5:15:08stylewarningoof ppc allocates complex doubles like crazy
5:23:46stylewarningkarlosz: ok, threading seems to work, cool
5:23:54stylewarningpegging 64 cores with a bunch of lisp
7:36:55aethoh, nice, if the PPC64 and RISC V ports happen, then SBCL will run on "everything"
7:39:49no-defun-allowedSorry to ask, but is threading on 32-bit ARM going to happen?
8:14:28Krystofnot unless there's a pressing need or a victim with lots of spare time
8:14:39Krystofmaybe "some spare time" rather than lots
8:15:42KrystofI'd like it to happen, and would be happy to help, but it needs someone who can dedicate some concentrated time to it
8:17:06no-defun-allowedI have school holidays now and I'd like to have a go at it, but I've never poked around at a compiler before, and I don't know how much time it would take to add threading support.
8:54:56Krystofit's a little tricky, because arm is register-starved. At the moment, the arm32 port dedicates a register to NIL, but there really aren't any other spare registers, I think
8:55:33Krystof(and a thread-enabled port needs a register to hold the current thread). so I think the strategy should be to have NIL-value be stored within the thread structure
8:55:53Krystofsomewhere I have a work-in-progress tree along these lines
10:50:58stassatsKrystof: register starvation hasn't really been a big deterrent for me, but having to write all the CAS loops and other synchronization primitives