カーネル実行時のイベントは以下のものがあります。これらは同時に4つまでしか設定できません。
Energy expenses are becoming an increasingly important
fraction of data center operating costs. At the same time,
the energy expense per unit of computation can vary significantly
between two different locations. In this paper,
we characterize the variation due to fluctuating electricity
prices and argue that existing distributed systems should be
able to exploit this variation for significant economic gains.
Electricity prices exhibit both temporal and geographic variation,
due to regional demand differences, transmission inefficiencies,
and generation diversity. Starting with historical
electricity prices, for twenty nine locations in the US, and
network traffic data collected on Akamai’s CDN, we use simulation
to quantify the possible economic gains for a realistic
workload. Our results imply that existing systems may be
able to save millions of dollars a year in electricity costs, by
being cognizant of locational computation cost differences.
Topics of particular interest include, but are not limited to:
| Paper Submission | Extended to November 25, 2009 - 12:00 pm, Hawaii time (GMT -10) |
| Author Notification | February 8, 2010 |
| Final Manuscript Due | March 15, 2010 |
General purpose computing on graphics processing units (GPUs) introduces the challenge of scheduling task parallel workloads on devices designed for data parallel applications. This paper analyzes the problem, and presents a method for merging workloads such that they can be run concurrently on an NVIDIA GPU. Some kernels do not fully utilize the processing power of the GPU, and thus overall throughput will increase when running two kernels alongside one another. Our approach scans a queue of independent CUDA kernels (i.e., code segments that will run on the GPU), across processes or from within the same process, and evaluates whether merging the kernels would increase both throughput on the device and overall efficiency of the computing platform. Using kernels from microbenchmarks and a Nearest Neighbor application we show that throughput is increased in all cases where the GPU would have been underused by a single kernel. An exception is the case of memory-bound kernels, seen in a Nearest Neighbor application, for which the execution time still outperforms the same kernels executed serially by 12-20%. It can also be benefitial to choose a merged kernel that over-extends the GPU resources, as we show the worst case to be bounded by executing the kernels serially. This paper also provides an analysis of the latency penalty that can occur when two kernels with varying completion times are merged
