CUDAapi.jl: DEPRECATED, use CUDA.jl instead!
Reusable components for CUDA development.
This package provides some reusable functionality for working with CUDA or NVIDIA APIs. It is intended for package developers, and does not provide concrete application functionality.
To check if a CUDA GPU is available, CUDAapi provides and exports the
has_cuda_gpu functions. These functions are useful to query whether you
will be able to import a package that requires CUDA to be available, such as
CuArrays.jl (CUDAapi.jl itself will always import without an error):
using CUDAapi # this will NEVER fail if has_cuda() try using CuArrays # we have CUDA, so this should not fail catch ex # something is wrong with the user's set-up (or there's a bug in CuArrays) @warn "CUDA is installed, but CuArrays.jl fails to load" exception=(ex,catch_backtrace()) end end
src/discovery.jl defines helper methods for discovering the NVIDIA
driver and CUDA toolkit, as well as some more generic methods to find libraries
and binaries relative to, e.g., the location of the driver or toolkit.
src/compatibility.jl contains hard-coded databases with software and hardware
compatibility information that cannot be queried from APIs.
CUDA version update
When a new version of CUDA is released, CUDAapi.jl needs to be updated accordingly:
discovery.jl: update the
compatibility.jl: update each
_dbvariable (refer to the comments for more info)
travis.osx: provide a link to the installers
appveyor.ps1: provide a link to the installer, and list the components that need to be installed
appveyor.yml: add the version to the CI rosters
GCC version update
gcc_minor_versions ranges in
discovery.jl to cover the new version.