You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm not sure that this is true for any number and any type of fields but I think this could be nonetheless investigated.
If proved beneficial in at least some useful situations, this could be allowed along with the mutable version, but I think that integrating this version will require many updates to the codebase.
The text was updated successfully, but these errors were encountered:
Actually some experimentation tells that this doesn't give almost any edge even in elementary models where the number of fields is low (almost same time for Schelling and Flocking).
See https://github.com/Tortar/Agents.jl/tree/agents_structs for how structs can be used instead of mutables one, this requires hacky code (this still doesn't pass all tests) and it doesn't even change that much the performance, so I think it's a dead-end
In a blogpost by Bogumił Kamiński at https://bkamins.github.io/julialang/2020/10/16/gctime.html, he describes that using
struct
's instead ofmutable struct
's as agents can be beneficial even when fields are updated with Accessors.jl.I'm not sure that this is true for any number and any type of fields but I think this could be nonetheless investigated.
If proved beneficial in at least some useful situations, this could be allowed along with the mutable version, but I think that integrating this version will require many updates to the codebase.
The text was updated successfully, but these errors were encountered: