Data-centric execution of speculative parallel programs

Abstract

Abstract-Multicore systems must exploit locality to scale, scheduling tasks to minimize data movement. While localityaware parallelism is well studied in non-speculative systems, it has received little attention in speculative systems (e.g., HTM or TLS), which hinders their scalability.We present spatial hints, a technique that leverages program knowledge to reveal and exploit locality in speculative parallel programs. A hint is an abstract integer, given when a speculative task is created, that denotes the data that the task is likely to access. We show it is easy to modify programs to convey locality through hints. We design simple hardware techniques that allow a state-of-the-art, tiled speculative architecture to exploit hints by: (i) running tasks likely to access the same data on the same tile, (ii) serializing tasks likely to conflict, and (iii) balancing tasks across tiles in a locality-aware fashion. We also show that programs can often be restructured to make hints more effective.Together, these techniques make speculative parallelism practical on large-scale systems: at 256 cores, hints achieve nearlinear scalability on nine challenging applications, improving performance over hint-oblivious scheduling by 3.3× gmean and by up to 16×. Hints also make speculation far more efficient, reducing wasted work by 6.4× and traffic by 3.5× on average.

Optimizing ordered graph algorithms with GraphIt

“…However, speculative execution in software incurs significant performance overheads as a commit queue has to be maintained, conflicts need to be detected, and values are buffered for potential rollback on conflicts. Hardware solutions have been proposed to reduce the overheads of speculative execution [2,[24][25][26]42], but it is costly to build customized hardware. Furthermore, some ordered graph algorithms, such as approximate set cover and k-core, cannot be easily expressed with speculative execution.…”

A Hardware Runtime for Task-Based Programming Models

“…The CP is used for task dependence analysis and schedule tasks when they are ready. Swarm [25] uses the co-design of the execution model and micro-architecture to exploit ordered irregular task parallelism. It relies on speculative task execution and conflict detection to preserve dependences, which requires hardware transactional memory support.…”

Architectural Support for Task Dependence Management with Flexible Software Scheduling

“…Swarm [12] relies on speculative task execution and conflict detection to preserve dependences. Swarm requires hardware support for speculation instead of for dependence management and uses either a FIFO or a spatial scheduler fixed in the architecture [13]. Fractal [14] extends Swarm to allow nested parallelism by means of task domains, that can be ordered or unordered to avoid over-serialization.…”

“…Considering task criticality [16], [17] or data locality [13] provides significant benefits in certain contexts. Moreover, the adaptability granted by software task schedulers is essential in modern high-performance computing systems with off-chip accelerators and multi-socket configurations, that can further improve performance and energy efficiency but require software intervention for task scheduling and data motion.…”