Change to LinearSolve.jl

[ChrisRackauckas]

Currently hits a weird error:

[ Info: Precompiling OrdinaryDiffEq [1dea7af3-3e70-54e6-95c3-0bf5283fa5ed]
ERROR: LoadError: syntax: SSAValue objects should not occur in an AST
Stacktrace:
  [1] top-level scope
    @ C:\Users\accou\.julia\dev\OrdinaryDiffEq\src\perform_step\rosenbrock_perform_step.jl:27
  [2] include(mod::Module, _path::String)
    @ Base .\Base.jl:418
  [3] include(x::String)
    @ OrdinaryDiffEq C:\Users\accou\.julia\dev\OrdinaryDiffEq\src\OrdinaryDiffEq.jl:4
  [4] top-level scope
    @ C:\Users\accou\.julia\dev\OrdinaryDiffEq\src\OrdinaryDiffEq.jl:148
  [5] include
    @ .\Base.jl:418 [inlined]
  [6] include_package_for_output(pkg::Base.PkgId, input::String, depot_path::Vector{String}, dl_load_path::Vector{String}, load_path::Vector{String}, concrete_deps::Vector{Pair{Base.PkgId, UInt64}}, source::String)
    @ Base .\loading.jl:1318
  [7] top-level scope
    @ none:1
  [8] eval
    @ .\boot.jl:373 [inlined]
  [9] eval(x::Expr)
    @ Base.MainInclude .\client.jl:453
 [10] top-level scope
    @ none:1
in expression starting at C:\Users\accou\.julia\dev\OrdinaryDiffEq\src\perform_step\rosenbrock_perform_step.jl:27
in expression starting at C:\Users\accou\.julia\dev\OrdinaryDiffEq\src\OrdinaryDiffEq.jl:1

@YingboMa do you know what could cause that?

[ChrisRackauckas]

Test case is:

using OrdinaryDiffEq, StaticArrays, BenchmarkTools
function rober(du,u,p,t)
    y₁,y₂,y₃ = u
    k₁,k₂,k₃ = p
    @inbounds begin
    du[1] = -k₁*y₁+k₃*y₂*y₃
    du[2] =  k₁*y₁-k₂*y₂^2-k₃*y₂*y₃
    du[3] =  k₂*y₂^2
    end
    nothing
end
prob = ODEProblem(rober,[1.0,0.0,0.0],(0.0,1e5),[0.04,3e7,1e4])
@btime solve(prob,Rosenbrock23()) # 761.578 μs (633 allocations: 53.05 KiB)

Only the Rosenbrock and Extrapolation methods will be tedious, the rest should already be handled by the Newton part.

[ChrisRackauckas]

Need to not forget to undo this 😅

[ChrisRackauckas]

need to do this to the whole file, and the whole extrapolation caches file

[vpuri3]

sounds cumbersome. vim macros are might speed it up

[ChrisRackauckas]

This adds one copyto! operation, but it shouldn't be bad.

[YingboMa]

It's the first time I see this error message. Does it still occur?

[ChrisRackauckas]

Tests are passing, so now I'm just going to refactor a bit to condense the code before merging 🎉

[ChrisRackauckas]

Benchmarks look to be comfortably a wash even though it allocates more.

using OrdinaryDiffEq, SnoopCompile, ForwardDiff

lorenz = (du,u,p,t) -> begin
        du[1] = 10.0(u[2]-u[1])
        du[2] = u[1]*(28.0-u[3]) - u[2]
        du[3] = u[1]*u[2] - (8/3)*u[3]
end

u0 = [1.0;0.0;0.0]; tspan = (0.0,100.0);
prob = ODEProblem(lorenz,u0,tspan); alg = Rodas5();
tinf = @snoopi_deep ForwardDiff.gradient(u0 -> sum(solve(ODEProblem(lorenz,u0,tspan),alg)), u0)
tinf = @snoopi_deep ForwardDiff.gradient(u0 -> sum(solve(ODEProblem(lorenz,u0,tspan),alg)), u0)

# Before
InferenceTimingNode: 1.219960/3.583559 on Core.Compiler.Timings.ROOT() with 34 direct children
InferenceTimingNode: 1.064226/1.756057 on Core.Compiler.Timings.ROOT() with 13 direct children

# After
InferenceTimingNode: 1.257616/3.774263 on Core.Compiler.Timings.ROOT() with 33 direct children
InferenceTimingNode: 1.006660/1.756162 on Core.Compiler.Timings.ROOT() with 13 direct children

using OrdinaryDiffEq, SnoopCompile, ForwardDiff, BenchmarkTools

lorenz = (du,u,p,t) -> begin
        du[1] = 10.0(u[2]-u[1])
        du[2] = u[1]*(28.0-u[3]) - u[2]
        du[3] = u[1]*u[2] - (8/3)*u[3]
end

u0 = [1.0;0.0;0.0]; tspan = (0.0,100.0);
prob = ODEProblem(lorenz,u0,tspan); alg = Rodas5();
tinf = @snoopi_deep solve(ODEProblem(lorenz,u0,tspan),alg)
tinf = @snoopi_deep solve(ODEProblem(lorenz,u0,tspan),alg)
@btime solve(ODEProblem(lorenz,u0,tspan),alg)

@btime solve(ODEProblem(lorenz,u0,(0.0,10.0)),alg)

# Before
InferenceTimingNode: 1.060946/2.763417 on Core.Compiler.Timings.ROOT() with 8 direct children
InferenceTimingNode: 0.002333/0.002333 on Core.Compiler.Timings.ROOT() with 0 direct children
2.055 ms (14069 allocations: 781.14 KiB)
175.000 μs (1341 allocations: 75.97 KiB)

# After
InferenceTimingNode: 1.129433/3.072010 on Core.Compiler.Timings.ROOT() with 7 direct children
InferenceTimingNode: 0.002182/0.002182 on Core.Compiler.Timings.ROOT() with 0 direct children
2.221 ms (22155 allocations: 2.24 MiB)
162.900 μs (1917 allocations: 184.00 KiB)

ROBER Before after

Pollu Before after

[ChrisRackauckas]

SciML/DelayDiffEq.jl#222
SciML/StochasticDiffEq.jl#450

Downstream PRs are already ready.