marcv81/mandelbrot

Mandelbrot

This project renders the Mandelbrot fractal to experiment with parallelism.

Compile

Compile for CPU using standard library threads.

g++ -std=c++11 -fPIC -o libmandelbrot.so --shared mandelbrot-threads.cpp

Compile for GPU using CUDA.

nvcc --compiler-options '-fPIC' -o libmandelbrot.so --shared mandelbrot.cu

Compile for CPU using OpenACC.

pgc++ -std=c++11 -fPIC -o libmandelbrot.so -shared -ta=multicore mandelbrot-openacc.cpp

Compile for GPU using OpenACC.

pgc++ -std=c++11 -fPIC -o libmandelbrot.so -shared -ta=nvidia mandelbrot-openacc.cpp

Run

Benchmark rendering and save the image to JPEG.

python3 benchmark.py

Explore the fractal interactively.

python3 explore.py

Benchmarks

Best render time of 5 attempts at different resolutions. CPU: Intel i7-8700, GPU: NVIDIA GTX 1060.

	CPU (threads)	GPU (CUDA)	CPU (OpenACC)	GPU (OpenACC)
1080p	25.7 ms	6.2 ms	24.5 ms	5.8 ms
4k	103.4 ms	22.0 ms	96.6 ms	22.6 ms
16k	1.614 s	269.7 ms	1.519 s	287.9 ms