Siever jobs failing on 64-bit Arm SoC

[technik007.cz]

That's why it is important, to report issues - otherwise they can't be fixed at all.

Michael.

Alright, lets cook this pig while water is still boiling !

[yoyo]

Please download this file http://www.rechenkraft.net/yoyo/downloa ... 04-4989.in and store it with the name "in".
Afterwards run sr2sieve with it:
sr2sieve -i in -f factors.txt --pmin 348200000000000 --pmax 348300000000000 -v

This is basically what is done in the workunit and what failed above on your computer.

[technik007.cz]

Please download this file http://www.rechenkraft.net/yoyo/downloa ... 04-4989.in and store it with the name "in".
Afterwards run sr2sieve with it:
sr2sieve -i in -f factors.txt --pmin 348200000000000 --pmax 348300000000000 -v

This is basically what is done in the workunit and what failed above on your computer.

It has been running fine for a few minutes without any issues, so guessing there must be a different problem there. It was always crashing within a few seconds, do you understand?
EDIT: Within few seconds few work units crashed.

[technik007.cz]

root@rpi5:/var/lib/boinc-client/projects/www.rechenkraft.net_yoyo# ./sr2sieve -i in -f factors.txt --pmin 348200000000000 --pmax 348300000000000 -v
sr2sieve 1.9.3 -- A sieve for multiple sequences k*b^n+/-1 or b^n+/-k.
Compiled on Feb 2 2018 with GCC 5.4.0 20160609.
L1 data cache 16Kb (default), L2 cache 256Kb (default).
Reading `in' ...
WARNING: --pmin=348200000000000 from command line overrides pmin=1000000000000 from `in'
Read 370352 terms for 23 sequences from ABCD format file `in'.
Sieving standard form sequences k*795^n+/-1.
Split 23 base 795 sequences into 463 base 795^60 subsequences.
Using 376Kb for subsequence bitmaps.
Using 661Kb for subsequence congruence tables.
Using 4196Kb for Legendre symbol lookup tables.
Using 16Kb for the baby-steps giant-steps hashtable, maximum density 0.20.
Baby step method gen/8, giant step method gen/2, ladder method gen/2.
Using 128Kb for the Sieve of Eratosthenes bitmap.
Expecting to find factors for about 3.18 terms in this range.
sr2sieve 1.9.3 started: 100003 <= n <= 499999, 348200000000000 <= p <= 348300000000000
p=348202054815983, 2861036 p/sec, 0 factors, 2.1% done, 0 sec/factor, ETA 24 Jul 04:54

[yoyo]

It seems that the app is basically running and the difference is just a diferent environment as mentioned by JagDoc in the referenced thread.

[technik007.cz]

This is a log of one of these errors. It is caused by the wrapper or something around it.

stderr out

<core_client_version>7.20.5</core_client_version>
<![CDATA[
<message>
process exited with code 193 (0xc1, -63)</message>
<stderr_txt>
19:30:52 (511261): wrapper (7.7.26016): starting
19:30:52 (511261): wrapper (7.7.26016): starting
terminate called after throwing an instance of 'std::logic_error'
what(): basic_string::_M_construct null not valid
SIGABRT: abort called
Stack trace (3 frames):
[0x49741c]
[0x7fff7f4187b0]
[0x502598]

Exiting...

</stderr_txt>
]]>

[technik007.cz]

BTW Looking for good or bad work units on your web server is very painful. There is no filtering available so everything from all apps is together and my about 50 Arm devices have generated guessing hundreds pages of errors. So you feel like clicking monkey clicking next, next, etc.. 20 times before validated results appear. Then again 20 mouse clicks. Crazy...

I was just watching 8core NanoPC-T6 which failed 16 tasks within 2seconds

[technik007.cz]

...
paused network on this particular machine and going through files sr2...*xml and sr2..*.in left in main project folder.
Then opened one xml file and why is there on 4th line application with exe ending?

<job_desc>
<task>
<application>sr2sieve.exe</application>
<command_line>-i in -f factors.txt --pmin 364200000000000 --pmax 364300000000000 -v</command_line>
<stdout_filename>stdout.txt</stdout_filename>
<stderr_filename>stderr.txt</stderr_filename>
<checkpoint_filename>checkpoint.txt</checkpoint_filename>
<fraction_done_filename>fractiondone.txt</fraction_done_filename>
</task>
<zip_output>
<zipfilename>factor.zip</zipfilename>
<filename>.*(txt|log)</filename>
</zip_output>
</job_desc>

Because when I run it manually with right aarch64 app it runs fine.

[technik007.cz]

This is xml from the finished ecm task. No 'exe' in the app name. But perhaps this is not important because it depends on how wrapper translates this information.

<job_desc>
<task>
<application>ecm</application>
<stdout_filename>out</stdout_filename>
<command_line>-param 0 -v -timestamp -chkpnt checkpnt -inp in</command_line>
<checkpoint_filename>checkpnt</checkpoint_filename>
</task>
<task>
<application>ecm</application>
<stdout_filename>out</stdout_filename>
<command_line>-param 0 -v -timestamp -chkpnt checkpnt -inp in</command_line>
<checkpoint_filename>checkpnt</checkpoint_filename>
</task>
<task>
<application>ecm</application>
<stdout_filename>out</stdout_filename>
<command_line>-param 0 -v -timestamp -chkpnt checkpnt -inp in</command_line>
<checkpoint_filename>checkpnt</checkpoint_filename>
</task>
<task>
<application>ecm</application>
<stdout_filename>out</stdout_filename>
<command_line>-param 0 -v -timestamp -chkpnt checkpnt -inp in</command_line>
<checkpoint_filename>checkpnt</checkpoint_filename>
</task>
<task>
<application>ecm</application>
<stdout_filename>out</stdout_filename>
<command_line>-param 0 -v -timestamp -chkpnt checkpnt -inp in</command_line>
<checkpoint_filename>checkpnt</checkpoint_filename>
</task>
</job_desc>

[technik007.cz]

Today I tried to run a failed work unit via wrapper. So I reused files from the working slot and put them in the project folder it also runs fine and the siever creates checkpoints.

/boinc-client/projects/www.rechenkraft.net_yoyo# ./sievewrapper_145.03_aarch64-unknown-linux-gnu sr2_460500000000000-460600000000000-S820-1721549794-4595.xml
11:50:39 (50579): wrapper (7.7.26016): starting

-rwxr-xr-x 1 boinc boinc 628K Jun 11 20:21 sr2sieve_1.9.3.1_aarch64
-rwxr-xr-x 1 root root 628K Jul 24 11:50 sr2sieve.exe
-rw-r--r-- 1 root root 290 Jul 24 11:51 sr2sieve.log
-rw-r--r-- 1 boinc boinc 82 Jul 24 11:39 stat_icon
-rw-r--r-- 1 root root 775 Jul 24 11:50 stderr.txt
-rw-r--r-- 1 root root 3.5K Jul 24 12:04 stdout.txt
-rw-r--r-- 1 root root 24 Jul 24 12:01 wrapper_checkpoint.txt

[technik007.cz]

I just did check on running wrapper and found it completed the job as standalone. Just reminding this job failed when it was called by boinc client.
Also it is good to mention I stop and started this job as you can see in stdout output.
This is file list of project folder and output of stdout.txt :

-rw-r--r-- 1 root root 2 Jul 26 12:21 boinc_finish_called
-rw-r--r-- 1 boinc boinc 681 Jun 11 20:21 default_stat_icon_01.png
-rw-r--r-- 1 boinc boinc 11020 Jun 11 20:21 ecm_01.png
-rwxr-xr-x 1 boinc boinc 1410416 Jun 14 21:07 ecm-svn3013-aarch64
-rwxr-xr-x 1 boinc boinc 1086024 Jun 14 21:07 ecmwrapper_705.01_aarch64-unknown-linux-gnu
-rw-r--r-- 1 boinc boinc 12574 Jun 11 20:21 evo_01.png
-rw-r--r-- 1 root root 111 Jul 25 10:51 factors.txt
-rw-r--r-- 1 root root 31774 Jul 25 21:17 factor.zip
-rw-r--r-- 1 root root 9 Jul 25 21:16 fractiondone.txt
-rw-r--r-- 1 boinc boinc 1997685 Jul 24 11:39 in
-rw-r--r-- 1 root root 1797 Jul 24 11:49 initial_file_list
-rw-r--r-- 1 boinc boinc 124 Jul 24 11:40 job.xml
-rw-r--r-- 1 boinc boinc 9464 Jun 11 20:21 muon_01.png
-rw-r--r-- 1 boinc boinc 6677 Jun 11 20:21 ogr_01.png
-rw-r--r-- 1 boinc boinc 10749 Jun 11 20:21 rkn_01.png
-rwxr-xr-x 1 boinc boinc 2001704 Jun 11 20:21 sievewrapper_145.03_aarch64-unknown-linux-gnu
-rw-r--r-- 1 boinc boinc 68 Jul 24 11:39 slideshow_00
-rw-r--r-- 1 boinc boinc 68 Jul 24 11:39 slideshow_01
-rw-r--r-- 1 boinc boinc 68 Jul 24 11:39 slideshow_02
-rw-r--r-- 1 boinc boinc 69 Jul 24 11:39 slideshow_03
-rw-r--r-- 1 boinc boinc 68 Jul 24 11:39 slideshow_04
-rwxr-xr-x 1 boinc boinc 634840 Jun 11 20:21 sr1sieve_1.4.5_aarch64
-rw-r--r-- 1 boinc boinc 1997685 Jul 24 11:39 sr2_460500000000000-460600000000000-S820-1721549794-4595.in
-rwxr-xr-x 1 boinc boinc 643040 Jun 11 20:21 sr2sieve_1.9.3.1_aarch64
-rwxr-xr-x 1 root root 643040 Jul 24 11:50 sr2sieve.exe
-rw-r--r-- 1 root root 1008 Jul 25 21:17 sr2sieve.log
-rw-r--r-- 1 boinc boinc 82 Jul 24 11:39 stat_icon
-rw-r--r-- 1 root root 1525 Jul 26 12:21 stderr.txt
-rw-r--r-- 1 root root 185711 Jul 25 21:17 stdout.txt
-rw-r--r-- 1 root root 30 Jul 25 21:17 wrapper_checkpoint.txt


root@tinkerboard-2s:/boinc-client/projects/www.rechenkraft.net_yoyo# cat stdout.txt
sr2sieve 1.9.3 -- A sieve for multiple sequences k*b^n+/-1 or b^n+/-k.
Compiled on Feb 2 2018 with GCC 5.4.0 20160609.
L1 data cache 16Kb (default), L2 cache 256Kb (default).
Reading `in' ...
ERROR: Failed to open input file `in'.
sr2sieve 1.9.3 -- A sieve for multiple sequences k*b^n+/-1 or b^n+/-k.
Compiled on Feb 2 2018 with GCC 5.4.0 20160609.
L1 data cache 16Kb (default), L2 cache 256Kb (default).
Reading `in' ...
WARNING: --pmin=460500000000000 from command line overrides pmin=1000000000000 from `in'
Read 566569 terms for 21 sequences from ABCD format file `in'.
Sieving standard form sequences k*820^n+/-1.
Split 21 base 820 sequences into 699 base 820^60 subsequences.
Using 568Kb for subsequence bitmaps.
Using 741Kb for subsequence congruence tables.
Using 11970Kb for Legendre symbol lookup tables.
Using 16Kb for the baby-steps giant-steps hashtable, maximum density 0.27.
Baby step method gen/8, giant step method gen/4, ladder method gen/2.
Using 128Kb for the Sieve of Eratosthenes bitmap.
Expecting to find factors for about 3.64 terms in this range.
sr2sieve 1.9.3 started: 100000 <= n <= 500000, 460500000000000 <= p <= 460600000000000
sr2sieve 1.9.3 -- A sieve for multiple sequences k*b^n+/-1 or b^n+/-k.TA 25 Jul 21:57
Compiled on Feb 2 2018 with GCC 5.4.0 20160609.
L1 data cache 16Kb (default), L2 cache 256Kb (default).
Reading `in' ...
WARNING: --pmin=460500000000000 from command line overrides pmin=1000000000000 from `in'
Read 566569 terms for 21 sequences from ABCD format file `in'.
Sieving standard form sequences k*820^n+/-1.
Split 21 base 820 sequences into 699 base 820^60 subsequences.
Using 568Kb for subsequence bitmaps.
Using 741Kb for subsequence congruence tables.
sr2sieve 1.9.3 -- A sieve for multiple sequences k*b^n+/-1 or b^n+/-k.
Compiled on Feb 2 2018 with GCC 5.4.0 20160609.
L1 data cache 16Kb (default), L2 cache 256Kb (default).
Reading `in' ...
WARNING: --pmin=460500000000000 from command line overrides pmin=1000000000000 from `in'
Read 566569 terms for 21 sequences from ABCD format file `in'.
Sieving standard form sequences k*820^n+/-1.
Split 21 base 820 sequences into 699 base 820^60 subsequences.
Using 568Kb for subsequence bitmaps.
Using 741Kb for subsequence congruence tables.
Using 11970Kb for Legendre symbol lookup tables.
Resuming from checkpoint pmin=460501675157063 in `checkpoint.txt'.
Using 16Kb for the baby-steps giant-steps hashtable, maximum density 0.27.
Baby step method gen/8, giant step method gen/2, ladder method gen/8.
Using 128Kb for the Sieve of Eratosthenes bitmap.
Expecting to find factors for about 3.64 terms in this range.
sr2sieve 1.9.3 started: 100000 <= n <= 500000, 460501675157063 <= p <= 460600000000000
460515544793693 | 25827*820^448331+1
460558066860431 | 25314*820^267811+1
460568642961277 | 17889*820^349978+1
p=460599969589007, 835241 p/sec, 3 factors, 100.0% done, 33728 sec/factor, ETA 25 Jul 21:17
sr2sieve 1.9.3 stopped: at p=460600000000000 because range is complete.
Found factors for 3 terms in 119983.167 sec. (expected about 3.64)