Accelerator Cores Research Articles

Memory Allocation for Window-Based Image Processing on Multiple Memory Modules with Simple Addressing Functions

Accelerator cores in low-power embedded processors have on-chip multiple memory modules to increase the data access speed and to enable parallel data access. When large functional units such as multipliers and dividers are used for addressing, a large power and chip area are consumed. Therefore, recent low-power processors use small functional units such as adders and counters to reduce the power and area. Such small functional units make it difficult to implement complex addressing patterns without duplicating data among multiple memory modules. The data duplication wastes the memory capacity and increases the data transfer time significantly. This paper proposes a method to reduce the memory duplication for window-based image processing, which is widely used in many applications. Evaluations using an accelerator core show that the proposed method reduces the data amount and data transfer time by more than 50%.

Exploiting SPMD Horizontal Locality

In this paper, we analyze a particular spatial locality case (called horizontal locality) inherent to manycore accelerator architectures employing barrel execution of SPMD kernels, such as GPUs. We then propose an adaptive memory access granularity framework to exploit and enforce the horizontal locality in order to reduce the interferences among accelerator cores memory accesses and hence improve DRAM efficiency. With the proposed technique, DRAM efficiency grows by 1.42X on average, resulting in 12.3% overall performance gain, for a set of representative memory intensive GPGPU applications.

Lattice QCD Applications on QPACE

QPACE is a novel massively parallel architecture optimized for lattice QCD simulations. A single QPACE node is based on the IBM PowerXCell 8i processor. The nodes are interconnected by a custom 3-dimensional torus network implemented on an FPGA. The compute power of the processor is provided by 8 Synergistic Processing Units. Making efficient use of these accelerator cores in scientific applications is challenging. In this paper we describe our strategies for porting applications to the QPACE architecture and report on performance numbers.

Implementing molecular dynamics on hybrid high performance computers – short range forces

The use of accelerators such as graphics processing units (GPUs) has become popular in scientific computing applications due to their low cost, impressive floating-point capabilities, high memory bandwidth, and low electrical power requirements. Hybrid high-performance computers, machines with more than one type of floating-point processor, are now becoming more prevalent due to these advantages. In this work, we discuss several important issues in porting a large molecular dynamics code for use on parallel hybrid machines – (1) choosing a hybrid parallel decomposition that works on central processing units (CPUs) with distributed memory and accelerator cores with shared memory, (2) minimizing the amount of code that must be ported for efficient acceleration, (3) utilizing the available processing power from both multi-core CPUs and accelerators, and (4) choosing a programming model for acceleration. We present our solution to each of these issues for short-range force calculation in the molecular dynamics package LAMMPS, however, the methods can be applied in many molecular dynamics codes. Specifically, we describe algorithms for efficient short range force calculation on hybrid high-performance machines. We describe an approach for dynamic load balancing of work between CPU and accelerator cores. We describe the Geryon library that allows a single code to compile with both CUDA and OpenCL for use on a variety of accelerators. Finally, we present results on a parallel test cluster containing 32 Fermi GPUs and 180 CPU cores.

Hera-JVM

Heterogeneous multi-core processors, such as the IBM Cell processor, can deliver high performance. However, these processors are notoriously difficult to program: different cores support different instruction set architectures, and the processor as a whole does not provide coherence between the different cores' local memories. We present Hera-JVM, an implementation of the Java Virtual Machine which operates over the Cell processor, thereby making this platforms more readily accessible to mainstream developers. Hera-JVM supports the full Java language; threads from an unmodified Java application can be simultaneously executed on both the main PowerPC-based core and on the additional SPE accelerator cores. Migration of threads between these cores is transparent from the point of view of the application, requiring no modification to Java source code or bytecode. Hera-JVM supports the existing Java Memory Model, even though the underlying hardware does not provide cache coherence between the different core types. We examine Hera-JVM's performance under a series of real-world Java benchmarks from the SpecJVM, Java Grande and Dacapo benchmark suites. These benchmarks show a wide variation in relative performance on the different core types of the Cell processor, depending upon the nature of their workload. Execution of these benchmarks on Hera-JVM can achieve speedups of up to 2.25x by using one of the Cell processor's SPE accelerator cores, compared to execution on the main PowerPC-based core. When all six SPE cores are exploited, parallel workloads can achieve speedups of up to 13x compared to execution on the single PowerPC core.

Chameleon

Heterogeneous multicore processors have emerged as an energy- and area-efficient architectural solution to improving performance for domain-specific applications such as those with a plethora of data-level parallelism. These processors typically contain a large number of small, compute-centric cores for acceleration while keeping one or two high-performance ILP cores on the die to guarantee single-thread performance. Although a major portion of the transistors are occupied by the acceleration cores, these resources will sit idle when running unparallelized legacy codes or the sequential part of an application. To address this underutilization issue, in this article, we introduce Chameleon, a flexible heterogeneous multicore architecture to virtualize these resources for enhancing memory performance when running sequential programs. The Chameleon architecture can dynamically virtualize the idle acceleration cores into a last-level cache, a data prefetcher, or a hybrid between these two techniques. In addition, Chameleon can operate in an adaptive mode that dynamically configures the acceleration cores between the hybrid mode and the prefetch-only mode by monitoring the effectiveness of the Chameleon cache mode. In our evaluation with SPEC2006 benchmark suite, different levels of performance improvements were achieved in different modes for different applications. In the case of the adaptive mode, Chameleon improves the performance of SPECint06 and SPECfp06 by 31% and 15%, on average. When considering only memory-intensive applications, Chameleon improves the system performance by 50% and 26% for SPECint06 and SPECfp06, respectively.

Task Allocation with Algorithm Transformation for Reducing Data-Transfer Bottlenecks in Heterogeneous Multi-Core Processors: A Case Study of HOG Descriptor Computation

Heterogeneous multi-core processors are attracted by the media processing applications due to their capability of drawing strengths of different cores to improve the overall performance. However, the data transfer bottlenecks and limitations in the task allocation due to the accelerator-incompatible operations prevents us from gaining full potential of the heterogeneous multi-core processors. This paper presents a task allocation method based on algorithm transformation to increase the freedom of task allocation. We use approximation methods such as CORDIC algorithms to map the accelerator-incompatible operations to accelerator cores. According to the experimental results using HOG descriptor computation, the proposed task allocation method reduces the data transfer time by more than 82% and the total processing time by more than 79% compared to the conventional task allocation method.

Application Acceleration with the Cell Broadband Engine

The Cell Broadband Engine is a heterogeneous chip multiprocessor that combines a PowerPC processor core with eight single-instruction multiple-data accelerator cores and delivers high performance on many computationally intensive codes.

StreamRay

The wide availability of commodity graphics processors has made real-time graphics an intrinsic component of the human/computer interface. These graphics cores accelerate the z-buffer algorithm and provide a highly interactive experience at a relatively low cost. However, many applications in entertainment, science, and industry require high quality lighting effects such as accurate shadows, reflection, and refraction. These effects can be difficult to achieve with z-buffer algorithms but are straightforward to implement using ray tracing. Although ray tracing is computationally more complex, the algorithm exhibits excellent scaling and parallelism properties. Nevertheless, ray tracing memory access patterns are difficult to predict and the parallelism speedup promise is therefore hard to achieve. This paper highlights a novel approach to ray tracing based on stream filtering and presents StreamRay, a multicore wide SIMD microarchitecture that delivers interactive frame rates of 15-32 frames/second for scenes of high geometric complexity and exhibits high utilization for SIMD widths ranging from eight to 16 elements. StreamRay consists of two main components: the ray engine, which is responsible for stream assembly and employs address generation units that generate addresses to form large SIMD vectors, and the filter engine, which implements the ray tracing operations with programmable accelerators. Results demonstrate that separating address and data processing reduces data movement and resource contention. Performance improves by 56% while simultaneously providing 11.63% power savings per accelerator core compared to a design which does not use separate resources for address and data computations.

StreamRay

The wide availability of commodity graphics processors has made real-time graphics an intrinsic component of the human/computer interface. These graphics cores accelerate the z-buffer algorithm and provide a highly interactive experience at a relatively low cost. However, many applications in entertainment, science, and industry require high quality lighting effects such as accurate shadows, reflection, and refraction. These effects can be difficult to achieve with z-buffer algorithms but are straightforward to implement using ray tracing. Although ray tracing is computationally more complex, the algorithm exhibits excellent scaling and parallelism properties. Nevertheless, ray tracing memory access patterns are difficult to predict and the parallelism speedup promise is therefore hard to achieve. This paper highlights a novel approach to ray tracing based on stream filtering and presents StreamRay, a multicore wide SIMD microarchitecture that delivers interactive frame rates of 15-32 frames/second for scenes of high geometric complexity and exhibits high utilization for SIMD widths ranging from eight to 16 elements. StreamRay consists of two main components: the ray engine, which is responsible for stream assembly and employs address generation units that generate addresses to form large SIMD vectors, and the filter engine, which implements the ray tracing operations with programmable accelerators. Results demonstrate that separating address and data processing reduces data movement and resource contention. Performance improves by 56% while simultaneously providing 11.63% power savings per accelerator core compared to a design which does not use separate resources for address and data computations.

Portable library development for reconfigurable computing systems: A case study

Portable libraries of highly-optimized hardware cores can significantly reduce the development time of reconfigurable computing applications. This paper presents the tradeoffs and challenges in the design of such libraries. A set of library development guidelines is provided, which has been validated with the RCLib case study. RCLib is a set of portable libraries with over 100 cores, targeting a wide range of applications. RCLib portability has been verified in three major High-Performance reconfigurable computing architectures: SRC6, Cray XD1 and SGI RC100. Compared to full-software implementations, applications using RCLib hardware acceleration cores show speedups ranging from one to four orders of magnitude.

Integer-pixel motion estimation H.264/AVC accelerator architecture with optimal memory management

This paper presents an improved accelerator core for H.264/AVC video-coding motion estimation. The proposed hardware architecture meets the integer-pixel, full-search block-matching algorithm requirements with an optimal memory management and an effective data-path. Performance characteristics like low latency, high processing speed and efficiency near 100% are achieved without a high control overhead. The core calculates the 41 best motion vectors using a pipeline process. It is composed of a systolic 16 × 16 processor elements array, a sum of absolute differences adder tree and a Lagrangian rate/distortion cost optimizer. Implementation results based on FPGA devices in a system on chip (SoC) structure using VHDL are included.

EXOCHI

Future mainstream microprocessors will likely integrate specialized accelerators, such as GPUs, onto a single die to achieve better performance and power efficiency. However, it remains a keen challenge to program such a heterogeneous multicore platform, since these specialized accelerators feature ISAs and functionality that are significantly different from the general purpose CPU cores. In this paper, we present EXOCHI: (1) Exoskeleton Sequencer (EXO), an architecture to represent heterogeneous acceleratorsas ISA-based MIMD architecture resources, and a shared virtual memory heterogeneous multithreaded program execution model that tightly couples specialized accelerator cores with generalpurpose CPU cores, and (2) C for Heterogeneous Integration (CHI), an integrated C/C++ programming environment that supports accelerator-specific inline assembly and domain-specific languages. The CHI compiler extends the OpenMP pragma for heterogeneous multithreading programming, and produces a single fat binary with code sections corresponding to different instruction sets. The runtime can judiciously spread parallel computation across the heterogeneous cores to optimize performance and power. We have prototyped the EXO architecture on a physical heterogeneous platform consisting of an Intel® Core™ 2 Duo processor and an 8-core 32-thread Intel® Graphics Media Accelerator X3000. In addition, we have implemented the CHI integrated programming environment with the Intel® C++ Compiler, runtime toolset, and debugger. On the EXO prototype system, we have enhanced a suite of production-quality media kernels for video and image processing to utilize the accelerator through the CHI programming interface, achieving significant speedup (1.41X to10.97X) over execution on the IA32 CPU alone.

Modelling and control strategy of the Italian LBE-XADS

In this paper the dynamic behaviour of the Italian 80 MW th XADS (eXperimental Accelerator Driven System) is investigated, by means of a simple model for its main components as well as the most relevant interactions among them, namely the accelerator–core coupling and the corresponding neutronic features, the primary system with lead bismuth eutectic (LBE) as primary coolant, the secondary circuit with an organic diathermic fluid, and the associated air cooling system for an effective rejection of thermal power to the environment as a final heat sink. The simulation and control of the system dynamics took into account the singular features of the components, which show very different time constants, and the wide time interval needed by the organic fluid to flow into the secondary loop of the reactor. A control strategy based on a simultaneous feedforward–feedback scheme has been adopted, the main requirement being to keep constant the average temperature of the secondary fluid. The dynamic response of the whole system has been evaluated, significant transient initiators have been considered: the results show that the proposed zero-dimensional model, in spite of its simplicity, is sufficiently accurate to predict the plant response in terms of the most relevant variables. The results also confirm that the Italian LBE-XADS exhibits a large operational margin, owing to favourable design choices and protective engineered actions as well as to the adopted control strategy of the system.

Transient analyses to support the definition of trade experiment start-up procedures

The TRADE experiment, where a TRIGA reactor in subcritical configuration is coupled with a proton accelerator (cyclotron) through a spallation target interface, is conceived to demonstrate the feasibility of the coupling of the different components of an ADS (accelerator, target and subcritical core) at a thermal power level able to trigger the reactivity thermal feedbacks. Among the main purposes of the TRADE experiments there is the investigation of the start-up and shutdown procedures, to be defined as much as possible according to the TRIGA-RC1 safety approach. In this paper a preliminary analysis of TRIGA transients induced by accidental jumps of the cyclotron proton beam current impinging on the target assuming different subcriticality levels of the core has been carried out by means of TIESTE-MINOSSE and PARCS/RELAP5 codes, both able to treat coupled neutron kinetic and thermal-hydraulics.

Multimedia Signal Processors: An Architectural Platform with Algorithmic Compilation

Novel algorithmic features of multimedia applications and advances in VLSI technologies are driving forces behind the new multimedia signal processors. We propose an architecture platform which could provide high performance and flexibility, and would require less external I/O and memory access. It is comprised of array processors to be used as the hardware accelerator and RISC cores to be used as the basis of the programmable processor. It is a hierarchical and scalable architecture style which facilitates the hardware-software codesign of multimedia signal processing circuits and systems. While some control-intensive functions can be implemented using programmable CPUs, other computation-intensive functions can rely on hardware accelerators. To compile multimedia algorithms, we also present an operation placement and scheduling scheme suitable for the proposed architectural platform. Our scheme addresses data reusability and exploits local communication in order to avoid the memory/communication bandwidth bottleneck, which leads to faster program execution. Our method shows a promising performance: a linear speed-up of 16 times can be achieved for the block-matching motion estimation algorithm and the true motion tracking algorithm, which have formed many multimedia applications (e.g., MPEG-2 and MPEG-4).