Search
17:47:23
drmeister
A source-location structure? What would that look like?
17:47:41
drmeister
If I build cclasp with -j1 then inlining should work?
17:48:02
Bike
if it's in the serial order it should be fine.
17:48:36
Bike
the structure would have what we use for source info, i.e.e a file handle, a line number, a column number, and a file position.
17:49:00
drmeister
We have one of those.
17:49:26
drmeister
https://github.com/clasp-developers/clasp/blob/dev/include/clasp/core/sourceFileInfo.h#L83
17:49:31
drmeister
We use it everywhere.
17:50:14
drmeister
Well, it's there - I used to use it a lot - then source tracking degraded to almost useless. Now it's alive and better than ever.
17:50:22
drmeister
Will that class work for you?
17:51:22
Bike
by not using it i meant like, for example, the function description doesn't
17:52:17
drmeister
Right - it doesn't yet. Is that all that you would want to add to the function-description?
17:54:02
Bike
I don't want to add it, I want it to replace the source info that's already in it
17:59:09
Bike
i mean, it wouldn't make sense to have both, right?
17:59:30
Bike
there's nothing particularly i want to add. it's got the name, the docstring, the lambda list, the source position. that's fine
18:06:45
Bike
cmp:make-function-info should be cmp:make-function-description, right?
18:29:49
Bike
oh, hang on, only ClosureWithSlots has a FunctionDescription
18:29:55
Bike
so there's still a separate NamedFunction
18:33:45
Bike
i see the scraper uses define_source_info which uses ext:annotate for docstrings... so i guess not actuallytouching thefunction
18:40:41
Bike
i think we should just rip out the function hierarchy
21:09:48
drmeister
The function hierarchy should be straightforward to simplify now that we have done the merge.
23:44:47
drmeister
Bike: Do you get a lot of boehm warnings when building cclasp in serial mode?
23:44:49
drmeister
https://gist.github.com/drmeister/e1c63ccab3307569e301f3444ead8ba0
23:45:10
drmeister
Do you have any idea what might have changed that gives rise to them?
23:46:43
Bike
i traced them a bit and it seems to be during a rehash.
23:46:59
Bike
while mapping instructions in cleavir, was my guess
23:50:06
drmeister
Yeah - that's where I've seen it before - rehashing.
0:05:06
drmeister
Inlining works great with the serial build - but it sure takes an ugly bat to the llvm-ir.
0:10:03
Bike
Inlining takes a bat to the llvm-ir, you mean? Because it increases the size?
0:11:02
drmeister
It makes it a LOT more complicated.
0:11:08
drmeister
Lots of basic blocks.
0:11:17
drmeister
Branches - whatnot.
0:12:19
drmeister
fibn without inlining...
0:12:37
drmeister
https://usercontent.irccloud-cdn.com/file/hoCLtnbN/cfg.FIBN%5ECOMMON-LISP-USER%5EFN%5E%5E.dot.pdf
0:12:58
Bike
mm. it's the type dispatch.
0:12:59
drmeister
https://usercontent.irccloud-cdn.com/file/876GeQhN/cfg.FIBN%5ECOMMON-LISP-USER%5EFN%5E%5E.dot.pdf
0:13:49
drmeister
How were things today? I hardly bothered you about inlining - what's up with that?
0:14:21
drmeister
Did you get any closer to the problem in asdf?
0:16:07
Bike
i was busy with the functions and stuff so i haven't been working on inlining. i'm pretty sick of it. but i can focus on that tomorrow.
0:20:39
Bike
i can do like, ./waf build_bboehm -j18 && ./waf build_cboehm -j1 to sequence right, right?
2:50:55
drmeister
I'm sorry about the inlining - that is a slog.
2:53:35
Bike
it's not like it's your fault
3:53:44
beach
Good morning everyone!