加入CODE CHINA

· 不限速    · 不限空间    · 不限人数    · 私仓免费

免费加入
README.md

CUDA_Test

The main role of the project:

  • CUDA 10.0'usage(each test code gives the implementation of C ++ and CUDA, respectively, and gives the calculation time for each method)
  • TensorRT's usage
  • Simd's usage Simd GitHub
  • OpenMP's usage
  • SIMD's usage
  • Assembly Language's usage(MASM, NASM)

CUDA 10.0 test code:

  • simple
    • vector add: C = A + B
    • matrix multiplication: C = A * B
    • dot product
    • Julia
    • ripple
    • green ball
    • ray tracking
    • heat conduction
    • calculate histogram
    • streams' usage
  • layer(approximate)
    • channel normalize(mean/standard deviation)
    • reverse
    • prior_vbox
  • image process
    • bgr to gray
    • bgr to bgr565
    • gray image histogram equalization(only C++ implementation)
    • gray image edge detection: Laplacian(only C++ implementation)

TensorRT 2.1.2 test code:

  • MNIST
  • MNIST API(use api produce network)
  • GoogleNet
  • CharRNN
  • Plugin(add a custom layer)
  • MNIST Infer(serialize TensorRT model)

The project support platform:

  • windows10 64 bits: It can be directly build with VS2013 in windows10 64bits(Except TensorRT).
  • Linux:
    • CUDA supports cmake build(file position: prj/linux_cuda_cmake)
    • TensorRT support cmake build(file position: prj/linux_tensorrt_cmake)
    • Simd_Test support cmake build(file position: prj/linux_simd_cmake)
    • AssemblyLanguage_Test support cmake build(file position: prj/linux_assembly_language_cmake)

Screenshot:

Blog: fengbingchun

项目简介

🚀 Github 镜像仓库 🚀

源项目地址

https://github.com/fengbingchun/cuda_test

发行版本

当前项目没有发行版本

贡献者 2

fengbingchun @fengbingchun1

开发语言

  • C++ 89.5 %
  • C 6.4 %
  • HTML 2.5 %
  • CSS 0.9 %
  • CMake 0.5 %