Quincey Koziol

We're planning to discuss this issue at the next I/O WG meeting, on Oct. 3, at 2:30pm CT. Probably will be @edgargabriel , @devreal , Neil Fortner, and myself, at...

> My understanding from reading the Win32 docs and stackoverflow is that if we switch to using Fls* (fiber local storage) functions instead of Tls* we'll be able to use...

> https://devblogs.microsoft.com/oldnewthing/20191011-00/?p=102989 I'm reading https://learn.microsoft.com/en-us/windows/win32/procthread/fibers and have some concerns about using fiber local storage. What happens (in the probably rare case) when a Windows application that uses multiple fibers within...

> > In addition, the thread has to be converted to a fiber first, via ConvertThreadToFiber. This seems to be a requirement: https://stackoverflow.com/questions/19613016/fls-vs-tls-can-i-use-fiber-local-storage-in-place-of-tls . I don't think it's a good...

> Still debugging, but I can't seem to enable subfiling with these changes. CMake always fails when trying to find `MPI_Comm_split_type` from MPI. Reverting the patch I can enable the...

> Also, just a thought but it might be nice if the threading tests all printed out what threading model is being used for debugging purposes. That's already in the...

> > > Also, just a thought but it might be nice if the threading tests all printed out what threading model is being used for debugging purposes. > >...

> > Still debugging, but I can't seem to enable subfiling with these changes. CMake always fails when trying to find `MPI_Comm_split_type` from MPI. Reverting the patch I can enable...

> > @jhendersonHDF - I just noticed a place in the subfiling CMake configuration that I missed earlier. Can you please try again now? > > Still running into the...

> Needs to address performance regression. Thanks for running these tests, I'll start looking into the performance tomorrow.