I would like to let run the stock S@h v7 apps with app_info.xml file.

The stock S@h v7 6.98 for CPU and 6.98_cuda23 for GPU.

How should look the app_info.xml file entries?

I tried it, but failed.

Thanks.


- Best regards! :-) - Sutaru Tsureku, team seti.international founder. - Optimize your PC for higher RAC (@ SETI@home Main). - SETI@home needs your help. -

---

Please don't, not yet. The time and place for running optimised applications is on the Main project - after testing and launch - where the improved efficiency is valuable for the science.

Here, we need to test out the back-end plumbing for work supply, return, validation and - yes - credits, before we can move on to the next steps.

I would like to let run still the stock S@h v7 apps.

My machine got the cuda22, 23 & 32 apps.
The fastest is the cuda23 app, 22 need double time, 32 little bit slower.

I would like to increase the output of S@h v7 results on my machine, so I would need to use an app_info.xml file with cuda23 app/entries (also 6.98 CPU).

If it's not for public - to now, I'm also happy about a private message with the entries for an app_info.xml file.


- Best regards! :-) - Sutaru Tsureku, team seti.international founder. - Optimize your PC for higher RAC (@ SETI@home Main). - SETI@home needs your help. -

---

OK - to explain in more detail.

The trouble is, when you use an app_info.xml file, every speed report you submit, whatever app is actually running behind the anonymous mask, gets lumped into one great big pot.

But when you allow the project to allocate stock applications, the speed reports that you and everybody else submits are kept apart, version by version. That's the way that the project database will 'learn' which application is fastest - both for the project as a whole, and for your machine in particular.

Please allow stock to run for long enough to populate your fair share of the statistical base record.

Please don't, not yet. The time and place for running optimised applications is on the Main project - after testing and launch - where the improved efficiency is valuable for the science.

Here, we need to test out the back-end plumbing for work supply, return, validation and - yes - credits, before we can move on to the next steps.

Are you positive that this is a true statement? When Eric Korpela said we could use the anonymous platform a few weeks ago after the server problems were solved, I beleive I told him of my plans to use app_info.xml, but that I would not do so if he objected. He did not object. I use app_info.xml and have been for weeks. It is such a waste of resources (electricity, time, etc.) to have all these CPU architectural enhancements and not be able to use them.

The server has to parse the <stderr> section of the result anyway, because there is nothing to prevent the user from changing the plan class from CPU to GPU or VV.

---

Please don't, not yet. The time and place for running optimised applications is on the Main project - after testing and launch - where the improved efficiency is valuable for the science.

Here, we need to test out the back-end plumbing for work supply, return, validation and - yes - credits, before we can move on to the next steps.

Are you positive that this is a true statement? When Eric Korpela said we could use the anonymous platform a few weeks ago after the server problems were solved, I beleive I told him of my plans to use app_info.xml, but that I would not do so if he objected. He did not object. I use app_info.xml and have been for weeks. It is such a waste of resources (electricity, time, etc.) to have all these CPU architectural enhancements and not be able to use them.

The server has to parse the <stderr> section of the result anyway, because there is nothing to prevent the user from changing the plan class from CPU to GPU or VV.

Well, obviously whatever Eric says is definitive - only he has the complete overview of what needs testing, observing, studying, filling in, and correcting. And those needs may change from time to time.

I was merely pointing out to the OP that, in order to study and test the system as a whole - which Eric is clearly having to do, given his comments about the state of some of the server code yesterday - we need to have volunteers to populate the statistical tables for the stock workflow path: see my comments in the News area just now concerning what happens to runtime estimates when that data is incomplete.

With regard to your comments on inefficiency - I haven't looked to see what range of hardware you're running, and what applications you may have installed under app_info to run on it. But for the newer nVidia architectures - Fermi and Kepler - I think you'll find that the new stock applications installed this week are more efficient than anything you yet have (that statement doesn't apply to older 8xxx/9xxx/2xx cards). And - I honestly don't know the answer to this yet - you have checked that the workunits we search as part of this beta test project are genuinely new science, and added to the master ET search database and the NTPCKR? Unless the results are used, any question of efficiency rather goes out of the window.

...
The server has to parse the <stderr> section of the result anyway, because there is nothing to prevent the user from changing the plan class from CPU to GPU or VV.

BOINC does not support that kind of change. If it has sent a task to be done with a particular application, the result affects that application's statistics regardless of what was actually used.

For setiathome_v7, the stderr part of the metadata is parsed to see if it contains "result_overflow", in which case the validator flags the task as a runtime_outlier. That is the only use made of the stderr by the servers.

                                                                    Joe

And - I honestly don't know the answer to this yet - you have checked that the workunits we search as part of this beta test project are genuinely new science, and added to the master ET search database and the NTPCKR? Unless the results are used, any question of efficiency rather goes out of the window.

It's VERY long-standing question I tried to get answered too...
Eric, can you say something definitive on this point please?
Do beta results go in common scientific pool or not ?