Dynamic Scheduling
How an out-of-order core uses scoreboards, reservation stations, and the wakeup-select loop to hide latency while preserving dataflow semantics.
How an out-of-order core uses scoreboards, reservation stations, and the wakeup-select loop to hide latency while preserving dataflow semantics.
How unrolling, software pipelining, tiling, and interchange expose ILP and locality, and when dependence analysis forbids them.
Overlapping instruction stages: the five-stage schedule, the clock period set by the slowest stage, and the speedup formula that governs every later page.
Structural, data and control hazards, the forwarding and stall rules that resolve them, and the in-order commit a precise exception requires.
How register renaming removes false dependencies and the reorder buffer restores precise state at commit.