Dimitris, 2017-11-27 17:10 GMT-05:00 Dimitris Ntogkas <di...@math.umd.edu>:
> Thanks for your quick response! You are right about the l2 vs max norm. > However, the error is 1e-4 in the l2 norm too. Just a clarification to make > sure I understand your response. I was indeed thinking of the condition > number, that's why I checked it, but in my case the 1e-11 should lose up to > 5 more digits, which is still better than 1e-4. However, probably your > point is that since I was using cg with tolerance of 1e-8, this is already > a loss of accuracy that I did not take into account in the above > calculation. Is this correct? > Here is what I am thinking. The tolerance is 1e-8 so if the residual of serial is 1e-9 and the residual of parallel is 3e-9, they both satisfy the tolerance. Because the condition number is 1e5, a difference of 1e-9 in the residual will give you a difference of 1e4 in the error (which is what you have). So now the question is: does a difference of 1e-11 in the matrix plus round-off errors can create a difference of 1e-9 in the residual? This sounds possible but you have to check by looking what is going on at each step of the algorithm. Best, Bruno -- The deal.II project is located at http://www.dealii.org/ For mailing list/forum options, see https://groups.google.com/d/forum/dealii?hl=en --- You received this message because you are subscribed to the Google Groups "deal.II User Group" group. To unsubscribe from this group and stop receiving emails from it, send an email to dealii+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/d/optout.
