DifferentialEquations.jl多线程

Question

下面的代码是求解一个具有不同变量的ODE，PA对应于不同的Δ，这是我在每个循环之后唯一感兴趣的数据。由于循环是独立的，我假设我可以多线程这些循环，所以我在for循环前面添加了for：

#example
using DifferentialEquations
using Plots;
gr(grid = false, ms = 3, legend = false, msw = 1.5 ,mc = :white)
using Findpeaks

#parameters
P = 1.5
κ = 0.5
Γ = 0.0005

#ODE settings
fs = 400
T = 2*π
y0 = [1, 1]
t0 = [0, 100 *T]
t1 = [0, 50 *T]
function route(dx, x, p, t)
    dx[1] =  1im * (Δ * x[1] - 2 * real(x[2]) * x[1] - 0.5) - κ * x[1]
    dx[2] = -1im * (0.5 * P * abs(x[1])^2 + x[2]) - Γ * x[2]
end
Plots.scatter(1,1) #create new figure

Threads.@threads for i =  -1.3: 0.1 :-0.4
    global Δ = i
    global p =[P, Δ, κ, Γ]
    global prob = ODEProblem(route, collect(Complex{Float64}, y0), t0, p)
    global y1 = solve(prob)
    global y10 = last(y1);#omit unstable parts

    global prob2 = ODEProblem(route, collect(Complex{Float64}, y10), t1, p)
    global y2 = solve(prob2) #calculate steady state data

    global abss = abs.(y2)
    global a = abss[1,:].^2 

    global PA = begin
            global peaka = findpeaks(real(a), y2.t)
            global peaka = real(a)[peaka] #peak values
            global pa = Vector{Float64}(undef, length(peaka))
            pa[:] .= Δ #peak locations corresponding to Δ
            pa .+ 1im*peaka
    end
    global fig = scatter!(real(PA), imag(PA))
    display(fig)
end

但这将导致julia终端崩溃(使用vscode)，似乎与GR有关。

​

​

这是怎么回事？

Answer 1

看起来像是从GR扔出来的东西。我不确定GR是否安全，我会再次检查它，而不是微分方程解。