GPUArrays.jl
GPUArrays.jl copied to clipboard
max pool
An implementation of maxpool. Here's a sample benchmarking (CPU v/s GPU): https://gist.github.com/americast/95358d972647adf5c7ebcde7c58db51f
Tests were failing due to getindex is disabled error. I have made a small change in src/indexing.jl as a workaround.
Thanks.
Lol, the assertslow is there for a reason, since scalar indexing into a gpuarray is very slow and should not be done!
You can do allowslow(true) to disable that error. Since your benchmarks seem to indicate good speed, I'm guessing the problem happens on the other side of the test code. I'll check.
Ouch! The benchmarks don't include padding. I should have mentioned that. Sorry, my bad. I'll create a separate benchmark with padding.
Updated. Thank you @SimonDanisch for PR #111 and commit 1e1104e. I've updated the benchmarks at https://gist.github.com/americast/95358d972647adf5c7ebcde7c58db51f.