Skip to content

add sparsediff #202

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 2 commits into from
Jul 21, 2019
Merged

add sparsediff #202

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
60d705f
add sparsediff
huanglangwen Jul 19, 2019
c6373f6
fix splatting, Val type and jac usage
huanglangwen Jul 20, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions Project.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,8 +21,10 @@ RandomNumbers = "e6cf234a-135c-5ec9-84dd-332b85af5143"
RecursiveArrayTools = "731186ca-8d62-57ce-b412-fbd966d074cd"
Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
SparseDiffTools = "47a9eef4-7e08-11e9-0b38-333d64bd3804"

[compat]
SparseDiffTools = ">= 0.3.0"
julia = "1"

[extras]
Expand Down
2 changes: 2 additions & 0 deletions src/StochasticDiffEq.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -47,6 +47,8 @@ module StochasticDiffEq

import DiffEqBase: iip_get_uf, oop_get_uf, build_jac_config

using SparseDiffTools: forwarddiff_color_jacobian!, ForwardColorJacCache


const CompiledFloats = Union{Float32,Float64}

Expand Down
44 changes: 28 additions & 16 deletions src/derivative_wrappers.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,7 +5,7 @@ function derivative(f, x::Union{Number,AbstractArray{<:Number}},
if get_current_alg_autodiff(integrator.alg, integrator.cache)
d = ForwardDiff.derivative(f, x)
else
d = DiffEqDiffTools.finite_difference_gradient(f, x, alg.diff_type, eltype(x), Val{false})
d = DiffEqDiffTools.finite_difference_gradient(f, x, alg.diff_type, eltype(x), Val(false))
end
d
end
Expand All @@ -20,45 +20,57 @@ function derivative!(df::AbstractArray{<:Number}, f, x::Union{Number,AbstractArr
nothing
end

jacobian_autodiff(f,x,_,_)=ForwardDiff.derivative(f,x)
jacobian_autodiff(f,x::AbstractArray,_,hascolor::Val{false})=ForwardDiff.jacobian(f,x)
function jacobian_autodiff(f,x::AbstractArray,integrator,hascolor::Val{true})
colorvec=integrator.f.colorvec
jac=integrator.f.jac_prototype
J=jac isa SparseMatrixCSC ? similar(jac) : fill(0.,size(jac))
forwarddiff_color_jacobian!(J,f,x,color=colorvec,sparsity=jac)
J
end
jacobian_finitediff(f,x,difftype,_,_)=DiffEqDiffTools.finite_difference_derivative(f, x, difftype, eltype(x))
jacobian_finitediff(f,x::AbstractArray,difftype,_,hascolor::Val{false})=DiffEqDiffTools.finite_difference_jacobian(f, x, difftype, eltype(x), Val{false})
jacobian_finitediff(f,x::AbstractArray,difftype,integrator,hascolor::Val{true})=
DiffEqDiffTools.finite_difference_jacobian(f, x, difftype, eltype(x), Val(false),color=integrator.f.colorvec,sparsity=integrator.f.jac_prototype)

function jacobian(f, x,
integrator::DiffEqBase.DEIntegrator)
local J
alg = unwrap_alg(integrator, true)
isarray = typeof(x) <: AbstractArray
if get_current_alg_autodiff(integrator.alg, integrator.cache)
if isarray
J = ForwardDiff.jacobian(f,x)
else
J = ForwardDiff.derivative(f,x)
end
J = jacobian_autodiff(f,x,integrator,Val(DiffEqBase.has_colorvec(integrator.f)))
else
if isarray
J = DiffEqDiffTools.finite_difference_jacobian(f, x, alg.diff_type, eltype(x), Val{false})
else
J = DiffEqDiffTools.finite_difference_derivative(f, x, alg.diff_type, eltype(x))
end
J = jacobian_finitediff(f,x,alg.diff_type,integrator,Val(DiffEqBase.has_colorvec(integrator.f)))
end
J
end

jacobian_autodiff!(J,f,fx,x,jac_config::ForwardColorJacCache)=forwarddiff_color_jacobian!(J,f,x,jac_config)
jacobian_autodiff!(J,f,fx,x,jac_config::ForwardDiff.JacobianConfig)=ForwardDiff.jacobian!(J, f, fx, x, jac_config)

function jacobian!(J::AbstractMatrix{<:Number}, f, x::AbstractArray{<:Number}, fx::AbstractArray{<:Number}, integrator::DEIntegrator, jac_config)
if alg_autodiff(integrator.alg)
ForwardDiff.jacobian!(J, f, fx, x, jac_config)
jacobian_autodiff!(J, f, fx, x, jac_config)
else
DiffEqDiffTools.finite_difference_jacobian!(J, f, x, jac_config)
end
nothing
end

jac_cache_autodiff(alg,f,uf,du1,uprev,u,hascolor::Val{false})=ForwardDiff.JacobianConfig(uf,du1,uprev,ForwardDiff.Chunk{determine_chunksize(u,alg)}())
jac_cache_autodiff(alg,f,uf,du1,uprev,u,hascolor::Val{true})=ForwardColorJacCache(uf,uprev,color=f.colorvec,sparsity=f.jac_prototype)

function DiffEqBase.build_jac_config(alg::StochasticDiffEqAlgorithm,f,uf,du1,uprev,u,tmp,du2)
if !has_jac(f)
if alg_autodiff(alg)
jac_config = ForwardDiff.JacobianConfig(uf,du1,uprev,ForwardDiff.Chunk{determine_chunksize(u,alg)}())
jac_config = jac_cache_autodiff(alg,f,uf,du1,uprev,u,Val(DiffEqBase.has_colorvec(f)))
else
colorvec= f.colorvec isa Nothing ? Base.OneTo(length(u)) : f.colorvec
if alg.diff_type != Val{:complex}
jac_config = DiffEqDiffTools.JacobianCache(tmp,du1,du2,alg.diff_type)
jac_config = DiffEqDiffTools.JacobianCache(tmp,du1,du2,alg.diff_type,color=colorvec,sparsity=f.jac_prototype)
else
jac_config = DiffEqDiffTools.JacobianCache(Complex{eltype(tmp)}.(tmp),Complex{eltype(du1)}.(du1),nothing,alg.diff_type)
jac_config = DiffEqDiffTools.JacobianCache(Complex{eltype(tmp)}.(tmp),Complex{eltype(du1)}.(du1),nothing,alg.diff_type,color=colorvec,sparsity=f.jac_prototype)
end
end
else
Expand Down
60 changes: 60 additions & 0 deletions test/sparsediff_tests.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,60 @@
using Test
using StochasticDiffEq
using SparseArrays
using LinearAlgebra

#https://github.com/JuliaDiffEq/SparseDiffTools.jl/blob/master/test/test_integration.jl
function f(dx,x,p,t)
for i in 2:length(x)-1
dx[i] = x[i-1] - 2x[i] + x[i+1]
end
dx[1] = -2x[1] + x[2]
dx[end] = x[end-1] - 2x[end]
nothing
end

function g(dx,x,p,t)
dx .= 0
nothing
end

function second_derivative_stencil(N)
A = zeros(N,N)
for i in 1:N, j in 1:N
(j-i==-1 || j-i==1) && (A[i,j]=1)
j-i==0 && (A[i,j]=-2)
end
A
end

function generate_sparsity_pattern(N::Integer)
dl = repeat([1.0],N-1)
du = repeat([1.0],N-1)
d = repeat([-2.0],N)
return Tridiagonal(dl,d,du)
end

jac_sp = sparse(generate_sparsity_pattern(10))
#jac = second_derivative_stencil(10)
colors = repeat(1:3,10)[1:10]
u0=[1.,2.,3,4,5,5,4,3,2,1]
tspan=(0.,10.)
sdefun_sp= SDEFunction(f,g,colorvec=colors,jac_prototype=jac_sp)
prob_sp = SDEProblem(sdefun_sp,g,u0,tspan)
prob_std = SDEProblem(f,g,u0,tspan)

sol_sp=solve(prob_sp,SKenCarp(autodiff=false))
@test sol_sp.retcode==:Success#test sparse finitediff
sol=solve(prob_std,SKenCarp(autodiff=false))
@test sol_sp.u[end]≈sol.u[end] atol=1e-4
@test length(sol_sp.t)==length(sol.t)

sol_sp=solve(prob_sp,SKenCarp())
sol=solve(prob_std,SKenCarp())
@test sol_sp.u[end]≈sol.u[end] atol=1e-4
@test length(sol_sp.t)==length(sol.t)

#sol_sp=solve(prob_sp,SKenCarp(autodiff=false),abstol=1e-10,reltol=1e-10)
#sol=solve(prob_std,SKenCarp(autodiff=false),abstol=1e-10,reltol=1e-10)
#@test sol_sp.u[end]≈sol.u[end]
#@test length(sol_sp.t)==length(sol.t)