Checkpoint allocation and release

Abstract

Out-of-order speculative processors need a bookkeeping method to recover from incorrect speculation. In recent years, several microarchitectures that employ checkpoints have been proposed, either extending the reorder buffer or entirely replacing it. This work presents an in-dept-study of checkpointing in checkpoint-based microarchitectures, from the desired content of a checkpoint, via implementation trade-offs, and to checkpoint allocation and release policies. A major contribution of the article is a novel adaptive checkpoint allocation policy that outperforms known policies. The adaptive policy controls checkpoint allocation according to dynamic events, such as second-level cache misses and rollback history. It achieves 6.8% and 2.2% speedup for the integer and floating point benchmarks, respectively, and does not require a branch confidence estimator. The results show that the proposed adaptive policy achieves most of the potential of an oracle policy whose performance improvement is 9.8% and 3.9% for the integer and floating point benchmarks, respectively. We exploit known techniques for saving leakage power by adapting and applying them to checkpoint-based microarchitectures. The proposed applications combine to reduce the leakage power of the register file to about one half of its original value.

Mower

“…al. [9] propose a similar checkpointing scheme. They take checkpoints more frequently when mispredictions are frequent, and vice versa.…”

Mitigating the Effect of Misspeculations in Superscalar Processors

“…Their mechanism can reuse the results of instructions and the outcome of branches obtained during the first run. They then designed an enhanced checkpoint allocation algorithm [41] which would allocate checkpoints according to dynamic events, such as second-level cache misses and rollback history.…”

“…When a misprediction is detected, the flush tag of the mispredicted branch is broadcast through the entire pipeline to selectively eliminate the dependent instructions. Golander et al uses checkpoint tags, instead of branch flush tags, to selectively flush misprediction dependent instructions [52]. Their mechanism has to broadcast all of the checkpoint tags following the misprediction sequentially as each tag can only control the instructions in its own section.…”

Mower : A New Design for Non-Blocking Misprediction Recovery

“…If predictions are going well, the window size is dynamically expanded and the misprediction recovery scheme switches to using checkpoints.Golander et. al [17]. propose a similar scheme without an ROB utilizing checkpointing.…”