Skip to content

stdio2016/useless-cuda

Repository files navigation

useless-cuda

Experiment with CUDA and to test my GPU speed

  • collective-test: Output variable numbers of items from each input item.
  • cuda-related: Try to know something about GPU hardware, but not actually run CUDA kernel.
  • experiment: Well, experiments...
  • fft: Fast Fourier transform
  • gameoflife: Conway's Game of Life simulation
  • haltkernel: Stop GPU kernels from host side to get the intermediate result. Warning! These programs may hang your computer.
  • hamilton: Counting Hamiltonian cycles of a directed graph.
  • lightout: Counting the solutions of n*m Lights Out game. Actually this game can be solved by Gausssian elimination, but I choose to brute force it by GPU.😃
  • memspeed: Test memory speed of various cases.
  • nqueen: Counting solutions to N-Queens puzzle. I have many kinds of implementations: AVX2, ARM64 NEON, CUDA, and Metal.
  • sparse-subgraph: Compute the number of edges of sparsest k-subgraph.
  • persist.cu: I wrote this because one of my lab computer is slow at initializing CUDA. (It took about 0.5 seconds) I found out that CUDA initialization will be faster if some process is already using GPU, so all the program does is initialize CUDA context and then sleep indefinitely.

About

useless cuda programs/benchmarks

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published