Unique function argument names for no recompilation #272
Conversation
Fixes https://discourse.julialang.org/t/function-compiles-every-time/63273/5 by giving the arguments unique names. MWE: ```julia using ModelingToolkit, OrdinaryDiffEq @variables t x(t) RHS(t) # independent and dependent variables @parameters τ # parameters D = Differential(t) # define an operator for the differentiation w.r.t. time # your first ODE, consisting of a single equation, indicated by ~ @nAmed fol_separate = ODESystem([ RHS ~ (1 - x)/τ, D(x) ~ RHS ]) sys = structural_simplify(fol_separate) @time begin prob = ODEProblem(sys, [x => 0.0], (0.0,10.0), [τ => 3.0]) sol = solve(prob, Tsit5()) end ``` ``` 0.001047 seconds (2.20 k allocations: 126.812 KiB) 0.001023 seconds (2.20 k allocations: 126.812 KiB) 0.001021 seconds (2.20 k allocations: 125.672 KiB) 0.001045 seconds (2.20 k allocations: 125.672 KiB) ``` while it used to be 99.9% compile time.
| linenumbers = true) | ||
|
|
||
| dargs = map(arg -> destructure_arg(arg, !checkbounds), [args...]) | ||
| dargs = map((x) -> destructure_arg(x[2], !checkbounds, Symbol("ˍ₋arg$(x[1])")), enumerate([args...])) |
There was a problem hiding this comment.
two unicode underscores are better than one?
|
Codecov ReportAll modified and coverable lines are covered by tests ✅
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## master #272 +/- ##
==========================================
- Coverage 74.44% 74.36% -0.09%
==========================================
Files 19 19
Lines 1761 1759 -2
==========================================
- Hits 1311 1308 -3
- Misses 450 451 +1 ☔ View full report in Codecov by Sentry. |
|
|
||
| function numbered_expr(de::Equation,varnumbercache,args...;varordering = args[1], | ||
| lhsname=gensym("du"),rhsnames=[gensym("MTK") for i in 1:length(args)],offset=0) | ||
| lhsname=:du,rhsnames=[Symbol("MTK$i") for i in 1:length(args)],offset=0) |
There was a problem hiding this comment.
These are only used in C and MATLAB outputs iirc
There was a problem hiding this comment.
Might as well make the codegen stable there too.
There was a problem hiding this comment.
And in that case, we really don't need to gensym since it's not going to take globals from Julia 😅
| function destructure_arg(arg::Union{AbstractArray, Tuple}, inbounds, name) | ||
| if !(arg isa Arr) | ||
| DestructuredArgs(map(value, arg), inbounds=inbounds) | ||
| DestructuredArgs(map(value, arg), name, inbounds=inbounds) |
|
This might not solve the root issue. MTK is still regenerating the same function again. |
|
That function generation is rather cheap though. This is orders of magnitude better than before at least. |
Fixes https://discourse.julialang.org/t/function-compiles-every-time/63273/5 by giving the arguments unique names. MWE:
while it used to be 99.9% compile time. Tested on <1 and rebased.