https://hgpu.org/?p=7033
TAP: A TLP-Aware Cache Management Policy for a CPU-GPU Heterogeneous Architecture