性能

Deep learning (DL) workloads are moving towards accelerators for faster processing and lower cost. Modern DL accelerators are good at handling the large-scale multiply-accumulate operations that dominate DL workloads; however, it is…

It is well known that building analytical performance models in practice is difficult because it requires a considerable degree of proficiency in the underlying mathematics. In this paper, we propose a machine-learning approach to derive…

We consider the load balancing problem in large-scale heterogeneous systems with multiple dispatchers. We introduce a general framework called Local-Estimation-Driven (LED). Under this framework, each dispatcher keeps local (possibly…

Domain-specific software and hardware co-design is encouraging as it is much easier to achieve efficiency for fewer tasks. Agile domain-specific benchmarking speeds up the process as it provides not only relevant design inputs but also…

In neural network topologies, algorithms are running on batches of data tensors. The batches of data are typically scheduled onto the computing cores which execute in parallel. For the algorithms running on batches of data, an optimal batch…

We consider assignment policies that allocate resources to users, where both resources and users are located on a one-dimensional line. First, we consider unidirectional assignment policies that allocate resources only to users located to…

The authors have uploaded their artifact on Zenodo, which ensures a long-term retention of the artifact. The code is suitably documented, and some examples are given. A minimalistic overall description of the engine is provided. The…

Researchers within the Human Brain Project and related projects have in the last couple of years expanded their needs for high-performance computing infrastructures. The needs arise from a diverse set of science challenges that range from…

Sparse linear algebra is central to many scientific programs, yet compilers fail to optimize it well. High-performance libraries are available, but adoption costs are significant. Moreover, libraries tie programs into vendor-specific…

We consider the problem of preemptively scheduling jobs to minimize mean response time of an M/G/1 queue. When we know each job's size, the shortest remaining processing time (SRPT) policy is optimal. Unfortunately, in many settings we do…

We investigate the duet measurement procedure, which helps improve the accuracy of performance comparison experiments conducted on shared machines by executing the measured artifacts in parallel and evaluating their relative performance…

The softmax (also called softargmax) function is widely used in machine learning models to normalize real-valued scores into a probability distribution. To avoid floating-point overflow, the softmax function is conventionally implemented in…

While the cost of computation is an easy to understand local property, the cost of data movement on cached architectures depends on global state, does not compose, and is hard to predict. As a result, programmers often fail to consider the…

Networks-on-chip (NoCs) have become the standard for interconnect solutions in industrial designs ranging from client CPUs to many-core chip-multiprocessors. Since NoCs play a vital role in system performance and power consumption,…

We study the performance of non-adaptive scheduling policies in computing systems with multiple servers. Compute jobs are mostly regular, with modest service requirements. However, there are sporadic data intensive jobs, whose expected…

We consider the problem of scheduling serverless-computing instances such as Amazon Lambda functions, or scheduling microservices within (privately held) virtual machines (VMs). Instead of a quota per tenant/customer, we assume demand for…

Open-access neuroimaging datasets have reached petabyte scale, and continue to grow. The ability to leverage the entirety of these datasets is limited to a restricted number of labs with both the capacity and infrastructure to process the…

Performance variability has been acknowledged as a problem for over a decade by cloud practitioners and performance engineers. Yet, our survey of top systems conferences reveals that the research community regularly disregards variability…

Motivated by interest in providing more efficient services in customer service systems, we use statistical learning methods and delay history information to predict the conditional distribution of the customers' waiting times in queueing…

Serverless computing is increasingly popular because of the promise of lower cost and the convenience it provides to users who do not need to focus on server management. This has resulted in the availability of a number of proprietary and…