Page 2 of 2

Re: EPW crashes while reading ephmat files

Posted: Thu Jan 28, 2021 11:09 pm
by mdogan
Dear H. Lee,

You are right, the file is corrupted when about 12% of it has been written, if my interpretation is correct, which is a bad sign. I'm currently looping over the other 127 ephmat files with your test script to see if they have any I/O errors (so far they are fine). I guess if only 1 file was corrupted out of 128, I can have some hope that next time that number might be 0. Do you have any other suggestions? Recompiling the code, reducing the #procs, etc.? Thank you for all your help!

Best,
Mehmet

Re: EPW crashes while reading ephmat files

Posted: Fri Jan 29, 2021 12:01 am
by hlee
Dear Mehmet:

Could you let me know the compilers, the libraries, and their versions you used in building QE and EPW on Frontera?

PS) When checking all of your ephmat files, you had better remove the line of "PRINT *, i, g2" in the code above since Frontera has very strict rules on I/O; you might be blocked due to the heavy I/O.

Sincerely,

H. Lee

Re: EPW crashes while reading ephmat files

Posted: Fri Jan 29, 2021 1:16 am
by mdogan
Dear H. Lee,

Thank you for reminding me about the I/O issues. Luckily I had commented out that line. Also, all the ephmat files (except ephmat2) passed the test.

I added the make.inc file I used to the attachments [https://www.dropbox.com/sh/zwjvc2hq600k ... 7bYxa?dl=0]. I think it should have all the relevant information regarding the compilers and libraries, but if you need any other information, please let me know how I can get it. Thank you!

Best,
Mehmet

Re: EPW crashes while reading ephmat files

Posted: Fri Jan 29, 2021 6:36 pm
by hlee
Dear Mehmet:

I can't find anything in your make.inc that seems to be directly linked to your issue.
You can try to rerun, but as I said, basically as far as I/O issue is concerned the success of your runs is unpredictable.

I have a few plans and am doing something to enhance the reliability and efficiency of heavy I/O in EPW, but for the moment, we have little to do.

Sincerely,

H. Lee

Re: EPW crashes while reading ephmat files

Posted: Sat Jan 30, 2021 5:42 pm
by mdogan
Dear H. Lee,

Thank you very much for providing greatly valuable information, as always. I'm going to retry the calculation and give you an update. Thank you also for working on improving the reliability of heavy I/O. Please let me know when there are any updates to the code relevant to this issue.

Best,
Mehmet