Skip to content

A CUDA tutorial to make people learn CUDA program from 0

Notifications You must be signed in to change notification settings

RussWong/CUDATutorial

Repository files navigation

CUDATutorial

A CUDA tutorial to make people learn CUDA program from 0

test enviroment

Turing T4 GPU

compile command

  1. compile by hand

nvcc xxx.cu -o xxx

if that does not work, pls try:

nvcc xxx.cu --gpu-architecture=compute_yy -o xxx

xxx is file name, yy is GPU compute capability, ep.A100's compute capability is 86.

  1. one-click compile and run

please ensure:

1.cmake version >= 3.8

2.you have CUDA TOOLKIT installed in system root directory, downloaded link is https://developer.nvidia.com/cuda-downloads.

 mkdir build 
 cd build 
 cmake .. && make -j8 
 cd bin 
 ./xxx

remark

  • related performance data is attached at the top of code file.
  • the performance data is diverse and diverse on different GPU platforms and NVCC compiler, so some counter-intuitive result is normal, we should only explore and debug the result.
  • welcome all comments and pull requests.

update notes

v2.0

  • add cuda stream
  • add quantize

v2.1

  • add fp32/fp16 gemv(vec * mat,mat is col major)

v2.2

  • add fp32/fp16 gemv(vec * mat,mat is row major)
  • add some code explaination(WIP)

v2.6

  • add fp32 dropout

About

A CUDA tutorial to make people learn CUDA program from 0

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published