Parallel algorithm design for processor in memory (PIM) architectures

内存处理器 (PIM) 架构的并行算法设计

• 批准号: 239741-2006
• 负责人: Thulasiraman, Parimala
• 金额: $ 1.18万
• 依托单位: University of Manitoba
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2008
• 资助国家: 加拿大
• 起止时间: 2008-01-01 至 2009-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=394289
• 关键词: Parallel; algorithm; design; processor; memory

Parallel computing involves using  many computers to cooperatively solve a given problem fast. Parallel computing has played a significant role in application areas such as weather forecasting, physics and chemistry.Applications  as broadly based as finance, biology and medicine are bringing new to high performance parallel computing research.One of the fundamental problems in parallel computer architecture is that while processor speed has increased dramatically over the past few years, increaes in memory speed have been only modest.To bridge the processor-memory gap,computer architects have proposed integrating memory logic, processor and interconnect support on a single chip, called Processor-in-Memor(PIM) chips.Many such PIM chips can be connected creating a cellular architecture to achieve the next generation of supercomputing machines.The question one asks is what are we going to do with these powerful supercomputers?One of the most important applications that can profit from leveraging the powerful resources of parallel computing is medical imaging. Determining the internal structure of an organism with sufficient detail to yield diagnostic information of a patient,is a first step before administering any procedure.Medical imaging is the process by which physicians evaluate an area of the patient's body that is not normally visible.There are many imaging techiniques such Computed Tomography (CT), magnetic resonance imaging (MRI) and positron emission tomgraphy (PET). This research concentrates on (CT).These techniques detect abnormalities in organs such as breasts.  The images in any of the imaging  techniques are large requiring large amount of storage space and need to be processed fast to make real time decisions on the patient. Initially, the proposed research will focus on CT and expand it to other imaging techniques. Therefore, the focus of this proposed research will be on the design and performance evaluation of sophisticated algorithms for medical imaging techniques on PIM arcitectures .

并行计算涉及使用多台计算机协作快速解决给定问题。并行计算在天气预报、物理和化学等应用领域发挥了重要作用。金融、生物和医学等广泛的应用为高性能并行计算研究带来了新的机遇。并行计算机体系结构的一个基本问题是，虽然处理器速度在过去几年中有了显著提高，但内存速度的提高却非常有限。为了弥合处理器和内存之间的鸿沟，计算机架构师们提出将存储逻辑、处理器和互连支持集成在一个芯片上，称为内存处理器（PIM）芯片。许多这样的PIM芯片可以连接起来，形成一个蜂窝架构，以实现下一代超级计算机。问题是我们要用这些强大的超级计算机做什么？可以从并行计算的强大资源中获益的最重要的应用之一是医学成像。在实施任何程序之前，确定生物体的内部结构并提供足够的细节以获得患者的诊断信息是第一步。医学成像是医生评估病人身体正常情况下不可见区域的过程。有许多成像技术，如计算机断层扫描（CT），磁共振成像（MRI）和正电子发射断层扫描（PET）。本研究集中于（CT）。这些技术可以检测诸如乳房等器官的异常。任何成像技术中的图像都是大的，需要大量的存储空间，并且需要快速处理以对患者做出实时决策。初步研究将集中于CT，并扩展到其他成像技术。因此，本研究的重点将放在基于PIM架构的医学成像技术的复杂算法的设计和性能评估上。