- For truly effective RAM releasing, we would almost certainly need to
make major changes, to release RAM at an OS page level. 256K arenas were already too fat a granularity.
We can approximate that closely right now by using 4K pools _and_ 4K arenas: one pool per arena, and mmap()/munmap() are then called on one page at a time.
Sorry about this: there was another exceedingly subtle "there's more than one pool in an arena" assumption in the code, which didn't cause anything to fail, but did cause it to hold on to empty arenas in some cases. Repaired that, and then:
For memcrunch.py, using 200x the original initial objects, this works quite well! ... So, at the end, space utilization is over 90%:
1,968,234,336 / 2,165,366,784 = 0.90896117
Which changed to the slightly better:
1,968,235,360 / 2,143,465,472 = 0.91824916
OTOH, an even nastier version of the other program I posted isn't helped much at all, ending like so after phase 10: ... 232,861,440 / 4,073,746,432 = 0.0571615
And that one enjoyed a relatively huge improvement, to:
232,861,440 / 2,334,990,336 = 0.09972694
But none of that changes any of the bottom-level observations or speculations.