Method for performing data normalization processing by use of DMA (direct memory access) controller

Abstract

The invention provides a method for performing data normalization processing by use of a DMA (direct memory access) controller. The method comprises an arithmetic operation processing step and a number move normalization processing step, wherein the number move normalization processing step comprises the following steps of: configuring the value of a normalization deflection bit register in the DMA controller according to a normalization shift factor output by the arithmetic operation processing, and configuring a DMA number move control register; reading the original bit width data from a data source memory by the DMA controller through reading control, and storing the original bit width data into a DMA internal cache register; performing data normalization processing according to the value of the normalization deflection bit register to obtain the deflected bit width data; sending the deflected bit width data to a destination address for storage; if the number move is finished, sending a DMA number move interrupt signal; otherwise, repeating the processes. With the adoption of the method, complete streamline operation can be realized in the whole arithmetic operation processing process, the time overhead is reduced, and the hardware operation processing efficiency can be improved without causing more additional overhead.

Description

technical field [0001] The invention relates to the field of data processing, in particular to a method for normalizing data by using a direct memory access DMA controller. Background technique [0002] For many algorithms, calculations using maximum precision are required to ensure correct results. For example, when using a very small fraction, there may not be enough binary digits in the accumulator to accurately represent it. In this case, it is necessary to use normalization processing to solve this kind of problem. Normalization processing is to scale a number to a known order of magnitude. Normalized data has no significant digits to the left of the decimal point. A normalized data in an accumulator contains a sign bit and data bits of the specified bit width. A normalized data retains its original order of magnitude. After the calculation is performed, the result can be rescaled to the original order of magnitude. . [0003] In the process of hardware arithmetic o...

AI Technical Summary

Problems solved by technology

This solution adds a normalized operation to the reading process, changes the logic or increases the delay of the entire process, resulting in unsatisfactory timing.

It will cause the time sequence from reading data to the next operation (storage or output) to be relatively tight. In order to meet the improvement of the process in the future, this method has great limitations.

Images