GCiM: A Near-Data Processing Accelerator for Graph Construction
Hosted in Virtual Platform
Near-Memory and In-Memory Computing
DescriptionGraph is widely applied as a key data structure in practical applications. Real-world graph construction involves random and massive memory accesses, resulting in considerable processing time and energy consumption on CPUs and GPUs. In this work, we present GCiM, a specialized processing-in-memory architecture for graph construction and update. By directly deploying computing units on the logic layer of 3D-stacked memory cube, GCiM benefits from memory-level parallelism and further improves memory access efficiency with optimized processing ordering and data layout. According to our experiments, GCiM shows 634.64X and 53.29X performance speedup, 1470.7X and 442.56X energy redunction over CPU and GPU respectively.