GPUArrays.jl
GPUArrays.jl copied to clipboard
JuliaGPU
Reusable array functionality for Julia's various GPU backends.
See [here](https://github.com/JuliaImages/ImageFiltering.jl/issues/52) for a description of the operations involved. I expect that pooling will be relatively easy, convolutions are also straightforward enough in principle but hard to make really fast....
Currently `CuArray(zeros(3,3)) .+ CuArray(ones(3,3)) .* ones(3)` fails.
Since we're overriding Base all the time (and similarly packages like CuArrays override generic functionality from GPUArrays) we need some test infrastructure to make sure certain specialized methods are executed...
This is a fun one, not sure about the solution but it hurts usability. Take the following MWE: ```julia using CuArrays import CUDAnative using ForwardDiff @inline function CUDAnative.erf(input::ForwardDiff.Dual{T}) where T...
Aside from `adapt`, this is probably the biggest blocker to running Flux with CLArrays right now.
This is with CuArrays own indexing disabled. Strangely plain arrays seem to work, but not cuarrays. ``` julia> cu([1:5;])[[5,2,3]] 3-element CuArray{Float32,1}: 5.0 2.0 3.0 julia> cu([1:5;])[cu[5,2,3]] ERROR: scalar getindex is...
I got the following error when I tried simple transposing a gpuarray. I was using the most recent GPUArray and CUDAnative, Julia 0.6 and mac os 10.12.6. julia> b =...
Hello, dense CuArrays are recognized as AbstractGPUArray type. Indeed, if I do ```julia using CUDA using GPUArrays A = CUDA.rand(10, 10) A isa AbstractGPUArray ``` it returns true. However, it...
It looks like this is falling to the standard `generic_matmatmul!`, which errors on scalar indexing ```julia julia> using LinearAlgebra, Metal, StaticArrays julia> Metal.GPUArrays.allowscalar(false) julia> A = one(SMatrix{3,3}); B = MtlArray(ones(Float32,...
