EM searches, BRP Raidiopulsar and FGRP Gamma-Ray Pulsar

mikey
mikey
Joined: 22 Jan 05
Posts: 12679
Credit: 1839080161
RAC: 3934

Ian&Steve C. wrote: I get

Ian&Steve C. wrote:

I get the impression that the older/slower systems will be segregated to the BRP4/G project, while BRP7 will be reserved for the faster GPUs. But it’s all still in flux I think. If slower GPUs can still crunch BRP7 within whatever deadline they set, then it’ll be fine I think. Really just depends on how large the tasks are. 

Then it would work similarly to how they excluded gpu's with less than 4gb of ram from some of the tasks, I'm okay with that.

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6434
Credit: 9563307790
RAC: 9352651

mikey wrote: Ian&Steve C.

mikey wrote:

Ian&Steve C. wrote:

I get the impression that the older/slower systems will be segregated to the BRP4/G project, while BRP7 will be reserved for the faster GPUs. But it’s all still in flux I think. If slower GPUs can still crunch BRP7 within whatever deadline they set, then it’ll be fine I think. Really just depends on how large the tasks are. 

Then it would work similarly to how they excluded gpu's with less than 4gb of ram from some of the tasks, I'm okay with that.

How would the server divi them up? Or is that a user selects self issue?

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3945
Credit: 46670692642
RAC: 64175321

Einstein’s scheduler has a

Einstein’s scheduler has a bit more logic built into it than most projects. They can make distribution decisions based on the host details, like how much VRAM it has, or in the case of Nvidia cards, by the compute capability and card generation. It can withhold work from GPUs without enough VRAM (as far as BOINC displays, based on the first/best GPU), and also use plan classes to differentiate what kinds of devices get which kind of work, much like how they keep ARM devices to BRP4 and faster x86 CPUs and intel iGPUs to BRP4G

they have a lot of options to fine tune data distribution. 

_________________________________________________________________________

mikey
mikey
Joined: 22 Jan 05
Posts: 12679
Credit: 1839080161
RAC: 3934

Ian&Steve C.

Ian&Steve C. wrote:

Einstein’s scheduler has a bit more logic built into it than most projects. They can make distribution decisions based on the host details, like how much VRAM it has, or in the case of Nvidia cards, by the compute capability and card generation. It can withhold work from GPUs without enough VRAM (as far as BOINC displays, based on the first/best GPU), and also use plan classes to differentiate what kinds of devices get which kind of work, much like how they keep ARM devices to BRP4 and faster x86 CPUs and intel iGPUs to BRP4G

they have a lot of options to fine tune data distribution. 

I thought so!! One thing they might have gotten wrong, but it does work, is they will send me 15 BRP4G tasks to my Intel cpu based machines, and they all run at the same time, when one would think it should only run 1, or 2 if dual cpu's, at a time. I mean it does work and all the tasks DO get credits but it's a single cpu with multiple cores not multiple physical cpu's.

Tom M
Tom M
Joined: 2 Feb 06
Posts: 6434
Credit: 9563307790
RAC: 9352651

mikey wrote: Ian&Steve C.

mikey wrote:

Ian&Steve C. wrote:

Einstein’s scheduler has a bit more logic built into it than most projects. They can make distribution decisions based on the host details, like how much VRAM it has, or in the case of Nvidia cards, by the compute capability and card generation. It can withhold work from GPUs without enough VRAM (as far as BOINC displays, based on the first/best GPU), and also use plan classes to differentiate what kinds of devices get which kind of work, much like how they keep ARM devices to BRP4 and faster x86 CPUs and intel iGPUs to BRP4G

they have a lot of options to fine tune data distribution. 

I thought so!! One thing they might have gotten wrong, but it does work, is they will send me 15 BRP4G tasks to my Intel cpu based machines, and they all run at the same time, when one would think it should only run 1, or 2 if dual cpu's, at a time. I mean it does work and all the tasks DO get credits but it's a single cpu with multiple cores not multiple physical cpu's.

Do  you have your profile setup so you accept gpu tasks that could be run on a cpu?

Tom M

A Proud member of the O.F.A.  (Old Farts Association).  Be well, do good work, and keep in touch.® (Garrison Keillor)  I want some more patience. RIGHT NOW!

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3945
Credit: 46670692642
RAC: 64175321

BRP4G are not exclusively

BRP4G are not exclusively “GPU” tasks anymore. They send to x86 CPUs now. That’s why they changed the name from (Arecibo, GPU), to (Arecibo, Large) 
 

They reduced the size of the BRP4G from 16x BRP4 to only 8x BRP4. Each task runs on a single thread. They are not a multithreaded application for the CPU. The app used when it’s sent to an Intel GPU is not the same as when it’s sent to a CPU. 

_________________________________________________________________________

mikey
mikey
Joined: 22 Jan 05
Posts: 12679
Credit: 1839080161
RAC: 3934

Tom M wrote: mikey

Tom M wrote:

mikey wrote:

Ian&Steve C. wrote:

Einstein’s scheduler has a bit more logic built into it than most projects. They can make distribution decisions based on the host details, like how much VRAM it has, or in the case of Nvidia cards, by the compute capability and card generation. It can withhold work from GPUs without enough VRAM (as far as BOINC displays, based on the first/best GPU), and also use plan classes to differentiate what kinds of devices get which kind of work, much like how they keep ARM devices to BRP4 and faster x86 CPUs and intel iGPUs to BRP4G

they have a lot of options to fine tune data distribution. 

I thought so!! One thing they might have gotten wrong, but it does work, is they will send me 15 BRP4G tasks to my Intel cpu based machines, and they all run at the same time, when one would think it should only run 1, or 2 if dual cpu's, at a time. I mean it does work and all the tasks DO get credits but it's a single cpu with multiple cores not multiple physical cpu's.

Do  you have your profile setup so you accept gpu tasks that could be run on a cpu?

Tom M 

Yes I do, and that could be exactly it except for Ian&Steve answer

mikey
mikey
Joined: 22 Jan 05
Posts: 12679
Credit: 1839080161
RAC: 3934

Ian&Steve C. wrote: BRP4G

Ian&Steve C. wrote:

BRP4G are not exclusively “GPU” tasks anymore. They send to x86 CPUs now. That’s why they changed the name from (Arecibo, GPU), to (Arecibo, Large) 
 

They reduced the size of the BRP4G from 16x BRP4 to only 8x BRP4. Each task runs on a single thread. They are not a multithreaded application for the CPU. The app used when it’s sent to an Intel GPU is not the same as when it’s sent to a CPU. 

That makes sense. I wonder why they never changed it in the Project Preferences page?

Run only the selected applications

Binary Radio Pulsar Search (Arecibo)

Binary Radio Pulsar Search (Arecibo, GPU)

Gamma-ray pulsar binary search #1

Gamma-ray pulsar search #5

Gamma-ray pulsar binary search #1 (GPU)

Gravitational Wave search O2 Multi-Directional

Gravitational Wave search O2 Multi-Directional GPU

Gravitational Wave search O3 All-Sky

 

It's like they also never added (GPU) to the O3 All_Sky tasks either.

Ian&Steve C.
Ian&Steve C.
Joined: 19 Jan 20
Posts: 3945
Credit: 46670692642
RAC: 64175321

yeah that was the only place

yeah that was the only place they didnt change it.

if you look at the task results, you'll see it: https://einsteinathome.org/task/1301112058

_________________________________________________________________________

mikey
mikey
Joined: 22 Jan 05
Posts: 12679
Credit: 1839080161
RAC: 3934

Ian&Steve C. wrote: yeah

Ian&Steve C. wrote:

yeah that was the only place they didnt change it.

if you look at the task results, you'll see it: https://einsteinathome.org/task/1301112058

Yes I see it there

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.