CapitalEx — Today at 2:17 PM
When I was trying to make modal program fast, the number one thing I tried to do was make my program size stay extremely small
rosin — Today at 2:17 PM
I was curious to see if it would be faster to do a single large expansion of size n, rather than n small expansions 
in fact, as far as I can tell the do-row version of the program has 1 fewer re-write per cell, but it's still slower
so what I am learning here is that indeed it is better to keep things small
CapitalEx — Today at 2:19 PM
Devines implementation flips your program between two buffers. It writes your changes first, then copies the rest of the old buffer. So the long that copy has to be the slower things run.
rosin — Today at 2:19 PM
aha
Thank you for the explanation
CapitalEx — Today at 2:19 PM
No problem
rosin — Today at 2:19 PM
it was worth the attempt >:)
(wryl) — Today at 2:19 PM
The original Python implementation uses a single buffer, a queue, which means we don't need to spend time copying.
But it's not fast because.. Python.
CapitalEx — Today at 2:20 PM
I've been meaning to write a compile to Lua for my implementation cause the C impl is still leagues faster lol
(Well, outside of some cases where the rolling queue is a major win and the program is small)