The first task has completed - took about 50% longer than previous guppi tasks.

Unfortunately ended as Inconclusive but at least it finished without restarting and without a memory leak.

http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24264444

Thanks, I think -use_sleep can be removed.

---

News about SETI opt app releases: https://twitter.com/Raistmer

2nd one was another inconclusive but no restarts and no memory leak

http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24263674

I have now removed -use_sleep

I aborted all the non-guppi tasks but then fell foul of This computer has completed a daily quota .... and it refuses to get any fresh tasks :-)

Unfortunately I couldn't find any download packages for the CPU tasks with an aistub file I could vandalise. Never mind, I'll leave this client running the guppi tasks overnight and will restart stock in the morning, unless you want me to test something else?

The first task has completed - took about 50% longer than previous guppi tasks.

Unfortunately ended as Inconclusive but at least it finished without restarting and without a memory leak.

http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24264444

Thanks, I think -use_sleep can be removed.

Yeah, I tried -use_sleep on r3430 on main too, with my iGPU HD4600, driver version: 10.18.10.3907. No memory leaks/counting up working memory, or any other memory.

I'll remove -use_sleep, and returns to my original commandline, which was:

-cpu_lock -sbs 256 -period_iterations_num 40 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 512 -instances_per_device 1 -total_GPU_instances_num 4

---

WARNING!! "THIS IS A SIGNATURE", of the "IT MAY CHANGE AT ANY MOMENT" type. It may, or may not be considered insulting, all depending upon HOW SENSITIVE THE VIEWER IS, to certain inputs to/from the nervous system.

Well, next guess: for some reason queues withprofiling enabled result in memory leak on some iGPU/driver combos.
So I disable all parts of adaptation algorithm that rely onto kernel execution times. Will see if it will fix memory leak.
New build comes soon.

---

News about SETI opt app releases: https://twitter.com/Raistmer

Here is r3484 https://cloud.mail.ru/public/LqtQ/YBsLcr8nf.
Please test, especially iGPU one.

---

News about SETI opt app releases: https://twitter.com/Raistmer

EDIT-
From the Release notes for v5.24
- Fixed a memory leak in server side of remote sensor monitoring.

Ummm, maybe. Was still seeing problems on 5.30, running 5.32 now and will track more closely. Thanks for the release info, I had missed that when you initially posted. l8r

---

If I can help out by testing something, please let me know.
Available hardware and software is listed in my profile here.

Ok, testing the new app.

No iGPU on my machines.

Here is r3484 https://cloud.mail.ru/public/LqtQ/YBsLcr8nf.
Please test, especially iGPU one.

I can't test it now. I was fully asleep when you released it, and now I'm at work. Maybe later Friday evening, or during the weekend.

One might almost think that we're too few Beta testers with iGPU's running :-)

---

WARNING!! "THIS IS A SIGNATURE", of the "IT MAY CHANGE AT ANY MOMENT" type. It may, or may not be considered insulting, all depending upon HOW SENSITIVE THE VIEWER IS, to certain inputs to/from the nervous system.

One might almost think that we're too few Beta testers with iGPU's running :-)

A lot will be after release to main as usual...

---

News about SETI opt app releases: https://twitter.com/Raistmer

SETI@Home Informational message -9 result_overflow
WU true angle range is : 0.429549
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24283038


App will use system sleep quantum of size 3ms
SETI@Home Informational message -9 result_overflow
WU true angle range is : 0.429487
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24283058

No commandline
WU true angle range is : 0.429487
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24283335

-use_sleep -v 6
WU true angle range is : 0.429549
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24282349

App will use system sleep quantum of size 3ms
WU true angle range is : 0.429487
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24283310

No commandline
WU true angle range is : 0.008168
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24282797

-use_sleep -v 6
WU true angle range is : 0.008168
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24282429

App will use system sleep quantum of size 3ms
WU true angle range is : 0.008175
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24283818

No High angle work units, sorry

Edit..
Modified to show new commandline

SETI@Home Informational message -9 result_overflow
WU true angle range is : 0.429549
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24283038

WU true angle range is : 0.429549
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24282349

WU true angle range is : 0.008168
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24282429

No High angle work units, sorry

Sleep() & wait for event loops will be used in some places

Take a note, you currently have use_sleep enabled, FYI.

---

News about SETI opt app releases: https://twitter.com/Raistmer

Yes I know, I thought you wanted that installed as well as -v 6

I just added the -high_perf to the commandline just now.

Do you want test samples without any commandlines?

And for mid-range AR (WU true angle range is : 0.429549) your host able to sleep enough only on few PulseFind configs:

Fftlength=8,pass=3:Tune: sum=4793.56(ms); min=5.198(ms); max=20.38(ms); mean=14.31(ms); s_mean=12.27; sleep=15(ms); delta=97; N=335; usual
Fftlength=8,pass=4:Tune: sum=3108.09(ms); min=2.929(ms); max=19.45(ms); mean=14.06(ms); s_mean=13.41; sleep=15(ms); delta=128; N=221; usual
Fftlength=8,pass=5:Tune: sum=2351.24(ms); min=3.735(ms); max=16.07(ms); mean=13.59(ms); s_mean=13.52; sleep=15(ms); delta=147; N=173; usual
Fftlength=16,pass=3:Tune: sum=2609.91(ms); min=1.509(ms); max=16.84(ms); mean=11.45(ms); s_mean=13.75; sleep=15(ms); delta=186; N=228; usual
Fftlength=16,pass=4:Tune: sum=1876(ms); min=1.473(ms); max=16.26(ms); mean=9.822(ms); s_mean=13.58; sleep=15(ms); delta=233; N=191; usual

So, with default config + use_sleep we have here:

Время выполнения	4 мин. 23 сек.
Время ЦП	3 мин. 47 сек.

CPU almost same as GPU.

Try to add this option to your current tune line:
-sleep_quantum 3

How this will affect elapsed and CPU times for midrange ARs?

---

News about SETI opt app releases: https://twitter.com/Raistmer

Here is r3484 https://cloud.mail.ru/public/LqtQ/YBsLcr8nf.
Please test, especially iGPU one.

Just started the Intel app running - no command line switches at the moment - hope that's what you wanted?

Just started the Intel app running - no command line switches at the moment - hope that's what you wanted?

Exactly.
First check how defaults work then tuning attempt.
App intended to become "stock" one so it means it will operate unattended on most hosts. Do you still see memory leak?

---

News about SETI opt app releases: https://twitter.com/Raistmer

WU true angle range is : 0.008168
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24282429

And for GUPPI task:

Время выполнения	7 мин. 50 сек.
Время ЦП	4 мин. 44 сек.

little less than 50% savings on CPU time.

Looking into adaptation statistics:

Fftlength=32,pass=3:Tune: sum=34192.2(ms); min=5.368(ms); max=23.42(ms); mean=14.89(ms); s_mean=12.89; sleep=15(ms); delta=174; N=2296; usual
Fftlength=32,pass=4:Tune: sum=30779.1(ms); min=5.652(ms); max=22.7(ms); mean=14.85(ms); s_mean=13.26; sleep=15(ms); delta=176; N=2072; usual
Fftlength=32,pass=5:Tune: sum=27283.8(ms); min=6.132(ms); max=23.91(ms); mean=14.74(ms); s_mean=13.18; sleep=15(ms); delta=177; N=1851; usual
Fftlength=64,pass=3:Tune: sum=21673.7(ms); min=2.447(ms); max=19.37(ms); mean=14.41(ms); s_mean=13.69; sleep=15(ms); delta=275; N=1504; usual
Fftlength=64,pass=4:Tune: sum=16998.7(ms); min=2.809(ms); max=17.03(ms); mean=13.74(ms); s_mean=14.12; sleep=15(ms); delta=310; N=1237; usual
Fftlength=64,pass=5:Tune: sum=10864.4(ms); min=3.003(ms); max=16.38(ms); mean=13.03(ms); s_mean=14.81; sleep=15(ms); delta=438; N=834; usual
Fftlength=128,pass=3:Tune: sum=11556.3(ms); min=1.353(ms); max=21.02(ms); mean=11.42(ms); s_mean=13.59; sleep=15(ms); delta=680; N=1012; usual
Fftlength=128,pass=4:Tune: sum=9384.34(ms); min=1.457(ms); max=18.52(ms); mean=11.11(ms); s_mean=15.82; sleep=15(ms); delta=618; N=845; usual
Fftlength=128,pass=5:Tune: sum=8291.81(ms); min=1.532(ms); max=15.07(ms); mean=10.29(ms); s_mean=14.21; sleep=15(ms); delta=552; N=806; usual
Fftlength=256,pass=3:Tune: sum=11076.6(ms); min=0.6827(ms); max=19.64(ms); mean=10.41(ms); s_mean=13.81; sleep=15(ms); delta=680; N=1064; usual
Fftlength=256,pass=4:Tune: sum=8906.34(ms); min=0.7239(ms); max=16.38(ms); mean=9.639(ms); s_mean=15.36; sleep=15(ms); delta=621; N=924; usual
Fftlength=256,pass=5:Tune: sum=7841.21(ms); min=0.7803(ms); max=14.95(ms); mean=8.91(ms); s_mean=13.57; sleep=15(ms); delta=670; N=880; usual

That constitutes ~50% of PulseFind configs (though they have different impact so different weights).

---

News about SETI opt app releases: https://twitter.com/Raistmer

Edited my original post to show default, -use_sleep and lastly -sleep_quantum 3

Do you want -high_perf added?

Время выполнения	4 мин. 23 сек.
Время ЦП	3 мин. 47 сек.

CPU almost same as GPU.

And w/o sleep:

Время выполнения	4 мин. 22 сек.
Время ЦП	4 мин. 17 сек.

elapsed almost same but CPU considerably increased. So, CPU savings app provides have quite low impact on performance. At least in this test case.

---

News about SETI opt app releases: https://twitter.com/Raistmer

WU true angle range is : 0.008168
https://setiweb.ssl.berkeley.edu/beta//result.php?resultid=24282429

And for GUPPI task:

Время выполнения	7 мин. 50 сек.
Время ЦП	4 мин. 44 сек.

little less than 50% savings on CPU time.

and w/o sleep:

Время выполнения	7 мин. 38 сек.
Время ЦП	7 мин. 22 сек.

here sleep impact more noticeable (with bigger overhead at default config).

---

News about SETI opt app releases: https://twitter.com/Raistmer

Do you want -high_perf added?

IMHO It will have almost no impact but try it as part of best performance tuning. Now app switches to high-perf path automatically in more places.

---

News about SETI opt app releases: https://twitter.com/Raistmer