|
01:48
ilbot3 joined
05:49
brrt joined
06:09
brrt joined
06:17
domidumont joined
06:21
domidumont joined
|
|||
| brrt | good * | 06:25 | |
| yesterday when i couldn't sleep i figured out a solution to my issue | |||
| i'm not sure if you are going to laugh or cry, because it is the same solution as the others | |||
| nevertheless | |||
| it does have the twin advantages of being robust (well, a bit more robust anyway) and simple | 06:26 | ||
| we do assign a ordering number to the tiles-that-are-to-be-inserted | 06:27 | ||
| however, we do this explicitly | |||
| and with just relative meaning | 06:28 | ||
| e.g. | |||
| suppose we find that we must spill a register to load a value | |||
| we must do three things | |||
| a): find the register | |||
| b): find a place in the bytecode stream to spill it (by default just after its creation) | 06:29 | ||
| c): find a place in the bytecode strema to insert the load (by default just before its use) | |||
| if the place of b) and c) is the same, first insert the spill, and assign it relative order nr $x; then insert the load, and assign it relative order nr $x + 1 | 06:30 | ||
| this way i can always ensure that pairs of inserted tiles have a relative order, and they also have a relative order to tiles that are already present in the bytecode stream | 06:31 | ||
| bytecode stream ~~ tile list, in this case | |||
| so yes, i'm faced with difficulty, and i just pick another integer | |||
| in this case, explicit integers are better than implicit | 06:32 | ||
|
07:05
brrt joined
07:36
ilbot3 joined
07:54
lizmat joined
|
|||
| brrt | \o lizmat | 07:59 | |
|
09:05
FROGGS joined
09:21
brrt joined
13:42
domidumont joined
13:56
brrt joined
14:02
zakharyas joined
14:18
domidumont joined
15:15
domidumont joined
19:05
zakharyas joined
19:11
zakharyas joined
|
|||