@acl so jealous an NSolve regression is making a lot of my code so slow it is not usable at the moment. Back to v7 for some work. Need to figure out if I can make it run faster. Already miss the new v8 features!
@Gabriel it probably uses LAPACK and ARPACK internally so I doubt that would make much difference (except perhaps in passing stuff around). but I don't really know of course
@acl A simple question about OS X: how do I search the Library/Mathematica directory for *cuda* and get the paths of the results, not their names, in a nicely browseable list? Windows Explorer can do this. Can I use the Finder, or do I need to drop down to Terminal
Hi @Arnoud! Have you heard any complaints about CUDALink not working in 9 if 8 is also installed (and CUDALink has also been used in 8)?
Too many try to use Matlab for symbolics, and have hard time. Wonder why they suffer like this when they can use Mathematica for that. Poor person here having hard time with 3 by 3 symbolic matrix eigenvalues stackoverflow.com/questions/13624360/…
@Szabolcs You can do it via finder... Just go to the directory and type "cuda" in the search bar. Then in the bar underneath that says "Search: This Mac, Mathematica, Shared", click the second one
I don't know why it didn't yesterday... I didn't change anything.
It's nice to have a good graphics card and machine... things are so smooth now. With my older laptop, I had to reduce the size to 100x100 and even then, it was choppy
Honestly, I should one day get to read attentively the rules of stackexchange. I learned to behave, and some rules, by experience and by negative reinforcement of the "serious people"
What I really meant was that it isn't a strict rule book that you read from cover to cover... we aren't attorneys/judges who need to know the law thoroughly and apply it by the letter. You mostly "learn" the rules around here by participating regularly in the site and you've been around since day 1, so you're pretty fine there. There'll always be something that no one knows and we either cook up something on the spot or search Meta Stack Overflow long and hard :P
fyi, just finished making table of all symbols in all contexts for version 9. useful, to allow one to search for a symbol in all of version 9 from one page. 12000.org/my_notes/compare_mathematica/V9/main.html over 7,000 symbols. Added Developer` and Internal` contexts also and all standard packages
seems like most people don't use the template autocompletion stuff, since I seem to be one of the few who noticed (to me it was immediately obvious because I use it constantly)
also is this a duplicate or am I missing something?
@halirutan Thanks. Re: formatter - I did some work there, but did not commit yet, it is not stable. I probably could do some quick update to make the tabification customizable, so that we can start using it for SE. That should be easy to do.
@halirutan Frankly, my current priorities are to launch the Github-based code exchange stuff first, for a number of reasons. I think that we really need a mechanism to start exhanging code, that seems more important than refining the formatter, which is a next thing.
@halirutan Yes, I will make a branch, planned to do that on weekend anyway. Things were a bit hectic recently
@halirutan Re: things to do: for formatter, it needs more than just bug-fixes, it needs serious refactoring. I want to make a DSL so that people will be able to write their own formatting templates
@halirutan Re: code-echange I am working on a set of tools which would allow people to publish code easily on a small scale from within M, and have versioning based on Github gists without the need to actually learn Git
@halirutan Once I do the core refactoring for the formatter, it will be much easier to collaborate on it. Right now, while the core formatter design seems fine, it is stil half-baked
@halirutan Re: box structure no, I already have teh pure string version, although it does go to the boxes. But boxes are harder, so once boxes are fine, strings are more or less automatic
@halirutan Re: parser - no, I have not. Couldyou give a link?
@ArnoudBuzing Arnoud, do you know whether one of the guys from WRI which are often around use Linux? As I see this, we have you, Andy, Brett, John the Futz Fultz, Ruebenkoe
Is there some function that lets me find all (at least most) roots in a bounded region of a real-valued function of several variables besides NSolve? (NSolve doesn't even try, it stays unevaluated.)
yea, FindRoot has a bit odd error mesage, FindRoot[x+y, {x,y}] gives error about number of equations and variables while: FindRoot[{x + y, 0.}, {{x, 0}, {y, 0}}] runs just fine
seems like the sane default would be to just pad with zeros if the algorithms need it
Perhaps the only reason for me to try to learn emacs...
I do like vimlatex though... I'm fairly comfortable with visualizing the layout and equations in my head, and as you said, having a large monitor helps
@rm-rf judging from online and offline discussions with lots of people, I must be the only person in the world who's incapable of visualising my equations from latex
@Gabriel Yeah, that's what I've been using for quite a while. The only drawback of that is a general vim drawback in that you can't do anything else while it compiles
First of all, I'd like to apologize to anyone affected. We had an extensive prerelease trial period where this problem did not turn up and we are still examining why this is the case.
We're still debating to pull the Windows version from the portal until we can provide a new full version with this fix incorporated.
For this it is important to understand (and this community can help a lot) how many people were affected and how many people (Windows users) were not affected by this.
I saw almost exactly the same problem when I first installed Mathematica version 8 on a machine that was already running version 7. I don't remember what I did to fix it back then (should have written it down), but it involved a lot of uninstallation, re-installation, reboots, and miscellaneous ...
@Gabriel We're making that decision today. But the more information we have about this problem, the better it is. At the moment we have precise information on who was affected, but no information on Windows users that were unaffected.
If you have any such information, I kindly request to mention it here or email me (arnoudb@wolfram.com).
Also, if you were affected it would be very useful to know which versions of Mathematica you had installed previously. I think most of you have either Mathematica 8 or Mathematica 7, but if older versions existed on the machine that would be good to know.
So far we have reason to believe all fresh installations of Mathematica 9 are ok and that it is the interaction with previous installations that are causing the problem.
@ArnoudBuzing I installed 9 on a Windows 7 x64 machine where I have 8 installed, but I didn't experience any BSOD... Is there a test I can do to make it happen (if it's not too dangerous :).
@ArnoudBuzing OK. Well, no BSODs for me. I had & have 8 installed with all default preferences, except I did enable version-specific preferences before installing 9.
What does CUDAFluidDynamics[] really do? From the docs, it looks like a demo of fluid mechanics using CUDA, but if I call it on my machine, I don't get the neat vortices that is shown in the docs. Changing options doesn't help either...
Would there be any reason that having both v8 and v9 installed could cause problems? Seems like v9 has font issues with the wolfram alpha integration now :-(
but I need to have v8 installed as a bunch of my old notebooks don't work correctly in v9
@Gabriel Mathematica 9 uses a font character for this. You can see the difference between Mathematica 8 and 9 in this screen shot: i.imgur.com/7Jbl8.png
@ArnoudBuzing okay so it suggest I still have some left over font problems ... I installed v8 after v9 wonder if it overwrote one of the new font packages
Once I finish some work today ... I will do a clean install again ... I can send you guys an email to see if the issue still persists. I will do it in the same order to see if I get the font in v9 and then lose it after I install v8
@AndrewMoylan It looks like the letters are actually bigger (wider) in the second screenshot. (When you increase the font size in Windows, sometimes they only grow horizontally by a pixel before they grow vertically.)
But it's very subtle, difficult to tell when looking at a single character.
@Szabolcs I suspected it might be the character size too, but after looking at individual pixels with anti-aliasing off, it appears to me to be an explicit decision to increase the spacing:
In 9 they seem to be 13 point (not shown with a checkbox in the menu), while in 8 they're 12 point. If you set them to 12 point in 9 too, they'll look the same.
How amazing will some of the antialiased smoothed 3D graphics look?! I'm thinking of the nice exampels like Plot3D[Im[ArcSin[(x + I y)^4]], {x, -2, 2}, {y, -2, 2}, Mesh -> None, PlotStyle -> Directive[Yellow, Specularity[White, 20], Opacity[0.8]], ExclusionsStyle -> {None, Red}]
Of course, when HiDPI comes, the default PlotPoints may need to be revised ... makes sense to do that in parallel with multithreading all those function evaluations perhaps.
@Szabolcs well at least they answered; I asked why autocomplete doesn't work and got no answer. I guess they're overloaded anyway (or maybe I should have switched from the trial to the premier license to elicit a response)
Thanks @acl, that's why I was asking. Frankly I do not see any good reason for downloading the trial (which is limited in Export,ect anyway). Better to get directly the real McCoy :-)
CorrelationFunction[ARProcess[{a}, s^2], t] /. t -> -10
and, from the Details:
CorrelationFunction of the process proc is the CovarianceFunction c normalized by the outer product of the standard deviation function [Sigma] at times s and t:
and in CovarianceFunction, Expectation[(x[s]-[Mu][s])(x[t]-[Mu][t])]
Perhaps there's some idea of processes starting at time 0, who knows. But it is at least suspicious that in the Scope part of the docs, that same example gives a^Abs@h
@halirutan I just realized that one problem with a pre-generated palette is that if it is made using 9, 8 will complain that the notebook might be incompatible.
No BSOD here, but it is a little unstable. Mathematica crashed 2 or 3 times during the day; sometimes freezes for 5 to 10 seconds (like program not responding). Is this familiar to any of you? (XP SP3)
@rm-rf CudaFluidDynamics is just a toy demo of some sort of free mesh CFD (I imagine like SPH). You are suposed to see a similar image to the one of the documentation center, at first with all points static; when you move your mouse over it (don't recall if clicking) the points start moving
I wished we had a cuda code like dualSPHysics or similar (fluid inlet would be great; not the periodic stuff) integrated into mathematica. Also, mathematica as an integration tool would be quite good also to manage inporting 3D models, launching meshing tools (for fluids; not the tetgen already linked), generating config files for codes like in OpenFoam, and eventualy analysing the results (although there are good tools on the market for that)
@P.Fonseca I don't think that is related. It would be good to report if you are seeing something specific ... It would be good if Technical Support can help diagnose this.
@P.Fonseca If you do CUDAFluidDynamics[NumberOfParticles -> 17000], then it automatically starts moving (I wonder if there are preset forcings). Similar results for 16, 17, 18k (and inbetween numbers) and at 19k, it breaks down... 15k points seem to be the upper limit for static points. I can't pump it up to 20k. If I set to 20k, I can't get it to work again (for a lower number) without restarting the kerne;
@rm-rf I don't have CUDA here, but will let you know later
@ArnoudBuzing I still can't repeat the problem in a systematic way. It has happened several times when on the documentation center first page, and it seems to be more related with dynamics. But things like shift return on the simplest Plot command, and waiting for more than 20 seconds have happened (I actually canceled before the "end"). And yet, launching exactly the same input 5 seconds later, outputs instantly. So, I can only classify as unstable, but can't help much on finding the reason.
@rm-rf In this answer of yours, if instead of commenting out you just move the v9 key to the second line and you can run both without changing the file in between, can you confirm and if it works for you update the post?
there was a question about citations so I was wandering if anyone knows if the citation support in V9 is the same on Macs and Windows. In v8 Macs got the documentation but didn't get the functionality.
@ssch Ah... hmm. I haven't tried it one at a time. I think I tried it simultaneously. I'll give it a go again later in the day, after saving all my work :D
@ssch I don't know, but I find it even weird, that v9 does not recognize when CUDA is installed for version 8. It just returns False for CUDAQ and messages for all other functions.
@Searke it was a question not a statement. In V8 an old version of author tools exists but you won't get a hit if you type "AuthorTools" into the help guide, i.e. undocumented. Whereas in V9: reference.wolfram.com/mathematica/AuthorTools/guide/…
...so i am assuming the package has been updated ???
...but now noticing a lot of "page not found" pages when I click on links so maybe this documentation appears by accident -- it got captured in some sort of automatic build of the docs when it actually is supposed to be excluded (if so this remains disappointing)