An accelerated NTT to compute the product of two polynomials, based CUDA and multi-threaded CPU.
The project uses CUDA by default, additional options are as follows:
-normal: Force single-threaded NTT on the CPU.-cpu: Force multi-threaded NTT on the CPU instead of GPU.-cuda: Force execute NTT on the GPU.-test: Enable python tests, which are enabled by default when BUILD TYPE is Debug.-ntest: Disenable python tests, which are disenabled by default when BUILD TYPE is Release.-n <number>: Degree(log) of the first polynomial.-m <number>: Degree(log) of the second polynomial.-iter <number>: The number of iterations to run the program.
The project has the following build dependencies:
- NVIDIA CUDA 11.5 Toolkit (or higher) for CUDA.
- OpenMP for CPU multi-threading.
- python 3.6 (or higher) for optional test.
git clone https://github.com/KMushaL/ParallelNTT.git
mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_CUDA_ARCHITECTURES="your_cuda_compute_capability"
make -j number_of_cores