Quincey Koziol
Quincey Koziol
Possibly the change I made to cuda_hmem_verify_devices() in this libfabric PR would be a helpful direction to explore: https://github.com/ofiwg/libfabric/pull/9170
PR is up: https://github.com/open-mpi/ompi/pull/11486
New PR, with correct signed-off-by and cleaner history: https://github.com/open-mpi/ompi/pull/11531
> I feel like the memory alloc sanity check code is unnecessary and we should just have a GitHub action that uses -fsanitize or valgrind or whatnot. We shouldn't be...
> > I feel like the memory alloc sanity check code is unnecessary and we should just have a GitHub action that uses -fsanitize or valgrind or whatnot. We shouldn't...
How about breaking the isnan change into a separate PR and also the H5S changes into another PR?
> I made some progress. `H5detect` creates a file `H5Tinit.c` and its contents differ based on build machine introspection. Yes, that is its purpose. :-)
Sorry for the delay, this has been a tough week. :-/ H5detect is indeed the only introspection tool that HDF5 runs. We do support cross-compiling on other systems, usually by...
This may not work for cross-compiles+cross-execution environment situations, where the build system is different from the targeted runtime system. This is the benefit of h5detect - you are not just...
As @gnuoyd says, this _could_ be done at runtime, within an application, but the extra overhead of executing it each time would be frustrating for applications to accept, since it...