Optimized cuda FIR filter implementation for complex sample data. Generic CPU implementation for reference. Python code for generating and testing vectors.

You will need a working install of the Cuda Development Enviroment with the binary directory in your path.

1. Follow post install steps on nvidia's website (if this is the first time using cuda): Post install instructions

2. Checkout the code

git clone https://github.com/tylerjw/cuda_fir.git

3. Create build directory

cd cuda_fir
mkdir build
cd build

4. Run cmake

cmake ../

5. Build

make

In the scripts directory there are a few python scripts. You will need python3 with scipy and numpy to run them.

• generateExampleData.py - creates noise.cap and filtered.cap and plots fft of them into png files
• samples.py - plots fft of cpuout.cap and cudaout.cap in png files
• testout.py - compares cpuout.cap and cudaout.cap

1. Use scripts/generateExampleData.py to generate sample data.
2. Run the code to generate outputs.
3. Use scripts/samples.py and scripts/testout.py to compare.

python3 scripts/generateExampleData.py
build/bin/main noise.cap cpuout.cap cudaout.cap
python3 scripts/samples.py
python3 scripts/testout.py