Messages

That rule was created automatically and it appears to match the name of the variable that was coded in the program.  ie: appliation (sic) is correct as it seems to work.

Not sure why David did that.  From BoincTasks I select the BOINC website to get to the forum or to check for latest updates.  After getting redirected it is a PITA to find those sites.

[EDIT]  After I asked about this at the boinc Q&A forum they fixed the redirect.  It now goes where it always went.  Hopefully they wont change it again.

After editing TThrottle.xml to get the order of my pair of HD-7950s temperatures to correspond, I decided to swap cards in respective pcie slots.  Unaccountably, I lost the temperatures for one of the cards.  The outer card was having its temperature reported for the card closet to the CPU.  The card closest was not being reported at all.  On a hunch, I pulled the HDMI dongle (dummy load) off the outer card and put it on the inner one and that fixed the problem after a reboot.  That is probably the problem I have on another ATI system that is not connected to any monitor.  I need to put an HDMI load on the "primary" card to force the driver to set an order. I am just guessing but clearly having a monitor or dummy load connected to an ATI card makes a difference in how the temperature is reported.  I have not seen this on nVidia cards on some of my headless systems with no dummy load or monitor.

[EDIT] - HMM--cannot do an attachment.  Dont see any way under Chrome.  I will put on my website
http://stateson.net/rules.zip maybe someone can tell me how to add a zip file attachment.

I need to implement a rule and I first tried 1.73 then the beta.  I was successful in implementing two somewhat similar rules, but there was a problem (bug) that I would like to report and also a feature request.

Some history:  Occasionally, one or another of my systems gets an nVidia or ATI kernel error and recovers usually immediately with no bad effects.  Unfortunately, one of my video board, an HD7950 occasionally does not recover.  I discovered that simply suspending and resuming any project on that board will fix the problem.  A sleep command for 10 seconds works fine and is easy to implement at the win10x64 command prompt with "boinccmd --set_gpu_mode never 10".  I thought I would implement a rule to do this for me.  This board processes MilkyWay tasks that take no more than 3 minutes.  At 5 minutes elapsed time there must be a real problem so I based my rule on MilkyWay.  I installed boinctasks on that system, made sure that history was enabled and configured and tested the rule which worked (after some time tinkering of course).  Attached to this post is a zip that contains the rule, and two command files.  One FixGPU.bat causes boinc to sleep, the other, CallMe.vbs sends me a text message about the problem and is called by FixGPU.

All seemed to be fine until I decided I wanted to do this from my main desktop system and not have boinctasks run on the dedicated boinc research system.  I ran to two problems.

1.  Feature request:  It is not possible to run a local program and also issue a sleep command to a remote system.  The bottom right corner box is where the sleep parameter goes.  That is the same box that is supposed to have the path to the program that I wanted to run so as to send a text message to my cell.

2.  Both 1.73 and 1.75 improperly handle the sleep parameter.  Where the rules.xml file has <ivalue0>300</ivalue0> that puts "00d,00:05:00" into the dialog box "value" which is correct.  I used 5 seconds for the time for the 5 minute MilkyWay threshold and I put 10 into the bottom right corner box.  Unaccountably, when clicking on "check" the 5 minutes is changed to 10 seconds.  One or more of the fields in the dialog box are accessed improperly.  After "OK" and exiting, if one does a file compare of the rules.xml with a backup you can spot that the 300 seconds was changed to 10 seconds which is incorrect.  I have tested my rules.xml file and it works but you must edit it to put in "yes" to activated it because if you ever bring it up in the rules editor dialog box it will get corrupted.  I am including two pictures that demonstrate the problem.

Here is my working rules.xml file

<rules>
<rule>
    <active>yes</active>
    <name>7950</name>
    <computer>z400-3-RX570</computer>
    <project>Milkyway@Home</project>
    <appliation>1.46%20MilkyWay@Home%20(opencl_ati_101)</appliation>
    <type0>5</type0>
    <type1>0</type1>
    <type2>0</type2>
    <operator0>2</operator0>
    <operator1>0</operator1>
    <operator2>0</operator2>
    <ivalue0>300</ivalue0>
    <ivalue1>-1</ivalue1>
    <ivalue2>-1</ivalue2>
    <dvalue0>0.000000</dvalue0>
    <dvalue1>-1.000000</dvalue1>
    <dvalue2>-1.000000</dvalue2>
    <itime>5</itime>
    <color>6569215</color>
    <event_show>1</event_show>
    <event_internal>7</event_internal>
    <event_external>0</event_external>
    <event_parameters>10</event_parameters>
</rule>
</rules>

This is what happens if it is brought up in the rules dialog box.  The file compare is under it.  Note that the 5 minutes was changed to 10 seconds.

Radeon driver update 18.1.1 fixed problem.  I was running 17.

I made that change in tthrottle.xml to get my second HD-7950 recognized but I have the same problem as my other pair of HD-7850.  The 2nd video boards temperature is not reported:  Instead, the first boards temperature is reported in its place.  By "first board" I mean the ATI board that is furthermost from the CPU.  Unfortunately, the one closest to the CPU always seems to run hotter and if that was reported "twice" I would be happy as I know the other one is always cooler (unless the fan stops of course).

Anyway, I can get the 2nd board's temp using Radeon software's overclock option or through gpu-z but neither of those programs report back to boinctasks.  Lemme know if I can help debug this.

Thanks for looking!

[EDIT] I also tried using 2 and then 1 for "inactive" but both of those options caused crashes and setting that ATI "inactive" to 0 stopped the crashes.

I replaced a Q6600 (2.4ghz socket 775) with a Xeon X5470  (3.33ghz socket 771) and the displayed temps were almost 15deg higher than actual.  Looking around, I found this list of Tj Max values for various Intel CPUs:
https://forums.overclockers.co.uk/threads/official-tjmax-figures.17936945
X54xx was 85 deg and when I put that value into TThrottle's tjunction setting the displayed temperatures dropped to where they matched AIDA64 and SpeedFan.

There might be a better list somewhere else, but the change from 100 to 85 allowed all BOINC tasks on my MS-7380 motherboard to run at full speed.

HTH someone!

"batch" files don't seem to execute, instead they show up in notepad.  ie:  putting "beep.bat" in that command window seems, at least on my win7x64 system, to run "notepad.exe beep.bat" instead.  However, if I use "beep.cmd" then all works fine and the file is executed as if it ended in ".bat".  Not sure what is happening.  I even looked in my antivirus history to see if something was happening there.  I didn't see a bug list for this but I might have missed it.

The GPUs ID number does not match that reported by boinc on the boinctasks program.  For example, the following LOG from tthrottle does not list an ID. 

nvidia: found 3 logical devices
nvidia: found 3 physical devices
nvidia: Temperature 66 °C, max Temperature 127 °C
nvidia: Temperature 58 °C, max Temperature 127 °C
nvidia: Temperature 52 °C, max Temperature 127 °C

nvidia: GeForce GTS 250, GeForce GTS 250, GeForce GTX 650 Ti

The following screen shot shows the temperature display from tthrottle


I am guessing that the ID numbers in the above log should correlate to 2,1,0 since the message log (boinc) shows the following

2/6/2012 9:46:08 PM |  | NVIDIA GPU 0: GeForce GTX 650 Ti (driver version 310.90, CUDA version 5.0, compute capability 3.0, 1024MB, 8381368MB available, 1425 GFLOPS peak)
2/6/2012 9:46:08 PM |  | NVIDIA GPU 1: GeForce GTS 250 (driver version 310.90, CUDA version 5.0, compute capability 1.1, 1024MB, 974MB available, 705 GFLOPS peak)
2/6/2012 9:46:08 PM |  | NVIDIA GPU 2: GeForce GTS 250 (driver version 310.90, CUDA version 5.0, compute capability 1.1, 1024MB, 950MB available, 705 GFLOPS peak)

I wish that the log in tthrottle show the same GPU ID number that BOINC shows.  The reason for this is that I restrict some GPU's from crunching certain projects because some overheat.  Yesterday, I had a gts250 up at 92c on primegrid.  This one, with the high temperature, was in the middle of 3 PCIe slots and does not get enough air even with eVga precision 3.04 setting the fan at %100.  It runs just fine on collatz, poem, and other projects that use opencl.  However, primegrid easily overtaxes this board and I would  have to set the gpu and memory clocks down to minimum to prevent overheating.  It is easier just to exclude the gpu from processing.  To do that, I need to know the device number that boinc uses. 

Put this down at low priority because boinctasks does show the temperature and the device (gpu) number.  However, I do not always have boinctasks available and if you have only tthrottle and boincmgr then there seems no way to properly id the gpu.   Obviously, this affects only those systems with more than 1 gpu.

Fred:  I am currently using 7.0.28 for all except linux and not doing much linux anymore.

The ability to exclude certain gpu's for various projects (or even certain apps run by a project) was implement in 6.13 as indicated in the Client Reference wiki

I never messed with this before because there are a lot of projects but the project POEM seems usefull and I wanted to run GPU apps but unaccountably if more than 1 video board (of same manufacturer) then they have a problem.

I think einstein as a somewhat similar problem and app_config.xml was useful for them (although WCG requested app_config).

IANE, but it seem that app_config and app_info are used only by projects but cc_config is used by boinc so boinc says there are only "1", "2", etc GPUs available and the project does not know any better so only one app gets run at a time.

On your BT program, it would seem to me that if I check that box  [ x ] for ignore syntax error, that the entire contents should be transmitted over to the target computer "as is".  I dont understand how the text excluding that POEM projects was missing when I checked at the target.

Anyway, thanks for your effort and for looking that this problem. 

I though I could use your program to make extensive changes to that cc_config file and not have to put a monitor and keyboard on the system.

I do not see a way to exclude the GPU for a specific project in the boinc preferences  For example, at POEM I was told that their project is unable to run on a system with 2 or more video boards of the same manufacturer.  The following post (message 7731) spelled out what I had to add to  the cc_config file in order to use their app on a system with two nvidia boards.

<exclude_gpu>
<url>http://boinc.fzk.de/poem/</url>
<device_num>1</device_num>
<app>poemcl</app>
</exclude_gpu>

The above worked fine as I used notepad to edit cc_config.xml and add those lines in.  BT's edit config did not allow those lines to be added.  Those lines exclude device 1 from running application poemcl on project bonc.fzk.de/poem.  On a system with 3 gtx460's I had to add 1 and 2.  Excluding using the app and project allows the excluded GPUs to continue to be used on milkyway, collatz, etc.

AFAIK this has to go into cc_config.xml although a thread over on boinc indicates that the project admin need to  "rtfm" and fix this problem and not rely on users to manually do a workaround.

I am confused about how the edit cc_config works.  I have been using it recently (BT 1.40) and it seemed to be working but when I added (just yesterday) some project related directives they were not passed on to the project.

I added the following to one of my systems using the edit cc_config feature

<exclusive_app>DVDFab.exe</exclusive_app>
<exclude_gpu>
<url>http://boinc.fzk.de/poem/</url>
<device_num>1</device_num>
<app>poemcl</app>
</exclude_gpu>

What happened was the DVDFab exclusion showed up but the POEM related project stuff didnt.  Also, the entire set of XML directives, I guess those "default" ones, then showed up on my system including the one I didn't want that disables remote RPC calls.   When I issued the "read config" (actually it was read automatically) command BT stopped reporting results and I put a monitor on the system to discover the problem.

Observation:  If I create (using text editor at the system) a really simple cc_config file that has exactly what I need such or even an empty one such as

<cc_config>
    <log_flags>
    </log_flags>
    <options>
    </options>
</cc_config>


When I call up that system and "Edit config" using BT, that simple cc_config gets replaced by the huge default one shown in the BT edit config (V7) dialog box.  If I replace the entire contents of that BT edit config dialog box with the empty one shown above, it is ignored and the huge default one is left on the disk drive.  I did not put a port sniffer on to see exactly what is going on (yes, it is easier to just ask).

1.  Is this behavior by design?  ie: BOINC or BT

2. Is BOINC supplying the default cc_config when the RPC message to get the config file is received?

3.  Are you sending the project related parameters such as <exclude_gpu> , etc?  If so, they are not showing up on the target system.

4.  Replacing 1.40 with 1.44 I see a syntax error for <vbox_windows> that was not there before.
  I think fixing that is on your wish list as I saw that somewhere.  I also have 7.0.28 which may
  not support vbox_window (??) as I see it listed as a bad tag in events.
  However, project specific stuff is not being sent (or does not show up) with 1.44 either.

5. Clicking "apply" on the edit config dialog box also causes BOINC to read cc_config.  I assume this is boinc doing the reading w/o being told explicitly to do so.

observation on shutdown:  I have problems on shutdown but they are related to network problems.  On one system that has both WIFI and hardwired ethernet, I had to disable the WIFI to force the system to use the hardwired ethernet.  This solved the a shutdown or BT "closings" as the WIFI connection was poor where I had the system placed.  I thought win7 would automatically use the hardwired connection but it seems not.

I also have occasional problems with my powerline ethernet adapter.  It does not have much bandwidth and on rare occasions BT takes a long time to finish a scan.

Actually, I put in a plug for BOINCTASKS and asked that the core client (BOINC) be modified to support updating app_config and app_info xml files as shown here

If they dont implement my suggestion, maybe they will implement a get and set rpc call to allow easy upload/download of app_config and app_info.  At a minimum, the core client rpc call could include the name of the xml file allowing "app_info" to be used instead of "app_config" since they both go into the same project folder.

If they do implement my suggestion then your cc_config tool would allow easy edit of app_config and app_info else you would have to add an "edit app_config" to boinctasks.

Thanks for reading and thank you for your BT program.

Unaccountably, the gadget is popping up unexpectedly.  This usually happens shortly after booting the system up and is NOT associated with the mouse cursor moving over the icon.  It happens 3 or 4 times and then it rarely occurs again unless I do move the mouse over the icon.

I have a 2 monitor system.  I can be working in the 2nd monitor browsing with chrome or IE and suddenly, the gadget pops up in the primary monitor.  I think this started with beta 1.43 as I don't remember seeing this before.  I have malwarebytes pro and m$oft's security essentials.   I do not see how to debug this and it is a nuisance.  For example, while writing this, I brought up IE, dragged it into the primary monitor and the instant I released the mouse from the "drag"  the gadget popped up near where the mouse was released.  However, I cannot repeat the problem as it wont happen the 2nd time I do it.

Thanks for looking.