You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I ran into this error when using OrdinaryDiffEq.jl with julia --compiled-modules=no:
┌ Warning: Error requiring `ForwardDiff` from `LoopVectorization`
│ exception =
│ LoadError: UndefVarError: Dual not defined
│ in expression starting at ~/.julia/packages/LoopVectorization/09Hhf/src/simdfunctionals/vmap_grad.jl:6
└ @ Requires ~/.julia/packages/Requires/7Ncym/src/require.jl:49
The most minimal case I managed to narrow it down to is having a new project with only ForwardDiff and LoopVectorization as dependencies, then starting Julia with --compiled-modules=no and first importing LoopVectorization, then ForwardDiff. When compiled modules are activated or when ForwardDiff is imported first, the error does not appear.
The text was updated successfully, but these errors were encountered:
My use case for --compiled-modules=no is that the Julia precompilation mechanism doesn’t interact well with MPI, so if you’re running an MPI program you can get errors if multiple instances of Julia are trying to precompile a package simultaneously. Running without compiled modules avoids this issue and is recommended in the MPI.jl documentation.
I ran into this error when using OrdinaryDiffEq.jl with
julia --compiled-modules=no
:The most minimal case I managed to narrow it down to is having a new project with only
ForwardDiff
andLoopVectorization
as dependencies, then starting Julia with--compiled-modules=no
and first importingLoopVectorization
, thenForwardDiff
. When compiled modules are activated or whenForwardDiff
is imported first, the error does not appear.The text was updated successfully, but these errors were encountered: