arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1912.02165cs.DCcs.LG
收藏

L3 Fusion:CPU上的快速变换卷积

L3 Fusion: Fast Transformed Convolutions on CPUs

NeuralMagic Inc(NeuralMagic公司);MIT(麻省理工学院);Facebook AI Research(脸书人工智能研究院)

Rati Gelashvili, Nir Shavit, Aleksandar Zlateski

摘要:

通过Winograd或FFT变换的快速卷积已经成为执行卷积层计算的首选方式,因为它大大减少了所需操作的数量。最近的工作表明,对于许多层结构,设计良好的快速卷积实现可以极大地利用现代CPU,显着减少计算时间。然而,现代CPU上存在的大量共享L3缓存经常被忽视,并且算法仅针对私有L2缓存进行优化。在本文中,我们提出了一个有效的'L3融合'算法,是专门为CPU与大量的共享L3缓存。使用分层屋顶线模型,我们表明,在许多情况下,特别是对于具有较少通道的层,“L3融合”方法可以大大优于由大供应商(如Intel)提供的标准3阶段。我们验证了我们的理论研究结果,通过基准测试我们的'L3融合'实现对公开可用的最先进的。

英文摘要:

Fast convolutions via transforms, either Winograd or FFT, had emerged as a preferred way of performing the computation of convolutional layers, as it greatly reduces the number of required operations. Recent work shows that, for many layer structures, a well--designed implementation of fast convolutions can greatly utilize modern CPUs, significantly reducing the compute time. However, the generous amount of shared L3 cache present on modern CPUs is often neglected, and the algorithms are optimized solely for the private L2 cache. In this paper we propose an efficient `L3 Fusion` algorithm that is specifically designed for CPUs with significant amount of shared L3 cache. Using the hierarchical roofline model, we show that in many cases, especially for layers with fewer channels, the `L3 fused` approach can greatly outperform standard 3 stage one provided by big vendors such as Intel. We validate our theoretical findings, by benchmarking our `L3 fused` implementation against publicly available state of the art.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑