00:11 colomon joined 01:31 FROGGS__ joined 06:20 FROGGS[mobile] joined 06:28 cognominal joined 06:30 cognominal joined 06:35 vincent joined
dalek arVM: 18c6079 | jimmy++ | src/io/ (2 files):
small cleanups, and removed a needless MVMROOT
12:41
jnthn JimmyZ: uh, put that setup flag back, please 12:50
I need that for somehting
Rest looks fine
oh, I see why you don't need it 12:51
But that's not a cleanup. That's an obfuscation.
JimmyZ ah.... 12:54
dalek arVM: f318012 | jimmy++ | src/io/eventloop.c:
re-add setup var
12:56
jnthn Thanks
JimmyZ++
JimmyZ np
jnthn oh argh
wait
You didn't fix what bothered me 12:57
The duplicate call the queue poll
JimmyZ what's wrong with it? 12:58
jnthn Mine had one call and it was really easy to see we were just polling for values in a loop
Now you have to read the if and then the thing at the end of the loop
JimmyZ I think it's the same except mine avoids multi-assign setup=1 in the loop 13:01
jnthn Fine, I'll just make it how I want myself. It's better than wasting time arguing with you. :/ 13:02
JimmyZ hehe :{ 13:04
:P
dalek arVM: 5889985 | jimmy++ | src/io/eventloop.c:
revert cleanups in setup_work function
JimmyZ jnthn: I don't quite understand when needs MVM_ASSIGN_REF ... 13:11
jnthn JimmyZ: Any assignment of one MVMCollectable into memory managed by another. 13:14
JimmyZ Thanks, I think I get it now 13:15
jnthn The only common exception is when assigning to a register 13:16
Because a MVMFrame is not collectable 13:17
Taht's why all the register writes don't have to be write-barriered.
JimmyZ That seems to reduce the gc pressure. 13:18
jnthn Well, it avoids a ton of checks on every single register assignment, for sure. 13:19
Which will make JIT a bunch less painful
13:37 dalek joined 14:23 salv0 joined 16:23 cognominal joined 17:21 daxim joined 21:38 LLamaRider joined
dalek arVM: 6cc9c86 | jnthn++ | / (5 files):
Stub various async ops.
22:47
jnthn My plan for tomorrow: implement those :) 23:19