对数函数ln(x)在SDTA指令集结构上的实现

Implementation of the Natural Logarithm Function ln(x) on the SDTA Instruction Set Architecture

SDTA指令集体系结构是一种基于传输触发的VLIW体系结构。本文结合SDTA指令集结构的特点,经过循环展开和循环化简、强度消弱、过程集成、机器方言和指令归并等指令调度优化技术,高效实现了自然对数函数ln(x)。实验结果表明,在Neuron处理器上,ln(x)不但数据精度高,而且运行周期数只有gcc3.2.2数学库中自然对数函数运行周期数的33%左右。

Based on the transport triggering architecture, the SDTA （Synchronized Data Triggering Architecture） instruction set architecture is a VLIW（Very Long Instruction Word） architecture. According to the SDTA instruction set architecture, taking advantage of the instruction scheduling and optimization techniques,such as loop unrolling and loop sim- plification, strength reduction, procedure integration, machine idiom and instruction merge, ln（x） is implemented successfully in this paper. The experimental results show that the running circles of ln（x） is decreased to about 33%, compared with the natural logarithm function in the mathematical library of gcc3.2. 2 on the Neuron microprocessor, while the precision of ln（x） is adequately high.