
Cauldron Development LLC

Mod Edit: Updated Links - PantherX
Cauldron Development LLC
http://cauldrondevelopment.com/
User avatar
jcoffland
Pande Group Member
 
Posts: 1001
Joined: Fri Oct 10, 2008 6:42 pm
Location: San Jose, CA

Re: FAHClient V7.4.2 (Open-Beta)

Postby 7im » Sat Jan 25, 2014 12:42 am

Can I get a tarball version of this?
How to provide enough information to get helpful support
Tell me and I forget. Teach me and I remember. Involve me and I learn.
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: FAHClient V7.4.2 (Open-Beta)

Postby Fadamor » Sat Jan 25, 2014 12:46 am

Loaded it up instead of 7.3 and one change I noticed on startup is the following log entry:

Code: Select all
00:38:59:WARNING:WU01:FS01:Changed SMP threads from 12 to 11 this can cause some work units to fail


7.3 used to let me use all 12 threads. Is this change by design?
Image
Image
Fadamor
 
Posts: 312
Joined: Thu Mar 13, 2008 7:21 pm
Location: Northern VA

Re: FAHClient V7.4.2 (Open-Beta)

Postby 7im » Sat Jan 25, 2014 12:57 am

It will still let you use all 12 threads, but the installer sets cores to N-1 by default when a GPU is detected.

If not GPU folding, change the core setting on the CPU slot from -1 to 12.
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: FAHClient V7.4.2 (Open-Beta)

Postby 7im » Sat Jan 25, 2014 1:13 am

Hey, the PPD is way more accurate!!! Well done.
User avatar
7im
 
Posts: 14648
Joined: Thu Nov 29, 2007 4:30 pm
Location: Arizona

Re: FAHClient V7.4.2 (Open-Beta)

Postby Fadamor » Sat Jan 25, 2014 1:44 am

OK. I am GPU folding as well, so I'll leave it as-is.
Fadamor
 
Posts: 312
Joined: Thu Mar 13, 2008 7:21 pm
Location: Northern VA

Re: FAHClient V7.4.2 (Open-Beta)

Postby PantherX » Sat Jan 25, 2014 2:12 am

Fadamor wrote:Loaded it up instead of 7.3 and one change I noticed on startup is the following log entry:

Code: Select all
00:38:59:WARNING:WU01:FS01:Changed SMP threads from 12 to 11 this can cause some work units to fail


7.3 used to let me use all 12 threads. Is this change by design?

Please note that if the value is 11, FahCore_a3 and FahCore_a4 will automatically remap it to 10 since 11 is considered a bad prime number. You can check the log file to see the output from the FahCore itself.

Depending on the GPU Model and Driver version installed, you may or may not require 1 CPU to run FahCore_17 so do test it on your system and optimize it if needed.
ETA:
Now ↞ Very Soon ↔ Soon ↔ Soon-ish ↔ Not Soon ↠ End Of Time

Welcome To The F@H Support Forum Ӂ Chrome Folding App (Beta) Ӂ Troubleshooting "Bad WUs" Ӂ Troubleshooting Server Connectivity Issues
User avatar
PantherX
Site Moderator
 
Posts: 6321
Joined: Wed Dec 23, 2009 9:33 am

Re: FAHClient V7.4.2 (Open-Beta)

Postby netblazer » Sat Jan 25, 2014 2:59 am

Updated all 3 machines.

2 out of 3 machine had this error message during the update (and I had had it at least once during the previous beta) :


Failed to install files. Most likely some software, possibly Folding@home is currently using one or more files that the installer is trying to upgrade...... take a while after the application has closed.


The last file visible in the list was cufft32_30_14.dll.

I waited just long enough to take the snapshot and send it to paint before hitting retry and it worked right away on all machines. So maybe a 3-5 secs retry window before failing the installer???


Very small annoyance >> I always used the advanced GUI, never the web interface. So shooting me there after an update is out of place (perfect for a first install tho). I think I saw an accepted ticket about this but I'm not 100% on this one. Would be nice to remember the preferred interface and just resend you there.


Very big annoyance >> I have the client set to checkpoint every 3 minutes. I had a 8900 project running and it was right around 99.7% complete when I updated. After restarting fah it started back down at 98%. With 59 min TPF, it's really annoying. I was expecting to lose that last 40 minutes or so but not 2 frames...ish


I understand that it's not always safe to checkpoint the work but after more or less 100 minutes, that would mean that 33 checkpoints in a row were NOT savable. I find that incredibly hard to believe. Even if this is somehow only a 50/50 shot, that would make this a 8 billion to 1 odds... Still well over 1 Million to 1 if you ignore that last partial frame.

Not a show stopper, but not small either once you consider 100K ??? upgrades.
netblazer
 
Posts: 88
Joined: Sun Dec 22, 2013 8:36 am

Re: FAHClient V7.4.2 (Open-Beta)

Postby ChristianVirtual » Sat Jan 25, 2014 3:38 am

netblazer wrote:Very big annoyance >> I have the client set to checkpoint every 3 minutes. I had a 8900 project running and it was right around 99.7% complete when I updated. After restarting fah it started back down at 98%. With 59 min TPF, it's really annoying. I was expecting to lose that last 40 minutes or so but not 2 frames...ish

Core 17 (to my understanding) is not using your checkpoint setting; each project has its internal setting; for 8900 that is every other frame. So at frame 98 was the last valid checkpoint; when you restart at 99.7% you go back to 98.
ImageImage
Please contribute your logs to http://ppd.fahmm.net
User avatar
ChristianVirtual
 
Posts: 1535
Joined: Tue May 28, 2013 12:14 pm
Location: 日本 東京

Re: FAHClient V7.4.2 (Open-Beta)

Postby netblazer » Sat Jan 25, 2014 3:57 am

ChristianVirtual wrote:
netblazer wrote:Very big annoyance >> I have the client set to checkpoint every 3 minutes. I had a 8900 project running and it was right around 99.7% complete when I updated. After restarting fah it started back down at 98%. With 59 min TPF, it's really annoying. I was expecting to lose that last 40 minutes or so but not 2 frames...ish

Core 17 (to my understanding) is not using your checkpoint setting; each project has its internal setting; for 8900 that is every other frame. So at frame 98 was the last valid checkpoint; when you restart at 99.7% you go back to 98.



Seriously??

If so, how hard is it to put that little tidbit of info in the darn gui interface right next to the checkpoint setting? or at the very least say that not all cores use that setting with a link to an updatable wiki.

Stop hiding all that )(&(*. This software shouldn't be THAT hard to use!!!!!!!!!!!!

I'm a software programmer, DBA, I work with network admins and I have 40 000+ paid hours exp with computers and well, well over 100K hours of personal usage experience on top of that spanning almost 30 years.

Even with all that I find your stuff unbelievably annoying to use & understand. You think a NORMAL power user would go that far? NO.
Your mom? HELL NO.


Final Edit, I went back re-read the whole message. While wording would be slightly improved it's clear enough and I missed it. I still think a wiki link with short link like "Click here for list of Excluded Cores" (blue and underlined, would make this much clearer).
Last edited by netblazer on Sat Jan 25, 2014 4:07 am, edited 2 times in total.
netblazer
 
Posts: 88
Joined: Sun Dec 22, 2013 8:36 am

Re: FAHClient V7.4.2 (Open-Beta)

Postby netblazer » Sat Jan 25, 2014 3:59 am

Moreover, that WU reached 99% and, again, nothing downloaded. And I mean 99% reported in the logs.

I have 3 CPU n units finishing up today. I'll report on those later.

Project 8900 has 2.5M steps, so it shouldn't even have been affected by that non-100K steps thinggy.
netblazer
 
Posts: 88
Joined: Sun Dec 22, 2013 8:36 am

Re: FAHClient V7.4.2 (Open-Beta)

Postby ChristianVirtual » Sat Jan 25, 2014 4:23 am

jcoffland wrote:Fixed QRB calculation, again. #1044


looks really good ! very close to historical credits :D
User avatar
ChristianVirtual
 
Posts: 1535
Joined: Tue May 28, 2013 12:14 pm
Location: 日本 東京

Re: FAHClient V7.4.2 (Open-Beta)

Postby PantherX » Sat Jan 25, 2014 5:47 am

Fresh installation on Windows 8 and Ubuntu LTS without any issues. Everything looks good so far including the updating of GPUs.txt file on Linux.

Regarding the TPF, I had GTX 660 Ti and GTX 650 in the same system and they both downloaded different WUs from the same Project. It initially showed 0.11% until GTX 660 Ti reached 5% completion and used those values for GTX 650 until GTX 650 also reached 5%. After that, both Slots operate independently from each other.
User avatar
PantherX
Site Moderator
 
Posts: 6321
Joined: Wed Dec 23, 2009 9:33 am

Re: FAHClient V7.4.2 (Open-Beta)

Postby iBozz » Sat Jan 25, 2014 9:53 am

When will the MacOSX (Mavericks?) installation guide be ready. I don't mind beta versions provided I know exactly how to upgrade from the current version.
27” late 2009 iMac running under MacOSX 10.11.3 (15D21) El Capitan with 2.8GHz Intel quad-core i7 processor; 16GB 1067MHz DDR3 RAM; ATI Radeon HD 4850 512MB graphics
User avatar
iBozz
 
Posts: 181
Joined: Wed Nov 26, 2008 7:01 pm
Location: NW England, UK

Re: FAHClient V7.4.2 (Open-Beta)

Postby davidcoton » Sat Jan 25, 2014 11:08 am

Thanks Joseph

The PPD calculation is a big plus! Looks spot on now. :D
Pause persistence over restart now fixed, both Win Vista and Ubuntu. :D

Not so happy with the Web Control, no improvements and too confusing for general release :(
    The "tabs" are unclear
    The points total qualified by passkey -- different to every other occurrence of total points in the system
    The bug report being indexed by IP address -- not fit for purpose because ADSL does not assign fixed IP addresses

Everything else can wait for 7.5.x.

David
Image
davidcoton
 
Posts: 940
Joined: Wed Nov 05, 2008 3:19 pm
Location: Cambridge, UK

Next

Return to V7.4.x Public Beta Windows/Linux/MacOS X

Who is online

Users browsing this forum: No registered users and 1 guest

cron