/x86-moving-average-filter

A signal-filtering function implemented in x86 Assembly. It deals with signals consisting of 32-Bit Integers.

Primary LanguageC

x86 Moving Average Filter

This is a filter for cleaning up noisy signals using a moving-average filter with free weighting. The main part of the filtern is implemented in x86 Assembler with a framework programm written in C. By using the rotation feature of the FPU Stack, this filter function saves on memory accesses and outperforms the standard c-library functions when calculating this specific filtering formula.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

At this point in time, the project only runs on 32-Bit Linux systems (tested on Ubuntu).

You will need the commandline tools nasm, make and gcc (also, gcc-multilib seems to solve some compatibility issues). Use the following command to install them:

sudo apt install nasm make gcc gcc-multilib

Building

After cloning the repository, simply run the following commands:

cd x86-moving-average-filter
make

Usage

The filter uses the following formula to adjust the :

gk = (alpha⋅gk-1+beta⋅fk)/(alpha+beta) with g0 = s (starting value)

(f refers to the unfiltered value, while g is an already filtered value)

To use the filtering function, you first need to save your source signal in a textfile as integers seperated by newline characters. After that, you can call the function with your specified parameters:

./movafilter [float alpha] [float beta] [int starting value] [source filepath] [output filepath]

Running the tests

There are built-in test cases to cross validate the asm routine with a c-function. Use the following command to run the tests:

movafilter runtest

Note: Small differences between the filtered values and the validation are not necessarily a sign of the filter malfunctioning. They are probably caused by differences in the handling of floating point between the c standard library and the assembler routine (especially big floating-point values seem to be affected here). It is best to look at the actual output of the validation function when deciding if the asm filter is working correctly.

Authors