Search
Wednesday, 16th of May 2018, 21:01:52 UTC
22:25:42
Bicyclidine
karlosz: declaring a standard function notinline is prohibited, i think
22:26:16
Bike
but you must have an actual function somewhere, so just generate a call to that
22:27:27
drmeister
Bike: It got to the last file and I had to run to catch the train.
22:28:03
drmeister
Someone is going to have to rock me to sleep tonight.
22:29:05
drmeister
Did you push the latest changes to cst-catch?
22:33:05
drmeister
Bike: Yeah - cando used operator< and operator<=
22:33:10
drmeister
where did you delete them from?
22:33:32
Bike
what on earth does cando do with them
22:34:09
Bike
or do you just mean on numbers
22:34:17
Bike
https://github.com/clasp-developers/clasp/commit/6cd1e493760edd13d6c606e2ec84f0f3936d5bbe this is the commit
22:34:23
drmeister
It has a units package - remember when the Mars Orbiter skipped off the Martian atmosphere and was lost in space?
22:34:29
drmeister
This is to prevent that.
22:34:42
Bike
can the operators just be defined in the particular class?
22:35:40
Bike
comparison between generals seems like a bad idea
22:35:59
Bike
like, you don't want to allow comparing a dimensionless number with a unit-ed thing
22:38:08
drmeister
It makes sure that you are comparing quantities to quantities.
22:38:22
Bike
but then it can just be defined for the unit-ed class, right?
22:40:39
drmeister
Well, I've got clasp_number_compare(x,y)
22:41:35
Bike
so it can't just be defined for the class?
0:42:20
RickyRat
** NICK Nephromancer
2:14:21
DVSSA
drmeister: Should we set a time for that skype?
4:15:53
drmeister
DVSSA: Catch me tomorrow and we can sort it out
4:15:59
drmeister
loke: Are you still online?
4:17:03
drmeister
::notify loke The dev branch built - I pushed it to the master branch
4:17:03
Colleen
drmeister: Got it. I'll let loke know as soon as possible.
5:44:08
karlosz
its actually realtively easy to tell which <- instructions are actually needed in the flowgraph. if a <- instructions's target node has only 1 def, and either the source node has 1 def or the target has 1 use, then the <- instruction can be eliminated, with the source and target nodes merging into one node
5:45:00
karlosz
that's the algorithm i'm using to only produce the minimum amount of runtime store and loads from <- instructions
5:45:04
karlosz
it seems to work well
5:45:34
karlosz
perhaps that can be the basis for a general purpose HIR pass that cleans up the <- nodes
Thursday, 17th of May 2018, 9:01:52 UTC