Reproduce “Shared Memory-contention-aware Concurrent DNN Execution for Diversely Heterogeneous System-on-Chips” 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
Ismet Dagli, Mehmet Belviranli
HaX-CoNN schedules layer groups of concurrent inferences across GPUs and domain-specific accelerators that share memory. It profiles transition costs, estimates memory-contention slowdowns, and uses a constraint-based solver to choose mappings. On Xavier AGX, concurrent VGG-19 and ResNet152 achieved 13.01 ms and 77 frames per second, versus 16.05 ms and 62 for the baseline; gains apply to this setting.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.