It comes from the hwloc API. It doesn't use integers because some users want to provide their own distance matrix that was generated by benchmarks. Also we normalize the matrix to have latency 1 on the diagonal (for local memory access latency ) and that causes non-diagonal items not to be integers anymore (Linux and ACPI SLIT report 10 for local memory latency and custom values > 10 for non-local latency).
I am actually revisiting that hwloc API right now. I am open to comments and suggestion about all this. By the way, I talked to Jeff about this recently: the BTL should use the distance in the hwloc tree first, instead of these latency values. I'll try to send patches one day. Brice Le 28/04/2016 20:00, dpchoudh . a écrit : > Hello all > > I am wondering about the rationale of using floating point numbers for > calculating 'distances' in the openib BTL. Is it because some > distances can be infinite and there is no (conventional) way to > represent infinity using integers? > > Thanks for your comments > > Durga > > > The surgeon general advises you to eat right, exercise regularly and > quit ageing. > > > _______________________________________________ > devel mailing list > de...@open-mpi.org > Subscription: https://www.open-mpi.org/mailman/listinfo.cgi/devel > Link to this post: > http://www.open-mpi.org/community/lists/devel/2016/04/18830.php
