kinsol interface for GMRES missing

[jguterl]

The MWE

using NonlinearSolve
using Sundials
p = [2.0,3.0]
f(u,p) = u .* u .- p
u0 = [1.0, 1.0]

probN = NonlinearProblem(f, u0, p)
s = KINSOL(;linear_solver=:GMRES)
solver = solve(probN, s, reltol = 1e-7)

fails
with

UndefVarError: LS not defined
Stacktrace:
 [1] kinsol(f::Function, y0::Vector{Float64}; userdata::Nothing, linear_solver::Symbol, jac_upper::Int64, jac_lower::Int64)
   @ Sundials ~/.julia/packages/Sundials/4gP3E/src/simple.jl:15
 [2] __solve(prob::NonlinearProblem{Vector{Float64}, false, Vector{Float64}, NonlinearFunction{false, SciMLBase.FullSpecialize, typeof(f), LinearAlgebra.UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED_NO_TIME), Nothing, Nothing}, Base.Pairs{Symbol, Union{}, Tuple{}, NamedTuple{(), Tuple{}}}, SciMLBase.StandardNonlinearProblem}, alg::KINSOL{:GMRES}, timeseries::Vector{Any}, ts::Vector{Any}, ks::Vector{Any}, recompile::Type{Val{true}}; kwargs::Base.Pairs{Symbol, Float64, Tuple{Symbol}, NamedTuple{(:reltol,), Tuple{Float64}}})
   @ Sundials ~/.julia/packages/Sundials/4gP3E/src/common_interface/solve.jl:80

[jguterl]

I believe the issue is the lack of interface in the method kinsol

[ChrisRackauckas]

Yes that case is just missing the handling

https://github.com/SciML/Sundials.jl/blob/master/src/simple.jl#L70-L76

It's not hard to read the Sundials docs and add the right call into there though. Someone just needs to do it.

[jguterl]

Ok. I will look into it next week. Is there a set of benchmarks for the preconditioners?

…

On Fri, May 26, 2023 at 2:37 PM Christopher Rackauckas < ***@***.***> wrote: Yes that case is just missing the handling https://github.com/SciML/Sundials.jl/blob/master/src/simple.jl#L70-L76 It's not hard to read the Sundials docs and add the right call into there though. Someone just needs to do it. — Reply to this email directly, view it on GitHub <#405 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AEESMZHGORTCKJTMVVU7OZTXIEO7NANCNFSM6AAAAAAYPDYQ5E> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[ChrisRackauckas]

There's no benchmarks specifically on the nonlinear solvers yet. We need to set that up in the SciMLBenchmarks.

[ViralBShah]

This works now:

julia> s = KINSOL(;linear_solver=:GMRES)
KINSOL{:GMRES}(0, 0, nothing, 0, 0, :None, 0)

julia> solver = solve(probN, s, reltol = 1e-7)
retcode: Success
u: 2-element Vector{Float64}:
 1.4142135623730954
 1.7320508075688774