Wolfram Mathematica

« first day (1375 days earlier) ← previous day next day → last day (3410 days later) »

Pragabhava

01:05

@JacobAkkerboom Thanks a lot Jacob!

3 hours later…

xslittlegrass

03:38

@Szabolcs I played with your LibraryLink example a little bit. I have no idea why the second call takes extra time. And it seems that this only happens when using "Shared" passing, the default automatic passing, though much slower, doesn't have this problem. Moreover, if we pass the unpacked array first, but then change the value of the packed array, there is no slowing in the pass:

In[13]:= Do[o@"clone"[arrayPacked], {100}]; // Timing
Out[13]= {0.071653, Null}

In[14]:= Do[o@"clone"[arrayUnPacked], {100}]; // Timing
Out[14]= {0.593709, Null}

In[15]:= Do[o@"clone"[arrayPacked], {100}]; // Timing
Out[15]= {0.19213, Null}

In[16]:= arrayPacked = Table[RandomReal[], {1000}, {1000}];

In[17]:= Do[o@"clone"[arrayPacked], {100}]; // Timing
Out[17]= {0.077167, Null}

So I guess I agree with your hypothesis that Mathematica somehow obtained a copy (for memorizing ?)before passing to the library.

I also tested in version 9. It seems the behavior is the same as in version 10.

3 hours later…

kirma

06:52

Hmmh... how to implement a stack efficiently on Mma?

Patrick Stevens

07:23

@kirma I use a linked list. To push, {a, stack}. To pop, Last@stack.

`pop := (stack = Last@stack;)

push[n_] := (stack = {n, stack};)

stack = {};` isn't too slow

kirma

Hmm. Sounds good enough I guess.

4 hours later…

Jacob Akkerboom

11:08

Q: In 10.1.0 BenchmarkPlot doesn't work?

Bug introduced in 10.1 and fixed in 10.2.0 In Mathematica 10.1.0 BenchmarkPlot from the GeneralUtilities package doesn't work; copying Taliesin Beynon's initial use of the function on this site: Needs["GeneralUtilities`"] myPosIdx[x_] := <|Thread[x[[#[[All, 1]]]] -> #]|> &@GatherBy[Range@Len...

@Szabolcs what you do you suggest we do with the standardized header in the Q&A above?

Meta

Q: Should I also notify Wolfram about a feature request?

I've just posted a feature request (More efficient method to compute moments of the Johnson $S_B$ distribution). What is the practice, is it usually considered enough or for efficiency I should also post my feature request somewhere on the Wolfram site? PS I've used feature request tag here alth...

discussion feature-request tagging

Murta

12:02

Mathematica is in the 47th position in the Tiobe index!... that is great! Tks to @Rod about this info. I believe this is a record.

Jacob Akkerboom

@Szabolcs because BenchmarkPlot has broken again in 10.3.0 anymore, that is. I have edited the question now.

Ah yes, I see your edit. That makes sense, thanks

Szabolcs

@Jacob The <br> trick ensures that Pickett's report generator will extract the second line too, even tough it is shown on a separate line on the website.

Jacob Akkerboom

12:27

@Szabolcs, it looks as if it is not displaying properly (yet?) in Picketts bugs overview.

Oh I suppose it only searches for the word "fixed", so I guess it wasn't going to show up anyway.

3 hours later…

Patrick Stevens

15:24

@J.M. Always a pleasure to see someone else who uses true ellipses … rather than three dots

Pickett

16:21

@JacobAkkerboom I don't keep that updated, which is why I wrote in large letters "regularly updated" (I shouldn't have written that)

but the code's there for anyone to run.

Jacob Akkerboom

@Pickett Haha, yes, well it is a bug list after all :)

Pickett

16:35

A cron job on Wolfram Cloud + Wolfram data drop would probably work in order to create a timeline of added/fixed bugs, but would take some time to create :/

1 hour later…