Search
14:13:27
usha
@Shiho - did you see my new results - the one where I add hydrogen to the lysozyme system with tleap and then generate the parameters?
15:10:30
shiho
usha: Yes, I saw them. I'm working with charges, so I'll check it after that. Thank you!
15:10:58
usha
:) sure just wanted to check to see if I could help
17:20:41
Bike
looking at a fairly complex function, each instruction was copied 11.3 times on average
17:20:54
Bike
might as well do a histogram
17:29:14
Bike
no obvious shape. 180 instructions were onlycopied once, 165 21 times (the max), 512 14 times (the mode)
17:33:48
Shinmera
Bike: do you know if clasp signals conditions on floating point exceptions/signals/traps?
17:34:58
Bike
Yes. It dies. We have a bug for it somewhere
17:36:05
Shinmera
Guess I won't have to implement float-features:with-float-traps-masked for now then
17:55:56
Bike
well, not definitely a problem, but there's definitely overcopying
19:13:02
kpoeck
@shinmera do you have an example for a calculation where a float trap would be produced?
19:14:36
kpoeck
I see that (progn (+ most-positive-long-float most-positive-long-float) nil) works, just the printing dies
19:15:08
kpoeck
I a compiling a fix for the printing, still 100 files to go
19:16:18
kpoeck
I just fixed (/ 1 0) to return DIVISION-BY-ZERO
19:17:02
kpoeck
(/ (coerce 1 'short-float) (coerce 0.0 'short-float)) also works
19:17:13
Shinmera
return? Surely you mean signal
19:17:15
kpoeck
and used to calculate the inifinities
19:17:43
Shinmera
Infinities shouldn't need to be calculated. They're a precise bit value of a float.