Posts by Virtual Boss*


log in
1) Questions and Answers : GPU applications : Open CL ...... WTF (Message 1441566)
Posted 311 days ago by Profile Virtual Boss*
Thanks Ageless

I think I understand the gist of it.

I guess my Vivante GPU (Freescale i.MX6) units with Open CL capability will be useless unless someone with the know-how compiles for them.

Although to me the concept seems the same as if MOBO manufacturers made boards for Windows 7 (or any other OS) and did not provide a BIOS for the OS to communicate with the hardware, forcing developers to come up with thier own for each brand.
2) Questions and Answers : GPU applications : Open CL ...... WTF (Message 1441172)
Posted 312 days ago by Profile Virtual Boss*
I have read a number of posts about Open CL and now I am confused!!!!

It seems from the posts that the apps have to be "tailored" for specific manufacturers.

I thought the idea of Open CL was that ONE app would run on ANY Open CL capable device.

If not ..... what is the point??????
3) Message boards : Number crunching : Impossible Deadline (Message 1403758)
Posted 400 days ago by Profile Virtual Boss*
I think the server may expire tasks if it thinks your host won't get them done before deadline.

That does not seem likely, there was still plenty of time.

What our hosts have in common is that both are on their first v7 tasks. Maybe the server doesn't trust them enough yet or something like that. Although I'm not quite satisfied with that explanation either.

Not true. My host has previously completed V7's.
4) Message boards : Number crunching : Impossible Deadline (Message 1403531)
Posted 400 days ago by Profile Virtual Boss*
would have to walk the code to ascertain why/if such a failed contact times them out and the reasoning behind it (if any).

I take it your question is answered?


Hmmm ... maybe a change in thinking

That rig just asked for more tasks, and the AP that was not in its task list got timed out by the server at the time of the scheduler request.

It now seems to me that the server no longer resends task that have not been received by the client - it cancels them by expiring the deadline instead!

5) Message boards : Number crunching : Impossible Deadline (Message 1403491)
Posted 400 days ago by Profile Virtual Boss*
Actually, that rig only has CPU...

Do you have the log entries from the time the units were timed out? that must have been another scheduler contact.

Such deadlines mean the server timed those tasks out - must have decided it couldn't send them to you any more, which is strange - none of the usual reasons for that apply.


Ah ha!

Event log at expiry time.

15/08/2013 7:59:47 AM | SETI@home | Sending scheduler request: To fetch work.
15/08/2013 7:59:47 AM | SETI@home | Requesting new tasks for CPU
15/08/2013 8:04:54 AM | SETI@home | Scheduler request failed: Timeout was reached
15/08/2013 8:05:22 AM | | Project communication failed: attempting access to reference site
15/08/2013 8:10:28 AM | | BOINC can't access Internet - check network connection or proxy configuration.


Signal on modem was still poor - happens sometimes as I am 44km (28mile) from the cell tower, and behind a hill!


So what you're saying is the server is [heuristically challenged]?


Looks like it. :)

The server must have decided it was unable to send me MB's so it would cancel them and send me an AP instead.
6) Message boards : Number crunching : Impossible Deadline (Message 1403453)
Posted 400 days ago by Profile Virtual Boss*
I had a batch of 11 cpu tasks allocated to one of my machines that all had a deadline of 1 Hr 7 min and 20 Secs. One is - Here

I checked my event log and the tasks were never downloaded.
There was also 1 AP allocated that is missing from my task list.

Here are the event log entries around the time the server allocated them.

15/08/2013 6:49:40 AM | SETI@home | Requesting new tasks for CPU
15/08/2013 6:49:56 AM | SETI@home | Scheduler request failed: Couldn't resolve host name
15/08/2013 6:51:03 AM | SETI@home | Sending scheduler request: To fetch work.
15/08/2013 6:51:03 AM | SETI@home | Requesting new tasks for CPU
15/08/2013 6:51:19 AM | SETI@home | Scheduler request failed: Couldn't resolve host name
15/08/2013 6:52:14 AM | SETI@home | Sending scheduler request: To fetch work.
15/08/2013 6:52:14 AM | SETI@home | Requesting new tasks for CPU
15/08/2013 6:54:13 AM | SETI@home | Scheduler request failed: Failure when receiving data from the peer
15/08/2013 6:55:09 AM | SETI@home | Sending scheduler request: To fetch work.
15/08/2013 6:55:09 AM | SETI@home | Requesting new tasks for CPU
15/08/2013 6:55:25 AM | SETI@home | Scheduler request failed: Couldn't resolve host name
15/08/2013 6:56:16 AM | SETI@home | Sending scheduler request: To fetch work.
15/08/2013 6:56:16 AM | SETI@home | Requesting new tasks for CPU
15/08/2013 6:56:32 AM | SETI@home | Scheduler request failed: Couldn't resolve host name
15/08/2013 6:57:12 AM | SETI@home | Sending scheduler request: To fetch work.
15/08/2013 6:57:12 AM | SETI@home | Requesting new tasks for CPU
15/08/2013 6:57:28 AM | SETI@home | Scheduler request failed: Couldn't resolve host name


It must have been the request that resulted in data failure where the 12 tasks went missing as for about +/- 15 minutes all other requests show as Couldn't resolve host name due to loss of signal on my wireless data modem.

However, that does not explain why the deadline was impossibly short!
7) Message boards : Number crunching : Pause work if outside temp is over xx degrees? (Message 1391211)
Posted 430 days ago by Profile Virtual Boss*
I have seen a similar ducting system a long time ago, but as a Power Supply exhaust duct - it fed into a 6" plastic pipe which went up two storeys.
The draft created by 2 machines running at the bottom was so strong the computers ran for years with no PS fans.

BTW: The guy who wrote the article did not check his facts! Condensation only occurs when hot (and moist) air is sucked through cold hardware, not the other way round.
8) Message boards : Number crunching : Changing BOINC computer ID's (Message 1063042)
Posted 1355 days ago by Profile Virtual Boss*
Are you speaking about the hostID (as Metod, S56RKO does) or about the host CPID? The latter indeed seems to be changed by 6.10.58 on every IP address change, but that only affect statistics history and doesn't produce double host entries in the project database.

Gruß,
Gundolf
[edit]CPID = Cross Project ID
There are account CPIDs and host CPIDs[/edit]



AHAA ... must be the host CPID, I think. (Confusing isn't it)

I am fairly sure the ID shown on the project page doesn't change.

The ID shown on stats sites certainly does, and as the comp reports in to each project, the stats site transfers existing credit tallys over to the new ID until the old ID disappears (when credit reaches zero?) (When this happens it seems variable amounts of credit just dissappear into a black hole).

6.10.58 has been around for quite awhile, and this seems to be only affecting that version and hasn't been fixed. (Oh, 6.10.56 as well)

I think I will just roll-back to an earlier version than 6.10.56, as it is driving me crazy when I go to a stats site and see that my computer only "appeared" a couple of days before, and the total BOINC credit shown is considerably less than the the sum of the totals for each project.

Thanks to all for your input.

@Link - Yep, same problem on 3 of my machines. 2x 6.10.58, 1x 6.10.56 Interesting that one of my 6.10.56 hosts is changing ID's and the other hasn't (yet).
9) Message boards : Number crunching : Changing BOINC computer ID's (Message 1062972)
Posted 1356 days ago by Profile Virtual Boss*
It seems they change ID intermittently during a shutdown/reboot. Not every time though (~10-20%)


My observation is that ID only changes when client and server disagree about number of connections made.

On the client side that number is in client_state.xml in line
<rpc_seqno>xxx</rpc_seqno>

and on server side it's in line
Number of times BOINC has contacted server: xxx

Note that client_state.xml bears information about all projects that host participates in so you need to find project section for correct project (if there are multiple).
Client ID only gets changed when client thinks it contacted server fewer times than server thinks it did. It's fine if client thinks it contacted server more times than server does.

If there are no problems with connection from client to server, then these two numbers won't differ. However, they could if BOINC client crashes or is forcibly stopped and doesn't refresh information in client_state.xml. I guess it could happen if shutdown procedure doesn't give enough time for BOINC to stop its clients and then refresh the state file.
State file only gets written now and then and is not in sync with client's internal state all the time. It could also happen that status of a task (or upload or report status) gets out of sync, but I guess that would be handled a bit more gracefully by upload and/or scheduling server.

All other information (OS type, IP address, BOINC version, coprocessor/GPU type, ...) is irrelevant in this scope.


Interesting ... but wouldn't that apply to all boinc versions?
I have only had the issue with 6.10.58, and never with any of the other versions I am running.
10) Message boards : Number crunching : Changing BOINC computer ID's (Message 1062251)
Posted 1358 days ago by Profile Virtual Boss*
3 of my machines have developed neuroses :) over the last months.

It seems they change ID intermittently during a shutdown/reboot. Not every time though (~10-20%)

All 3 machines are running 6.10.58 but different OS (1x Win7-64, 2x XP-32)

I have not found any other threads about this problem so is it just me?


I have seen the ID's change if the IP of that computer changes. If you are using DHCP and the lease expires which causes the machine to get a new IP.


Hmmm ... They are using DHCP, but I am fairly certain they are still using their originally allocated IP's.

But thanks, that is certainly something to check next time it happens.
11) Message boards : Number crunching : Changing BOINC computer ID's (Message 1062239)
Posted 1358 days ago by Profile Virtual Boss*
3 of my machines have developed neuroses :) over the last months.

It seems they change ID intermittently during a shutdown/reboot. Not every time though (~10-20%)

All 3 machines are running 6.10.58 but different OS (1x Win7-64, 2x XP-32)

I have not found any other threads about this problem so is it just me?
12) Message boards : Number crunching : BOINC 6.10.43/44 pulled (Message 995957)
Posted 1590 days ago by Profile Virtual Boss*
Unless you're suffering from this particular problem, I wouldn't advise anybody to change their BOINC version just because of this status change. v6.10.43 is much better for general use than v6.10.18

Of course, if you have a CUDA card with limited RAM or an OS which uses a lot of VRAM, you may need to change, but the VRAM requirements of SETI/CUDA shouldn't ever be a problem if your card has 512MB or above.


I have not had any problems with 6.10.43 with CUDA

But I am experiencing a weird problem with the CPU processing.

With 4 WUs running simultaneously on a very busy quad, when a task completion time gets close to the deadline, that task does not go into high priority, but the task with the shortest completion time does start running high priority!

If the time situation worsens, then the task with the next shortest completion time goes high priority as well!

And same again!

Has anyone else seen this?
13) Questions and Answers : GPU applications : MemPitch value? (Message 992296)
Posted 1606 days ago by Profile Virtual Boss*
My 8600GT seems to have a much larger MemPitch value than other 8600GT cards.

My card:
setiathome_CUDA: Found 1 CUDA device(s):
Device 1 : GeForce 8600 GT
totalGlobalMem = 1073414144
sharedMemPerBlock = 16384
regsPerBlock = 8192
warpSize = 32
memPitch = 2147483647
maxThreadsPerBlock = 512
clockRate = 1570000
totalConstMem = 65536
major = 1
minor = 1
textureAlignment = 256
deviceOverlap = 1
multiProcessorCount = 4

Other cards:
setiathome_CUDA: Found 1 CUDA device(s):
Device 1 : GeForce 8600 GT
totalGlobalMem = 268107776
sharedMemPerBlock = 16384
regsPerBlock = 8192
warpSize = 32
memPitch = 262144
maxThreadsPerBlock = 512
clockRate = 1188000
totalConstMem = 65536
major = 1
minor = 1
textureAlignment = 256
deviceOverlap = 1
multiProcessorCount = 4

Could this be due to my card having 1GB vs 256MB on most cards?

Is this a value that can be changed?

If YES, would changing this value affect crunch speed?

Just curious ..... Does anyone know?

Bruce
14) Message boards : Cafe SETI : Arizona Moon`s coconut posting thread IV ;-) (Message 941748)
Posted 1794 days ago by Profile Virtual Boss*
I have never had anyone try to sell me a photo of my own house. Sounds like a really stupid job to me.


Hei Carlos,

In photography, the 'photographer's eye' is more important than the camera. This is what made photogs like Ansel Adams so famous. Moon has a beautiful house in a vista setting so maybe this photog got a great pic of it in just the right light?

I hope she won't upgrade to a Windows 7 setting. It looks very nice as it is.


ROFLMAO :)
15) Message boards : Cafe SETI : Beet's Give Us a Caption XIV (Message 941746)
Posted 1794 days ago by Profile Virtual Boss*
He hadn't crunched since the classic days, didn't realize how long an AP would take on the old 386SX16 without a coprocessor!

16) Message boards : Cafe SETI : Arizona Moon`s coconut posting thread IV ;-) (Message 939656)
Posted 1802 days ago by Profile Virtual Boss*
Congrats on your million, James.

17) Message boards : Cafe SETI : Aphorisms (Message 936008)
Posted 1819 days ago by Profile Virtual Boss*
Safety is no accident!
18) Message boards : Cafe SETI : Beet's Give Us a Caption XIV (Message 935983)
Posted 1819 days ago by Profile Virtual Boss*
OK, the time for judgement is nigh!

So many good ones it was hard to select the top three, but the final decision is ...

2nd Runner up: HAL - Sorry the Caption is late - Got tied up in traffic!

Runner up: Uli - Yikes. My boss got to them, because if I am late again, he can tell me goodbye.

Winner: Dirk Villareal Wittich - The nightmare vision of a workaholic!!

Over to you Dirk!
19) Message boards : Cafe SETI : Arizona Moon`s coconut posting thread IV ;-) (Message 935776)
Posted 1821 days ago by Profile Virtual Boss*
Very nice avatar James, I like it. :)
20) Message boards : Cafe SETI : Arizona Moon`s coconut posting thread IV ;-) (Message 935645)
Posted 1821 days ago by Profile Virtual Boss*
Hei Moon, and all.

Just dropped in to say hello.

So thats why I can never find the pot ... Moons already got it.

Thats a great avatar klc53.

ZZZZ's time now for me.

Bye for now.


Next 20

Copyright © 2014 University of California