Hi John,
It is very strange to see the
< and
> bugs survive because they were fixed as early as on August 7. You can see it by the date of
this message on the FBSL forum. Perhaps something went wrong that night when we were pushing the code up and down the repo prior to working out a common plan.
Thanks for the
ATAN fix. It slipped me 'cause of SB not barking up the ATN tree.
1. Now I'll be dowloading your code from the repo; I want to see how you split the arrays. I'm glad it works but
for the sake of maximum unification I suggest doing it my way because
I did it for SB, FBSL and O2 already.
2. We have yet to see a case where garbage collection interferes with calc speed, functionality, or precision. So my suggestion is to subside to
20 000 for the heap, and
10 000, for the stack,
in all the 3 languages. Under these conditions, the recursive stress test runs in
8 sec for SB,
9 sec for FBSL, and
X millisec for O2 (still crashes, in fact). The memory footprints at these settings are also very similar between SB and FBSL.
Please give me an explicit answer if you (dis)agree with these two propositions.
I'll be working to fix OxyLISP to full usability today and see how it behaves under this stress test. Once through with it, we'll discuss where to go further.
