CADO-NFS error (exit code -9)
 

Hi all, I am factoring a C143 but the program stopped in error, with no mention of what kind. Here are the last few lines diplayed.

[CODE]Info:Square Root: Creating file of (a,b) values
Warning:Command: Process with PID 1917 finished with return code -9
Error:Square Root: Program run on server failed with exit code -9
Error:Square Root: Command line was: /home/ubuntu/cado-nfs/build/ip-172-31-36-46/sqrt/sqrt -poly nfsdata/c145.poly -prefix nfsdata/c145.dep.gz -dep 0 -t 8 -side0 -side1 -gcd > nfsdata/c145.sqrt.stdout.4 2> nfsdata/c145.sqrt.stderr.4
Error:Square Root: Stderr output (last 10 lines only) follow (stored in file nfsdata/c145.sqrt.stderr.4):
Error:Square Root: Alg(1): starting level 3 at cpu=1962.67s (wct=257.15s), 8 values to multiply
Error:Square Root: Alg(7): level 1 took cpu=147.07s (wct=18.52s)
Error:Square Root: Alg(7): starting level 2 at cpu=1964.10s (wct=257.33s), 16 values to multiply
Error:Square Root: Alg(4): level 2 took cpu=167.89s (wct=21.14s)
Error:Square Root: Alg(4): starting level 3 at cpu=1964.98s (wct=257.44s), 8 values to multiply
Error:Square Root: Alg(2): level 2 took cpu=167.94s (wct=21.15s)
Error:Square Root: Alg(2): starting level 3 at cpu=1967.93s (wct=257.81s), 8 values to multiply
Error:Square Root: Alg(3): level 2 took cpu=166.18s (wct=20.93s)
Error:Square Root: Alg(3): starting level 3 at cpu=1982.80s (wct=259.69s), 8 values to multiply
Error:Square Root:
Traceback (most recent call last):
File "cado-nfs/cado-nfs.py", line 122, in <module>
factors = factorjob.run()
File "cado-nfs/scripts/cadofactor/cadotask.py", line 5957, in run
last_status, last_task = self.run_next_task()
File "cado-nfs/scripts/cadofactor/cadotask.py", line 6049, in run_next_task
return [task.run(), task.title]
File "cado-nfs/scripts/cadofactor/cadotask.py", line 4940, in run
raise Exception("Program failed")
Exception: Program failed[/CODE]

I resumed the computation and had the same issue. Can it just be that not enough memory is available? Thank you in advance.

Yes, this looks like running out of memory. By default CADO runs one dependency per thread which makes the square root very memory-intensive.

Try adding a line
[code]tasks.sqrt.threads = 1[/code]
to your parameter file and resuming - this will cause sqrt to run single-threaded so memory use is lower.

Thank you, it worked.

Another way that I've used with larger runs is to add a swap file so I could still use full threads.

Both methods seem to work fine. Since I am running remote instances with limited disk space, and also considering that the sqrt phase takes a negligible time compared to the rest of the factorization, I'll stick with the thread tweaking for now.

I work with 8 threads so I expect the default is tasks.sqrt.threads = 8 and indeed I monitored the memory usage in semi-real time and it overflows. I tried a C140 with tasks.sqrt.threads = 4 and it completed fine. To make it work I edited the <workdir>/c140.parameters_snapshot.<x> file after the error fired so I was able to save the previous computation.

Is there a way to make this change permanent? Should I edit the files in cado-nfs/parameters/factor/ ? I assume for example that for a C140 the relevant file is cado-nfs/parameters/factor/params.c140, is that correct?

Thank you again.

You can edit the params files or you can add the option in your command line. I do the latter on a continual basis with my scripts. Just add "tasks.sqrt.threads=4" within the command:
[code]
./cado-nfs.py <comp> . . . tasks.sqrt.threads=4 . . .
[/code]I pull the spaces around the "=" out when I add an option to the command line, but I probably don't need to. Options added to the command line will show up in the snapshots.

Ah, that's useful! Thank you very much!