Mill isn't stack or register based, it's closer to a dataflow machine. They use the bypass network to forward results from the FU skid buffers, and have some CAM/renaming scheme that makes the whole thing *act* like a giant addressable shift register.
Seriously excited about this development. I'd like to see a stack architecture vs register architecture design get the same attention.
Like the Mill?
Mill isn't stack or register based, it's closer to a dataflow machine. They use the bypass network to forward results from the FU skid buffers, and have some CAM/renaming scheme that makes the whole thing *act* like a giant addressable shift register.
Also it's very much a CISC (function call instruction?!) so maybe Dr Patterson should update his slides!
Watch a Teach-able moment @ 22:25 ....