Huge amount of dynamic dispatch

[nsajko]

julia> using JET, Groebner, AbstractAlgebra

julia> _, x = QQ["x"]
(Univariate polynomial ring in x over rationals, x)

julia> @report_opt groebner([x^3-x^2-1])

The @report_opt command prints so many entries that I can't even scroll up to the beginning. I'm guessing that improving the situation here would tremendously improve performance.

[sumiya11]

Hi @nsajko , thanks for reporting this.

Wow, this is a lot of dynamic dispatch !

Previously, I have tried to reduce dynamic dispatch in Groebner a bit. My experience at that time was that if the types in the main "number-crunching" routines were inferred successfully, then dynamic dispatch in other parts of the code did not hinder runtime performance too much.

I may be wrong about that. Seeing the number of errors reported by @report_opt, it is possible that I missed something, and revisiting this issue in a systematical manner can be useful.

[sumiya11]

By the way, as I understand, some of the runtime dispatch comes from other modules. For example, if I do

using JET, Groebner, AbstractAlgebra
R, x = QQ["x"]
@report_opt target_modules=(Groebner,) groebner([x + 1])

the list seems to be much shorter

EDIT: but still, @report_opt groebner([x + 1]) contains useful info that is not included when using target_modules=(Groebner,). So, thanks for bringing this to my attenrion