On Monday, 24 November 2025 at 15:27:33 UTC, Jabba Laci wrote:
On Monday, 24 November 2025 at 15:07:19 UTC, Serg Gini wrote:
Do you use LDC with -O3?
With this setting, it's 2 seconds instead of 3. Still very
slow. In this case a matrix would have been a better choice
(I'll rewrite with that), but I find it strange that the
dictionary solution is so slow.
rules of thumb from the data oriented design crowd, every layer
of memory is 100x slower then the one before it; cause of the
speed of light
Inside a cpu you got a very few ints that asm plays with
then "L0" thru "L3" caches
then ram
(then harddrives)
(then the internet)
If a million ints is enough to go into L3 or ram, your leaving
allot of 0's on the table.
Optimizers dont improve your types, they can do allot of work on
your functions but you asked for an int instead of a bool for
each light, your going to get 32 lights inside your computer for
each, and then your probaly on 64 bit os so your aa will be using
some 64 bits for parts of that; You asked for *allot* of overhead.
