@KCPower If your code is only going to display plots of the results then it may be quicker to strive for less accuracy in the numeric integrations. For example you could try evalf(Int( ..., epsilon =1e-5 )) instead of  int(..., numeric).  You could also try forcing a particular  method, and I have a hazy recollection that method=_d01akc might do faster as long as the accuracy is not too strict and the max intervals is high enough for convergence at the higher frequencies. Fiddling with this is what I lack time for now, sorry.

Time and time again it has turned out that the most effective way to speed up such code is focus on improving the efficiency of the serial (unparalellized) version.

I see several serious innefficiencies, offhand. I expect at least a few orders of ten speedup are possible. Unfortunately I have almost no time for it right now.

@taro It's the layout that I mostly have problems with. I described some of the issues in an earlier comment above. Another problem is that the central panel is too narrow. The Editor window is also too narrow.

Also, the layout changes are mostly just a re-skinning of the layout -- a degradation IMO, and nothing I've ever seen anyone here request. But several bugfixes and improvements that have been asked for previously (sometime multiple times) have generally not been made.

Inappropriate or misspelled tags cannot be deleted.

The Search is weak, in multiple ways.

Here is some Math typeset by the "Maple Math" button in the Editor, which inlines a generated image with coarse rendering and other problems: Even five years ago there were better solutions for math on web pages.

2D plots in uploaded worksheets still get gridlines added, unless the person explicitly adds the option to turn it off in the plotting command. Is that bug still present in MapleNet, or is this site still using a version that's 10 years old?

Plots in uploaded worksheets look very poor and coarse (innappropriate jpeg export?), and lose modern lighting effects that make them look good in Maple itself.

@eclark Sorry I don't know details of size restrictions on Maple Cloud uploads. I don't think that it allows uploading a sheet compressed (zip, gzip, etc).

There may be a small chance that the plot structure could be made smaller (eg. float[8] data in GRID or MESH instead of listlists) but we'd have to see it, or code that made it preferably. Are the frames from displayed Graphs?

If your target audience is small then they could try the free Maple Player. It allows manual rotation of a free-standing inlined 3D plot. (But it also allows you to even set things up so that others can execute some code behind embedded components, though the code can't be entirely secured from inspection.)

@tsunamiBTP I cleared off all cookies and things and (for that reason or some other) I can usually now upload, yes, also using Chrome.

However sometimes my Windows 7 Chrome cannot even submit a Comment/Answer, at all, even without uploaded attachment. The Submit button just does nothing. This doesn't always go wrong, though. It dosn't happen on my up-to-date Firefox on Windows 7 though, or on my Chrome on Ubuntu.

I really find it hard on my eyes to read anything here, in the new format. Perhaps I'll have to stop.

@mmcdara  I think that the responders here all understood the math, up front. You asked why it went wrong, and so I tried to explain how it does so. I submitted a bug report earlier yesterday.

@Carl Love Quite correct. 

Quite a few commands in the plots package accept strings as the (rhs) value for keyword options. But not this one.

The (lhs) keywords themselves can also benefit from using the quote-protected global name. Eg,

':-fieldstrength'=':-log'

The colon-minus is to prevent accidentally picking up a local (or rebinding), and the uneval quotes are to protect against an assigned value. (In the current example :-log is protected, so the uneval quotes are less pertinent for the rhs value. But this is still the more robust way to write "production code").

@Christopher2222 You are missing at least these (for which I have only some readily available to run and query kernelopts(version) ).

I used 64bit Linux to make the following queries, and corroborated all on Solaris.

MapleV 5.1  Jan 11 1998? (5.51 in your terminology, on 1998.03? )
6.01
6.02
7.01
8.01
9.01
9.02   
9.03        Oct  1 2003 Build ID 141050
9.04   
9.50        Apr  7 2004 Build ID 155251 (= 2004.37 ?)
9.51        Aug  9 2004 Build ID 163356 (= 2004.83 ?)
9.52   
16.02       Nov 18 2012 Build ID 788210
16.02a                  Build ID 788570
17.02       Sep  5 2013 Build ID 872941

@asa12 You seem to have failed to notice the following effect of your initial substitution of `a` for `a(t)`.

subs( a(t)=a, diff(a(t), t) ):

> %;

                            0

Would you care to give us your defn of Lerror?

You'll probably want to change `sum` to `add`, and `int` to `Int` (and throw in a wrapping calls to `evalf` inside `c` and `Fourierf`. And it seems like you'll want `c` to be a proc...end proc rather than just an operator, so that you can give it option remember and avoid needless recomputation.

acer

@Kitonum You're welcome.

I considered leaving the Sign(P) call out of Sign(P)*IntegralPart(P), while also leaving IntegralPart=0 out of the conditional, which would be simpler. I left it the way it is because it reminds me of why the bug can occur.

Of course, even if one lacked the newer NumberTheory routine the individual strings could still be computed (with some additonal code) and stuck together in the manner shown using Typesetting.

@Preben Alsholm I would simply delete this as a Duplicate, except that it now has a response.

Have you considered making the default File->New sheet to be Worksheet instead of Document, and the default input mode to be 1D Maple Notation instead of 2D Math?

acer

There is far too much wasted space. For example, the margins between the left and right edges and any text content (especially in Replies/Answerws). The switch to what appears to be double-spacing is another example. Dead space at the borders of the rectangular image boxes is another example. This problem is severe.

Having the author's member-icon and details appear inside the body of the reply (instead of in the title-box) adds to the wasted space problem.

The Green-Arrow upload doesn't work at all in my Linux Chrome. (I tried inlining a sheet, or just link insertion, with and without Cancel, etc.)

I have to struggle to read the text on the grey background boxes.

On my android phone the whole experience makes the site unusable and unreadable, even if I turn it sideways. The display does not resize properyly and content bleeds offscreen to the right. With my phone upright I often see paragraphs of text formatted in with one word per line (yes, sentences displayed in a huge column one word wide).

acer

@artfin Could you say how much faster the full 50-frame run was, using what I posted above vs the original, for you?

Do you have a rough idea of how fast (relatively speaking) you'd want it?

If you want it very fast then I suggest devising a single procedure that computes the results needed for a whole frame, the numerical computations of which could all be done under evalhf. 

That should be the only proc which takes argument `t`. Better still cut out almost all the function calls and use local variables for the intermediate results.