<mattip>
it seems like it would be a good fit for the RPython GC too
<fijal>
that's a complex incref/decref
<fijal>
5-10% seems very mellow, I wonder if it's true
<fijal>
I think if people can agree what "python with no gil" means, it'll be much easier to do for us
<mattip>
using a compile time switch and an ABI mark for c-extensions means that c-extension authors will have to opt-in to providing compatible wheels,
<mattip>
which avoids a whole class of pain for c-extension authors who may not want to support no-gil
<arigato>
it talks about one thread modifying the reference count of an object while another thread frees it, and how to hack at the memory allocator to somehow make it work
<arigato>
I cannot see how it can be made completely reliable, but then I don't know everything
jcea has joined #pypy
otisolsen70 has quit [Quit: Leaving]
dmalcolm has joined #pypy
Atque has quit [Remote host closed the connection]