Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 9
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 5695 times and has 8 replies Next Thread
cw64
Advanced Cruncher
Joined: Oct 6, 2007
Post Count: 120
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
More efficiency please

In the spirit of squeezing out every last processor cycle possible I have some questions about the HCC GPU app.

My experience with GPU processing has been that they don't require a whole CPU core to run (leaving a core free). The HCC GPU app runs on CPU then GPU then CPU. During the GPU phase the assigned CPU is mostly not in use. Can anything be done here?

Can the current CPU phases be made to run on the GPU?
Can the app release the CPU to BOINC to work temporarily on another CPU job during the GPU phase?
Even better can the app work on a CPU part of another HCCGPU task while the GPU is processing? Keeping the assigned CPU at 100% output.
TaskA CPU1->	TaskA GPU ->TaskA CPU2->TaskB GPU ->TaskB CPU2->Task C GPU -> etc.
TaskB CPU1 TaskC CPU1 Task D CPU1
Concurrent processes within app.
Can the CPU phases be done off the grid (i.e. in house) leaving a pure GPU WU?
Can the current WU be reworked into 3 (one for each phase. Like DDDT one result being the input of the next task)? You could stack several jobs together to offset the extra bandwidth this would necessitate.
----------------------------------------

----------------------------------------
[Edit 1 times, last edit by cw64 at Oct 13, 2012 2:03:38 PM]
[Oct 13, 2012 2:01:32 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: More efficiency please

My 4 core runs 4 CPU WUs (DSFL) and 1 GPU simultaneously..... biggrin
[Oct 13, 2012 2:09:45 PM]   Link   Report threatening or abusive post: please login first  Go to top 
nanoprobe
Master Cruncher
Classified
Joined: Aug 29, 2008
Post Count: 2998
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: More efficiency please

In the spirit of squeezing out every last processor cycle possible I have some questions about the HCC GPU app.

My experience with GPU processing has been that they don't require a whole CPU core to run (leaving a core free). The HCC GPU app runs on CPU then GPU then CPU. During the GPU phase the assigned CPU is mostly not in use. Can anything be done here?

Can the current CPU phases be made to run on the GPU?
Can the app release the CPU to BOINC to work temporarily on another CPU job during the GPU phase?
Even better can the app work on a CPU part of another HCCGPU task while the GPU is processing? Keeping the assigned CPU at 100% output.
TaskA CPU1->	TaskA GPU ->TaskA CPU2->TaskB GPU ->TaskB CPU2->Task C GPU -> etc.
TaskB CPU1 TaskC CPU1 Task D CPU1
Concurrent processes within app.
Can the CPU phases be done off the grid (i.e. in house) leaving a pure GPU WU?
Can the current WU be reworked into 3 (one for each phase. Like DDDT one result being the input of the next task)? You could stack several jobs together to offset the extra bandwidth this would necessitate.

The WCG HCC GPU app is designed to use I CPU core and I doubt that will change. What you're asking would require an enormous amount of very intricate coding if it is even viable.
----------------------------------------
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.


[Oct 13, 2012 3:03:47 PM]   Link   Report threatening or abusive post: please login first  Go to top 
LUFTY
Cruncher
Joined: Apr 27, 2007
Post Count: 25
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: More efficiency please

Hi Nanoprobe

Your config file works really well so I am settling for that however it does mean that 4 of more cores are not being used for other WCG tasks which is a shame and also I agree my other 4 CPU cores are not fully utilised.

Kind regards


Lufty
[Oct 13, 2012 3:44:40 PM]   Link   Report threatening or abusive post: please login first  Go to top 
cw64
Advanced Cruncher
Joined: Oct 6, 2007
Post Count: 120
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: More efficiency please

Can the CPU usage be set to not be exclusive in BOINC? E.g on a 4 core system 4 CPU WUs are running at the same time as a GPU. The GPU app runs like any other application would outside of BOINC, taking CPU priority from one of the CPU tasks but not stopping it.
----------------------------------------

[Oct 14, 2012 10:06:02 AM]   Link   Report threatening or abusive post: please login first  Go to top 
cw64
Advanced Cruncher
Joined: Oct 6, 2007
Post Count: 120
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: More efficiency please

My 4 core runs 4 CPU WUs (DSFL) and 1 GPU simultaneously..... biggrin

Hmm I seemed to have ignored this previously

How are you doing this exactly?
----------------------------------------

[Oct 14, 2012 10:31:10 AM]   Link   Report threatening or abusive post: please login first  Go to top 
johncmacalister2010@gmail.com
Veteran Cruncher
Canada
Joined: Nov 16, 2010
Post Count: 799
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: More efficiency please

My 4 core runs 4 CPU WUs (DSFL) and 1 GPU simultaneously..... biggrin

Hmm I seemed to have ignored this previously

How are you doing this exactly?


I'd like to know, too, please. Both my 6 cores run 5 CPU tasks and 1 GPU task simultaneously....is more possible??

Thanks!
----------------------------------------


crunching, crunching, crunching.

AMD Ryzen 5 2600 6-core Processor with Windows 11 64 Pro.

AMD Ryzen 7 3700X 8-Core Processor with Windows 11 64 Pro (part time)


smile
[Oct 14, 2012 11:42:21 AM]   Link   Report threatening or abusive post: please login first  Go to top 
cw64
Advanced Cruncher
Joined: Oct 6, 2007
Post Count: 120
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: More efficiency please

My 4 core runs 4 CPU WUs (DSFL) and 1 GPU simultaneously..... biggrin

Hmm I seemed to have ignored this previously

How are you doing this exactly?


I'd like to know, too, please. Both my 6 cores run 5 CPU tasks and 1 GPU task simultaneously....is more possible??

Thanks!


I think this is without HCC on the GPU, another project like Milkyway e.g.
----------------------------------------

[Oct 14, 2012 5:06:13 PM]   Link   Report threatening or abusive post: please login first  Go to top 
nanoprobe
Master Cruncher
Classified
Joined: Aug 29, 2008
Post Count: 2998
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: More efficiency please

Hi Nanoprobe

Your config file works really well so I am settling for that however it does mean that 4 of more cores are not being used for other WCG tasks which is a shame and also I agree my other 4 CPU cores are not fully utilised.

Kind regards


Lufty

There are some other worthy non WCG projects that you could use your remaining 4 cores. Look at SIMAP and docking@home. I run SIMAP now and then although they don't always have work. Docking always has work when I've run it.
----------------------------------------
In 1969 I took an oath to defend and protect the U S Constitution against all enemies, both foreign and Domestic. There was no expiration date.


[Oct 14, 2012 6:39:26 PM]   Link   Report threatening or abusive post: please login first  Go to top 
[ Jump to Last Post ]
Post new Thread