The auction algorithm has been widely used to solve problems in several areas. One important characteristic of this algorithm is that its distributive nature simplifies the adoption of parallel implementations. With i...
详细信息
ISBN:
(纸本)9783031042096;9783031042089
The auction algorithm has been widely used to solve problems in several areas. One important characteristic of this algorithm is that its distributive nature simplifies the adoption of parallel implementations. With its various processing cores and 512-bit vectorized instructions, many-core and multicore machines have the potential to considerably increase the performance of this algorithm. The aim of this work is to efficiently execute the auction algorithm in these architectures. The main contributions of this work are: proposal and implementation of a scalable vectorized version for AVX-512 instructions;detailed evaluation of vectorized and parallel versions. Results show that vectorization in conjunction with parallelism is capable of harnessing the full potential of many-core and multicore machines. Moreover, the best scalability of a many-core processor, allowed over 750-fold speedup when comparing the proposed parallel vectorized version against the original sequential one.
暂无评论