For the purpose of learning zig and a couple of other things, i have written a NN implementation. To increase the speed of the Matrix Vector products i have used Blas. Since i am developing on Fedora i am calling Openblas-serial through the Flexiblas library/wrapper. The single threaded performance of the whole thing is fine, but for “reasons” i want to do the entire thing multi threaded. The plan was to accomplish that by creating multiple instances of the same NN, distribute those to multiple threads and then have these threads execute the individual workloads in parallel.

For the purpose of learning zig and a couple of other things, i have written a NN implementation. To increase the speed of the Matrix Vector products i have used Blas. Since i am developing on Fedora i am calling Openblas-serial through the Flexiblas library/wrapper. The single threaded performance of the whole thing is fine, but for “reasons” i want to do the entire thing multi threaded. The plan was to accomplish that by creating multiple instances of the same NN, distribute those to multiple threads and then have these threads execute the individual workloads in parallel.