Processing-in-Memory (PIM) is a promising approach to overcoming the memory-wall bottleneck. However, the PIM community has largely treated its two fundamental data layouts, Bit-Parallel (BP) and Bit-Serial (BS), as if they were interchangeable. This implicit "one-layout-fits-all" assumption, often hard-coded into existing evaluation frameworks, creates a critical gap: architects lack systematic, workload-driven guidelines for choosing the optimal data layout for their target applications. To address this gap, this paper presents the first systematic, workload-driven characterization of BP and BS PIM architectures. We develop iso-area, cycle-accurate BP and BS PIM architectural models and conduct a comprehensive evaluation using a diverse set of benchmarks. Our suite includes both fine-grained microworkloads from MIMDRAM to isolate specific operational characteristics, and large-scale applications from the PIMBench suite, such as the VGG network, to represent realistic end-to-end workloads. Our results quantitatively demonstrate that no single layout is universally superior; the optimal choice is strongly dependent on workload characteristics. BP excels on control-flow-intensive tasks with irregular memory access patterns, whereas BS shows substantial advantages in massively parallel, low-precision (e.g., INT4/INT8) computations common in AI. Based on this characterization, we distill a set of actionable design guidelines for architects. This work challenges the prevailing one-size-fits-all view on PIM data layouts and provides a principled foundation for designing next-generation, workload-aware, and potentially hybrid PIM systems.
翻译:暂无翻译