Reproduce “Heterogeneous Memory Benchmarking Toolkit” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Golsana Ghaemi, Gabriel Franco, Kazem Taram, Renato Mancuso

MEMSCOPE, an open-source kernel-level toolkit, provides precise characterization of heterogeneous memory subsystems in multicore ARM embedded systems. It reveals the temporal behavior and contention across different memory types and shared interconnects, demonstrating significant performance impacts from cross-memory interference and cache bank contention.

Key insights

  • Performance bottlenecks in heterogeneous memory systems often lie in shared interconnects (e.g., Cache Coherent Interconnect) rather than individual memory controllers, leading to significant cross-memory type interference.
  • Cache bank-level contention can cause substantial performance degradation even for cache hits and is not fully mitigated by L2 cache partitioning.
  • Memory allocation decisions for real-time applications must consider not only the target memory module's raw performance but also the system-wide impact of contention on shared bus infrastructure.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.