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
Could we have an AoT compiling mechanism based on copy-pasting chunks of the interpreter to limit branching and allow optimization across bytecode op sequences? It could be useful for some standard functions (though just rewriting them in C++ might be easier) but also for some snippets that we could match by a bytecode hash of some sort.
Unless we go the preprocessing route in which case it could be possible to AoT compile entire cartridges.
It appears that the orbys demo uses 70KiB worth of Lua heap before even executing code. About 20KiB of that is the y8 runtime. Maybe some bytecode optimization would be worth it? See Opeth.
There is a number of things that could be done to reduce RAM usage, at varying levels of complexity:
The text was updated successfully, but these errors were encountered: