Search for: All records

The abstraction of a shared memory space over separate CPU and GPU memory domains has eased the burden of portability for many HPC codebases. However, users pay for ease of use provided by system-managed memory with a moderate-to-high performance overhead. NVIDIA Unified Virtual Memory (UVM) is currently the primary real-world implementation of such abstraction and offers a functionally equivalent testbed for in-depth performance study for both UVM and future Linux Heterogeneous Memory Management (HMM) compatible systems. The continued advocacy for UVM and HMM motivates improvement of the underlying system. We focus on UVM-based systems and investigate the root causes of UVM overhead, a non-trivial task due to complex interactions of multiple hardware and software constituents and the desired cost granularity. In our prior work, we delved deeply into UVM system architecture and showed internal behaviors of page fault servicing in batches. We provided quantitative evaluation of batch handling for various applications under different scenarios, including prefetching and oversubscription. We revealed that the driver workload depends on the interactions among application access patterns, GPU hardware constraints, and host OS components. Host OS components have significant overhead present across implementations, warranting close attention. This extension furthers our prior study in three aspects: fine-grain cost analysis and breakdown, extension to multiple GPUs, and investigation of platforms with different GPU-GPU interconnects. We take a top-down approach to quantitative batch analysis and uncover how constituent component costs accumulate and overlap, governed by synchronous and asynchronous operations. Our multi-GPU analysis shows reduced cost of GPU-GPU batch workloads compared to CPU-GPU workloads. We further demonstrate that while specialized interconnects, NVLink, can improve batch cost, their benefits are limited by host OS software overhead and GPU oversubscription. This study serves as a proxy for future shared memory systems, such as those that interface with HMM, and the development of interconnects. 

Cambrian–Devonian sedimentary rocks of the northern Canadian Cordillera record both the establishment and demise of the Great American Carbonate Bank, a widespread carbonate platform system that fringed the ancestral continental margins of North America (Laurentia). Here, we present a new examination of the deep-water Road River Group of the Richardson Mountains, Yukon, Canada, which was deposited in an intra-platformal embayment or seaway within the Great American Carbonate Bank called the Richardson trough. Eleven detailed stratigraphic sections through the Road River Group along the upper canyon of the Peel River are compiled and integrated with geological mapping, facies analysis, carbonate and organic carbon isotope chemostratigraphy, and new biostratigraphic results to formalize four new formations within the type area of the Richardson Mountains (Cronin, Mount Hare, Tetlit, and Vittrekwa). We recognize nine mixed carbonate and siliciclastic deep-water facies associations in the Road River Group and propose these strata were deposited in basin-floor to slope environments. New biostratigraphic data suggest the Road River Group spans the late Cambrian (Furongian) – Middle Devonian (Eifelian), and new chemostratigraphic data record multiple global carbon isotopic events, including the late Cambrian Steptoean positive carbon isotope excursion, the Late Ordovician Guttenberg excursion, the Silurian Aeronian, Valgu, Mulde (mid-Homerian), Ireviken (early Sheinwoodian), and Lau excursions, and the Early Devonian Klonk excursion. Together, these new data not only help clarify nomenclatural debate centered around the Road River Group, but also provide critical new sedimentological, biostratigraphic, and isotopic data for these widely distributed rocks of the northern Canadian Cordillera.