Locality-Based Cache Management and Warp Scheduling for Reducing Cache Contention in GPU

GPGPUs has gradually become a mainstream acceleration component in high-performance computing. The long latency of memory operations is the bottleneck of GPU performance. In the GPU, multiple threads are divided into one warp for scheduling and execution. The L1 data caches have little capacity, while multiple warps share one small cache. That makes the cache suffer a large amount of cache contention and pipeline stall. We propose Locality-Based Cache Management (LCM), combined with the Locality-Based Warp Scheduling (LWS), to reduce cache contention and improve GPU performance. Each load instruction can be divided into three types according to locality: only used once as streaming data locality, accessed multiple times in the same warp as intra-warp locality, and accessed in different warps as inter-warp data locality. According to the locality of the load instruction, LWS applies cache bypass to the streaming locality request to improve the cache utilization rate, extend inter-warp memory request coalescing to make full use of the inter-warp locality, and combine with the LWS to alleviate cache contention. LCM and LWS can effectively improve cache performance, thereby improving overall GPU performance. Through experimental evaluation, our LCM and LWS can obtain an average performance improvement of 26% over baseline GPU.

Memory and Cache Contention Denial-of-Service Attack in Mobile Edge Devices

In this paper, we introduce a memory and cache contention denial-of-service attack and its hardware-based countermeasure. Our attack can significantly degrade the performance of the benign programs by hindering the shared resource accesses of the benign programs. It can be achieved by a simple C-based malicious code while degrading the performance of the benign programs by 47.6% on average. As another side-effect, our attack also leads to greater energy consumption of the system by 2.1× on average, which may cause shorter battery life in the mobile edge devices. We also propose detection and mitigation techniques for thwarting our attack. By analyzing L1 data cache miss request patterns, we effectively detect the malicious program for the memory and cache contention denial-of-service attack. For mitigation, we propose using instruction fetch width throttling techniques to restrict the malicious accesses to the shared resources. When employing our malicious program detection with the instruction fetch width throttling technique, we recover the system performance and energy by 92.4% and 94.7%, respectively, which means that the adverse impacts from the malicious programs are almost removed.