| 10:47 |
jonsykkel |
its kinda distgusting to lok at memory dump where 70% of bytes are 00 |
| 10:47 |
jonsykkel |
im considerating memory model where each cell is 64bit then seprately, 4bit types packed togetehr in start of heap |
| 10:47 |
jonsykkel |
that would make car/cdr 32bit, limiting adres space to 32gb |
| 10:47 |
jonsykkel |
would that limitation make a script lang useless to anybudy else than me? |
| |
~ 16 minutes ~ |
| 11:03 |
asciilifeform |
jonsykkel: not useless, but potentially annoying (esp. if run w/out gc) |
| 11:10 |
jonsykkel |
indeed not so good for gcless operation |
| 11:13 |
asciilifeform |
( in asciilifeform's cosmography, 'scripting lang' is 1 where a typical proggy runs '1 shot' -- e.g. vtron -- and terminates. and so gc-less operation is potentially The Right Thing : simpler runtime, easily-predicted time, etc ) |
| 11:17 |
jonsykkel |
doesnt cache effects make time somwat unpredictable either way? |
| 11:19 |
jonsykkel |
in my toy tests prog total execution time is lower with simple gc than no gc |
| 11:19 |
jonsykkel |
presumably cuz of those |
| |
~ 57 minutes ~ |
| 12:17 |
asciilifeform |
jonsykkel: pretty strange (if a cell aint accessed, should not be dinging cache locality of proggy) |
| 12:22 |
jonsykkel |
well in gc version it collects a couple of times, so ends up accessing smaller amount of cells |
| |
~ 24 minutes ~ |
| 12:47 |
asciilifeform |
aa makes sense, it's a stop©/compacting gc, neh |
| |
~ 1 hours 50 minutes ~ |
| 14:37 |
jonsykkel |
yep |