Search

Dynamically scheduled high-level synthesis (HLS) enables the use of load-store queues (LSQs) which can disambiguate data hazards at circuit runtime, increasing throughput in codes with unpredictable memory accesses. However, the increased throughput comes at the price of lower clock frequency and higher resource usage compared to statically scheduled circuits without LSQs. The lower frequency often nullifies any throughput improvements over static scheduling, while the resource usage becomes prohibitively expensive with large queue sizes. This paper presents a method for achieving dynamically scheduled memory operations in HLS without significant clock period and resource usage increase. We present a novel LSQ based on shift-registers enabled by the opportunity to specialize queue sizes to a target code in HLS. We show a method to speculatively allocate addresses to our LSQ, significantly increasing pipeline parallelism in codes that could not benefit from an LSQ before. In stark contrast to traditional load value speculation, we do not require pipeline replays and have no overhead on misspeculation. On a set of benchmarks with data hazards, our approach achieves an average speedup of 11$\times$ against static HLS and 5$\times$ against dynamic HLS that uses a state of the art LSQ from previous work. Our LSQ also uses several times fewer resources, scaling to queues with hundreds of entries, and supports both on-chip and off-chip memory., Comment: To appear in the International Conference on Field Programmable Technology (FPT'23), Yokohama, Japan, 11-14 December 2023

Dynamically scheduled high-level synthesis (HLS) achieves higher throughput than static HLS for codes with unpredictable memory accesses and control flow. However, excessive dataflow scheduling results in circuits that use more resources and have a slower critical path, even when only a part of the circuit exhibits dynamic behavior. Recent work has shown that marking parts of a dataflow circuit for static scheduling can save resources and improve performance (hybrid scheduling), but the dynamic part of the circuit still bottlenecks the critical path. We propose instead to selectively introduce dynamic scheduling into static HLS. This paper presents an algorithm for identifying code regions amenable to dynamic scheduling and shows a methodology for introducing dynamically scheduled basic blocks, loops, and memory operations into static HLS. Our algorithm is informed by modulo-scheduling and can be integrated into any modulo-scheduled HLS tool. On a set of ten benchmarks, we show that our approach achieves on average an up to 3.7$\times$ and 3$\times$ speedup against dynamic and hybrid scheduling, respectively, with an area overhead of 1.3$\times$ and frequency degradation of 0.74$\times$ when compared to static HLS., Comment: To appear in the 33rd International Conference on Field-Programmable Logic and Applications (2023)

The current emissions from computing are almost 4% of the world total. This is already more than emissions from the airline industry and are projected to rise steeply over the next two decades. By 2040 emissions from computing alone will account for more than half of the emissions budget to keep global warming below 1.5$^\circ$C. Consequently, this growth in computing emissions is unsustainable. The emissions from production of computing devices exceed the emissions from operating them, so even if devices are more energy efficient producing more of them will make the emissions problem worse. Therefore we must extend the useful life of our computing devices. As a society we need to start treating computational resources as finite and precious, to be utilised only when necessary, and as effectively as possible. We need frugal computing: achieving our aims with less energy and material., Comment: IAB workshop on Environmental Impact of Internet Applications and Systems, 2022

For efficient hardware-accelerated simulations of quantum circuits, we can define hardware-specific quantum-circuit transformations. We use a functional programming approach to create a quantum-circuit analysis and transformation method implemented in Haskell. This tool forms a key part of our larger quantum-computing simulation toolchain. As an example of hardware acceleration, we discuss FPGA-based simulations of selected quantum arithmetic circuits, including the transformation steps to optimise the hardware utilisation. Future development steps in the Haskell-based analysis and transformation tool are outlined. The described toolchain can be found on GitHub: https://github.com/DevdudeSami/fqt.

Session Types offer a typing discipline that allows protocol specifications to be used during type-checking, ensuring that implementations adhere to a given specification. When looking to realise global session types in a dependently typed language care must be taken that values introduced in the description are used by roles that know about the value. We present Sessions, a Resource Dependent EDSL for describing global session descriptions in the dependently typed language Idris. As we construct session descriptions the values parameterising the EDSLs' type keeps track of roles and messages they have encountered. We can use this knowledge to ensure that message values are only used by those who know the value. Sessions supports protocol descriptions that are computable, composable, higher-order, and value-dependent. We demonstrate Sessions expressiveness by describing the TCP Handshake, a multi-modal server providing echo and basic arithmetic operations, and a Higher-Order protocol that supports an authentication interaction step., Comment: In Proceedings PLACES 2019, arXiv:1904.00396

There is a large body of legacy scientific code written in languages like Fortran that is not optimised to get the best performance out of heterogeneous acceleration devices like GPUs and FPGAs, and manually porting such code into parallel languages frameworks like OpenCL requires considerable effort. We are working towards developing a turn-key, self-optimising compiler for accelerating scientific applications, that can automatically transform legacy code into a solution for heterogeneous targets. In this paper we focus on FPGAs as the acceleration devices, and carry out our discussion in the context of the OpenCL programming framework. We show a route to automatic creation of kernels which are optimised for execution in a "streaming" fashion, which gives optimal performance on FPGAs. We use a 2D shallow-water model as an illustration; specifically we show how the use of \emph{channels} to communicate directly between peer kernels and the use of on-chip memory to create stencil buffers can lead to significant performance improvements. Our results show better FPGA performance against a baseline CPU implementation, and better energy-efficiency against both CPU and GPU implementations., Comment: Presented: HLPGPU at HiPEAC 2018 (Manchester, UK). arXiv admin note: text overlap with arXiv:1711.04471

Image convolution is widely used for sharpening, blurring and edge detection. In this paper, we review two common algorithms for convolving a 2D image by a separable kernel (filter). After optimising the naive codes using loop unrolling and SIMD vectorisation, we choose the algorithm with better performance as the baseline for parallelisation. We then compare the parallel performance of the optimised code using OpenMP, OpenCL and GPRM implementations on the Intel Xeon Phi. We also measure the effects of optimisation techniques and demonstrate how they affect both sequential and parallel performance. Apart from comparing the code complexity as well as the performance of the chosen parallel programming models, we investigate the impact of a parallelisation technique, task agglomeration in GPRM., Comment: This work has been included in the doctoral thesis of Ashkan Tousimojarad, "GPRM: a high performance programming framework for manycore processors", University of Glasgow, 2016

Massively parallel accelerators such as GPGPUs, manycores and FPGAs represent a powerful and affordable tool for scientists who look to speed up simulations of complex systems. However, porting code to such devices requires a detailed understanding of heterogeneous programming tools and effective strategies for parallelization. In this paper we present a source to source compilation approach with whole-program analysis to automatically transform single-threaded FORTRAN 77 legacy code into OpenCL-accelerated programs with parallelized kernels. The main contributions of our work are: (1) whole-source refactoring to allow any subroutine in the code to be offloaded to an accelerator. (2) Minimization of the data transfer between the host and the accelerator by eliminating redundant transfers. (3) Pragmatic auto-parallelization of the code to be offloaded to the accelerator by identification of parallelizable maps and reductions. We have validated the code transformation performance of the compiler on the NIST FORTRAN 78 test suite and several real-world codes: the Large Eddy Simulator for Urban Flows, a high-resolution turbulent flow model; the shallow water component of the ocean model Gmodel; the Linear Baroclinic Model, an atmospheric climate model and Flexpart-WRF, a particle dispersion simulator. The automatic parallelization component has been tested on as 2-D Shallow Water model (2DSW) and on the Large Eddy Simulator for Urban Flows (UFLES) and produces a complete OpenCL-enabled code base. The fully OpenCL-accelerated versions of the 2DSW and the UFLES are resp. 9x and 20x faster on GPU than the original code on CPU, in both cases this is the same performance as manually ported code., Comment: 12 pages, 5 figures, submitted to "Computers and Fluids" as full paper from ParCFD conference entry

In this paper I introduce a mechanism to derive program transforma- tions from order-preserving transformations of vector types. The purpose of this work is to allow automatic generation of correct-by-construction instances of programs in a streaming data processing paradigm suitable for FPGA processing. We show that for it is possible to automatically derive instances for programs based on combinations of opaque element- processing functions combined using foldl and map, purely from the type transformations., Comment: This work is supported by the EPSRC through the TyTra project (EP/L00058X/1)

High Performance Computing (HPC) platforms allow scientists to model computationally intensive algorithms. HPC clusters increasingly use General-Purpose Graphics Processing Units (GPGPUs) as accelerators; FPGAs provide an attractive alternative to GPGPUs for use as co-processors, but they are still far from being mainstream due to a number of challenges faced when using FPGA-based platforms. Our research aims to make FPGA-based high performance computing more accessible to the scientific community. In this work we present the results of investigating the acceleration of a particular atmospheric model, Flexpart, on FPGAs. We focus on accelerating the most computationally intensive kernel from this model. The key contribution of our work is the architectural exploration we undertook to arrive at a solution that best exploits the parallelism available in the legacy code, and is also convenient to program, so that eventually the compilation of high-level legacy code to our architecture can be fully automated. We present the three different types of architecture, comparing their resource utilization and performance, and propose that an architecture where there are a number of computational cores, each built along the lines of a vector instruction processor, works best in this particular scenario, and is a promising candidate for a generic FPGA-based platform for scientific computation. We also present the results of experiments done with various configuration parameters of the proposed architecture, to show its utility in adapting to a range of scientific applications., Comment: This is an extended pre-print version of work that was presented at the international symposium on Highly Efficient Accelerators and Reconfigurable Technologies (HEART2014), Sendai, Japan, June 911, 2014

We present the TyTra-IR, a new intermediate language intended as a compilation target for high-level language compilers and a front-end for HDL code generators. We develop the requirements of this new language based on the design-space of FPGAs that it should be able to express and the estimation-space in which each configuration from the design-space should be mappable in an automated design flow. We use a simple kernel to illustrate multiple configurations using the semantics of TyTra-IR. The key novelty of this work is the cost model for resource-costs and throughput for different configurations of interest for a particular kernel. Through the realistic example of a Successive Over-Relaxation kernel implemented both in TyTra-IR and HDL, we demonstrate both the expressiveness of the IR and the accuracy of our cost model., Comment: Pre-print and extended version of poster paper accepted at international symposium on Highly Efficient Accelerators and Reconfigurable Technologies (HEART2015) Boston, MA, USA, June 1-2, 2015

In this report we present a novel approach to model coupling for shared-memory multicore systems hosting OpenCL-compliant accelerators, which we call The Glasgow Model Coupling Framework (GMCF). We discuss the implementation of a prototype of GMCF and its application to coupling the Weather Research and Forecasting Model and an OpenCL-accelerated version of the Large Eddy Simulator for Urban Flows (LES) developed at DPRI. The first stage of this work concerned the OpenCL port of the LES. The methodology used for the OpenCL port is a combination of automated analysis and code generation and rule-based manual parallelization. For the evaluation, the non-OpenCL LES code was compiled using gfortran, fort and pgfortran}, in each case with auto-parallelization and auto-vectorization. The OpenCL-accelerated version of the LES achieves a 7 times speed-up on a NVIDIA GeForce GTX 480 GPGPU, compared to the fastest possible compilation of the original code running on a 12-core Intel Xeon E5-2640. In the second stage of this work, we built the Glasgow Model Coupling Framework and successfully used it to couple an OpenMP-parallelized WRF instance with an OpenCL LES instance which runs the LES code on the GPGPI. The system requires only very minimal changes to the original code. The report discusses the rationale, aims, approach and implementation details of this work., Comment: This work was conducted during a research visit at the Disaster Prevention Research Institute of Kyoto University, supported by an EPSRC Overseas Travel Grant, EP/L026201/1

Processors with large numbers of cores are becoming commonplace. In order to take advantage of the available resources in these systems, the programming paradigm has to move towards increased parallelism. However, increasing the level of concurrency in the program does not necessarily lead to better performance. Parallel programming models have to provide flexible ways of defining parallel tasks and at the same time, efficiently managing the created tasks. OpenMP is a widely accepted programming model for shared-memory architectures. In this paper we highlight some of the drawbacks in the OpenMP tasking approach, and propose an alternative model based on the Glasgow Parallel Reduction Machine (GPRM) programming framework. As the main focus of this study, we deploy our model to solve a fundamental linear algebra problem, LU factorisation of sparse matrices. We have used the SparseLU benchmark from the BOTS benchmark suite, and compared the results obtained from our model to those of the OpenMP tasking approach. The TILEPro64 system has been used to run the experiments. The results are very promising, not only because of the performance improvement for this particular problem, but also because they verify the task management efficiency, stability, and flexibility of our model, which can be applied to solve problems in future many-core systems., Comment: Final version as appeared in "dx.doi.org/10.1109/ISPDC.2014.11"

The emergence of multicore and manycore processors is set to change the parallel computing world. Applications are shifting towards increased parallelism in order to utilise these architectures efficiently. This leads to a situation where every application creates its desirable number of threads, based on its parallel nature and the system resources allowance. Task scheduling in such a multithreaded multiprogramming environment is a significant challenge. In task scheduling, not only the order of the execution, but also the mapping of threads to the execution resources is of a great importance. In this paper we state and discuss some fundamental rules based on results obtained from selected applications of the BOTS benchmarks on the 64-core TILEPro64 processor. We demonstrate how previously efficient mapping policies such as those of the SMP Linux scheduler become inefficient when the number of threads and cores grows. We propose a novel, low-overhead technique, a heuristic based on the amount of time spent by each CPU doing some useful work, to fairly distribute the workloads amongst the cores in a multiprogramming environment. Our novel approach could be implemented as a pragma similar to those in the new task-based OpenMP versions, or can be incorporated as a distributed thread mapping mechanism in future manycore programming frameworks. We show that our thread mapping scheme can outperform the native GNU/Linux thread scheduler in both single-programming and multiprogramming environments., Comment: ParCo Conference, Munich, Germany, 2013

With rapidly evolving technology, multicore and manycore processors have emerged as promising architectures to benefit from increasing transistor numbers. The transition towards these parallel architectures makes today an exciting time to investigate challenges in parallel computing. The TILEPro64 is a manycore accelerator, composed of 64 tiles interconnected via multiple 8x8 mesh networks. It contains per-tile caches and supports cache-coherent shared memory by default. In this paper we present a programming technique to take advantages of distributed caching facilities in manycore processors. However, unlike other work in this area, our approach does not use architecture-specific libraries. Instead, we provide the programmer with a novel technique on how to program future Non-Uniform Cache Architecture (NUCA) manycore systems, bearing in mind their caching organisation. We show that our localised programming approach can result in a significant improvement of the parallelisation efficiency (speed-up)., Comment: This work was presented at the international symposium on Highly- Efficient Accelerators and Reconfigurable Technologies (HEART2013), Edinburgh, Scotland, June 13-14, 2013

We present the Glasgow Parallel Reduction Machine (GPRM), a novel, flexible framework for parallel task-composition based many-core programming. We allow the programmer to structure programs into task code, written as C++ classes, and communication code, written in a restricted subset of C++ with functional semantics and parallel evaluation. In this paper we discuss the GPRM, the virtual machine framework that enables the parallel task composition approach. We focus the discussion on GPIR, the functional language used as the intermediate representation of the bytecode running on the GPRM. Using examples in this language we show the flexibility and power of our task composition framework. We demonstrate the potential using an implementation of a merge sort algorithm on a 64-core Tilera processor, as well as on a conventional Intel quad-core processor and an AMD 48-core processor system. We also compare our framework with OpenMP tasks in a parallel pointer chasing algorithm running on the Tilera processor. Our results show that the GPRM programs outperform the corresponding OpenMP codes on all test platforms, and can greatly facilitate writing of parallel programs, in particular non-data parallel algorithms such as reductions., Comment: In Proceedings PLACES 2013, arXiv:1312.2218

PLACES 2013 (full title: Programming Language Approaches to Concurrency- and Communication-cEntric Software) was the sixth edition of the PLACES workshop series. After the first PLACES, which was affiliated to DisCoTec in 2008, the workshop has been part of ETAPS every year since 2009 and is now an established part of the ETAPS satellite events. This year, PLACES was the best attended workshop at ETAPS 2013. The workshop series was started in order to promote the application of novel programming language ideas to the increasingly important problem of developing software for systems in which concurrency and communication are intrinsic aspects. This includes software for multi- and many-core systems, accelerators and large-scale distributed and/or service-oriented systems. The scope of PLACES includes new programming language features, whole new programming language designs, new type systems, new semantic approaches, new program analysis techniques, and new implementation mechanisms.