Brett (ggtgp.delete@this.yahoo.com) on February 21, 2021 11:59 pm wrote:
> ...
> I am a big fan of the Mill approach as it exposes chains of operations making opcode merging easier and
> does so in way that decreases opcode size. You get a potential critical operation chain length reduction
> which can get you 10% in some cases and the smaller opcodes get you another few percent. This is why I promoted
> a OoOe version of Mill that is slightly different, which is against the world view of the Mill founders.
> The in-order wide advantage of the Mill team is a loser against OoOe running typical low IPC code.

I do not pretend to have completely understood what Jim Keller is working on right now, but it looks like sending a bloc of operations to the processor managing that part of the data; that block of operation could be one "independant mill packet"...
Jim Keller system seems to be a very big array of Wormhole processor, itself described at the end of https://www.anandtech.com/show/16354/jim-keller-becomes-cto-at-tenstorrent-the-most-promising-architecture-out-there