Skip to content

victorygogogo/haDNN

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

57 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

haDNN

A proof-of-concept CNN implementation in Halide.

Convolution layers were implemented in both direct-conv and FFT. Single thread benchmark compared to Caffe (GEMM):

conv

fft

Conv is comparable to Caffe GEMM when k=3. FFT is significantly faster when k >= 5.

In multithread settings Halide will also scale better than Caffe.

About

Proof-of-Concept CNN in Halide

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 81.7%
  • Python 11.1%
  • Lua 3.5%
  • Shell 2.4%
  • Makefile 1.3%