SciML / DiffEqGPU.jl / 272 / 1

0%

master: 0%

push

travis-ci-com

<a href="https://github.com/SciML/DiffEqGPU.jl/commit/<a class=hub.com/SciML/DiffEqGPU.jl/commit/d1c69f0da8c86835f04aeeca3227964e1cc6ba5b">d1c69f0da<a href="https://github.com/SciML/DiffEqGPU.jl/commit/d1c69f0da8c86835f04aeeca3227964e1cc6ba5b">&quot;&gt;alternative norm

Oh thanks. That narrowed it down, and:

```julia
using OrdinaryDiffEq, DiffEqGPU, ForwardDiff, Test, Unitful

function lorenz(du,u,p,t)
 @inbounds begin
     du[1] = p[1]*(u[2]-u[1])
     du[2] = u[1]*(p[2]-u[3]) - u[2]
     du[3] = u[1]*u[2] - p[3]*u[3]
 end
 nothing
end

u0 = [ForwardDiff.Dual(1f0,(1.0,0.0,0.0));ForwardDiff.Dual(0f0,(0.0,1.0,0.0));ForwardDiff.Dual(0f0,(0.0,0.0,1.0))]
tspan = (0.0f0,100.0f0)
p = (10.0f0,28.0f0,8/3f0)
prob = ODEProblem(lorenz,u0,tspan,p)
prob_func = (prob,i,repeat) -&amp;gt; remake(prob,p=rand(Float32,3).*p)
monteprob = EnsembleProblem(prob, prob_func = prob_func)

@inline ODE_DEFAULT_NORM(u::AbstractArray{&amp;lt;:ForwardDiff.Dual},::Any) = sqrt(sum(DiffEqBase.UNITLESS_ABS2∘DiffEqBase.value,u) / length(u))
@inline ODE_DEFAULT_NORM(u::ForwardDiff.Dual,::Any) = abs(DiffEqBase.value(u))

@time sol = solve(monteprob,Tsit5(),EnsembleGPUArray(),trajectories=10_000,saveat=1.0f0,internalnorm=ODE_DEFAULT_NORM)
```

That fixed it. So the regression seems to be due to https://github.com/SciML/DiffEqBase.jl/commit/</a><a class="double-link" href="https://github.com/SciML/DiffEqGPU.jl/commit/<a class="double-link" href="https://github.com/SciML/DiffEqGPU.jl/commit/68daccd1b23563568699049488e9ba708efb96d8">68daccd1b</a>">68daccd1b</a><a href="https://github.com/SciML/DiffEqGPU.jl/commit/d1c69f0da8c86835f04aeeca3227964e1cc6ba5b">#diff-54090e4bd9da3cc6b0fbb839b246aeb3

0 of 220 relevant lines covered (0.0%)

0.0 hits per line