基于无线通信DSP中超越函数的设计
发布时间：2019-01-20 08:47
【摘要】：无线通信是指利用电磁波信号能在自由空间中传播的这一特性进行信息交换的通信方式。现今，对无线通信系统越来越深入的研究，要求驱动装置的处理器必须具有更高的性能。数字信号处理器(DSP)高速，高集成度，低功耗，并行结构，完善的开发环境，并拥有仿真模块和仿真调试接口的特点，正是新型的无线通信所需要的。因而，，无线通信与DSP芯片相结合的技术得到了越来越广泛的应用。本文研究的Tricore电路就是将DSP和微控制器(MCU)集成到一个基于精简指令集计算机(RISC)的32位CPU内核上的电路。 超越函数在数学领域中与代数函数相反，指的是变量之间的关系式不能用有限次的加、减、乘、除、乘方、开方表示的函数。例如指数函数，对数函数，三角函数，反三角函数就属于超越函数，而且属于初等超越函数。本文研究的就是初等超越函数中的指数函数和对数函数。 指数、对数运算是Turbo码的MAP算法和Log-MAP算法中不可避免的运算，且实验证明，指数、对数运算的复杂程度直接影响了整个算法的执行效率。现有的研究人员多采用改变算法或函数近似的方式避免进行指数、对数运算，从而达到简化运算的效果。本文针对这个问题，在Tricore电路原有的指令集上扩展了专门用于指数和对数运算的指令，并从电路上实现其功能，可供MAP算法和Log-MAP算法的调用，以降低运算复杂度。 首先，本文研究了指数函数和对数函数的算法原理，包括完全查表法、扩展查表法、泰勒展开法、线性近似法及查表与线性近似相结合的方法，对各种算法的精度和占用存储空间进行比较；接着，基于Tricore电路的结构和指令集定义格式提出了用于指数和对数运算的专用扩展指令，给出具体的指令格式编码；然后基于这些扩展指令，用Verilog HDL语言实现了协处理器内核的基本系统结构，并针对协处理器的精度优化提出了增加一个标志信号的方法；最后针对指数函数和对数函数的各种算法及设计的协处理器进行了功能的仿真与验证。
[Abstract]:Wireless communication means to exchange information by using the characteristic that electromagnetic wave signal can propagate in free space. Nowadays, more and more research on wireless communication system requires that the processor of driving device must have higher performance. Digital signal processor (DSP) has the characteristics of high speed, high integration, low power consumption, parallel structure, perfect development environment, and has the characteristics of simulation module and simulation debugging interface. Therefore, the technology of combining wireless communication with DSP chip has been applied more and more widely. The Tricore circuit studied in this paper is the integration of DSP and microcontroller (MCU) into a 32-bit CPU kernel based on reduced instruction set computer (RISC). Transcendental function is the opposite of algebraic function in the field of mathematics. It refers to the function in which the relation between variables can not be expressed by finite order addition, subtraction, multiplication, division, multiplier and square. For example, exponential function, logarithmic function, trigonometric function and anti-trigonometric function belong to transcendental function and elementary transcendental function. In this paper, we study the exponential function and the logarithmic function in the elementary transcendental function. Exponential and logarithmic operation is the inevitable operation in MAP algorithm and Log-MAP algorithm of Turbo code. The experiment shows that the complexity of exponential and logarithmic operation directly affects the efficiency of the whole algorithm. Most of the present researchers avoid exponential and logarithmic operations by changing the algorithm or function approximation, so as to simplify the operation. Aiming at this problem, this paper extends the instruction specially used for exponent and logarithmic operation on the original instruction set of Tricore circuit, and realizes its function from the circuit, which can be used to call MAP algorithm and Log-MAP algorithm, so as to reduce the computational complexity. First of all, the algorithm principle of exponential function and logarithmic function is studied in this paper, including complete look-up table method, extended look-up table method, Taylor expansion method, linear approximation method and the combination of look-up table and linear approximation. The accuracy and storage space of various algorithms are compared. Then, based on the structure of Tricore circuit and instruction set definition format, a special extended instruction for exponential and logarithmic operation is proposed, and the specific instruction format coding is given. Then, based on these extended instructions, the basic system structure of coprocessor kernel is implemented with Verilog HDL language, and a method of adding a sign signal to the precision optimization of coprocessor is proposed. Finally, various algorithms of exponential function and logarithmic function and the designed coprocessor are simulated and verified.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP368.1
本文编号：2411867
[Abstract]:Wireless communication means to exchange information by using the characteristic that electromagnetic wave signal can propagate in free space. Nowadays, more and more research on wireless communication system requires that the processor of driving device must have higher performance. Digital signal processor (DSP) has the characteristics of high speed, high integration, low power consumption, parallel structure, perfect development environment, and has the characteristics of simulation module and simulation debugging interface. Therefore, the technology of combining wireless communication with DSP chip has been applied more and more widely. The Tricore circuit studied in this paper is the integration of DSP and microcontroller (MCU) into a 32-bit CPU kernel based on reduced instruction set computer (RISC). Transcendental function is the opposite of algebraic function in the field of mathematics. It refers to the function in which the relation between variables can not be expressed by finite order addition, subtraction, multiplication, division, multiplier and square. For example, exponential function, logarithmic function, trigonometric function and anti-trigonometric function belong to transcendental function and elementary transcendental function. In this paper, we study the exponential function and the logarithmic function in the elementary transcendental function. Exponential and logarithmic operation is the inevitable operation in MAP algorithm and Log-MAP algorithm of Turbo code. The experiment shows that the complexity of exponential and logarithmic operation directly affects the efficiency of the whole algorithm. Most of the present researchers avoid exponential and logarithmic operations by changing the algorithm or function approximation, so as to simplify the operation. Aiming at this problem, this paper extends the instruction specially used for exponent and logarithmic operation on the original instruction set of Tricore circuit, and realizes its function from the circuit, which can be used to call MAP algorithm and Log-MAP algorithm, so as to reduce the computational complexity. First of all, the algorithm principle of exponential function and logarithmic function is studied in this paper, including complete look-up table method, extended look-up table method, Taylor expansion method, linear approximation method and the combination of look-up table and linear approximation. The accuracy and storage space of various algorithms are compared. Then, based on the structure of Tricore circuit and instruction set definition format, a special extended instruction for exponential and logarithmic operation is proposed, and the specific instruction format coding is given. Then, based on these extended instructions, the basic system structure of coprocessor kernel is implemented with Verilog HDL language, and a method of adding a sign signal to the precision optimization of coprocessor is proposed. Finally, various algorithms of exponential function and logarithmic function and the designed coprocessor are simulated and verified.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP368.1
【参考文献】
相关会议论文 前1条
1 李睿;胡剑浩;李少谦;;指数“拟和”的FPGA实现[A];2006中国西部青年通信学术会议论文集[C];2006年
本文编号：2411867
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2411867.html
