English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

GPUs exploit a high degree of thread-level parallelism to hide long-latencystalls. Due to the heterogeneous compute requirements of differentapplications, there is a growing need to share the GPU across multipleapplications in large-scale computing environments. However, while CPUs offerrelatively seamless multi-application concurrency, and are an excellent fit formultitasking and for virtualized environments, GPUs currently offer onlyprimitive support for multi-application concurrency. Much of the problem in acontemporary GPU lies within the memory system, where multi-applicationexecution requires virtual memory support to manage the address spaces of eachapplication and to provide memory protection. In this work, we perform adetailed analysis of the major problems in state-of-the-art GPU virtual memorymanagement that hinders multi-application execution. Existing GPUs are designedto share memory between the CPU and GPU, but do not handle multi-applicationsupport within the GPU well. We find that when multiple applications spatiallyshare the GPU, there is a significant amount of inter-core thrashing on theshared TLB within the GPU. The TLB contention is high enough to prevent the GPUfrom successfully hiding stall latencies, thus becoming a first-orderperformance concern. We introduce MASK, a memory hierarchy design that provideslow-overhead virtual memory support for the concurrent execution of multipleapplications. MASK extends the GPU memory hierarchy to efficiently supportaddress translation through the use of multi-level TLBs, and usestranslation-aware memory and cache management to maximize throughput in thepresence of inter-application contention.

Improving Multi-Application Concurrency Support Within the GPU Memory System