I feel that users should be able to define their own queue for the tasks in progress. This would also solve the problem of switching between projects. Some users complain when the program switches between tasks minutes before they are finished. If we could define our own queue, we could tell the program which task to perform first.

The default queue should be an option between first-in-first-out and the traditional deadline date. After that, we should be able to customize the queue by line item. For the GPU tasks, the program would skip over non GPU tasks when a GPU slot is available.

I'd like to second that idea, or something even more in depth. The global 'resource share' preferences seem like a pretty inefficient and clunky way of managing how tasks run. To make the best use of resources on local machines it would be best to have local preferences for this.

On my home desktop for example I primarily run Einstein which can crank through BRP tasks very quickly using the GPU - so I want the GPU running at all times. If I try and run Rosetta on the same machine though, it will get stuck in long periods where it is only running Rosetta (CPU only) tasks and nothing for Einstein (or vice versa).

This is just an inefficient use of my resources - what I want is to be able to dedicate the GPU at all times to Einstein and dedicate a couple of CPU cores at all times to Rosetta, which would leave them both crunching away uninterrupted. From what I have gathered reading around here though it doesn't sound like this is planned to ever be implemented. Has there ever been a rationale given for this? The only real way to do what I want right now is to physically separate these projects on different machines, which isn't a great solution.

what I want is to be able to dedicate the GPU at all times to Einstein and dedicate a couple of CPU cores at all times to Rosetta, which would leave them both crunching away uninterrupted.

And you haven't tried setting that up that way then? In the Einstein Project preferences, set "Use CPU" to No, set whichever "Use GPU" to Yes. Set the work you want to receive to only the BRP4s and BRP5s. Set "If no work for these, accept work for CPU" to No.

At Rosetta project preferences, set only "Use CPU" to Yes and whichever work you want to run on it. (I see Rosetta still uses very old back-end software, so no choice of what to run. It'll use the CPU automatically then)

I have had such a setup for the past year where Einstein runs only on my GPU and Seti runs only on 3 of my CPU cores. The latter core is free to supply the GPU.
If you can't get that to work that way, you either have an old BOINC version that does things differently, or you have set the project preferences wrong.

Einstein project preferences wrote:

Resource share 800
Use CPU no
Use ATI GPU yes
Use NVIDIA GPU no
Use INTEL GPU no
Run only the selected applications
Binary Radio Pulsar Search (Arecibo): no
Binary Radio Pulsar Search (Arecibo, GPU): yes
Binary Radio Pulsar Search (Perseus Arm Survey): yes
Gravitational Wave S6 Directed Search (CasA) : no
Gamma-ray pulsar search #2: no
Run CPU versions of applications for which GPU versions are available no

Seti project preferences wrote:

Resource share 800
Use CPU yes
Use ATI GPU no
Use NVIDIA GPU no
Run only the selected applications
SETI@home Enhanced: yes
SETI@home v7: yes
AstroPulse v6: yes
If no work for selected applications is available, accept work from other applications? no

---

This is just an inefficient use of my resources - what I want is to be able to dedicate the GPU at all times to Einstein

At Rosetta deselect 'Use ATI GPU' or 'Use NVIDIA GPU' as appropriate.

and dedicate a couple of CPU cores at all times to Rosetta, which would leave them both crunching away uninterrupted.

Use the app_config.xml mechanism to limit the max current tasks running at Einstein and Rosetta:

Application configuration

Claggy

Doh - I didn't realize there was a "Use CPU" flag for Einstein. I'll try that and see what happens. Will be very happy if I can get it working this way.

Seems like the .xml files offer a lot more control than is in the boinc manager GUI - is any of that planned to be in the GUI at some point?

Seems like the .xml files offer a lot more control than is in the boinc manager GUI - is any of that planned to be in the GUI at some point?

If you're talking about the app_config.xml file, then you're better off using the option for it in the Einstein project preferences:

GPU utilization factor of BRP apps
DANGEROUS! Only touch this if you are absolutely sure of what you are doing!
Wrong setting might even damage your computer! Use solely on your own risk!
Min: 0.0 / Max: 1.0 / Default: 1.0

You can't use this one at Rosetta anyway, as Rosetta doesn't have a GPU application, nor is their back-end code up-to-date enough to know what you mean with an app_config.xml setup.
So then you could use it at Einstein, but in this case, Einstein's own solution is safer. Less to fiddle with and get wrong.
When you set the GPU utilization factor to 0.5, it means that two tasks can run on the GPU simultaneously.

And no, for the foreseeable future, none of that will make it into BOINC Manager.

---