Open dylanede opened 3 weeks ago
I also have found that the workaround of performing
scalarize
on everything prior tolinear_expansion
can be incredibly slow and memory-hungry for moderately sized systems.
Thanks for pointing this out and for the PR! I agree this is a performance bottleneck. I have an idea on how to improve this and will aim to have a fix pushed soon this year.
It looks like there's no support for array and matrix operations in
linear_expansion
, e.g.broadcast
.I also have found that the workaround of performing
scalarize
on everything prior tolinear_expansion
can be incredibly slow and memory-hungry for moderately sized systems. For example for a particular system of 407 equations in 407 unknowns,@time
for scalarizing equations and variables and then callinglinear_expansion
giveson my machine.