There is some discussion about this going on at comp.soft-sys.matlab. It can be accessed via http://www.mathworks.com/matlabcentral/newsreader/. The shipping version of X11.app in Leopard is rife with problems, many of which have been addressed with the quality work of Ben Byer. These changes can be read about on the X11 mailing list of Apple's and in various spots around the web. By and large, they've solved my problems with MATLAB. Not sure about Simulink specifically (are the problems java related, possibly?), but the newsgroup seems to have some info about that as well.
I expected to see some speed increase with the underlying optimization of Leopard but instead Matlab 2007a felt slower, and running Bench on my Macbook (2.0GHz, 2G RAM) showed it's just above a G4.
There is a problem with the fortran interface to the mex/mx functions
I'm able to build a mexmaci, but when run
dyld: lazy symbol binding failed: Symbol not found:
_mxgetnumberofelements_
Checking the Matlab libraries linked to gives
nm libmx.dylib
[...]
000026f8 T _mxGetNumberOfElements
0003ba18 S _mxGetNumberOfElements.eh
[...]
00012428 T _mxgetnumberofelements700_
000417a8 S _mxgetnumberofelements700_.eh
000129f2 T _mxgetnumberofelements730_
00000000 A _mxgetnumberofelements730_.eh
[...]
nm libmex.dylib
[...]
U _mxGetNumberOfElements
[...]
nm libmat.dylib
[...]
U _mxGetNumberOfElements
[...]
It looks like there are two problems. The first is that the symbol is exported as lowercase and it has a trailing underscore. Both things that are common in FORTRAN compiled code. If it's possible to pass arguments at compile time you may have luck with:
-fno-underscoring
-fcase-preserve (not sure if this one works with gfortran or not)
Unfortunately, -fcase-preserve isn't supported by gfortran.
Mathworks advices using g95 compiled on Tiger (doesn't work for me though) ... but have implied they hope to resolve this problem with a coming release. However, no word if that release will be 2008a or later.
The Parallel Computing Tool Box does not work reliably under Leopard. I have contacted the technical support department at MathWorks and here is their final response:
Hello Kent,
I am writing in reference to your Service Request # 1-69UZQJ regarding 'Only 3 workers open'.
After getting in touch with you the last time, we were still looking into the issue. This appears to be an issue in Apple OS X 10.5, and has been reported to Apple (Problem ID 5938846).
Currently there are no workarounds for this. Our developers as stated above have reported the bug to Apple and will try to resolve the issue once they have a solution from Apple.
Thanks for keeping me up-to-date.
If you need further assistance regarding this issue, please reply to this email preserving the THREAD ID listed below. If you have a new technical support question, please submit a new request here:
http://www.mathworks.com/contact_TS.html
Sincerely,
Mayank Advani
Application Support Engineer
Technical Support Department
The MathWorks, Inc.
Phone: (508) 647-7000 option 5
This problem remains unresolved as of July 17th, 2008.
yeah, I have R2007a installed on my 10.5 iMac, and the same version installed in an XP vmware image on the same machine. The version running in XP through the VM is at least twice as fast for most operations, and the GUI is much more responsive, despite the overhead involved in running through VMware etc...
Simulink does not work properly
X11 quits when you try to drag items from one window to another. Apparently Mathworks have solved this problem with a patch:
http://www.mathworks.com/support/solutions/data/1-4N26RI.html?solution=1-4N26RI
However, the link to the actual patch seems to be broke. Anyone successfully downloaded and unzipped the patch?
MatLab works well in Leopard
I've been using MatLab 2007a with the Developer versions of Leopard for months and have had no problems. I have not tested simulink.
Simulink does not work properly
Same problem here, tried to download the patch form Windows and Linux too, still no luck. Any ideas..? This really sucks...
There is some discussion
There is some discussion about this going on at comp.soft-sys.matlab. It can be accessed via http://www.mathworks.com/matlabcentral/newsreader/. The shipping version of X11.app in Leopard is rife with problems, many of which have been addressed with the quality work of Ben Byer. These changes can be read about on the X11 mailing list of Apple's and in various spots around the web. By and large, they've solved my problems with MATLAB. Not sure about Simulink specifically (are the problems java related, possibly?), but the newsgroup seems to have some info about that as well.
Slow
I expected to see some speed increase with the underlying optimization of Leopard but instead Matlab 2007a felt slower, and running Bench on my Macbook (2.0GHz, 2G RAM) showed it's just above a G4.
A bit disappointed, really.
Patch is now available
The download from mathworks.com is now working.
I installed the patch and it all works fine now.
Anyone successful with mex/matlab on Leopard?
Although I had no problem mexing on my PPC based Macs running Tiger, I've not been able to do the same with either my PPC or Intel Macs.
Before going into any detail, I thought I'd ask if anyone has been successful mexing using Leopard.
Ben
mex and the fortran interface
There is a problem with the fortran interface to the mex/mx functions
I'm able to build a mexmaci, but when run
dyld: lazy symbol binding failed: Symbol not found:
_mxgetnumberofelements_
Checking the Matlab libraries linked to gives
nm libmx.dylib
[...]
000026f8 T _mxGetNumberOfElements
0003ba18 S _mxGetNumberOfElements.eh
[...]
00012428 T _mxgetnumberofelements700_
000417a8 S _mxgetnumberofelements700_.eh
000129f2 T _mxgetnumberofelements730_
00000000 A _mxgetnumberofelements730_.eh
[...]
nm libmex.dylib
[...]
U _mxGetNumberOfElements
[...]
nm libmat.dylib
[...]
U _mxGetNumberOfElements
[...]
The symbol really isn't there.
Re: MATLAB 2007 in Leopard
Hi,
It looks like there are two problems. The first is that the symbol is exported as lowercase and it has a trailing underscore. Both things that are common in FORTRAN compiled code. If it's possible to pass arguments at compile time you may have luck with:
-fno-underscoring
-fcase-preserve (not sure if this one works with gfortran or not)
Regards,
Dave
Re: Re; MATLAB 2007 in Leopard
Thanks Dave,
Unfortunately, -fcase-preserve isn't supported by gfortran.
Mathworks advices using g95 compiled on Tiger (doesn't work for me though) ... but have implied they hope to resolve this problem with a coming release. However, no word if that release will be 2008a or later.
Regards,
Ben
Mex files for Matlab
I just realized I hadn't close the loop here.
The problem was the Mathworks changed their mex interface to include a cpp that is case sensitive.
All lower, all upper, and a SpecificCamelCase are supported.
Ben Abbott
small deeds > BIG INTENTIONS
Parallel Computing Tool Box
The Parallel Computing Tool Box does not work reliably under Leopard. I have contacted the technical support department at MathWorks and here is their final response:
Hello Kent,
I am writing in reference to your Service Request # 1-69UZQJ regarding 'Only 3 workers open'.
After getting in touch with you the last time, we were still looking into the issue. This appears to be an issue in Apple OS X 10.5, and has been reported to Apple (Problem ID 5938846).
Currently there are no workarounds for this. Our developers as stated above have reported the bug to Apple and will try to resolve the issue once they have a solution from Apple.
Thanks for keeping me up-to-date.
If you need further assistance regarding this issue, please reply to this email preserving the THREAD ID listed below. If you have a new technical support question, please submit a new request here:
http://www.mathworks.com/contact_TS.html
Sincerely,
Mayank Advani
Application Support Engineer
Technical Support Department
The MathWorks, Inc.
Phone: (508) 647-7000 option 5
This problem remains unresolved as of July 17th, 2008.
verrry slow
yeah, I have R2007a installed on my 10.5 iMac, and the same version installed in an XP vmware image on the same machine. The version running in XP through the VM is at least twice as fast for most operations, and the GUI is much more responsive, despite the overhead involved in running through VMware etc...
madness!