This is one of the series of talks about the mill architecture. The Mill is a family of CPU designs. The Mill has a 10x single-thread power/performance gain over conventional out of order superscalar architectures. This talk will explain:
- Review of the Belt, the replacement for general registers.
- phasing, a 3x average speedup of dependent operations
- ganging, when operations need more arguments
- sidekick predicates, to speed up conditional branches
- First Winner Rule, for N-way branching
- cascaded calls, phasing across call/return
