realitix has quit [Read error: Connection reset by peer]
k1nd0f has joined #pypy
nunatak has joined #pypy
commandoline has quit [Ping timeout: 268 seconds]
commandoline has joined #pypy
k1nd0f has quit [Ping timeout: 258 seconds]
kenaan has quit [Ping timeout: 244 seconds]
Ai9zO5AP has joined #pypy
realitix has joined #pypy
tsaka__ has quit [Ping timeout: 255 seconds]
<cfbolz>
arigato: are you around?
whitewolf has quit [Ping timeout: 246 seconds]
tsaka__ has joined #pypy
<arigato>
cfbolz: yes
<arigato>
at least until the next power cut due to the excess snow falling since yesterday
<cfbolz>
ouch
<cfbolz>
arigato: I am thinking about pycket (the racket/scheme in rpython). can I simulate an endless C stack with stacklets?
<cfbolz>
basically checking on recursion whether the C stack is getting close to full, and switching to a new stacklet if yes
<cfbolz>
(right now it's completely stackless using explicit trampolines and classes to store all the stack data, but that's only cool in the case of tail calls, where the JIT removes all the allocations)
<arigato>
so yes, mayeb
<cfbolz>
arigato: call/cc would be done differently, so that's not a concern
<cfbolz>
only care about running out of C stack for deep recursion
<arigato>
you still need a trampoline which runs with very little stack below it, and you start a stacklet for every chain of frames, and switch away from it and back to the trampoline if needed
<cfbolz>
yes, sure
<cfbolz>
it's going to be hairy to implement, I suppose, but you don't see any fundamental problems with the plan, do you?
<arigato>
no
<cfbolz>
cool :-)
<arigato>
honestly I would also consider just increasing the C stack limit, on systems where you can
<cfbolz>
arigato: but that's not really enough. a lot of things are really not tail recursive, and so if you, say, want to build a big enough linked list you are going to run into trouble
<arigato>
I'm not sure to see what's the difference
<cfbolz>
but yes, I suppose I should delay thinking about stacklets and try with a huge C stack limit to see whether it actually helps the problems we are having
<arigato>
we're comparing one huge stack and many small pieces which take exactly the same total
<arigato>
the issue with one huge stack is shadowstack, though
<cfbolz>
ok
<cfbolz>
yes, I saw the issue too
<arigato>
ok
<cfbolz>
anyway, will experiment
<cfbolz>
arigato: thanks :-)
adamholmberg has joined #pypy
adamholmberg has quit [Ping timeout: 268 seconds]
realitix_ has joined #pypy
dddddd has joined #pypy
realitix has quit [Ping timeout: 245 seconds]
realitix_ is now known as realitix
<mattip>
aws gave me some credits to see if I could build arm32 on aarch64
<mattip>
I can use a chroot or a docker to build, and translates