Open
Description
The system is a bit unusual (has a large number of similar variables). I am also not sure how interesting cases like this is, and how feasible they are to deal with. However, since the performance difference is large, I figured it could be interesting to know.
# Fetch packages.
using Catalyst, OrdinaryDiffEqDefault, OrdinaryDiffEqRosenbrock, FiniteStateProjection, BenchmarkTools
# Create the model.
rs = @reaction_network begin
(p,d), 0 <--> X
(kB,kD), 2X <--> X2
end
# Create the ODEProblem.
ps = [:p => 1.0, :d => 0.2, :kB => 2.0, :kD => 5.0]
u0 = zeros(25,25)
u0[1,1] = 1.0
fsp_sys = FSPSystem(rs, [:X, :X2])
oprob = ODEProblem(fsp_sys, u0, 50.0, ps)
# Simulate using default solver and Rodas5P.
@btime sol1 = solve($oprob, Rodas5P()) # 126.148 ms (3719 allocations: 6.96 MiB)
@btime sol2 = solve($oprob) # 2.689 s (309050 allocations: 138.95 MiB)Here, the state is a 25x25 matrix (so 625 variables in total).
Metadata
Metadata
Assignees
Labels
No labels