AssertionError: ctx === CuCurrentContext()

Specific Domains GPU
1

I am trying to setup Julia to use my GPU, out of curiosity as I am a new user.

After a few problems, I encoutered this one, where I have not found anything (apart from this which didn’t help me either):

using BenchmarkTools
using FFTW
using CuArrays
using Random

const nx = 1024  # do 1024 x 1024 2D FFT
xc = CuArray{ComplexF64}(CuArrays.randn(Float64, nx, nx))

AssertionError: ctx === CuCurrentContext()

Stacktrace:
 [1] context() at /home/benoit/.julia/packages/CUDAnative/Phjco/src/init.jl:61
 [2] generator() at /home/benoit/.julia/packages/CuArrays/rNxse/src/rand/CURAND.jl:33
 [3] #randn#113(::Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tuple{}}}, ::Function, ::Type{Float64}, ::Int32, ::Int32) at /home/benoit/.julia/packages/CuArrays/rNxse/src/rand/random.jl:176
 [4] randn(::Type{Float64}, ::Int32, ::Int32) at /home/benoit/.julia/packages/CuArrays/rNxse/src/rand/random.jl:176
 [5] top-level scope at In[8]:7
2

Interesting; is it always reproducible? Was this in a fresh REPL session? Does it happen if you just call CUDAnative.context()? Could you try the following (and paste its output):

julia> using CUDAdrv

julia> CuCurrentContext()

julia> using CUDAnative

julia> CUDAnative.initialize()

julia> CuCurrentContext()
CuContext(Ptr{Nothing} @0x0000556265bd5720)

julia> CUDAnative.thread_contexts
1-element Array{Union{Nothing, CuContext},1}:
 CuContext(Ptr{Nothing} @0x0000556265bd5720)

julia> CUDAnative.context()
CuContext(Ptr{Nothing} @0x0000556265bd5720)
3

I had this problem when running the code in a fresh Jupyter Notebook, and it was reproducible. Now that I wrote it in the REPL

julia> using BenchmarkTools

julia> using FFTW

julia> using CuArrays

[ Info: CUDAnative.jl failed to initialized, GPU functionality unavailable (set JULIA_CUDA_SILENT or JULIA_CUDA_VERBOSE to silence or expand this message)

julia> using Random

julia> const nx = 1024  # do 1024 x 1024 2D FFT
1024

julia> xc = CuArray{ComplexF64}(CuArrays.randn(Float64, nx, nx))
ERROR: AssertionError: ctx === CuCurrentContext()
Stacktrace:
 [1] context() at /home/benoit/.julia/packages/CUDAnative/Phjco/src/init.jl:61
 [2] generator() at /home/benoit/.julia/packages/CuArrays/rNxse/src/rand/CURAND.jl:33
 [3] #randn#113(::Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tuple{}}}, ::Function, ::Type{Float64}, ::Int32, ::Int32) at /home/benoit/.julia/packages/CuArrays/rNxse/src/rand/random.jl:176
 [4] randn(::Type{Float64}, ::Int32, ::Int32) at /home/benoit/.julia/packages/CuArrays/rNxse/src/rand/random.jl:176
 [5] top-level scope at none:0

As for your commands,

julia> using CUDAdrv

julia> CuCurrentContext()

julia> 

julia> using CUDAnative

julia> CUDAnative.initialize()
ERROR: CUDA error: invalid device context (code 201, ERROR_INVALID_CONTEXT)
Stacktrace:
 [1] throw_api_error(::CUDAdrv.cudaError_enum) at /home/benoit/.julia/packages/CUDAdrv/aBgcd/src/error.jl:136
 [2] macro expansion at /home/benoit/.julia/packages/CUDAdrv/aBgcd/src/error.jl:149 [inlined]
 [3] cuCtxGetDevice at /home/benoit/.julia/packages/CUDAdrv/aBgcd/src/libcuda.jl:144 [inlined]
 [4] device at /home/benoit/.julia/packages/CUDAdrv/aBgcd/src/context.jl:145 [inlined]
 [5] device!(::CuDevice) at /home/benoit/.julia/packages/CUDAnative/Phjco/src/init.jl:114
 [6] initialize() at /home/benoit/.julia/packages/CUDAnative/Phjco/src/init.jl:32
 [7] top-level scope at none:0

julia> CuCurrentContext()

julia> CUDAnative.thread_contexts
0-element Array{Union{Nothing, CuContext},1}

julia> CUDAnative.context()
ERROR: AssertionError: ctx === CuCurrentContext()
Stacktrace:
 [1] context() at /home/benoit/.julia/packages/CUDAnative/Phjco/src/init.jl:61
 [2] top-level scope at none:0
4

Well, that message says it all: the package failed to initialize, so isn’t functional. Follow the advise of the message, run with JULIA_CUDA_VERBOSE=true to see additional details.