| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 876.1 | Tons of programs available | HPSTEK::XIA | | Sat May 21 1988 18:56 | 24 |
| There are tons of programs in the market for solving linear systems
since solving linear system is at the core of Numerical Analysis.
The popular packages are IMSL library, Linpack, and PCGpack. There
are essentially 2 methods of solving a linear system. Direct methods
and the Relaxation methods. Gaussian elimination LU decomposition
fall in to the first category. The Relaxation methods are iterative
methods that is they contruct a contraction map which converges
to the solution. Gauss-Seidel and Gauss-Jacobi methods are the
popular ones. There is also this the conjugate gradient method
(CG). PCGpack is the one that uses the CG method.
There are also the sparse matrix packages. These packages reconize
that most matrix equations have tons of 0's in it. The sparse matrix
packages takes advantage of that to reduce the amount of work and
the amount of memory involved in solving the linear system. The
above metioned packages do have the sparse matrix option (I believe)
Markowicz and to certain extend Barry were the big players in the
developmentof sparse matrix algorithm.
Hope that helps.
Eugene
P.S. There are matrix packages all over DEC. However, if you opt
to do it yourself, I can recommend some books. On the other hand,
if you want to solve large matrices especially when they are sparse,
I strongly recommend you to use one of the existing packages since
they are optimized.
|
| 876.2 | Where can I find these programs? | JRDV06::FUJITA | | Mon May 23 1988 08:31 | 19 |
| Hello, Eugene.
Thank you very much for your information. But where can I find a
program package which solves n-dimentional linear simultaneous
equation with Gauss method (LU decomposition)? I would like to run
the simulation program (it means the matrix size is rather large)
written by a professor of Tokyo Institute of Technology. It is using
other company's computer library. We should re-write the portion where
this library is called. If possible, I would like to install the
matrix package including Gaussian elimination LU decomposition
method through E-net. The document or something describing how to
call the library is also necessary. I have searched this note with
the keyword "matrix", but I could not find proper information.
P.S. If the popular packages are available, please inform me on
it, too.
Regards,
T.Fujita
|
| 876.3 | math software from SLATEC | CTCADM::ROTH | If you plant ice you'll harvest wind | Mon May 23 1988 13:11 | 9 |
| I have LINPACK (and a bunch of other stuff) on-line, but not the
documentation. The users guide is readily available from SIAM
on the programs though.
Look on RUSSEL::DUA1:[SLATEC]SLATEC.TLB, and the other programs
in that directory. This text library contains most of the useful
mathematical software you'll need.
- Jim
|
| 876.4 | Array or Orthogonal? | HPSTEK::XIA | | Mon May 23 1988 16:01 | 15 |
| Re. .2 .3
One important side note: Since your matix is large, you might want
to use a sparse matrix package. However, there are many ways one
can represent a sparse matrix. In Fortran, this is done by keeping
an extra index. In C or Pascal where pointers and records/struct
are available, it is usually done with an orthogonal representation.
Linpack PCGpack and IMSL library are all written in FORTRAN (I
believe). So if your matrix is in orthogonal form, you can either
convert it to an array form (very inefficient) or you can use a
package that deals directly with orthogonal representation. I believe
the matrix package in SPICE3 is of this nature (though I am not
absolutely sure). You might be able to get a package from Berkeley
for free by contacting [email protected] .
Eugene
|
| 876.5 | | SHOREY::SHOREY | a legend in his own mind... | Tue Jun 07 1988 12:19 | 15 |
| hello,
i'm also in need of a linear programming package, something
that can be linked (?) with other code and called from that code.
i'm writing a decision support system as a school project this
summer and need this for my optimization model.
i'd prefer not to spend the entire semester on linear programming,
bit i'd be grateful for anything anybody out there might have that
i could use.
thanks,
brian
|