time consuming function "addToResult!"

[johhell]

test case

instModel.result.n_w_invariant = 149
instModel.nsegments = 1
length(instModel.x_start) = 17
number of result points = 5509

timing

... Timings for simulation of SMtest (without initialization):
 ─────────────────────────────────────────────────────────────────────────────────────────
                                                               Time          Allocations  
                                                         ───────────────   ───────────────
                     Total measured:                          2.50s            1.03GiB    

 Section                                         ncalls     time    %tot     alloc    %tot
 ─────────────────────────────────────────────────────────────────────────────────────────
 Modia.simulate!                                      1    2.40s   96.2%   0.98GiB   94.8%
 DifferentialEquations.solve                          1    2.40s   96.2%   0.98GiB   94.8%
 Modia getDerivatives!                            17.6k    2.38s   95.2%   1.03GiB   99.5%*
*Modia addToResult!                               5.51k    1.55s   62.0%   0.97GiB   93.9%   <<====
 Modia LinearEquationsIteration!                  87.8k    167ms    6.7%   4.02MiB    0.4%
 Modia.init!                                          1   95.6ms    3.8%   55.1MiB    5.2%
 Modia LinearEquationsIteration! (solve A*x=b)    52.7k   78.1ms    3.1%   4.02MiB    0.4%
 Modia eventIteration!                                1   91.5μs    0.0%   2.56KiB    0.0%
 DifferentialEquations.ODEProblem                     1   1.23μs    0.0%     0.00B    0.0%
 ─────────────────────────────────────────────────────────────────────────────────────────
  2.498988 seconds (2.94 M allocations: 1.034 GiB, 6.49% gc time)

modified addToResult!

function addToResult!(m::SimulationModel{FloatType,TimeType}, x, time, w_invariant...)::Nothing where {FloatType,TimeType}
the result of getDerivatives is called w_invariant and has the type Tuple, which is immutable. So it can be used directly (without deepcopy) in:

#     push!(result.w_invariant[end], deepcopy(w_invariant))
    push!(result.w_invariant[end], w_invariant)

timing with modified addToResult!

... Timings for simulation of SMtest (without initialization):
 ─────────────────────────────────────────────────────────────────────────────────────────
                                                               Time          Allocations  
                                                         ───────────────   ───────────────
                     Total measured:                          943ms            88.5MiB    

 Section                                         ncalls     time    %tot     alloc    %tot
 ─────────────────────────────────────────────────────────────────────────────────────────
 Modia getDerivatives!                            17.6k    859ms   91.2%   83.7MiB   94.6%
 Modia.simulate!                                      1    853ms   90.6%   33.4MiB   37.8%
 DifferentialEquations.solve                          1    853ms   90.5%   33.3MiB   37.6%
 Modia LinearEquationsIteration!                  87.8k    160ms   17.0%   4.02MiB    4.5%
 Modia.init!                                          1   88.8ms    9.4%   55.1MiB   62.2%
 Modia LinearEquationsIteration! (solve A*x=b)    52.7k   74.4ms    7.9%   4.02MiB    4.5%
*Modia addToResult!                               5.51k   65.5ms    6.9%   23.8MiB   26.9%   <<====
 Modia eventIteration!                                1   85.6μs    0.0%   2.56KiB    0.0%
 DifferentialEquations.ODEProblem                     1   1.30μs    0.0%     0.00B    0.0%
 ─────────────────────────────────────────────────────────────────────────────────────────
  0.944182 seconds (1.02 M allocations: 88.559 MiB)

time consumption and storage allocation was reduced dramatically

postscript

in previous Modia versions (<=0.8.4) the same approach was used.

function addToResult!(m::SimulationModel, variableValues...)::Nothing
    push!(m.result_vars , variableValues)
    push!(m.result_der_x, deepcopy(m.der_x))
    return nothing
end

used versions

Modia version = "0.9.4-dev"
Julia = 1.7.1