A novel architecture for large windows processors

Several processor architectures with large instruction windows have been proposed. They improve performance by maintaining hundreds of instructions in flight to increase the level of instruction parallelism (ILP). Such architectures replace a re-order buffer (ROB) with a check-pointing mechanism and an out-of-order release of the processor resources. Check-pointing, however, leads to an imprecise state recovery on mispredicted branches and exceptions and frequent re-execution of current-path instructions during the state recovery. It also requires large register files complicating renaming, allocation and release of physical registers. This technical report proposes a new processor architecture that does not use either a traditional ROB or check-pointing, avoids the above-mentioned problems, and has a fast, distributed state recovery mechanism. Its novel register management architecture allows implementation of large register files with simpler and more scalable, register renaming and commit. It is also key to the precise recovery mechanism.Postprint (published version