I didn't look at your example code at all; it's just a general remark that if you allocate an array with Array(Float64, m, n), until you fill it with something it might contain nans. That can, of course, contaminate other arrays if used in computation. But this may not be what's happening here.
--Tim On Monday, January 12, 2015 10:33:24 AM Peter Brady wrote: > Tim, > > Are you saying that even though `rkvar2.w.f1` and `rkvar2.sol` do not > contain any NaNs, if the memory is uninitialized, it can still lead to > NaNs? I'll change some of my initializations to typemaxes to see if I've > missed something. > > I don't have any kind of self-contained example because the problem is > pretty random. Although, when it does show up, it's always on the first > timestep which makes me suspect my initialization routines rather than a > problem with my system configuration. > > On Mon, Jan 12, 2015 at 10:08 AM, Tim Holy <[email protected]> wrote: > > If you're using uninitialized memory, then NaNs will sometimes be present, > > depending on what was in that chunk of memory. > > > > --Tim > > > > On Monday, January 12, 2015 08:38:54 AM Steven G. Johnson wrote: > > > On Monday, January 12, 2015 at 11:26:31 AM UTC-5, Peter Brady wrote: > > > > So I take it that no one else has this issue? Maybe I should compile > > > > julia myself and see if the problem goes away. > > > > > > I've never seen this issue reported. Can you post a simple > > > > self-contained > > > > > test script that illustrates the problem?
