Study on Fine-Grained Synchronization in Many-Core Architecture

Abstract

The synchronization between threads has serious impact on the performance of many-core architecture. When communication is frequent, coarse-grained synchronization brings significant overhead. Thus, coarse-grained synchronization is not suitable for this situation. However, the overhead of fine-grained synchronization is still small when the communication is frequent. For the many-core architecture which supports fine-grained synchronization with on-chip storage, we propose fine-grained synchronization algorithms for scientific computation application 2-D wavefront and LU decomposition. At first, according to the memory access mode, an efficient method of data allocation is proposed. Then, way of thread partition and synchronization are discussed. Finally, we estimate the two algorithms based on Godson-T many-core architecture. The results of experiments show that the relative speedup is almost linear and the execution time is only 53.2 % of the coarse-grained synchronization. After the global barriers are eliminated, LU decomposition achieved 13.1% performance improvement. Moreover, the experiments prove that the fine-grained mechanism is able to improve the performance of processor and it has a good scalability.

Performance of a Hardware Scheduler for Many-core Architecture

“…In the latter, hardware steering logic allocates strands (chains of dependent instructions) to cores based on inter-strand dependencies. Hardware scheduling for Godson-T is described in [31] [14], comparing also fine grain to barrier synchronization and managing instance dependencies in addition to coarser task dependencies. Carbon [32] and ADM [33] use hardware task queues to support scheduling.…”

“…While solutions that incorporate static scheduling during compile time exist [9] [10] [11], they cannot dynamically schedule tasks according to run-time workload. Hardware schedulers, on the other hand, enable such advantages and hold more promise for efficient many-cores [4] [12] [13] [14] [15] [16].…”

Preliminary Investigation of Accelerating Molecular Dynamics Simulation on Godson-T Many-Core Processor

“…• Fine-grained parallelism [5]. Each core works as a lightweight hardware thread unit executing in a non-preemptive manner.…”

“…A dedicated synchronization manager (SM) is a centralized unit to collect and handle synchronization requests, which provides architectural support for fast mutual exclusion, barrier and signal/wait synchronization. In addition, an extremely efficient thread execution runtime system has been developed to manage thread execution [5,6].…”

“…Also assume that cell 2 interacts with {1, 4} and that cell 5 interacts with {4, 8}. Then we construct two cell pairs P C [2] = {1, 4} and P C [5] = {4, 8}. We then use the core-core communication to transfer the atomic data according to the cell pairs from core j to core i.…”