Performance

In general calculating the thermodynamic properties of a mixture scales with the number of constituent species. The key feature of IdealGasThermo.jl is the ability to represent a mixture of gases as a single composite_species which can provide noticable performance improvements.

Below is an example. We first create a function that sets the gas temp 100 times and calculates $c_p$, $h$, $\phi$, and $dc_p/dT$. We do this so that

TT = rand(200.:600., 100)

function benchmark_Gas(TT::AbstractVector, gas::AbstractGas)
    @views for i in eachindex(TT)
        gas.T = TT[i]; 
        gas.cp
        gas.ϕ
        gas.h
        gas.cp_T
    end
end

We now initialze a Gas and Gas1D instances and benchmark the above function.

gas = Gas()
gas1D = Gas1D()

using BenchmarkTools

julia> @benchmark benchmark_Gas($TT, $gas1D)
BenchmarkTools.Trial: 10000 samples with 9 evaluations.
 Range (min … max):  2.778 μs …  12.551 μs  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     2.884 μs               ┊ GC (median):    0.00%
 Time  (mean ± σ):   2.909 μs ± 202.135 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

  ▂█▅▅ ▄▇▇▅▃▃▃▃▂▂▁                                            ▂
  ████████████████▆▆▇▇▆▆▆██▆▆▇▇█▇▇▇▇▇▇▆▇▆▆▅▆▆▅▅▆▄▆▄▆▅▆▅▆▆▆▅▄▆ █
  2.78 μs      Histogram: log(frequency) by time      3.74 μs <

 Memory estimate: 0 bytes, allocs estimate: 0.

julia> @benchmark benchmark_Gas($TT, $gas)
BenchmarkTools.Trial: 10000 samples with 4 evaluations.
 Range (min … max):  7.604 μs …  13.541 μs  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     7.667 μs               ┊ GC (median):    0.00%
 Time  (mean ± σ):   7.922 μs ± 561.794 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

  ██▂ ▂▄▅▄▅▃▁▂▂▂                                              ▂
  ███▆███████████▇▇▇█▇▇▇▇█▇▆██▇███▇▆▇▆▇▇▇▆▇▇▇▆▇▆▇▆▆▇▇▆▆▇▇▇▆▅▅ █
  7.6 μs       Histogram: log(frequency) by time      10.2 μs <

 Memory estimate: 0 bytes, allocs estimate: 0.

We see that the Gas1D calculations are ~3x faster. If we repeat this with a mixture that has more species the time taken for the Gas type calculation roughly scales with the number of species where as the Gas1D type calculations are independent of the number of constituent species in the gas.

For 10 components:

julia> @benchmark benchmark_Gas($TT, $gas)
BenchmarkTools.Trial: 10000 samples with 1 evaluation.
 Range (min … max):  17.208 μs … 149.167 μs  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     17.875 μs               ┊ GC (median):    0.00%
 Time  (mean ± σ):   18.134 μs ±   1.755 μs  ┊ GC (mean ± σ):  0.00% ± 0.00%

  ▂▃▁  ▁█▇▅▃▂▃▄▄▃▂▁▁▁▂▃▁                                       ▂
  ███▆▅█████████████████▇▅▅▅▆▅▅▅▄▅▅▅▅▄▄▅▅▅▆▃▅▁▅▅▆▄▄▅▅▅▅▆▅▅▄▅▆▆ █
  17.2 μs       Histogram: log(frequency) by time      22.6 μs <

 Memory estimate: 0 bytes, allocs estimate: 0.

julia> @benchmark benchmark_Gas($TT, $gas1D)
BenchmarkTools.Trial: 10000 samples with 9 evaluations.
 Range (min … max):  2.792 μs …  13.907 μs  ┊ GC (min … max): 0.00% … 0.00%
 Time  (median):     2.903 μs               ┊ GC (median):    0.00%
 Time  (mean ± σ):   2.958 μs ± 233.528 ns  ┊ GC (mean ± σ):  0.00% ± 0.00%

  ▁▁   █▇▆▄▄▄▄▂▁▁▂▂▂▁                                         ▁
  ██▇▄███████████████▇▅▅▆▆▆▆▆▆▄▆▇▇▇▇▆▆▇▇▇▇▆▆▆▆▅▆▆▆▅▇▅▆▅▅▅▅▅▆▆ █
  2.79 μs      Histogram: log(frequency) by time      3.75 μs <

 Memory estimate: 0 bytes, allocs estimate: 0.