04:14 <tankf33der> abu[m]: morning

04:14 <tankf33der> i can write this code, give up.

04:58 <abu[m]> Good morning tankf33der! Which way did you try?

05:35 <tankf33der> Via wait inside co

05:36 <tankf33der> and (time) calculation as second attempt

05:36 <tankf33der> it always blocks

05:37 <abu[m]> 'wait' cannot be used, because every coroutine will wait then, adding up all the times

05:37 <abu[m]> Because coroutines all run in the same process

05:38 <tankf33der> Give me a hint then

05:38 <abu[m]> I have a solution, but I could not use more than 10000 co's because of stack space

05:38 <abu[m]> As I said, I think we need 'pause' from @lib/simul.l

05:39 <abu[m]> It is in the discrete event simulation library

05:39 <abu[m]> https://software-lab.de/doc/des.html

05:44 <abu[m]> Shall I post my solution? Perhaps you can get it running with more stack space.

05:45 <tankf33der> Post it

05:45 <abu[m]> I tried "ulimit -s unlimited" but without success for 100 k coroutines

05:46 <abu[m]> http://pb1n.de/?aa377e

05:46 <abu[m]> (stack 6) seems the minimum

05:47 <abu[m]> An empty coroutine also runs with (stack 2), but 'pause' needs more stack it seems

05:51 <abu[m]> Also, creating more coroutines lets it get slower and slower, because for each new coroutine all the stack frames need to be traversed searching for a free place

05:52 <tankf33der> Insane

05:52 <tankf33der> next level

05:54 <tankf33der> Afk.

06:03 <abu[m]> Anyway, now I understand why 'pause' needs quite some stack space. It uses 'idx' which is recursive.

13:55 <abu[m]> No, I'm wrong. 'idx' is not recursive.

13:55 <abu[m]> (stack 2) is simply too small for a usable stack segment

13:58 <abu[m]> I would recommend a minimum stack segment size of 8

14:11 <beneroth> yeah... I guess you can construct a recursive/circular idx, but it would be a bad idea...

14:13 <abu[m]> Yes, all internal binary trees in pil are non-recursive

14:14 <abu[m]> Recursive versions are a lot easier, but may use a lot more stack