Ok then use -log_summary and put the first solve in a separate stage (see PetscStageRegister()) and send the results of a run back demonstrating the slow first solver and we may be able to see what the story is.
Barry On Jan 28, 2014, at 5:23 PM, David Liu <[email protected]> wrote: > wow that is news to me. I always assumed that this is normal. > > I'm pretty certain it's not the preallocation. I'm using MatCreateMPI, and to > my knowledge I wouldn't even be able to set the values without crashing if I > didn't preallocate. (If I'm not mistaken, the setting values slowly without > preallocating is only possible if you create the Mat using MatCreate + > MatSetup). > > Also, I'm certain that the time is taken on the first solve, not the setting > of values, because I use the matrix in a MatMult first to get the RHS before > solving, and the MatMult happens before the long first solve. > > > On Tue, Jan 28, 2014 at 5:04 PM, Barry Smith <[email protected]> wrote: > > On Jan 28, 2014, at 1:36 PM, David Liu <[email protected]> wrote: > > > Hi, I'm writing an application that solves a sparse matrix many times using > > Pastix. I notice that the first solves takes a very long time, > > Is it the first “solve” or the first time you put values into that matrix > that “takes a long time”? If you are not properly preallocating the matrix > then the initial setting of values will be slow and waste memory. See > http://www.mcs.anl.gov/petsc/petsc-current/docs/manualpages/Mat/MatXAIJSetPreallocation.html > > The symbolic factorization is usually much faster than a numeric > factorization so that is not the cause of the slow “first solve”. > > Barry > > > > > while the subsequent solves are very fast. I don't fully understand what's > > going on behind the curtains, but I'm guessing it's because the very first > > solve has to read in the non-zero structure for the LU factorization, while > > the subsequent solves are faster because the nonzero structure doesn't > > change. > > > > My question is, is there any way to save the information obtained from the > > very first solve, so that the next time I run the application, the very > > first solve can be fast too (provided that I still have the same nonzero > > structure)? > >
