数字信号处理器(digital signal processor,DSP)通常采用超长指令字(very long instruction word,VLIW)和单指令多数据(single instruction multiple data,SIMD)的架构来提升处理器整体计算性能,从而适用于高性能计算、图像处理、嵌入...数字信号处理器(digital signal processor,DSP)通常采用超长指令字(very long instruction word,VLIW)和单指令多数据(single instruction multiple data,SIMD)的架构来提升处理器整体计算性能,从而适用于高性能计算、图像处理、嵌入式系统等各个领域.飞腾迈创数字处理器(FT-Matrix)作为国防科技大学自主研制的高性能通用数字信号处理器,其极致计算性能的体现依赖于对VLIW与SIMD架构特点的充分挖掘.不止是飞腾迈创系列,绝大多数处理器上高度优化的内核代码或核心库函数都依赖于底层汇编级工具或手工开发.然而,手工编写内核算子的开发方法总是需要大量的时间和人力开销来充分释放硬件的性能潜力.尤其是VLIW+SIMD的处理器,专家级汇编开发的难度更为突出.针对这些问题,提出一种面向飞腾迈创数字处理器的高性能的内核代码自动生成框架(automatic kernel code-generation framework on FT-Matrix),将飞腾迈创处理器的架构特性引入到多层次的内核代码优化方法中.该框架包括3层优化组件:自适应循环分块、标向量协同的自动向量化和细粒度的指令级优化.该框架可以根据硬件的内存层次结构和内核的数据布局自动搜索最优循环分块参数,并进一步引入标量-向量单元协同的自动向量化指令选择与数据排布,以提高内核代码执行时的数据复用和并行性.此外,该框架提供了类汇编的中间表示,以应用各种指令级优化来探索更多指令级并行性(ILP)的优化空间,同时也为其他硬件平台提供了后端快速接入和自适应代码生成的模块,以实现高效内核代码开发的敏捷设计.实验表明,该框架生成的内核基准测试代码的平均性能是目标-数字信号处理器(DSP)--的手工函数库的3.25倍,是使用普通向量C语言编写的内核代码的20.62倍.展开更多
针对浅地表频域电磁探测对接收信号采集、传输和现场高效处理的要求,提出基于FPGA+DSP的浅地表频域电磁探测数字处理系统.在FPGA中实现数据采集、控制和传输FIFO(First Input First Output)模块,采用新式通用并行端口UPP(Universal Para...针对浅地表频域电磁探测对接收信号采集、传输和现场高效处理的要求,提出基于FPGA+DSP的浅地表频域电磁探测数字处理系统.在FPGA中实现数据采集、控制和传输FIFO(First Input First Output)模块,采用新式通用并行端口UPP(Universal Parallel Port)实现大数据传输,基于TMS320C6748平台,采用正交锁定放大方法,设计高效率数据处理算法,利用上位机软件通过RJ45网口对系统进行控制并显示结果.实测结果表明:该架构数字处理系统,对不同金属有着较强探测能力,加快了数据传输速率,缩短了系统工作时间,提高了工作效率.展开更多
A total solution for Continuous Image Digital Processing System with High Precision (Digital Film System)has been introduced in this paper. The detailed information ,such as the architecture ,the work flow and cvcn th...A total solution for Continuous Image Digital Processing System with High Precision (Digital Film System)has been introduced in this paper. The detailed information ,such as the architecture ,the work flow and cvcn the construction of every part of the system,has been discussed. At last,a practical system has been given.展开更多
The membranes were prepared from Konjac Glucomannan at different stirring conditions.The stirring processes that affect the microstructure properties of membrane surface were preliminarily explored by means of two dig...The membranes were prepared from Konjac Glucomannan at different stirring conditions.The stirring processes that affect the microstructure properties of membrane surface were preliminarily explored by means of two digital parameters of fractal dimension and variance of gray level change of membrane surface SEM-image.Variance analysis of two factors of stirring and magnification was carried out,respectvely,to two parameters.展开更多
文摘数字信号处理器(digital signal processor,DSP)通常采用超长指令字(very long instruction word,VLIW)和单指令多数据(single instruction multiple data,SIMD)的架构来提升处理器整体计算性能,从而适用于高性能计算、图像处理、嵌入式系统等各个领域.飞腾迈创数字处理器(FT-Matrix)作为国防科技大学自主研制的高性能通用数字信号处理器,其极致计算性能的体现依赖于对VLIW与SIMD架构特点的充分挖掘.不止是飞腾迈创系列,绝大多数处理器上高度优化的内核代码或核心库函数都依赖于底层汇编级工具或手工开发.然而,手工编写内核算子的开发方法总是需要大量的时间和人力开销来充分释放硬件的性能潜力.尤其是VLIW+SIMD的处理器,专家级汇编开发的难度更为突出.针对这些问题,提出一种面向飞腾迈创数字处理器的高性能的内核代码自动生成框架(automatic kernel code-generation framework on FT-Matrix),将飞腾迈创处理器的架构特性引入到多层次的内核代码优化方法中.该框架包括3层优化组件:自适应循环分块、标向量协同的自动向量化和细粒度的指令级优化.该框架可以根据硬件的内存层次结构和内核的数据布局自动搜索最优循环分块参数,并进一步引入标量-向量单元协同的自动向量化指令选择与数据排布,以提高内核代码执行时的数据复用和并行性.此外,该框架提供了类汇编的中间表示,以应用各种指令级优化来探索更多指令级并行性(ILP)的优化空间,同时也为其他硬件平台提供了后端快速接入和自适应代码生成的模块,以实现高效内核代码开发的敏捷设计.实验表明,该框架生成的内核基准测试代码的平均性能是目标-数字信号处理器(DSP)--的手工函数库的3.25倍,是使用普通向量C语言编写的内核代码的20.62倍.
文摘A total solution for Continuous Image Digital Processing System with High Precision (Digital Film System)has been introduced in this paper. The detailed information ,such as the architecture ,the work flow and cvcn the construction of every part of the system,has been discussed. At last,a practical system has been given.
文摘The membranes were prepared from Konjac Glucomannan at different stirring conditions.The stirring processes that affect the microstructure properties of membrane surface were preliminarily explored by means of two digital parameters of fractal dimension and variance of gray level change of membrane surface SEM-image.Variance analysis of two factors of stirring and magnification was carried out,respectvely,to two parameters.