Towards a Programming Paradigm for Reconfigurable Computing: Asynchronous Graph Programming

Fryer, Joshua; Garcia, Paulo

doi:10.1109/etfa46521.2020.9211968

Cited by 8 publications

(6 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Like other existing parallelizing compilers that convert a source code to a destination code (e.g. sequential code to CUDA/OpenCL), our compiler also aim to convert C code to a new parallel programming model described in (AGP) [16]. Since some of these target programming models are not known worldwide and take time to be learnt and utilized, introducing a compiler to translate a well-known and general-purpose language to such a programming language is of great importance.…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Cross-Paradigm Compilation across Programming Models: from Imperative to Asynchronous Graph

Layeghi¹

View full text Add to dashboard Cite

Recent years have seen a rapid evolution in multi-core processor architectures. However, programming multi-core processors efficiently is a challenging endeavor [1], [2]. Either programmers empirically or ad hoc identify parallelization opportunities in their sequential code, then manually parallelize the implementation, or parallel programming paradigms (with appropriate tooling and support) must be employed to automate parallelization efforts. Despite encouraging efforts such as OpenMP [3] and OpenCL[4], we are far from automated parallelization.Asynchronous Graph Programming (AGP) is a novel parallel paradigm that is amenable to automated parallelization for multi-core processors. However, its semantics are both foreign to most programmers, and too low-level to support software development at scale. Thus, this thesis explores the development of a cross-paradigm compiler, that can translate code in an imperative language (C) to AGP, allowing existing code-bases to be re-deployed and automatically parallelized, despite the semantics of their original language being sequential. Towards this, we define semantic transformations from (a sub-set of) C to AGP, and demonstrate the implementation of a compiler that implements those transformations. Our results show that it is indeed possible to transform C code into AGP code, and that transformed parallel implementations running on a multi-core system, for a suite of testing benchmarks, i outperform sequential code substantially.First and foremost, I am grateful to Almighty God for leading me down the path of knowledge. I would like to express my deep and sincere gratitude to my research supervisor, Dr. Paulo Garcia for giving me the opportunity to do research and providing invaluable guidance throughout this research. I can't say thank you enough for his tremendous support and help. In every meeting we had together, I have been inspired by his continuance support, patience, and guidance which was vital for the completion of this project. I appreciate his help throughout the whole thesis at times when he was trying to put me in the right path. Special thanks to students of our research group for their valuable comments and suggestions especially to Sebastien with whom I have worked for a period of time before our thesis begins. I would like to express my heartfelt gratitude to my parents and my brother and sister, for their wise counsel and widespread support throughout my study in Canada. I would like to thank my fiance, Soroush, who has been supporting me all the time and he has always been source of inspiration and strength for me. Finally, I would like to thank my friends, Amirali, Parnia and Jasmine, for providing me with happy distractions to rest my mind outside of my research and having a group study sessions together on campus.

show abstract

Section: Discussionmentioning

confidence: 99%

“…In the following, we will explore some existing code translation techniques between parallel programming models on heterogeneous many-core architectures. We will then introduce our novel code translator which translates a C code to AGP (which is a parallel programming model has been described in detail in [16]).…”

Section: Parallel Programming Modelsmentioning

confidence: 99%

Cross-Paradigm Compilation across Programming Models: from Imperative to Asynchronous Graph

Layeghi¹

View full text Add to dashboard Cite

show abstract

“…Asynchronous Graph Programming (AGP) [20], is a novel programming model developed to offer high levels of parallelism and asynchronicity while requiring the programmer to spare little to no thought of parallelization. This is achieved as an intermediate representation, which compiles a higher-level language like C into a graph of single instructions.…”

Section: Our Approachmentioning

confidence: 99%

“…Should both prove fruitful, it will also provide insight on possible bottlenecks and speedups gained from different node allocations for future node allocation algorithms and hardware designs. AGP [20] is a novel programming model that offers interesting parallelization and asynchronicity possibilities, however, graph processing models are not very common. It takes specialized hardware to take full advantage of the capabilities a model offers.…”

Section: Agp Syntaxmentioning

confidence: 99%

A Message Passing Many-Core Architecture for an Asynchronous Graph Programming Model

Cook¹

View full text Add to dashboard Cite

Noticeable improvements in processor performance have been achieved by researching programming models, control flow parallelization, general architecture, memory access, and code compilation [53][4]. In this thesis, we seek to improve general processing by applying a many-core message passing (MPMC) architecture with a novel Asynchronous Graph Programming model (AGP).AGP abstracts higher-level languages into a graph of single instructions providing very high levels of parallelism and asynchronicity. The MPMC architecture utilizes a novel method of segmenting a graph among cores in tandem with the many-core model to exploit AGPs parallelism.We evaluate the MPMC architecture implementing a functional simulation that, although incapable of providing empirical measurements, provides an efficient method of evaluation that helps accelerate the development cycle. We found that the MPMC architecture can reach a 97% improvement in execution time from a single-core configuration, with room to improve given more cores and better node allocation strategies.

show abstract

“…[35] proposes a tool that analyzes source code to automatically identify sections that can be parallelized and annotates them with OpenMP annotations. Similarly, [50] converts sequential C code into a CUDA-compatible program that can be accelerated by a GPU with CUDA cores, and [36] proposes a compiler that completely translates a program written in a compatible subset of C into a program expressed in a graph-based paradigm known as Asynchronous Graph Programming (AGP) [51].…”

Section: Automated Parallelismmentioning

confidence: 99%

Arch-Lang : A DSE Tool for Flexible Hardware/Software Co-Design

Fryer¹

View full text Add to dashboard Cite

Multicore architectures have emerged as an avenue to continue the improvement of software performance even as growth in single-core performance has struggled over the past two decades [1] [2]. However, designing for such systems is a more complex task than for single-core, sequential computing. To effectively utilize multicore systems, designers must devise communication strategies, coordinate dataflow between processing elements in a way that avoids erroneous behaviour such as race conditions, and the software's general architecture must be designed in a way that takes advantage of advancements in hardware architecture [3] [4].In this thesis, we present an architecture and domain-specific language that allows software developers to rapidly prototype hardware architectures, software partitioning, and inter-core communication strategies at design time. We demonstrate an interpreter for this domain-specific language, and use it to illustrate a process of experimenting with inter-core communication and software partitioning strategies.

show abstract

Towards a Programming Paradigm for Reconfigurable Computing: Asynchronous Graph Programming

Cited by 8 publications

References 21 publications

Cross-Paradigm Compilation across Programming Models: from Imperative to Asynchronous Graph

Cross-Paradigm Compilation across Programming Models: from Imperative to Asynchronous Graph

A Message Passing Many-Core Architecture for an Asynchronous Graph Programming Model

Arch-Lang : A DSE Tool for Flexible Hardware/Software Co-Design

Contact Info

Product

Resources

About