DeNovo: Rethinking Hardware for Disciplined Parallelism
نویسندگان
چکیده
We believe that future large-scale multicore systems will require disciplined parallel programming practices, including data-race-freedom, deterministic-by-default semantics, and structured, explicit parallel control and side-effects. We argue that this software evolution presents far-reaching opportunities for parallel hardware design to greatly improve complexity, power-efficiency, and performance scalability. The DeNovo project is rethinking hardware design from the ground up to exploit these opportunities. This paper presents the broad research agenda of DeNovo, including a holistic rethinking of cache coherence, memory consistency, communication, and cache architecture.
منابع مشابه
Verification and Performance of the Denovo Cache Coherence Protocol By
With the advent of multicores, parallel programming has gained a lot of importance. For parallel programming to be viable for the predicted hundreds of cores per chip, shared memory programming languages and environments must evolve to enforce disciplined practices like “determinism-by-default semantics” and ban “wild shared-memory behaviors” like arbitrary data races and potential non-determin...
متن کاملN-gram language models for massively parallel devices
For many applications, the query speed of N -gram language models is a computational bottleneck. Although massively parallel hardware like GPUs offer a potential solution to this bottleneck, exploiting this hardware requires a careful rethinking of basic algorithms and data structures. We present the first language model designed for such hardware, using B-trees to maximize data parallelism and...
متن کاملExplorer N - gram language models for massively parallel devices
For many applications, the query speed of N -gram language models is a computational bottleneck. Although massively parallel hardware like GPUs offer a potential solution to this bottleneck, exploiting this hardware requires a careful rethinking of basic algorithms and data structures. We present the first language model designed for such hardware, using B-trees to maximize data parallelism and...
متن کاملAnalyses, Hardware/Software Compilation, Code Optimization for Complex Dataflow HPC Applications
The advent of parallelism in supercomputers and in more classical end-user computers increases the need for high-level code optimization and improved compilers. Until 2006, the typical power-consumption of a chip remained constant for a given area as the transistor size decreased (this evolution is referred to as Dennard scaling). In other words, energy efficiency was following an exponential l...
متن کاملRethinking Hardware Support for Network Analysis and Intrusion Prevention
The performance pressures on implementing effective network security monitoring are growing fiercely due to rising traffic rates, the need to perform much more sophisticated forms of analysis, the requirement for inline processing, and the collapse of Moore’s law for sequential processing. Given these growing pressures, we argue that it is time to fundamentally rethink the nature of using hardw...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2010