Rank: 1167 based on 115 downloads (last 30 days) and 4 files submitted
photo

Yuval

E-mail

Personal Profile:

 

Watch this Author's files

 

Files Posted by Yuval View all
Updated   File Tags Downloads
(last 30 days)
Comments Rating
08 Jul 2013 Screenshot MMX - Multithreaded matrix operations on N-D matrices N-D matrix operations. Like Boettcher's ndfun or Tursa's mtimesx, only faster. Author: Yuval multiply, backslash, multithreaded, multiple multiplicati..., multiprod 35 19
  • 5.0
5.0 | 8 ratings
10 Oct 2012 Screenshot Catenary - hanging rope between two points Computes the catenary shape (hanging rope) of a given length between two given points. Author: Yuval catenary, hanging rope 25 0
  • 5.0
5.0 | 1 rating
02 Nov 2010 Screenshot CapSim - the MATLAB physics engine CapSim is a general-topology planar physics simulator designed for academic purposes. Author: Yuval simulation, physics, collision detection, contact, friction 27 3
  • 4.75
4.8 | 4 ratings
04 Aug 2008 LCP / MCP solver (Newton-based) Solve the Linear Complementarity Problem. Author: Yuval linear algebra, lcp, mcp, mathematics, newtonbased, linear complementarit... 28 1
  • 5.0
5.0 | 1 rating
Comments and Ratings by Yuval View all
Updated File Comments Rating
08 Oct 2013 MMX - Multithreaded matrix operations on N-D matrices N-D matrix operations. Like Boettcher's ndfun or Tursa's mtimesx, only faster. Author: Yuval

Thanks Syed. Unfortunately I don't have time to add that feature. Just use the identity

(B/A)' = (A'\B')

Remember that transposition is a very cheap operation. You can avoid loops by using
my_T = @(x) permute(x,[2 1 3 4])

08 Jul 2013 MMX - Multithreaded matrix operations on N-D matrices N-D matrix operations. Like Boettcher's ndfun or Tursa's mtimesx, only faster. Author: Yuval

Tamar, I realized what the problem is. LCC is an ANSI C compiler. MMX uses some features of C++ and therefore cannot be compiled with LCC. Both Microsoft and Intel offer free versions of their compilers which matlab recognizes (mex -setup). Try using these.

06 Jul 2013 MMX - Multithreaded matrix operations on N-D matrices N-D matrix operations. Like Boettcher's ndfun or Tursa's mtimesx, only faster. Author: Yuval

Tamar,

I believe you found a bug. Thank you. The declaration of the variable temp in line 151 is in the middle of the function code. This is standard in C++ but not in C, where declarations are supposed to come before statements. I suppose my compiler (VS) was a little too lenient.

Move line 151 to 129 and you should be fine.

04 Jun 2013 MMX - Multithreaded matrix operations on N-D matrices N-D matrix operations. Like Boettcher's ndfun or Tursa's mtimesx, only faster. Author: Yuval

Alexandros,

Thank you so much for your feedback. This is not a bug. As described in the documentation, for rank deficient matrices mmx gives the least-norm (pseudoinverse) solution rather than the maximally-sparse solution. You didn't specify how your matrices were generated but clearly they are not 'in general position'. Try using random A's and you will see no discrepancy (to machine precision). Alternatively, compare mmx's output to pinv(A)*B.

12 Apr 2013 MMX - Multithreaded matrix operations on N-D matrices N-D matrix operations. Like Boettcher's ndfun or Tursa's mtimesx, only faster. Author: Yuval

Thanks Adam!
Please spread the word.
Adding complex number support should be pretty easy.
If I see that mmx gets more traction in the community and that more people want it i'd be happy to do it myself.

As it stands it seems that it's buried in a dark corner of file exchange...

Comments and Ratings on Yuval 's Files View all
Updated File Comment by Comments Rating
22 Nov 2014 Catenary - hanging rope between two points Computes the catenary shape (hanging rope) of a given length between two given points. Author: Yuval sam

28 Jul 2014 MMX - Multithreaded matrix operations on N-D matrices N-D matrix operations. Like Boettcher's ndfun or Tursa's mtimesx, only faster. Author: Yuval Robbert

Dear Yuval,
Thanks for the great tool, I would also very much appreciate complex number support.

10 Mar 2014 MMX - Multithreaded matrix operations on N-D matrices N-D matrix operations. Like Boettcher's ndfun or Tursa's mtimesx, only faster. Author: Yuval Evan

I was able to compile on OSX 10.8 on 2013b using XCode 5.0.2 by changing link_dir from '/extern/lib/maci64' to 'bin/maci64.' The comments say that the 'mmx_mkl_muli' build uses the BLAS/LAPACK libraries that come with Matlab, but it seems as though it's still trying to use intel mkl libraries. I also had to stop the program from cd-ing out of the mmx package directory so that it wouldn't fail to find mmx.ccp. Works now though!

09 Mar 2014 MMX - Multithreaded matrix operations on N-D matrices N-D matrix operations. Like Boettcher's ndfun or Tursa's mtimesx, only faster. Author: Yuval Evan

Hi Yuval,

Sorry if this is a naive question, but I'm trying to follow the instructions to download the Intel MKL, and it appears to be quite an expensive software package. Is there another configuration I can use? Thanks!

08 Oct 2013 MMX - Multithreaded matrix operations on N-D matrices N-D matrix operations. Like Boettcher's ndfun or Tursa's mtimesx, only faster. Author: Yuval Yuval

Thanks Syed. Unfortunately I don't have time to add that feature. Just use the identity

(B/A)' = (A'\B')

Remember that transposition is a very cheap operation. You can avoid loops by using
my_T = @(x) permute(x,[2 1 3 4])

Contact us