freenode/#clasp - IRC Chatlog
Search
14:41:18
drmeister
I added a -U command line option to unpack faso files into object files for debugging and eventual linking if we choose to do that
14:45:18
drmeister
So on macOS and Linux with boehm - all fasl files are fasp files and compile-file-parallel works.
14:46:55
drmeister
With mps compile-file-parallel is still broken - there may be a multi threading issue that may be MPS specific or may be hidden in boehm
14:55:00
drmeister
I haven't done the merge yet - I've been testing everything against Cando on linux and macOS and changing the wscript so that the default is CLASP_BUILD_MODE is faso and USE_COMPILE_FILE_PARALLEL is True
18:29:10
drmeister
I've put this back into stream_dispatch_table - it shouldn't screw anything up anymore.
18:30:23
drmeister
They do have pybind11 code - we should be able to change that to clbind - and if not - we can fix clbind to look more like pybind11
18:35:03
drmeister
I'm merging dev into dev-chris and building everything on linux and macOS - if everything builds clean I'll merge dev-chris back into dev and do the same thing. That's probably redundant - no?
18:36:35
drmeister
Does the new atomics code come with a way to do compare-and-swap with the new atomic slots?
18:37:12
Bike
i don't know that I can make that work in C++. (it's not that hard to make our compiler do it)
18:38:02
Bike
for a full implementation i'd also like to allow (cas whatever) function names, analogous to (setf whatever), but that seemed like a deep rabbit hole
19:53:51
Bike
drmeister: hey if i turn the variable on it does actually catch the "load must specify explicit alignment" problem
20:29:31
Bike
oh, and it's set up to dump every module it sees even if there are no problems... rejected
21:22:38
Bike
doesn't seem to take any more time but i am kind of inclined to leave it off until we alter the compiler again
0:41:05
drmeister
I'd expect the module verifier to call the function verifier - but probably not given what you see.
0:43:10
drmeister
Wild speculation - maybe the instructions are connected together as a linked list and the first or last instruction isn't pointing to NULL.
0:44:59
Bike
https://github.com/clasp-developers/clasp/blob/dev/src/lisp/kernel/cmp/jit-setup.lsp#L70-L71 mhm.
2:24:42
drmeister
I hit a little problem with what I think is a change you made cleaning up some code.
2:27:27
drmeister
is cons_car(x) and cons_cdr(x) non-reference taking versions of CONS_CAR(x) and CONS_CDR(x)?
2:28:00
Bike
you're actually the one who wrote them, i just used them because they're convenient and I like inline functions more than macros.
2:28:50
Bike
https://github.com/clasp-developers/clasp/blob/dev-chris/include/clasp/core/cons.h#L443-L448
2:29:20
drmeister
They should only be used in the context if (x.consp()) { cons_car(x); ... cons_cdr(x);}