The aim of this project is to measure the speedup obtained for the algorithm K Nearest Neighbors when implemented in OpenMP/CUDA with respect to the C++ sequential version.
- Intel Core i7-8750H @ 2.20Ghz (up to 4.10Ghz with Turbo Boost)
- 16 GiB RAM
- NVIDIA GeForce GTX 1050 Ti (Notebook)
- CUDA 10.0
Sift and siftsmall datasets can be found in http://corpus-texmex.irisa.fr, where they are called ANN_SIFT1M and ANN_SIFT10K respectively.
wget ftp://ftp.irisa.fr/local/texmex/corpus/sift.tar.gz
mkdir data
tar -xf sift.tar.gz -C data
git clone https://github.com/fedem96/Nearest_Neighbors.git
mkdir experiments
cd Nearest_Neighbors
nvcc -O3 src/main.cu --gpu-architecture=compute_61 -Xcompiler -fopenmp -o run
./run -d ../data -e ../experiments