PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise

Advanced search

Message boards : Number crunching : weird scheduling after going to mt apps

Author Message
Jim WilkinsProject donor
Send message
Joined: 4 Sep 07
Posts: 167
ID: 11784
Credit: 65,364,991
RAC: 11,157
321 LLR Ruby: Earned 2,000,000 credits (2,003,677)Cullen LLR Amethyst: Earned 1,000,000 credits (1,011,142)ESP LLR Ruby: Earned 2,000,000 credits (2,006,731)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,015,421)PPS LLR Ruby: Earned 2,000,000 credits (2,014,635)PSP LLR Ruby: Earned 2,000,000 credits (2,318,872)SoB LLR Ruby: Earned 2,000,000 credits (3,423,667)SR5 LLR Ruby: Earned 2,000,000 credits (2,002,072)SGS LLR Amethyst: Earned 1,000,000 credits (1,000,597)TRP LLR Amethyst: Earned 1,000,000 credits (1,008,278)Woodall LLR Amethyst: Earned 1,000,000 credits (1,239,938)321 Sieve Amethyst: Earned 1,000,000 credits (1,011,922)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (200,004)Generalized Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (2,001,945)PPS Sieve Sapphire: Earned 20,000,000 credits (31,981,732)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,010,599)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,124,676)AP 26/27 Ruby: Earned 2,000,000 credits (2,007,865)GFN Turquoise: Earned 5,000,000 credits (5,981,177)
Message 132998 - Posted: 19 Sep 2019 | 19:13:26 UTC

I have transitioned to mt apps on my LLRs. It appears to me that there are issues scheduling tasks since I have transitioned.

Running On iMac, OSX 10.14.6 (Mojave), Boinc 7.14.2.

I have a 4 CPU machine. I have PrimeGrid 75% of the CPUs and the 4th CPU is free to run other Boinc Tasks.

I have set PrimeGrid to use 2 CPUs for mt apps.

The effect I am seeing is that indeed, LLR apps are using 2 CPUs, but the other two CPUs are have no PrimeGrids tasks running. I expected that BOINC would use the 2 "empty" CPUs for other BOINC projects, I have 2 other BOINC tasks that could those empty CPUs, but they are not being tasked.

Is this the behavior that I should expect from BOINC in this scenario?


____________
Thanks,
Jim

Profile Michael GoetzProject donor
Honorary cruncher
Avatar
Send message
Joined: 21 Jan 10
Posts: 13240
ID: 53948
Credit: 221,809,280
RAC: 150,039
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de PrimesFound 1 prime in the 2020 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,822,730)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Jade: Earned 10,000,000 credits (12,872,444)PSP LLR Turquoise: Earned 5,000,000 credits (5,197,957)SoB LLR Sapphire: Earned 20,000,000 credits (34,291,181)SR5 LLR Jade: Earned 10,000,000 credits (10,007,110)SGS LLR Ruby: Earned 2,000,000 credits (2,276,011)TRP LLR Ruby: Earned 2,000,000 credits (2,790,862)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Jade: Earned 10,000,000 credits (10,061,196)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (22,875,008)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (72,383,585)PSA Jade: Earned 10,000,000 credits (12,445,029)
Message 133000 - Posted: 19 Sep 2019 | 20:22:38 UTC - in response to Message 132998.

I have transitioned to mt apps on my LLRs. It appears to me that there are issues scheduling tasks since I have transitioned.

Running On iMac, OSX 10.14.6 (Mojave), Boinc 7.14.2.

I have a 4 CPU machine. I have PrimeGrid 75% of the CPUs and the 4th CPU is free to run other Boinc Tasks.

I have set PrimeGrid to use 2 CPUs for mt apps.

The effect I am seeing is that indeed, LLR apps are using 2 CPUs, but the other two CPUs are have no PrimeGrids tasks running. I expected that BOINC would use the 2 "empty" CPUs for other BOINC projects, I have 2 other BOINC tasks that could those empty CPUs, but they are not being tasked.

Is this the behavior that I should expect from BOINC in this scenario?



How did you "give PrimeGrid 75% of the CPUs"?


____________
My lucky number is 75898524288+1
(I am NOT an administrator anymore, so please don't PM me with questions. I can't help.)

Jim WilkinsProject donor
Send message
Joined: 4 Sep 07
Posts: 167
ID: 11784
Credit: 65,364,991
RAC: 11,157
321 LLR Ruby: Earned 2,000,000 credits (2,003,677)Cullen LLR Amethyst: Earned 1,000,000 credits (1,011,142)ESP LLR Ruby: Earned 2,000,000 credits (2,006,731)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,015,421)PPS LLR Ruby: Earned 2,000,000 credits (2,014,635)PSP LLR Ruby: Earned 2,000,000 credits (2,318,872)SoB LLR Ruby: Earned 2,000,000 credits (3,423,667)SR5 LLR Ruby: Earned 2,000,000 credits (2,002,072)SGS LLR Amethyst: Earned 1,000,000 credits (1,000,597)TRP LLR Amethyst: Earned 1,000,000 credits (1,008,278)Woodall LLR Amethyst: Earned 1,000,000 credits (1,239,938)321 Sieve Amethyst: Earned 1,000,000 credits (1,011,922)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (200,004)Generalized Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (2,001,945)PPS Sieve Sapphire: Earned 20,000,000 credits (31,981,732)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,010,599)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,124,676)AP 26/27 Ruby: Earned 2,000,000 credits (2,007,865)GFN Turquoise: Earned 5,000,000 credits (5,981,177)
Message 133004 - Posted: 19 Sep 2019 | 21:21:47 UTC - in response to Message 133000.

Sorry...ambiguous term.

Well, it used to but not now, so let me try again.

PrimeGrid is allocated 75% of the available CPUs which is 3 CPUs on my Mac. The other two projects, World Community Grid and LHC@home are allocated 12.5% each, presumably the 4th CPU.

For PrimeGrid, I have set multitasking on and limited to 2 threads. It uses two CPUs and yes, I have confirmed that. The other two CPUs are idle when it appears to me that they could be used, at least by WCG or LHC.

So I guess my question is, why are these two CPUs not being used?

Thanks,
Jim

____________
Thanks,
Jim

Profile Michael GoetzProject donor
Honorary cruncher
Avatar
Send message
Joined: 21 Jan 10
Posts: 13240
ID: 53948
Credit: 221,809,280
RAC: 150,039
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de PrimesFound 1 prime in the 2020 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,822,730)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Jade: Earned 10,000,000 credits (12,872,444)PSP LLR Turquoise: Earned 5,000,000 credits (5,197,957)SoB LLR Sapphire: Earned 20,000,000 credits (34,291,181)SR5 LLR Jade: Earned 10,000,000 credits (10,007,110)SGS LLR Ruby: Earned 2,000,000 credits (2,276,011)TRP LLR Ruby: Earned 2,000,000 credits (2,790,862)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Jade: Earned 10,000,000 credits (10,061,196)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (22,875,008)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (72,383,585)PSA Jade: Earned 10,000,000 credits (12,445,029)
Message 133005 - Posted: 19 Sep 2019 | 21:26:54 UTC - in response to Message 133004.

PrimeGrid is allocated 75% of the available CPUs which is 3 CPUs on my Mac. The other two projects, World Community Grid and LHC@home are allocated 12.5% each, presumably the 4th CPU.


Please explain how you did this. There's no control I know of that does this, unless you're running multiple BOINC clients on your computer, or running multiple VMs.


____________
My lucky number is 75898524288+1
(I am NOT an administrator anymore, so please don't PM me with questions. I can't help.)

Jim WilkinsProject donor
Send message
Joined: 4 Sep 07
Posts: 167
ID: 11784
Credit: 65,364,991
RAC: 11,157
321 LLR Ruby: Earned 2,000,000 credits (2,003,677)Cullen LLR Amethyst: Earned 1,000,000 credits (1,011,142)ESP LLR Ruby: Earned 2,000,000 credits (2,006,731)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,015,421)PPS LLR Ruby: Earned 2,000,000 credits (2,014,635)PSP LLR Ruby: Earned 2,000,000 credits (2,318,872)SoB LLR Ruby: Earned 2,000,000 credits (3,423,667)SR5 LLR Ruby: Earned 2,000,000 credits (2,002,072)SGS LLR Amethyst: Earned 1,000,000 credits (1,000,597)TRP LLR Amethyst: Earned 1,000,000 credits (1,008,278)Woodall LLR Amethyst: Earned 1,000,000 credits (1,239,938)321 Sieve Amethyst: Earned 1,000,000 credits (1,011,922)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (200,004)Generalized Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (2,001,945)PPS Sieve Sapphire: Earned 20,000,000 credits (31,981,732)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,010,599)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,124,676)AP 26/27 Ruby: Earned 2,000,000 credits (2,007,865)GFN Turquoise: Earned 5,000,000 credits (5,981,177)
Message 133009 - Posted: 20 Sep 2019 | 2:44:42 UTC - in response to Message 133005.

Michael,

In BOINC preferences, I have allocated 75% of my CPUs to PrimeGrid. I have three BOINC programs that I run, PrimeGrid, LHC, and WCG. LHC and WCG are allocated 12.5% of my CPUs which for my machine is 1 CPU.

In Primegrid, I am running tasks that, under the user's control, can be multi-threaded. This preference is settable in the PrimeGrid preferences. I have set this to 2 threads which are using 2 of my CPUs.

This leaves me with 1 CPU that can PrimeGrid tasks and 1 CPU that can run the LHC and WCG, lets call that #4. The remaining CPU that can run PrimeGrid tasks I will denote as #3.

So, with the PrimeGrid task running 2-thread multitasking, I cannot run a second 2-thread multitask on #3. But, I should able to run either a LHC or a WCG task on both #3 and #4, correct. Right now those CPUs are idle.

Hope this helps.

As usual,
____________
Thanks,
Jim

Profile Michael GoetzProject donor
Honorary cruncher
Avatar
Send message
Joined: 21 Jan 10
Posts: 13240
ID: 53948
Credit: 221,809,280
RAC: 150,039
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de PrimesFound 1 prime in the 2020 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,822,730)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Jade: Earned 10,000,000 credits (12,872,444)PSP LLR Turquoise: Earned 5,000,000 credits (5,197,957)SoB LLR Sapphire: Earned 20,000,000 credits (34,291,181)SR5 LLR Jade: Earned 10,000,000 credits (10,007,110)SGS LLR Ruby: Earned 2,000,000 credits (2,276,011)TRP LLR Ruby: Earned 2,000,000 credits (2,790,862)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Jade: Earned 10,000,000 credits (10,061,196)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (22,875,008)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (72,383,585)PSA Jade: Earned 10,000,000 credits (12,445,029)
Message 133010 - Posted: 20 Sep 2019 | 3:09:59 UTC - in response to Message 133009.

Michael,

In BOINC preferences, I have allocated 75% of my CPUs to PrimeGrid.


You mean on the website? Under PrimeGrid preferences, the "Resource Share" setting? Or maybe "Max # of simultaneous PrimeGrid Tasks"?

Or do you mean in the BOINC client's "computing preferences", Under "usage limits", the "Use at most ### % of the CPUs" and/or "Use at most ### % of CPU time"?

Any of those? Something else?

The reason I'm having trouble answering your question is that none of those allow you to directly allocate 75% of your CPU cores to PrimeGrid., so it's not clear to me what you've done. I therefore can't tell you why it's not working.
____________
My lucky number is 75898524288+1
(I am NOT an administrator anymore, so please don't PM me with questions. I can't help.)

Jim WilkinsProject donor
Send message
Joined: 4 Sep 07
Posts: 167
ID: 11784
Credit: 65,364,991
RAC: 11,157
321 LLR Ruby: Earned 2,000,000 credits (2,003,677)Cullen LLR Amethyst: Earned 1,000,000 credits (1,011,142)ESP LLR Ruby: Earned 2,000,000 credits (2,006,731)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,015,421)PPS LLR Ruby: Earned 2,000,000 credits (2,014,635)PSP LLR Ruby: Earned 2,000,000 credits (2,318,872)SoB LLR Ruby: Earned 2,000,000 credits (3,423,667)SR5 LLR Ruby: Earned 2,000,000 credits (2,002,072)SGS LLR Amethyst: Earned 1,000,000 credits (1,000,597)TRP LLR Amethyst: Earned 1,000,000 credits (1,008,278)Woodall LLR Amethyst: Earned 1,000,000 credits (1,239,938)321 Sieve Amethyst: Earned 1,000,000 credits (1,011,922)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (200,004)Generalized Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (2,001,945)PPS Sieve Sapphire: Earned 20,000,000 credits (31,981,732)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,010,599)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,124,676)AP 26/27 Ruby: Earned 2,000,000 credits (2,007,865)GFN Turquoise: Earned 5,000,000 credits (5,981,177)
Message 133093 - Posted: 21 Sep 2019 | 20:37:10 UTC

Einstein said something like "if you can't explain your issue, you don't understand it well enough" or something close to that. I suspect that is my problem. So, I will try to articulate better.

In BM Options, I have set under Computing Preferences to use at most 75% of my CPUs and use no more than 50% of the CPUs. I take this to mean the over some given period of time, BOINC will try to usually use 75% of my CPUs, for me that is 3 CPUs, at an average 50% CPU utilization.

In PrimeGrid preferences, under Job Control and Multi-threading I have set multi-threading to on, and I have set no limit to the number of threads that may be used or the number of tasks that may be multi-threaded.

I have also set the resource share for PrimeGrid at 750. Note that I have set LHC and WCG each to 12.5%

What I see is that there is one PG task running 2 threads over 2 CPUs. The third and 4th CPUs are not being used by BOINC at all.

What I expected to see was 1 non mt task in the remaining CPU that PG reserved or a non PG task in the 4th CPU.

I feel like I am on a Mobius strip.

Jim


____________
Thanks,
Jim

Profile Michael GoetzProject donor
Honorary cruncher
Avatar
Send message
Joined: 21 Jan 10
Posts: 13240
ID: 53948
Credit: 221,809,280
RAC: 150,039
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de PrimesFound 1 prime in the 2020 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,822,730)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Jade: Earned 10,000,000 credits (12,872,444)PSP LLR Turquoise: Earned 5,000,000 credits (5,197,957)SoB LLR Sapphire: Earned 20,000,000 credits (34,291,181)SR5 LLR Jade: Earned 10,000,000 credits (10,007,110)SGS LLR Ruby: Earned 2,000,000 credits (2,276,011)TRP LLR Ruby: Earned 2,000,000 credits (2,790,862)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Jade: Earned 10,000,000 credits (10,061,196)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (22,875,008)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (72,383,585)PSA Jade: Earned 10,000,000 credits (12,445,029)
Message 133097 - Posted: 21 Sep 2019 | 23:08:26 UTC - in response to Message 133093.
Last modified: 21 Sep 2019 | 23:13:26 UTC

tl;dr: Getting BOINC to run multiple projects the way you want is very difficult. You're better off running only one project at a time. I avoid running multiple projects on the same computer unless absolutely necessary.

In BM Options, I have set under Computing Preferences to use at most 75% of my CPUs and use no more than 50% of the CPUs. I take this to mean the over some given period of time, BOINC will try to usually use 75% of my CPUs, for me that is 3 CPUs, at an average 50% CPU utilization.


Your understanding is slightly incorrect.

The two settings -- which are not at all related to each other -- are:

1) Use at most ### % of the CPUs
2) Use at most ### % of the CPU time

Let's take the second one first. #2 is a CPU throttle. It will keep Boinc's instantaneous use of the CPU below, in your case, 50%. It's not really telling it how many cores to use. It's telling it not to let BOINC get above 50% usage. This is good for keeping the heat down, or saving some CPU cycles for other programs.

Most people should leave that second setting at 100%. When a BOINC task (on one core) is running, let it run 100%.

#1 is the setting everyone changes. It determines how many CPU cores BOINC will use. It actually works on threads, not cores, so if you have an 4c/8t CPU, as far as BOINC is concerned, you would have an 8 core CPU. If you want to tell BOINC to run LLR only on the physical cores, for example, you would set #1 to 50% to make it use only 4 cores (really threads). If your CPU isn't hyperthreaded, this setting should be 100%

Also note that these two controls are for *every* task on that computer, not just PrimeGrid tasks.

In PrimeGrid preferences, under Job Control and Multi-threading I have set multi-threading to on, and I have set no limit to the number of threads that may be used or the number of tasks that may be multi-threaded.


Setting "no limit" to both controls is a good setting for most people.

Max_jobs determines how many total PrimeGrid jobs can be run simultaneously. This is useful for, say, running 2 PrimeGrid tasks and 2 SETI tasks simultaneously. Left to itself, BOINC will never do that correctly. But if you set Max_jobs to 2 on both websites (or in app_config), you can force it to run no more than many jobs at once.

If you're only running PrimeGrid tasks, this should always be set to "no limit".

I think you understand the multi-threading control. Max_threads sets the limit to how many threads each LLR task can use. It has no effect on other tasks.

I have also set the resource share for PrimeGrid at 750. Note that I have set LHC and WCG each to 12.5%


The only thing this does is help your computer decide which tasks it will download. 75% of the tasks should come from PrimeGrid. Once the tasks are on your computer, this setting has no effect. It doesn't control how they run, only how they're downloaded.


What I see is that there is one PG task running 2 threads over 2 CPUs. The third and 4th CPUs are not being used by BOINC at all.

What I expected to see was 1 non mt task in the remaining CPU that PG reserved or a non PG task in the 4th CPU.

I feel like I am on a Mobius strip.

Jim


Trying to micromanage BOINC into doing exactly what you want is analogous to herding cats -- provided the cats are wild Bengal tigers. It takes a lot of skill and practice, and you're going to get bit a couple of times.

What you want -- run 1 2-core CUl task, 1 1-core CUL task, and 1 task from either LHC or WCG -- isn't possible. When you set it to run N threads, it will do so on all tasks that support multi-threading. You can't make it run a 2-core CUL task and a 1-core CUL task.

This is what you *can* do, however.

The following applies to your desktop i5-4670 quad core. Your other i5 is only a dual-core CPU with hyperthreading (2c/4t). Mixing LLR, which should only be run on the full cores, and other tasks which can run fine on hyperthreads may produce mixed results. I'd use the same instructions for the i5 5250U, but be warned that your performance may suffer.

First of all, in your client, set BOTH OF THE CPU SETTINGS TO 100%. You want it using all the cores, and you want it to run BOINC tasks 100% of the time. This applies to all scenarios below. Remember, these are settings for this entire computer -- not just for PrimeGrid. Setting either number below 100% is literally telling BOINC that you want part of the CPU to be idle.

You can run 1 CUL task on three cores, and something else on the other core:

1) set Max_Tasks on the PG website to 1.
2) Set max_theads on the PG website to 3.
Those two settings will combine to limit PrimeGrid to running a single, 3 core task. That will leave a core free for another project.
Now, you have to set up LHC and WCG. Neither project has a max_tasks setting on their website, so you'll have to put <project_max_concurrent>1</project_max_concurrent> in both your LHC and WCG app_config.xml files to limit them to running only one task.

Even then, BOINC may decide to run *both* an LHC and WCG task together, which will only leave 2 cores free. That's not enough to run a PG 3-core task, so the other 2 cores will be idle.

You can get around this by only allowing LHC tasks, or WCG tasks, but not both, at any one time. Run one for a week, stop it, run the other, etc.

Or, try this instead:

Run 1 2-core PrimGrid CUL task, one LHC task, and one WCG task.

St it up exactly as above, except set max_threads to 2.

In both cases, you can set the resource share for all three projects to 100. Those values shouldn't matter if you're forcing BOINC to run only specific projects.
____________
My lucky number is 75898524288+1
(I am NOT an administrator anymore, so please don't PM me with questions. I can't help.)

Jim WilkinsProject donor
Send message
Joined: 4 Sep 07
Posts: 167
ID: 11784
Credit: 65,364,991
RAC: 11,157
321 LLR Ruby: Earned 2,000,000 credits (2,003,677)Cullen LLR Amethyst: Earned 1,000,000 credits (1,011,142)ESP LLR Ruby: Earned 2,000,000 credits (2,006,731)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,015,421)PPS LLR Ruby: Earned 2,000,000 credits (2,014,635)PSP LLR Ruby: Earned 2,000,000 credits (2,318,872)SoB LLR Ruby: Earned 2,000,000 credits (3,423,667)SR5 LLR Ruby: Earned 2,000,000 credits (2,002,072)SGS LLR Amethyst: Earned 1,000,000 credits (1,000,597)TRP LLR Amethyst: Earned 1,000,000 credits (1,008,278)Woodall LLR Amethyst: Earned 1,000,000 credits (1,239,938)321 Sieve Amethyst: Earned 1,000,000 credits (1,011,922)Cullen/Woodall Sieve (suspended) Silver: Earned 100,000 credits (200,004)Generalized Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (2,001,945)PPS Sieve Sapphire: Earned 20,000,000 credits (31,981,732)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,010,599)TRP Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,124,676)AP 26/27 Ruby: Earned 2,000,000 credits (2,007,865)GFN Turquoise: Earned 5,000,000 credits (5,981,177)
Message 133129 - Posted: 22 Sep 2019 | 20:51:00 UTC

Hey Michael,

Thanks for the explanation. I really appreciate it. As long as I understand what's happening, I'm OK with it so I will probably do what I am already doing, but now I understand why this is happening.

Thanks much,

Jim
____________
Thanks,
Jim

tazzduke
Send message
Joined: 19 Oct 10
Posts: 10
ID: 70546
Credit: 65,543,321
RAC: 890,003
Found 2 primes in the 2019 Tour de PrimesFound 3 primes in the 2020 Tour de Primes321 LLR Silver: Earned 100,000 credits (137,665)Cullen LLR Silver: Earned 100,000 credits (102,811)ESP LLR Silver: Earned 100,000 credits (325,190)Generalized Cullen/Woodall LLR Silver: Earned 100,000 credits (109,338)PPS LLR Gold: Earned 500,000 credits (645,256)PSP LLR Gold: Earned 500,000 credits (512,431)SoB LLR Silver: Earned 100,000 credits (465,303)SR5 LLR Silver: Earned 100,000 credits (192,384)SGS LLR Gold: Earned 500,000 credits (548,526)TRP LLR Gold: Earned 500,000 credits (515,602)Woodall LLR Silver: Earned 100,000 credits (113,032)321 Sieve Amethyst: Earned 1,000,000 credits (1,098,568)Cullen/Woodall Sieve (suspended) Bronze: Earned 10,000 credits (21,866)Generalized Cullen/Woodall Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,027,668)PPS Sieve Sapphire: Earned 20,000,000 credits (38,170,399)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Gold: Earned 500,000 credits (528,916)TRP Sieve (suspended) Silver: Earned 100,000 credits (145,706)AP 26/27 Turquoise: Earned 5,000,000 credits (5,268,029)GFN Jade: Earned 10,000,000 credits (15,078,832)PSA Gold: Earned 500,000 credits (548,526)
Message 133158 - Posted: 23 Sep 2019 | 11:43:14 UTC - in response to Message 133129.

Greetings Michael

That is one of the best explanations I have come across on the various BOINC websites.

Now here is one for you,

I wanted for my computer to do the following

AP26/27 - GPU only

Fermat Divisor Search LLR - CPU only

On the website preferences I set

Max Number of Task - 2
Max Number of Threads for LLR - 6 (4c8t machine)

This way, it could do 1 of each task at a time, but alas it did not, it kept on getting AP26/27 tasks only.

So I changed tasks to 3, I then got a CPU job, then when it finished then it got another AP26/27 gpu Workunit, even though I already had 2 workunits in the cache.

So I tried an app_config.xml, which we all used before the new website settings were put in place.

<app>
<name>llrDIV</name>
<fraction_done_exact/>
<report_results_immediately/>
<max_concurrent>1</max_concurrent>
</app>
<app_version>
<app_name>llrDIV</app_name>
<plan_class>mt</plan_class>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>

I also had my cache settings in boinc as follows

Store at least 0.1 days of work
Store up to an additional 0.1 days of work

Shutdown Boinc, then restarted Boinc.

And wham, it downloaded 56 tasks of Fermat Divisor Search LLR (CPU)

I thought the addition of max concurrent in the app_config file would have stopped it downloading so many, I was wrong.

Not to sure as to where to go with this one, but I have deselected the CPU job for the time being as I have a few to get through now.

Regards


Profile Michael GoetzProject donor
Honorary cruncher
Avatar
Send message
Joined: 21 Jan 10
Posts: 13240
ID: 53948
Credit: 221,809,280
RAC: 150,039
The "Shut up already!" badge:  This loud mouth has mansplained on the forums over 10 thousand times!  Sheesh!!!Discovered the World's First GFN-19 prime!!!Discovered 1 mega primeFound 1 prime in the 2018 Tour de PrimesFound 1 prime in the 2019 Tour de PrimesFound 1 prime in the 2020 Tour de Primes321 LLR Ruby: Earned 2,000,000 credits (2,822,730)Cullen LLR Ruby: Earned 2,000,000 credits (2,005,249)ESP LLR Turquoise: Earned 5,000,000 credits (5,009,577)Generalized Cullen/Woodall LLR Ruby: Earned 2,000,000 credits (2,145,754)PPS LLR Jade: Earned 10,000,000 credits (12,872,444)PSP LLR Turquoise: Earned 5,000,000 credits (5,197,957)SoB LLR Sapphire: Earned 20,000,000 credits (34,291,181)SR5 LLR Jade: Earned 10,000,000 credits (10,007,110)SGS LLR Ruby: Earned 2,000,000 credits (2,276,011)TRP LLR Ruby: Earned 2,000,000 credits (2,790,862)Woodall LLR Ruby: Earned 2,000,000 credits (2,195,123)321 Sieve Jade: Earned 10,000,000 credits (10,061,196)Cullen/Woodall Sieve (suspended) Ruby: Earned 2,000,000 credits (4,170,256)Generalized Cullen/Woodall Sieve (suspended) Turquoise: Earned 5,000,000 credits (5,059,304)PPS Sieve Sapphire: Earned 20,000,000 credits (22,875,008)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,035,522)TRP Sieve (suspended) Ruby: Earned 2,000,000 credits (2,051,121)AP 26/27 Jade: Earned 10,000,000 credits (10,114,260)GFN Emerald: Earned 50,000,000 credits (72,383,585)PSA Jade: Earned 10,000,000 credits (12,445,029)
Message 133161 - Posted: 23 Sep 2019 | 13:51:27 UTC - in response to Message 133158.

Greetings Michael

That is one of the best explanations I have come across on the various BOINC websites.

Now here is one for you,

I wanted for my computer to do the following

AP26/27 - GPU only

Fermat Divisor Search LLR - CPU only

On the website preferences I set

Max Number of Task - 2 <-- Change to "no limit"
Max Number of Threads for LLR - 6 (4c8t machine) <-- Change to "no limit"

This way, it could do 1 of each task at a time, but alas it did not, it kept on getting AP26/27 tasks only.

So I changed tasks to 3, I then got a CPU job, then when it finished then it got another AP26/27 gpu Workunit, even though I already had 2 workunits in the cache.

So I tried an app_config.xml, which we all used before the new website settings were put in place. <-- delete the entire app_config.xml file, then detach from PrimeGrid and add PrimeGrid back

<app>
<name>llrDIV</name>
<fraction_done_exact/>
<report_results_immediately/>
<max_concurrent>1</max_concurrent>
</app>
<app_version>
<app_name>llrDIV</app_name>
<plan_class>mt</plan_class>
<cmdline>-t 6</cmdline>
<avg_ncpus>6</avg_ncpus>
</app_version>

I also had my cache settings in boinc as follows

Store at least 0.1 days of work <-- these are fine, but I personally use 0, not 0.1
Store up to an additional 0.1 days of work

Shutdown Boinc, then restarted Boinc.

And wham, it downloaded 56 tasks of Fermat Divisor Search LLR (CPU)

I thought the addition of max concurrent in the app_config file would have stopped it downloading so many, I was wrong.

Not to sure as to where to go with this one, but I have deselected the CPU job for the time being as I have a few to get through now.


Explanation:

Glad you liked the explanation. It just kept getting longer and longer.. There was a lot to cover!

Your requirements are what I consider to be "normal" -- what 95% of users do. The good news is it's easy to set up. The bad news is BOINC will make it tricky to switch from what you're doing now to what you'll be doing tomorrow.

1) You don't need app_config.xml any more. Just delete the file.

2) BOINC WILL REMEMBER THE APP_CONFIG.XML SETTINGS EVEN AFTER YOU DELETE THE FILE!!! Yeah. I know. This has caused many a lot of grief. Set "no new work" on PrimeGrid and let all the PrimeGrid tasks finish and get reported to the server. Then DETACH PrimeGrid and ADD PRIMEGRID again. Now the app_config.xml file is *really* gone.

3) On our website preferences.set max_jobs to "no limit".

4) On our website preferences, also set max_threads to "no limit".

5) If you really want to be running on 6 of the 8 threads on your 4c/8t: In the BOINC manager on your computer, go into Computing Preferences and set "Use at most 75 % of the CPUs".

6) The second setting should be "Use at most 100 % of the CPU time". Most people should leave this at 100% all the time.

LLR does not like hyperthreads, so I strongly recommend that you use 4 threads, not 6, but it's your choice. You're not going any faster with 6 threads but you are drawing a lot more power. I suppose if it's snowing where you are, it can keep you warm! Otherwise, I recommend using 50% instead of 75% in step #5 above. This will cause the computer to run a single 4 threaded task.

Note that right now, the Fermat Divisor tasks are still only medium-sized, and you may get better performance running 2x 2-thread tasks, or even 4x 1-thread tasks. To change to 2x 2-thread tasks, all you have to do is set max_threads in the web preferences to "2" instead of "no limit". To change to 4x 1-thread tasks, all you have to do is set max_threads in the web preferences to "1" instead of "no limit". I am personally running the Fermat Divisor project on two computers. One is running 2x 2-threaded tasks, and the other is running 4x 1-threaded tasks. The tasks aren't yet large enough for running 4-thread tasks to be efficient.

If you're more interested in individual task speed than overall throughput (i.e., you would rather be 1st more often even if it means doing fewer tasks), then leave it set to "no limit" and run a single 4-thread task.


____________
My lucky number is 75898524288+1
(I am NOT an administrator anymore, so please don't PM me with questions. I can't help.)

tazzduke
Send message
Joined: 19 Oct 10
Posts: 10
ID: 70546
Credit: 65,543,321
RAC: 890,003
Found 2 primes in the 2019 Tour de PrimesFound 3 primes in the 2020 Tour de Primes321 LLR Silver: Earned 100,000 credits (137,665)Cullen LLR Silver: Earned 100,000 credits (102,811)ESP LLR Silver: Earned 100,000 credits (325,190)Generalized Cullen/Woodall LLR Silver: Earned 100,000 credits (109,338)PPS LLR Gold: Earned 500,000 credits (645,256)PSP LLR Gold: Earned 500,000 credits (512,431)SoB LLR Silver: Earned 100,000 credits (465,303)SR5 LLR Silver: Earned 100,000 credits (192,384)SGS LLR Gold: Earned 500,000 credits (548,526)TRP LLR Gold: Earned 500,000 credits (515,602)Woodall LLR Silver: Earned 100,000 credits (113,032)321 Sieve Amethyst: Earned 1,000,000 credits (1,098,568)Cullen/Woodall Sieve (suspended) Bronze: Earned 10,000 credits (21,866)Generalized Cullen/Woodall Sieve (suspended) Amethyst: Earned 1,000,000 credits (1,027,668)PPS Sieve Sapphire: Earned 20,000,000 credits (38,170,399)Sierpinski (ESP/PSP/SoB) Sieve (suspended) Gold: Earned 500,000 credits (528,916)TRP Sieve (suspended) Silver: Earned 100,000 credits (145,706)AP 26/27 Turquoise: Earned 5,000,000 credits (5,268,029)GFN Jade: Earned 10,000,000 credits (15,078,832)PSA Gold: Earned 500,000 credits (548,526)
Message 133177 - Posted: 23 Sep 2019 | 21:00:21 UTC - in response to Message 133161.

Greetings Michael

Thankyou for your wonderful feedback, I will take all onboard and fine tune after the AP27 challenge is over.

Thats where my is aim is, am more interested in individual task speed than overall throughput (i.e., you would rather be 1st more often even if it means doing fewer tasks), might find a prime or two, maybe :-)

Cheers

PS. No snow where I live in Australia (Perth, Western Australia).

Post to thread

Message boards : Number crunching : weird scheduling after going to mt apps

[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2020 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 1.24, 1.58, 1.65
Generated 3 Jul 2020 | 14:09:07 UTC