Browsing by Subject "GPU"

Heterogeneous processors with accelerators provide an opportunity to improve performance within a given power budget.

Many of these heterogeneous processors contain Graphics Processing Units (GPUs) that can perform graphics and embarrassingly parallel computation orders of magnitude faster than a CPU while using less energy. Beyond these obvious applications for GPUs, a larger variety of applications can benefit from a GPU's large computation and memory bandwidth. However, many of these applications are irregular and, as a result, require synchronization and scheduling that are commonly believed to perform poorly on GPUs. The basic building block of synchronization and scheduling is memory consistency, which is, therefore, the first place to look for improving performance on irregular applications. In this thesis, we approach the programmability of irregular applications on GPUs by thinking across traditional boundaries of the compute stack. We think about architecture, microarchitecture and runtime systems from the programmers perspective. To this end, we study architectural memory consistency on future GPUs with cache coherence. In addition, we design a GPU memory system

microarchitecture that can support fine-grain and coarse-grain synchronization without sacrificing throughput. Finally, we develop a task runtime that embraces the GPU microarchitecture to perform well

on fork/join parallelism desired by many programmers. Overall, this thesis contributes non-intuitive solutions to improve the performance and programmability of irregular applications from the programmer's perspective.

Trends in increasing web traffic demand an increase in server throughput while preserving energy efficiency and total cost of ownership. Present work in optimizing data center efficiency primarily focuses on using general purpose processors, however these might not be the most efficient platforms for server workloads. Data parallel hardware achieves high energy efficiency by amortizing instruction costs across multiple data streams, and high throughput by enabling massive parallelism across independent threads. These benefits are considered traditionally applicable to scientific workloads, and common server tasks like page serving or search are considered unsuitable for a data parallel execution model.

Our work builds on the observation that server workload execution patterns are not completely unique across multiple requests. For a high enough arrival rate, a server has the opportunity to launch cohorts of similar requests on data parallel hardware, improving server performance and power/energy efficiency. We present a framework---called Rhythm---for high throughput servers that can exploit similarity across requests to improve server performance and power/energy efficiency by launching data parallel executions for request cohorts. An implementation of the SPECWeb Banking workload using Rhythm on NVIDIA GPUs provides a basis for evaluation.

Similarity search is another ubiquitous server workload that involves identifying the nearest neighbors to a given query across a large number of points. We explore the performance, power and dollar benefits of using accelerators to perform similarity search for query cohorts in very high dimensions under tight deadlines, and demonstrate an implementation on GPUs that searches across a corpus of billions of documents and is significantly cheaper than commercial deployments. We show that with software and system modifications, data parallel designs can greatly outperform common task parallel implementations.