Today, an ever increasing number of natural scientists use computers for data analysis, modeling, simulation and visualization of complex problems. However, in the last decade the computer architecture has changed sig...Today, an ever increasing number of natural scientists use computers for data analysis, modeling, simulation and visualization of complex problems. However, in the last decade the computer architecture has changed significantly, making it increasingly difficult to fully utilize the power of the processor, unless the scientist is a trained programmer. The reasons for this shift include the change from single-core to multi-core processors, as well as the decreasing price of hardware, which allows researchers to build cluster computers made from commodity hardware. Therefore, scientists must not only be able to handle multi-core processors, but also the problems associated with writing distributed memory programs and handle communication between hundreds of multi-core machines. Fortunately, there are a number of systems to help the scientist e.g. Message Parsing Interface (MPI) [1] for handling communication, DistNumPy [2] for handling data distribution and Communicating Sequential Processes (CSP) [3] for handling concurrency related problems. Having said that, it must be emphasized that all of these methods require that the scientists learn a new method and then rewrite their programs, which mean more work for the scientist. A solution that does not require much work for the scientists is automatic parallelization. However, research dating back three decades has yet to find fully automated parallelization as a feasible solution for programs in general, but some classes of programs can be automatically parallelized to an extent. This paper describes an external library which provides a Parallel. For loop construct, allowing the body of a loop to be run in Parallel across multiple networked machines, i.e. on distributed memory architectures. The individual machines themselves may be shared memory nodes of course. The idea is inspired by Microsoft’s Parallel Library that supplies multiple Parallel constructs. However, unlike Microsoft’s Library our library supports distributed memory architectures. Preliminary tests have shown that simple problems may be distributed easily and achieve good scalability. Unfortunately, the tests show that the scalability is limited by the number of accesses made to shared variables. Thus the applicability of the library is not general but limited to a subset of applications with only limited communication needs.展开更多
Net Meta XL是一款基于贝叶斯理论研发的在Microsot Excel框架下执行网状Meta分析的宏。该宏于2014年正式推出,将数据提取与输入、结果输出和图形绘制集于一体。当前,该宏可供选择的模型丰富,所有操作均为菜单式且操作简单,但其仅适用...Net Meta XL是一款基于贝叶斯理论研发的在Microsot Excel框架下执行网状Meta分析的宏。该宏于2014年正式推出,将数据提取与输入、结果输出和图形绘制集于一体。当前,该宏可供选择的模型丰富,所有操作均为菜单式且操作简单,但其仅适用于二分类数据的网状Meta分析,仍有较大提升空间和需要完善之处。本文通过实例介绍应用Net Meta XL宏实现网状Meta分析的方法。展开更多
Microsoft.NET主要由Windows.NET、.NET框架、.NET企业服务器、Orchestration和Visual Studio.NET组成。Microsoft.NET框架是一个多语言组件开发和执行环境,由公共语言运行库(CLR Common Language Runtime);统一的编程类;ASP.NET三部分...Microsoft.NET主要由Windows.NET、.NET框架、.NET企业服务器、Orchestration和Visual Studio.NET组成。Microsoft.NET框架是一个多语言组件开发和执行环境,由公共语言运行库(CLR Common Language Runtime);统一的编程类;ASP.NET三部分组成。 Microsoft.NET主要特点。①基于.NET平台的完善软件服务。随着.NET平台的推进,软件将逐渐从产品形式向服务形式转化,软件应用是以Web服务的形式出现并在Internet上发布的,这也是整个IT行业的大势所趋。基于.NET平台的用户、开发人员只需要定制服务。展开更多
文摘Today, an ever increasing number of natural scientists use computers for data analysis, modeling, simulation and visualization of complex problems. However, in the last decade the computer architecture has changed significantly, making it increasingly difficult to fully utilize the power of the processor, unless the scientist is a trained programmer. The reasons for this shift include the change from single-core to multi-core processors, as well as the decreasing price of hardware, which allows researchers to build cluster computers made from commodity hardware. Therefore, scientists must not only be able to handle multi-core processors, but also the problems associated with writing distributed memory programs and handle communication between hundreds of multi-core machines. Fortunately, there are a number of systems to help the scientist e.g. Message Parsing Interface (MPI) [1] for handling communication, DistNumPy [2] for handling data distribution and Communicating Sequential Processes (CSP) [3] for handling concurrency related problems. Having said that, it must be emphasized that all of these methods require that the scientists learn a new method and then rewrite their programs, which mean more work for the scientist. A solution that does not require much work for the scientists is automatic parallelization. However, research dating back three decades has yet to find fully automated parallelization as a feasible solution for programs in general, but some classes of programs can be automatically parallelized to an extent. This paper describes an external library which provides a Parallel. For loop construct, allowing the body of a loop to be run in Parallel across multiple networked machines, i.e. on distributed memory architectures. The individual machines themselves may be shared memory nodes of course. The idea is inspired by Microsoft’s Parallel Library that supplies multiple Parallel constructs. However, unlike Microsoft’s Library our library supports distributed memory architectures. Preliminary tests have shown that simple problems may be distributed easily and achieve good scalability. Unfortunately, the tests show that the scalability is limited by the number of accesses made to shared variables. Thus the applicability of the library is not general but limited to a subset of applications with only limited communication needs.
文摘Net Meta XL是一款基于贝叶斯理论研发的在Microsot Excel框架下执行网状Meta分析的宏。该宏于2014年正式推出,将数据提取与输入、结果输出和图形绘制集于一体。当前,该宏可供选择的模型丰富,所有操作均为菜单式且操作简单,但其仅适用于二分类数据的网状Meta分析,仍有较大提升空间和需要完善之处。本文通过实例介绍应用Net Meta XL宏实现网状Meta分析的方法。
文摘Microsoft.NET主要由Windows.NET、.NET框架、.NET企业服务器、Orchestration和Visual Studio.NET组成。Microsoft.NET框架是一个多语言组件开发和执行环境,由公共语言运行库(CLR Common Language Runtime);统一的编程类;ASP.NET三部分组成。 Microsoft.NET主要特点。①基于.NET平台的完善软件服务。随着.NET平台的推进,软件将逐渐从产品形式向服务形式转化,软件应用是以Web服务的形式出现并在Internet上发布的,这也是整个IT行业的大势所趋。基于.NET平台的用户、开发人员只需要定制服务。
