|
01:08
FROGGS_ joined
03:02
btyler joined
06:35
FROGGS joined
|
|||
| FROGGS | so it looks like it is showed in here: lh.frame->work[lh.handler->block_reg].o | 06:57 | |
| I guess I can put my label in -> work also? | |||
| /* The temporary work space for this frame. After a call is over, this | |||
| * can be freed up. Must be NULLed out when this happens. */ | |||
| MVMRegister *work; | |||
| I guess I am safe to put something in there | 06:58 | ||
| jnthn | yes | 06:59 | |
| teaching, bbl | |||
| FROGGS | o7 | ||
| jnthn | but I note this may help elegantly solve the mapiter variable label issue efficiently too | ||
| FROGGS | nice :o) | 07:00 | |
| (very nice indeed) | |||
|
07:01
lue joined
07:15
zakharyas joined
|
|||
| jnthn wondered for a while if o7 was a curious form of o/ or one of those German things like "n8" where if you sounded out the number 7 in German with an "o" before it, you get a word :P | 08:39 | ||
| I'm pretty sure the answer to the second propositon is n1 though :P | |||
| nwc10 | jnthn: 3rd patch I no-pasted to convert reprconv.h macros to functions does seem to reliably slightly improve setting build times. | 08:42 | |
| (on "my" machine") | |||
| I'm a bit surprised. I didn't check the generated code, but I assume that it's causing some epxressions not to be evaluated twice | 08:43 | ||
| on "my" machine, for 1 version of 1 compiler | 08:45 | ||
| jnthn | nwc10: ooh | 08:49 | |
| OK. Can't apply it right nwo, though. | |||
| (at teaching) | 08:51 | ||
| nwc10 | jnthn: I suggest also benchmarking it | 08:55 | |
| FROGGS | jnthn: o7 was like a military greeting :o) | 09:07 | |
|
09:10
lue joined
|
|||
| jnthn | nwc10: ok | 10:36 | |
| Also need to bug-hunt some more tonight... | 10:37 | ||
| To see what bast Panda. | |||
| nwc10 | are you at home tonight? So have the services of both your laptop and desktop? | 10:42 | |
| jnthn | yeah | 10:45 | |
| here I have just the teaching machine | |||
| Thing is, my laptop these days is...very like my desktop | 10:46 | ||
| A 64-bit Windows install on an i7 | |||
| nwc10 | so, 2 to benchmark with :-) | ||
| jnthn | Just difference in number of cores. | ||
| nwc10 | (was more thinking that 1 can be busy benchmarking, while the other attends to the human) | ||
| jnthn | Oh :) | ||
| nwc10 | hopefully the teaching machine is not running XP | ||
| jnthn | win7 | 10:47 | |
| I'm only glad it's not using win8 | |||
| Last win8 machine I encountered, I got bored of trying to use it before I actaully got anything done on it :) | |||
| nwc10 | deskops - have core*s*. laptops - have core*s*. tablets - yes, core*s*. Phones, guess what? | 10:49 | |
| jnthn | My iPhone has 2, I think :) | ||
| If I was hardcore, I could use my multicore to play metalcore :P | 10:50 | ||
| nwc10 | ho ho, I don't think that my hack was the best: | 10:55 | |
| Stage mast : 44313.938 | |||
| but it did eventually finish and pass | |||
| 22.25h | 10:56 | ||
| jnthn | o.O | 10:57 | |
|
12:49
woolfy1 left
13:46
btyler joined
13:58
FROGGS joined
14:23
daxim joined
14:48
sergot_ joined
15:42
btyler joined
|
|||
| nwc10 | jnthn: which were the MAST structures that I was having fun with previously (for alignment) | 15:47 | |
| the ones that are written to from NQP code by treating them as instances P6Opaque | |||
| but are actually read by the C code as some sort of C structure for a different class | 15:48 | ||
| aha, most things in nodes_moar.h | 15:51 | ||
| jnthn | Yes, there. | 15:54 | |
|
15:54
btyler joined
|
|||
| nwc10 | OK, so the fun and games is because the size of the 6model object header changed | 15:54 | |
| such that sizeof(MVMP6opaque) % 8 is now 4 | 15:55 | ||
| was 0 before | |||
| jnthn | ah | ||
| nwc10 | so alignment is being worked out differently in two places in P6opaque.c | ||
| I've figured this out, not proven it :-) | 15:56 | ||
| going to sit with a beer to think about it | |||
| about how to fix it to be robust | |||
| jnthn | beer++ | ||
|
16:04
lue joined
16:48
lue joined
17:00
jnap joined
17:28
jnap1 joined
|
|||
| jnthn | Time to see if I can work out the panda breakage again... | 19:21 | |
| TimToady | someone should buy nwc10++ a really fast version of that cpu :) | 19:29 | |
| and overclock it in liquid nitrogen or so | |||
| [Coke] | liquid beertrogen. | 19:30 | |
| TimToady | splitting the atom? | 19:42 | |
|
20:19
jnap joined
|
|||
| dalek | arVM: 2360aaa | jnthn++ | src/ (4 files): Include SC identifiers in bytecode dump. |
21:35 | |
| jnthn | The seeming cause of the issue I'm seeing is that in the pre-comp of JSON::Tiny I have | 21:39 | |
| 01766 wval loc_34_obj, 3, 138 | |||
| 01767 setlexvalue loc_33_obj, '&to-json', loc_34_obj, 0 | |||
| But that is a routine declared in the file | 21:40 | ||
| So the first wval arg should be 0 | |||
| FROGGS | so is this the offending line? github.com/moritz/json/blob/master...iny.pm#L34 | 21:49 | |
| or is it the proto with 'is export' ? | |||
| jnthn | It explodes at 21 | ||
| Because the above thing shoved a Metamodel::PackageHOW into the lexical slot instead of the actual to-json routine | 21:50 | ||
| And the clone thus tries to clone the meta-object and explodes | |||
| timotimo | why would it do that :o | 22:01 | |
| jnthn | No darn clue :( | 22:02 | |
| FROGGS | gnight | 22:03 | |
| jnthn | wait, what... | 22:04 | |
| It compiles here JSON::Tiny before it compiles JSON::Tiny::Actions | |||
| How on earth does that even work... | 22:05 | ||
|
22:17
benabik joined
22:49
woolfy joined
22:50
lizmat joined
22:54
benabik joined
|
|||