search

naibaf7 / caffe

Caffe: a fast open framework for deep learning. With OpenCL and CUDA support.
http://caffe.berkeleyvision.org/
Other
86 stars 20 forks source link

Speed up convolution layers with batch-GEMM #10

Closed naibaf7 closed 8 years ago

naibaf7 commented 9 years ago

See the discussion at: https://github.com/clMathLibraries/clBLAS/issues/129#issuecomment-133151567

Alternative solutions using ViennaCL could also be evaluated.

naibaf7 commented 8 years ago

Will be solved differently with Caffe-libdnn (OpenCL).