Micro Collatz CPU Apps
log in

Advanced search

Message boards : News : Micro Collatz CPU Apps

Previous · 1 · 2 · 3 · 4 · Next
Author Message
Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19414 - Posted: 9 May 2014, 17:22:28 UTC - in response to Message 19400.

Ah. The preferences indeed does NOT show a "large" to uncheck. When you click "edit" it does then show "large".


That's the way BOINC works. It isn't like I've hidden it on the view page.

So I have to be a mind reader as well as anticipate exactly what large means (which was certainly larger than as described in the boinc notice) and have to poke about in the guts of this site to alter something I didn't choose in the first place.

Wonderful. Not!

Nick


No. You don't have to read minds. You see, someone who didn't accuse me of being mentally unstable or make wild accusations about me hijacking their computer suggested changing the preferences page so it describe the size difference of the work units. It sounded like a good idea and was as easy to implement, so I made that change on May 2nd. The preferences now state that each WU type is 16 times larger than its predecessor. I cannot state how long they will run because that is hardware dependent. GPUs range from 2.5 hours to 192 days to run a large work unit. I also asked for feedback on that change. No one has responded this far or offered a better solution.

So, if you have a suggestion, please post it in the "Wish List" section of the message board and it will be evaluated and discussed. If you just want to complain, please do so in the cafe section of the message board as I generally ignore it.

adrian
Avatar
Send message
Joined: 17 Mar 12
Posts: 8
Credit: 287,464,567
RAC: 0
Message 19428 - Posted: 10 May 2014, 13:41:48 UTC
Last modified: 10 May 2014, 13:56:11 UTC

So rising limit is different from my own selection on preference page?

I have plenty of Core2Duo units with

2570 floating point MIPS (Whetstone) per CPU
5224 integer MIPS (Dhrystone) per CPU
CAL: ATI GPU 0: ATI Radeon HD 2300/2400/3200/4200 (RV610) (CAL version 1.4.1734, 256MB, 224MB available, 128 GFLOPS peak)

I'd love to have margin rised to otherlap mine CPUs please

also another question
If I unticked CPU in preferences and requested only GPU WUs, will make sense for Micro-Solo selection for GPU?
Thanks

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19436 - Posted: 10 May 2014, 23:57:04 UTC - in response to Message 19428.

So rising limit is different from my own selection on preference page?

I have plenty of Core2Duo units with

2570 floating point MIPS (Whetstone) per CPU
5224 integer MIPS (Dhrystone) per CPU
CAL: ATI GPU 0: ATI Radeon HD 2300/2400/3200/4200 (RV610) (CAL version 1.4.1734, 256MB, 224MB available, 128 GFLOPS peak)

I'd love to have margin rised to otherlap mine CPUs please

also another question
If I unticked CPU in preferences and requested only GPU WUs, will make sense for Micro-Solo selection for GPU?
GPU would get solo and CPU would get mini. CPU is too fast for micro (4092 drystone max)
Thanks

Nick Warren
Send message
Joined: 16 Aug 12
Posts: 7
Credit: 1,121,077
RAC: 0
Message 19446 - Posted: 12 May 2014, 8:08:24 UTC - in response to Message 19414.

Ah. The preferences indeed does NOT show a "large" to uncheck. When you click "edit" it does then show "large".


That's the way BOINC works. It isn't like I've hidden it on the view page.

So I have to be a mind reader as well as anticipate exactly what large means (which was certainly larger than as described in the boinc notice) and have to poke about in the guts of this site to alter something I didn't choose in the first place.

Wonderful. Not!

Nick


No. You don't have to read minds. You see, someone who didn't accuse me of being mentally unstable or make wild accusations about me hijacking their computer suggested changing the preferences page so it describe the size difference of the work units. It sounded like a good idea and was as easy to implement, so I made that change on May 2nd. The preferences now state that each WU type is 16 times larger than its predecessor. I cannot state how long they will run because that is hardware dependent. GPUs range from 2.5 hours to 192 days to run a large work unit. I also asked for feedback on that change. No one has responded this far or offered a better solution.

So, if you have a suggestion, please post it in the "Wish List" section of the message board and it will be evaluated and discussed. If you just want to complain, please do so in the cafe section of the message board as I generally ignore it.


Gosh, you sound so awfully appreciative of what we do for you!

Nick Warren
Send message
Joined: 16 Aug 12
Posts: 7
Credit: 1,121,077
RAC: 0
Message 19447 - Posted: 12 May 2014, 8:22:18 UTC - in response to Message 19446.

Perhaps someone can also explain why I was given an estimated 130 hour WU and only 7 days to complete it in? Given that this PC is off at weekends and has other WUs for other projects to do the task was utterly impossible.

Now what do I do? Abort the thing and waste some 70 hours of computation?

Nick Warren
Send message
Joined: 16 Aug 12
Posts: 7
Credit: 1,121,077
RAC: 0
Message 19448 - Posted: 12 May 2014, 10:10:33 UTC - in response to Message 19447.

Bah. I've aborted the b**** thing.

I don't care how you dress things up and justify your actions. You've messed me about and you will have messed others about in a similar position to me which can hardly be unique.

Nick

Profile mikey
Avatar
Send message
Joined: 11 Aug 09
Posts: 3242
Credit: 1,691,802,693
RAC: 5,659,489
Message 19449 - Posted: 12 May 2014, 11:27:15 UTC - in response to Message 19448.

Bah. I've aborted the b**** thing.

I don't care how you dress things up and justify your actions. You've messed me about and you will have messed others about in a similar position to me which can hardly be unique.

Nick


You can smother them with kindness and get some help or bad mouth them and move on, sounds like the later is your choice. Have a great time and keep on crunching!

Nick Warren
Send message
Joined: 16 Aug 12
Posts: 7
Credit: 1,121,077
RAC: 0
Message 19451 - Posted: 12 May 2014, 12:34:06 UTC - in response to Message 19449.

Now I find that even though I've unchecked "large" boinc manager has downloaded another 130 hour WU.

So your recommended fix for stopping these ridiculous WUs doesn't even work.

I've suspended this project altogether. It just isn't worth the hassle trying to help you until you fix your bugs.

Nick

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19452 - Posted: 12 May 2014, 14:19:14 UTC - in response to Message 19451.

Don't you just love people with bad attitudes and who have an even worse reading comprehension that causes them jump to conclusions and make accusations without both knowing and understanding the facts?

But, what if they are right? Maybe there is a conspiracy. Maybe I added some code so that when their specific computers connect to the project that it purposely gives really bad work estimates just to make them angry and waste electricity. In fact, maybe this entire project is just a ruse so I can mess with people. Maybe I just work for big oil and this is just a dastardly plan get them to use more electricity so I can get rich and then use my wealth to control all the world governments.

Or maybe not.

Profile Conan
Avatar
Send message
Joined: 17 Jun 09
Posts: 40
Credit: 110,613,769
RAC: 0
Message 19457 - Posted: 12 May 2014, 22:06:21 UTC - in response to Message 19451.

Now I find that even though I've unchecked "large" boinc manager has downloaded another 130 hour WU.

So your recommended fix for stopping these ridiculous WUs doesn't even work.

I've suspended this project altogether. It just isn't worth the hassle trying to help you until you fix your bugs.

Nick


With my few years old AMD computers (Phenom type CPUs), the Mini will take from 5 to 7 hours and the Solo is going to take from 80 to 110 hours.
This is showing the 16x difference that is mentioned in the preferences.

I run 32 bit Windows and 64 bit Linux which is probably why I have such a spread of times.

Conan
____________

Kiska
Send message
Joined: 28 Mar 12
Posts: 8
Credit: 15,145,060
RAC: 0
Message 19487 - Posted: 17 May 2014, 11:27:35 UTC

How about the old 9300M GS? That old GPU I still have and still works, but can't do collatz fast enough before the deadline.

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19494 - Posted: 19 May 2014, 1:29:50 UTC - in response to Message 19487.

How about the old 9300M GS? That old GPU I still have and still works, but can't do collatz fast enough before the deadline.


Your 9300M GS has the same number of cuda cores as my 8400 GS. The solo and large WUs would timeout, but it runs the mini WUs just fine.

Kiska
Send message
Joined: 28 Mar 12
Posts: 8
Credit: 15,145,060
RAC: 0
Message 19509 - Posted: 20 May 2014, 10:25:31 UTC

Hi there. My core i5 3337U is grabbing micro units even though it has reported to be 7xxx million ops/sec. Is there a problem with scheduling? Or is it the server just sending them for fun/trolling?

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19516 - Posted: 20 May 2014, 14:02:50 UTC - in response to Message 19509.

Hi there. My core i5 3337U is grabbing micro units even though it has reported to be 7xxx million ops/sec. Is there a problem with scheduling? Or is it the server just sending them for fun/trolling?


Make that 5852 ops/sec. Collatz only looks at the integer operations. Regarless, it is not working properly if you are getting micro WUs. Are you by chance using an app_info.xml file? I haven't waded through all the server code but BOINC looks like it separates the logic for anonymous apps via app_info files vs standard apps. I think it is because so few people put the flops setting in the app_info file. Since the app info doesn't report integer operations, there's no way to implement the exclusion of the micro apps when using one. If that is the case, you probably want to leave the ARM device in the default venue with micro apps selected and move the computers to the home, school, or work venue and de-select the micro app for them.

Kiska
Send message
Joined: 28 Mar 12
Posts: 8
Credit: 15,145,060
RAC: 0
Message 19519 - Posted: 20 May 2014, 21:48:39 UTC - in response to Message 19516.

No I am not using an app_info.xml. SO I am using stock app

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19533 - Posted: 22 May 2014, 16:36:02 UTC

Host id 139003 has 2624.23 million ops/sec so it makes sense that it gets micro collatz WUs. The take about an hour to run so it will get full credit for each.

Host id 130309 has 5852.24 million ops/sec so it really shouldn't be getting micro WUs.
The scheduler log on the collatz server shows that the logic is working:

2014-05-20 16:25:10.4909 [PID=8167 ] [debug] [send WU#6414667 micro_collatz_2385543640912158794867_6442450944 HOST#130309] Limit i
s 4649999872.00 iops; host has 5852241206.35
2014-05-20 16:25:10.4910 [PID=8167 ] [debug] [send WU#6414732 micro_collatz_2385543987677080855667_6442450944 HOST#130309] Limit i
s 4649999872.00 iops; host has 5852241206.35
2014-05-20 16:25:10.4910 [PID=8167 ] [debug] [send WU#6415047 micro_collatz_2385544825614462887027_6442450944 HOST#130309] Limit i
s 4649999872.00 iops; host has 5852241206.35
2014-05-20 16:25:10.4911 [PID=8167 ] [debug] [send WU#6415081 micro_collatz_2385544855307719287923_6442450944 HOST#130309] Limit i
s 4649999872.00 iops; host has 5852241206.35


The above message is displayed when an app version is rejected for the host. The question now is why BOINC sends an app version when it clearly is marked as not the best version for that host.

FYI, the scheduler logic is the most buggy code in BOINC and it changes with every release. My guess is that the fancy logic that is supposed to tell it to learn which app is best is causing the problem. Rather than always sending the best app, BOINC randomly selects the apps and supposedly learns which runs best on your computer. The best method if you don't want Micro apps on your i5 is move it to a different venue and then edit those preferences so that it does not allow micro work (e.g. only check mini and solo for example).

steven
Send message
Joined: 12 Jan 14
Posts: 2
Credit: 57,749,196
RAC: 3
Message 19661 - Posted: 22 Jun 2014, 17:33:30 UTC
Last modified: 22 Jun 2014, 17:56:19 UTC

Sorry for being ignorant, but i've tried figuring this out myself. I just started running on my androids, and i have selected micro and mini wus to be used so i can run the micros on my androids. Problem is that i still get mini on them, and they take forever to complete. Am I missing something? How can i set this so that my laptop will continue to get mini and my androids will only get micros? I've been just canceling them because they take entirely too long, but that doesn't seem like the best solution to me. Any help would be appreciated.

Edit: I think i may have figured it out, I'll post back if not.

Profile Zydor
Avatar
Send message
Joined: 19 Aug 09
Posts: 364
Credit: 840,811,292
RAC: 0
Message 19663 - Posted: 22 Jun 2014, 21:14:26 UTC - in response to Message 19661.

How can i set this so that my laptop will continue to get mini and my androids will only get micros?


Use the Home/School/Work profiles to set different criteria for different machines. Set the option you want, then assign that profile to the machine.

Its Good Practice - but not essential - to leave Default set as a Default, then use two of the other profiles for your tasks.

steven
Send message
Joined: 12 Jan 14
Posts: 2
Credit: 57,749,196
RAC: 3
Message 19664 - Posted: 22 Jun 2014, 23:53:56 UTC - in response to Message 19663.


Use the Home/School/Work profiles to set different criteria for different machines. Set the option you want, then assign that profile to the machine.

Its Good Practice - but not essential - to leave Default set as a Default, then use two of the other profiles for your tasks.



That is, in fact, exactly what i did. Working perfectly fine now.

I initially tried to set the droids to mobile, but got an error saying invalid venue. Maybe that could be fixed, or allow us to make our own custom ones?

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 19665 - Posted: 23 Jun 2014, 2:33:59 UTC - in response to Message 19664.


Use the Home/School/Work profiles to set different criteria for different machines. Set the option you want, then assign that profile to the machine.

Its Good Practice - but not essential - to leave Default set as a Default, then use two of the other profiles for your tasks.



That is, in fact, exactly what i did. Working perfectly fine now.

I initially tried to set the droids to mobile, but got an error saying invalid venue. Maybe that could be fixed, or allow us to make our own custom ones?


It would be pretty hard to fix since the mobile venue has never existed. That's not to say that you can't have one if you want one. The boinc_dev email list is probably the place to request features or enhancements.

Previous · 1 · 2 · 3 · 4 · Next
Post to thread

Message boards : News : Micro Collatz CPU Apps


Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.