« first day (2878 days earlier)      last day (1603 days later) » 

b3m2a1
b3m2a1
12:45 AM
@CarlLange @ChrisK yeah Carl asked as question on YouTube about documentation and apparently that's what prompted them to stream it
 
 
8 hours later…
Carl Lange
Carl Lange
8:42 AM
@b3m2a1 What to do with this newfound power!
 
 
5 hours later…
1010011010
1010011010
2:01 PM
@kirkus Hi Kirkus, have you found any possibilities for improvement in the code?
I didn't hear from you anymore so I figured I'd ask :-)
 
 
1 hour later…
Feeds
Feeds
3:21 PM
Wolfram Blog: The Winners of the 2019 One-Liner Competition
posted on December 05, 2019 by Christopher Carlson

This year’s Wolfram Technology Conference was host to the eighth annual One-Liner Competition, an event where attendees show us the most astounding things they can accomplish with 128 or fewer characters of Wolfram Language code. Submissions included games, card tricks and yoga exercises, all implemented with less than one tweet’s worth of the Wolfram Language. [...]

 
 
1 hour later…
Feeds
Feeds
4:38 PM
 
b3m2a1
b3m2a1
4:50 PM
@CarlLange I think abusing it is the only reasonable option
 
kirkus
kirkus
@1010011010 hey sorry for the radio silence, I've still been focused on those Log[] calls since they are such a huge piece of the execution time. I actually submitted performance bug (feature request) to Wolfram. Right now, I'm trying a test to see if an implementation in C++ will help.
 
Feeds
Feeds
 
1010011010
1010011010
5:18 PM
@kirkus Wait, what? You're saying the compilation times are a bug?
 
Feeds
Feeds
5:59 PM
 
kirkus
kirkus
6:44 PM
@1010011010 no, I said the slow times of Log[] of negative numbers were a performance "bug".
 
 
2 hours later…
1010011010
1010011010
8:38 PM
Ha, fair
I bet this is also where much of the extra compilation time comes from then. I used logarithms to avoid overflow, which the system is very much prone to
 
kirkus
kirkus
@1010011010 so if you look at those original numbers I posted, doing a Log[] of the 196x196 was incredibly slow when negative numbers are in the mix, but in C++ using the built-in complex log function is ~100x faster.
so I'm doing a quick CreateLibrary with a "fastLog" function to see if there's a good overall speedup
 
1010011010
1010011010
How certain are we that this overhead is not caused by complicated branch cut machinery?
 
Nasser
Nasser
I think computer algebra systems like Mathematica are making people worst at math, not better. They throw in some code in expecting some magic to happen and just get a response from the computer and then wonder why they are not getting anything back.
 
kirkus
kirkus
@1010011010 not sure. as far as I could determine with profiling, on my machine, the Log[] call takes ~0.13s every time. Doing the same calculation in C++ takes ~0.001s
 
b3m2a1
b3m2a1
@Nasser I think they just make people more willing to try to do the math
Rather than just saying it's not worth even trying
And often it fails but at least they tried
 
Feeds
Feeds
9:26 PM
 
Nasser
Nasser
@b3m2a1 Yes, it makes them try the math. But what I mean, they become dependent on the computer to do everything for them, and so they do not know how to do the math themselves any more. I think CAS that shows step-by-step solutions are much more useful for learning from.
 
b3m2a1
b3m2a1
9:48 PM
I think the “any more” isn’t quite right. I think they never knew and never would anyway without it.
I will never be much of an integrator. Doesn’t matter if I didn’t have Mathematica or not.
I would hazard lots of people are in this boat
 
Feeds
Feeds
10:07 PM
 

« first day (2878 days earlier)      last day (1603 days later) »