repository·master·Indexed 27 days ago
https://github.com/nvidia/cuda-samplesA collection of demonstration programs for CUDA developers showcasing features of the CUDA Toolkit. The samples serve as a practical reference for learning and testing CUDA capabilities across Linux, Windows, and Tegra/Automotive Linux, covering topics such as asynchronous API usage, performance profiling with clock functions, multi-GPU applications via OpenMP, FP16 scalar products, and various matrix multiplication implementations using the CUDA Runtime API, Driver API, and libNVRTC.
SobolQRNG sample implements a Sobol Quasirandom Sequence Generator, which is primarily used in applications such as Computational Finance.BlackScholes_nvrtc sample implements Black-Scholes option pricing for European options. It demonstrates how to evaluate fair call and put prices by compiling CUDA kernels at runtime using the libNVRTC library.segmentationTreeThrust sample demonstrates an approach to constructing image segmentation trees. The implementation is based on Boruvka's Minimum Spanning Tree (MST) algorithm and utilizes data-parallel algorithms and performance strategies.immaTensorCoreGemm sample demonstrates integer General Matrix Multiply (GEMM) computations using the Warp Matrix Multiply and Accumulate (WMMA) API for integers, introduced in CUDA 10. It utilizes Tensor Cores (available from the Volta architecture onwards) to accelerate matrix operations. Additionally, it demonstrates how to use the cudaFuncAttributeMaxDynamicSharedMemorySize function attribute to reserve an extended amount of shared memory beyond the default limits.boxFilter sample implements a fast image box filter using CUDA. It demonstrates image processing techniques combined with OpenGL rendering for visualization.streamOrderedAllocationIPC sample demonstrates how to use Inter-Process Communication (IPC) pools with stream-ordered memory allocation. It specifically showcases the use of cudaMallocAsync and the cudaMemPool family of APIs to manage memory pools that can be shared across processes.quasirandomGenerator_nvrtc sample implements a Niederreiter Quasirandom Sequence Generator and Inverse Cumulative Normal Distribution functions to generate Standard Normal Distributions. It utilizes NVRTC (NVIDIA Runtime Compilation) to compile CUDA kernels at runtime.binomialOptions sample is a computational finance application that evaluates the fair call price for a given set of European options using a binomial model.quasirandomGenerator sample implements a Niederreiter Quasirandom Sequence Generator and Inverse Cumulative Normal Distribution functions. These are used for the generation of Standard Normal Distributions, primarily within the context of Computational Finance.