sparkdustjoe / norxmanaged Goto Github PK

TODO

This is ongoing, as doing Parallel loops in C++/CLR isn't as straight-forward as it is in C#. Not to mention, the Norx process does work division in an orchestrated round-robin manner among the threads/lanes, so that will take some time to work out.

Basically, if Parallelism = 4:
Lane 0 gets Block 0 => Block 4 => Block 8 => ...
Lane 1 gets Block 1 => Block 5 => Block 9 => ...
...
So determining which lane get's the end-block can get a little tricky, then they each get tagged with a "merge" byte, then _F processed, and then the end of each lane is XOR'd with every other lane when they're done. This produces the new state going forward through the Trailer/Footer data, and then the Finalize for the TAG output.

Still working it all out for the C++/CLR environment in .Net ...