Ok there have been maybe 3 display driver crashes over the last 2 months...
is a previous driver crash, one that has been fixed before AP launches, likely to effect AP...

I don't turn off my machine.. so it can be weeks between restarts, so if the display driver crashes and restarts is that going to screw with AP until I restart my PC?

---

Ok there have been maybe 3 display driver crashes over the last 2 months...
is a previous driver crash, one that has been fixed before AP launches, likely to effect AP...

I don't turn off my machine.. so it can be weeks between restarts, so if the display driver crashes and restarts is that going to screw with AP until I restart my PC?

I'm not sure how such old driver restart will affect. Perhaps not at all.
But if it was happened while app working - it will affect cause current AMD OpenCL runtime implementation doesn't return from OpenCL API call. Maybe same for NV OpenCL API, can't recall exactly.

EDIT: please do sequence of test actions described earlier. What result will be ?
EDIT2: and better if you will do this on current revision of app (that is, on SETI beta). If OpenCL MultiBeam shows same misbehavior as OpenCL AP on your host then no matter what of apps you will use. And to check if they behave differently or similarly is important part too.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

Also, if you can switch to anonymous platform for testing I can provide debug app to you that will print message to stderr each time it polls for BOINC exit request.
That way we will see directly if app does or doesn't check exit condition. if it does and continue execution then for some reason corresponding flag not set and it's subject of BOINC alpha list instead of this board. If it doesn't ask then it's app own failure and we continue debugging here. But this will require some work on your side too - to configure anonymous platform (correct app_info.xml file).

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

I'm about to head to bed so I can't run thru all of what you have asked.. but here is some of the info you asked about

I have boinc set to stop all tasks (cpu and gpu) on activity
so far I've seen seti7 beta (MB?) CPU and seti6 CPU tasks do as they should (stop when theres PC activity) and AP GPU and seti7 beta GPU tasks not stopping

when I say boinc appears to be running right I mean it says that the task has suspended but the task is still running in task manager... I can suspended the project, app or GPU/CPU in boinc and the task still continues to run in task manager EDIT: I was wrong.. if I force boinc to run tasks I can suspend them fine.. works correctly.. it appears its only when its using the stop on activity setting that things go screwy

I'm not sure what you mean by anonymous platform but I'm sure with some instructions I can set it up

ill come back tomorrow (maybe about 12 hours) and try the things in that list..

---

I'm about to head to bed so I can't run thru all of what you have asked.. but here is some of the info you asked about

Yeah,time difference makes things even harder.

So, app doesn't respond on suspend requests almost in all cases. Well, it becomes increasely interesting. Would be very nice if you setup anonymous platform (perhaps on SETI beta to not damage your production config on SETI main project).
Await download link via PM (to all who also reads this thread and experience same issue - I don't want to post link here cause debug build will provide quite big stderr and should be used only for debugging of this issue. PM me for link).

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

Do you reserve a CPU core?
I see you are running on AMD FX 8350.
Especially on high blanked APs this CPU gets unresponsive if no core is freed.

---

With each crime and every kindness we birth our future.

@Squish
and don't forget to post link to your SETi beta host after return please.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

@Squish
and don't forget to post link to your SETi beta host after return please.

This seems to be it:

Computer 63301

Claggy

@Squish
and don't forget to post link to your SETi beta host after return please.

This seems to be it:

Computer 63301

Claggy

Seems yes, thanks!

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

Example of additional debug output:

DEBUG: polled for exit/suspend request: exit NOT needed. Flags are: boinc_status.quit_request=0, boinc_status.abort_request=0, canRun=1

EDIT: PM with link sent. Awaiting results now.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

Do you reserve a CPU core?
I see you are running on AMD FX 8350.
Especially on high blanked APs this CPU gets unresponsive if no core is freed.

I have boinc set to use all cores (8) at 50% (to keep the CPU fan slow)
I see AP CL GPU uses one core at full speed when it isn't shutting down (thats how i can tell its still running the CPU fan is noisy)

---

Full CPU core load very resembles driver crash and restart. After such event GPU load will be zero and one CPU core load will be full....

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

if there's anything else I can help with let me know

1) you said app acknowledged exit on BOINC quit. It means that at least sometimes flag about exit passed and obeyed. So we need to find out why it's not always passed and/or obeyed. So, try next things:
a) snooze (and look into slot stderr/task manager - will app's process terminates and what delay between your snooze request and process disappearance from task manager's list).
b) snooze GPU (same)
c) suspend project (same)
d) suspend currently executing GPU task (same).

2) as I wrote earlier look into system log - were there any mentions about display driver restarts ?

3) did you configure BOINC to stop all excution on not idle state (user presents /keyboard activity state) or only GPU tasks should stop in this state? Have you any another GPU project active to check if it's only AP issue? Early you said OpenCL (or not?) MB app exits correctly, right ?
(OpenCL MB uses absolutely same style of BOINC critical section handling as OpenCL AP, that's why I'm inclined to consider critical section handling mostly irrelevant to this issue)

EDIT:

seti stops correctly its just astropulse ATI that doesn't

Lets start to be more precise with terminology and distinguish CPU tasks from OpenCL/GPU tasks at least. Both apps are "SETI" in broad sense so perhaps you spoke about SETI MultiBeam (or SETI Enchanced) app. But CPU one or GPU one ?

OK I ran through what you asked above with the non-beta versions as a quick test..
every time I told boinc to suspend tasks, either CPU or GPU by any method it worked without fail... so the only time it doesn't work is when the PC has been idle for 10+ mins and boinc has started everything up and I start using my PC the GPU task wont stop by any method (suspend GPU tasks in the menu or just the running task)

I got the file you sent but I'm not sure what to do for the app_info.xml to run it..

the CPU/GPU load thing you said.. the GPU is still going so I don't think its crashed

---

Hm... check for exit request is absolutely the same app-wise. In all cases it uses ExitCheck call and relies on BOINC to set corresponding flag for it.
And only in signle case BOINC doesn't set that flag...
For final conclusion it would be good if you would run debug build, but looks like it's BOINC API issue...

Regarding app_info.xml
just extract all files into seti beta project dir (BOINC turned OFF) then copy *.aistub into app_info.xml (or just rename it instead of copy).

you will get minimal app_info just for this particular app. Don';t do it on main cause you will lose all downloaded tasks this way.
On beta you should redownload 1-2 tasks after that (no more needed for our testing).
Then leave PC for 10 minutes as usual, come back, check if this debug app doesn't suspend too, copy whole stderr.txt file from slot to another location and send it to me via E-mail or by any other way.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

that didn't work it DLed and ran all the normal beta files not the one you supplied

---

that didn't work it DLed and ran all the normal beta files not the one you supplied

BOINC was switched OFF when you put all those files into project dir?
Not just suspended, full exit required.
Also, maybe project resen is needed too. I experienced some issue too when switched from stock to anonymous platform.
Maybe our more experienced in such transition friends will come with some advice.
For now try to exit BOINC fully, copy again, make sure that app_info.xml in place before starting BOINC and when started do project (SETI beta) reset.

EDIT: and after reset maybe one more BOINC restart.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

ok boinc complained in notices
11/05/2013 11:09:54 a.m. | SETI@home Beta Test | Message from server: Unknown app name in app_info.xml
11/05/2013 11:09:54 a.m. | SETI@home Beta Test | Message from server: Your app_info.xml file doesn't have a usable version of SETI@home v7.
11/05/2013 11:09:54 a.m. | SETI@home Beta Test | Message from server: Your app_info.xml file doesn't have a usable version of AstroPulse v6.

---

ok boinc complained in notices
11/05/2013 11:09:54 a.m. | SETI@home Beta Test | Message from server: Unknown app name in app_info.xml
11/05/2013 11:09:54 a.m. | SETI@home Beta Test | Message from server: Your app_info.xml file doesn't have a usable version of SETI@home v7.
11/05/2013 11:09:54 a.m. | SETI@home Beta Test | Message from server: Your app_info.xml file doesn't have a usable version of AstroPulse v6.

Doh, seti_v7 on beta.
Please edit app_info and change setiathome_enhanced to setiathome_v7
all occurencies.

app_info was for main as part of usual release pack. And we try to fit it to beta with V7 tasks instead of V6 ones...

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.

ok that worked...
so it still didn't quit when there was activity... I checked the stderr and it didn't get the quit message until I quit boinc..
http://temp-share.com/show/dPf3068hW

---

ok that worked...
so it still didn't quit when there was activity... I checked the stderr and it didn't get the quit message until I quit boinc..
http://temp-share.com/show/dPf3068hW

DEBUG: polled for exit/suspend request: exit NOT needed. Flags are: boinc_status.quit_request=0, boinc_status.abort_request=0, canRun=1
DEBUG: polled for exit/suspend request: exit needed. Flags are: boinc_status.quit_request=1, boinc_status.abort_request=0, canRun=1
GPU device synched
Termination request detected or computations are finished. GPU device synched, exiting...

So, app checks exit condition. Try to report this issue on BOINC alpha mail list or on BOINC support forums. There is definitely something to fix... but not on app side.

EDIT: maybe, there is even one another BOINC flag to check specially for this kind of exit? App aware only abouit listed flags. Will try to clarify this on BOINC dev list.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.