Mysterious Ready to Report

[David G. Pickett]

I have an Einstein task ready to report, but no delay showing on the project tab, and it persists even if I hit tools-retry pending.

[Glenn Jetton]

I have a Climate Prediction that was due on 11/1/2021.  It continues to run as though it is starting over and over.  This time it is up to 9 days.  I has been as high as 21 days and suddenly it starts over, still showing due on 11/1/2021.

--
You received this message because you are subscribed to the Google Groups "boinc_projects" group.
To unsubscribe from this group and stop receiving emails from it, send an email to boinc_project...@ssl.berkeley.edu.
To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/1726448417.657502.1645316024226%40mail.yahoo.com.

[Glenn Jetton]

To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/003501d825fd%244edc8ca0%24ec95a5e0%24%40cox.net.

[David G. Pickett]

My other PC has a Moo in the same state.  I am just saying, if they are ready to report, why aren't they reporting?  It's a self-contradictory status.

To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/006e01d82600%24e06a0690%24a13e13b0%24%40cox.net.

[David G. Pickett]

I put up a new computer and it did 2 things that seemed off:
 - 9 projects were downloading at once.  It seems more productive to limit this to N, perhaps 2, at once, so they get done downloading and start sooner.  Imagine if two people in a laundromat have 2 loads to dry each, and only 2 dryers are open.  If one says go ahead, the other gets done in one cycle and the other in 2 cycles.  If they both take one, they both get done in 2 cycles.  Sharing is a bad strategy.

 - The downloads stopped when the computer was in use.  There is no separate control for this, so I changed the setting to process when computer in use but set the suspend for not boinc use lower.  There should be better controls that allow downloads and uploads at all times, but as above maybe only 2 at a time, or 1 at a time if computer in use.

To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/992718146.661147.1645325538655%40mail.yahoo.com.

[Vitalii Koshura]

Hello David,

Thank you for your report.

I can't say anything about the first point, but regarding the second one, you can control it via the menu:

Best regards,

Vitalii Koshura

пт, 25 февр. 2022 г. в 19:08, 'David G. Pickett' via boinc_projects <boinc_p...@ssl.berkeley.edu>:

To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/786211044.2726594.1645812486649%40mail.yahoo.com.

[Eric Driver]

While we're knit-picking about the client, here is something that has bothered me for a while...

Is there a way to have the client process tasks based on the deadline and not the received time?  For example, when a task is resent with a reduced delay bound (i.e. earlier deadline), I would like my client to run that task with higher priority instead of running the lower priority tasks.  This behavior seems to defeat the purpose of the reduced delay bound.  I should point out that the "accelerated retry" mechanism on the server is working as expected, but it seems that the client ignores the shorter deadline.

To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/786211044.2726594.1645812486649%40mail.yahoo.com.

[Eric Korpela]

I believe this is a long standing bug.  When a result has been received and reported, but the report status is not successfully reported to the client, the client thinks the result has not been reported and the server reports a failure of subsequent attempts to report (maybe because the result has been deleted from the database? or maybe it just standard to report a failure when a result has already been reported?)

I see this every few months or so.  I reinstall too frequently to know how long it lasts.

To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/ef93781e-b4fd-b8b7-5610-0bf24be04201%40cox.net.

--

Eric Korpela

kor...@berkeley.edu
AST:7731^29u18e3

[David G. Pickett]

BOINC projects sometimes seem host fussy, as my older 2 core AMD Linux box with GPU has no tasks but my older win10 8 core Ryzen laptop and brand new Win 11 Intel Core I7 AIO are running tons of work (the I7 so far not finishing any, which goes back to my sharing comments).

Do BOINC tasks make good use of hyperthreading?  Most x86 hosts now have this!  I guess you need to have 2+ threads per process for this to work for you!  Hyperthreading so reminds me of the old H-800, running up to 8 programs in one core main memory with 8 sets of 32 registers (registers were in a relatively faster core memory)!

To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/CANGNY56ZvqOO4qvakR5bXeD0nBP%3D323E3WgOLP4K2p-cm_t%3DLw%40mail.gmail.com.

[Eric Korpela]

Unless something has changed BOINC doesn't distinguish between threads and cores.  If a processor has six dual threaded cores, BOINC treats it exactly the same as if it had 12 cores.  On systems with asymmetric processors or fewer NDPs or SIMD units than there are cores (i.e. some ARM processors used in phones have cores with different capabilities and different clock speeds or will shut down some cores when the phone is not in active use), BOINC assumes that all processors are equivalent and uses the processor count that the OS provides and assumes that they are all available all the time.

To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/1566688159.3046795.1645909528861%40mail.yahoo.com.

[David G. Pickett]

Hyperthreading works when a core has 2 threads in the same process, so they can share the VM view, caches, just have different registers, instruction buffer and compete for the same core CPU resources, many of which are now redundant or lightly used: integer units, float units, multipliers.  x86 dispatchers can only assign a second thread to a core if it is the same pid.  Some boinc devs do seem to do multi-threading.

It's smart not to leave money on the table!  :D

My empty boinc filled up overnight, which makes one wonder if the workload is making itself available promptly, or are users feeding work to the world gradually?

To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/CANGNY564Bzd0-2qa0RSyKOZytBmY%2BOBX9BcViQ5ZxiB3mnZHFw%40mail.gmail.com.

[Charlie Fenton]

FWIW, it's not just ARM processors used in phones that have cores with different capabilities. The current M1 Apple Silicon (ARM) Macs have an "8-core CPU with 4 performance cores and 4 efficiency cores", according to <https://www.apple.com/imac-24/specs/>,

Cheers,
--Charlie

> To view this discussion on the web visit https://groups.google.com/a/ssl.berkeley.edu/d/msgid/boinc_projects/CANGNY564Bzd0-2qa0RSyKOZytBmY%2BOBX9BcViQ5ZxiB3mnZHFw%40mail.gmail.com.