On Thursday, 3 August 2017 at 04:37:30 UTC, Matt wrote:
That seems like a good start. I'll probably start contributing
within a few weeks.
Great, http://gitter.im/libmir is probably the best way to get in
In the long run the goal should probably be to have a "go-to"
set of LA libraries, with appropriate data structures that have
friendly bindings to GPU LA computations with DCompute.
That is the idea, https://github.com/libmir/mir-glas already
provides a BLAS and I'll eventually get around to getting
DCompute integration for it and ndslice.