## [Libmesh-devel] I'm back...

 [Libmesh-devel] I'm back... From: Benjamin S. Kirk - 2003-03-17 01:30:46 Well, looks like things have been pretty busy while I was away! I will likely make a new release this week, but it will probably be after a talk I'm scheduled to give on Wednesday. I have made some rudimetary changes, removed some deprecated full specializations, etc... The complex LASPACK support is great! I think we should try and contact the author again and contribute it. I tried to talk to him two years ago but heard nothing... Daniel, here is some output from the MipsPro SGI compiler related to the complex support... I can look into this after my talk on Wednesday, but thought you might be interested... It seems at least in the first case that r should be a real number, no? cc-1278 CC: ERROR File = itersolv.c, Line = 700 No instance of overloaded function "sqrt" matches the argument list. The argument types are: (std::complex). r = sqrt(h[i][i] * h[i][i] + h[i][i+1] * h[i][i+1]); ^ cc-1278 CC: ERROR File = itersolv.c, Line = 1014 No instance of overloaded function "sqrt" matches the argument list. The argument types are: (std::complex). Gamma = 1.0 / sqrt(1.0 + Theta * Theta); ^ cc-1278 CC: ERROR File = itersolv.c, Line = 1083 No instance of overloaded function "sqrt" matches the argument list. The argument types are: (std::complex). Gamma = 1.0 / sqrt(1.0 + Theta * Theta); ^ I think the most pressing thing now is to get some testing in place. I like the idea of comparing solutions. I think it would also be worthwhile to read a "correct solution" from disk, feed it to the linear solver, and make sure that it converges in 0 iterations. That certainly makes sense for linear problems, and it is insensitive to irrelevant changes (like diffing two log files). Ideas?