CaLRS: A Critical-Aware Shared LLC Request Scheduling Algorithm on GPGPU

Figure 4

The memory divergence statistic of several benchmarks.