9:15:05fiddlerwoaroofexit70[m]: yeah, it's not as simple as that
9:16:00fiddlerwoaroofsbcl uses MAP_FIXED to allocate some memory regions at well-known addresses, and macOS doesn't like MAP_JIT|MAP_FIXED
11:35:50Krystoffiddlerwoaroof: this may not be optimal, but I *think* that all heaps are relocatable, and the heap is the only thing that needs to be MAP_JIT
11:58:18stassatsi don't think map_jit is going to help on its own
11:58:32stassatsas it seems to be truly not allowing wx
11:59:04stassatsit would've been all fine and dandy if we were running only one thread
11:59:42stassatsbut toggling between rx and rw is going to be a problem for the code already running in rx
12:10:29stassatsor it may be fine, if pthread_jit_write_protect_np is really per-thread
12:38:54stassatsflip214: i added :preserve-fds to run-program
13:04:42stassatshuh, div->mul is actually slower on my arm64 machine
13:04:52stassatswonder if it's due to constant construction
13:10:29stassatsgcc uses the same number of instructions
13:26:05stassatswell, if gcc and clang do it, i guess no reason not to, maybe this is just some slow arm core
13:31:24stassatsFAST-TRUNCATE-C/UNSIGNED=>UNSIGNED also does multiply-add for the remainder even if it's unused...
13:33:28stassatsok, div becomes slower when using more bits