Controlled Kernel Launch for Dynamic Parallelism in GPUs
Controlled Kernel Launch for Dynamic Parallelism in GPUs
复制标题
DOI：
10.1109/hpca.2017.14
复制
发表时间：
2017-02
期刊：
影响因子：
--
通讯作者：
Xulong Tang;Ashutosh Pattnaik;Huaipan Jiang;Onur Kayiran;Adwait Jog;Sreepathi Pai;M. Ibrahim;M. Kandemir;C. Das
中科院分区：
文献类型：
--
作者：
Xulong Tang;Ashutosh Pattnaik;Huaipan Jiang;Onur Kayiran;Adwait Jog;Sreepathi Pai;M. Ibrahim;M. Kandemir;C. Das
动态并行性（DP）是GPU的一个有前途的功能，它允许无需任何CPU干预即可在GPU上按需产卵。但是，此功能有两个主要缺点。首先，GPU内核的推出可以造成严重的性能罚款。其次，由于硬件限制限制，动态生成的内核并不总是能够有效地利用GPU内核。为了掌握这两个问题，我们建议Spawn是一个控制动态生成的内核的运行时框架，从而直接减少了相关的启动开销和排队延迟。此外，它允许将动态生成的和原始（父）内核更好地混合在一起，以使调度程序有效地隐藏其余的顶部开销并改善GPU资源的利用率。我们的结果表明，在13个基准测试中，产卵分别在平面（非DP）实施和基线DP上实现了69％和57％的速度。
Dynamic parallelism (DP) is a promising feature for GPUs, which allows on-demand spawning of kernels on the GPU without any CPU intervention. However, this feature has two major drawbacks. First, the launching of GPU kernels can incur significant performance penalties. Second, dynamically-generated kernels are not always able to efficiently utilize the GPU cores due to hardware-limits. To address these two concerns cohesively, we propose SPAWN, a runtime framework that controls the dynamically-generated kernels, thereby directly reducing the associated launch overheads and queuing latency. Moreover, it allows a better mix of dynamically-generated and original (parent) kernels for the scheduler to effectively hide the remaining overheads and improve the utilization of the GPU resources. Our results show that, across 13 benchmarks, SPAWN achieves 69% and 57% speedup over the flat (non-DP) implementation and baseline DP, respectively.