| Mathematics, MATLAB Version 7.3 (R2006b) Release Notes | |
New features and changes introduced in this version are:
max and min Now Use Magnitudes and Phase Angle for Complex Input
max and min No Longer Return Warning Messages for Inputs with Different Data Types
| Function | Description |
|---|---|
| amd | Interface to the amd algorithm. This interface is similar to that used in symamd, but is typically faster than symamd. |
| bvpxtend | Forms a guess structure for extending the boundary value problem solution |
| ldl | Full ldl factorization and solving for Hermitian matrices |
For complex input, min and max are computed using the magnitude, min(abs(x)) and max(abs(x)) respectively. In the case of equal magnitude elements, the phase angle, min(angle(x)) and max(angle(x)), is now used.
In previous versions, min and max were computed using the magnitude, and in the case of equal magnitude elements, the first element was used. This behavior was indeterministic and not consistent with the sort function.
Code in previous releases that relied on the output of this functionality for the case described above should be updated.
The lu function returns an additional output that helps improve numerical stability of sparse lu factorization.
Upper and lower triangular factors for chol and ldl improve performance for sparse chol. Referencing the lower triangle and returning the lower triangular factor is more memory efficient in the case of sparse chol.
Permutation vectors for lu, luinc, and ldl provide memory savings and, for large data, a noticeable performance improvement. You can now store permutation information in a single 1-by-N vector instead of an n-by-n matrix,
A new two-element threshold for lu and spparms gives you more control over sparse lu and sparse \ behavior.
Extra arguments exported for symbfact allow the return of the lower triangular symbolic factor.
MATLAB 7.3 supports new versions of the following libraries:
| Library Name | Version Supported in MATLAB 7.3 |
|---|---|
| AMD | 2.0 |
| COLAMD | 2.5 |
| CHOLMOD | 1.1 |
| UMFPACK | 5.0 |
The internal storage of sparse arrays on 64-bit systems has changed in the R2006b release. This change should be invisible to most MATLAB users. However, MEX-file programs that run on a 64-bit system and access sparse arrays must be modified and recompiled in order to run on MATLAB Version 7.3. This applies to existing MEX-files and to any new MEX-files that you write.
If you are affected by this change, you will need to take the following steps to make your code compatible with the new sparse array format:
Modify all sparse array declaration statements in MEX-files that are to run on a 64–bit system so that they now use the mwSize and mwIndex types instead of more specific type declarations such as int or INTEGER*4.
Change any MEX code that assigns a negative value to a variable used to hold a sparse array index or size. An example of this would be code that temporarily uses an array index as a flag, assigning a -1 to it to mark the index as unset. Do not try to cast negative int values to mwSize or mwIndex. Instead, change your code to avoid using negative values.
Recompile these MEX-files using the -largeArrayDims option
For a full description of what you will need to do, please read the section Sparse Arrays on 64-bit Systems in the MATLAB External Interfaces release notes.
Existing MEX-files that run on 64-bit systems and access sparse arrays in MATLAB will not operate correctly in MATLAB 7.3 unless the required changes outlined above are made and the files are recompiled with the -largeArrayDims option. New MEX-files that you create must also adhere to these guidelines.
The version of FFTW used in MATLAB has been upgraded from 3.0.2 to 3.1.1 Please read the Compatibility Considerations section, below.
Other changes in MATLAB FFTW support are:
The default planner method is now estimate. Previously, hybrid was the default.
There are new syntaxes for importing from and exporting to the library's internal single- and double-precision wisdom databases. These are as follows and are documented in the reference page for the MATLAB fftw function:
str = fftw('dwisdom')
str = fftw('swisdom')
fftw('dwisdom', str)
fftw('swisdom', str)
FFTW version 3.1.1 does not support wisdom produced by previous versions of the FFTW library. For this reason, FFTW wisdom that has been exported from a previous version of MATLAB cannot be imported successfully in MATLAB R2006b. Trying to import "old" wisdom results in a warning.
The betacore function generates a warning message in R2006b but completes successfully. This function will be made obsolete in a future release.
You should remove all instances of the betacore function from your M-file program code. This function will not be supported in a future release of MATLAB.
The following uses of MATLAB functions are obsolete as of this release. Attempts to use this functionality in the R2006a release generate a warning message but complete successfully. In R2006b, these operations fail and generate an error message.
The following functions in their entirety: bvpval, quad8, table1, table8, bessela
The sort function, when used with complex integer inputs
The gt, ge, lt, and le functions, when used with complex integer inputs
The beta function, when used with 3 inputs
The gamma function, when used with 2 inputs
The opts.cheb and opts.stagtol fields in the options structure of eigs
You will need to remove all instances that reference these functions at this time. These functions are no longer supported in MATLAB.
In MATLAB version 7.0 (Release 14), the functions max and min were changed to return results of a different data type than in previous releases. This behavior is described in max and min Now Have Restrictions on Inputs of Different Data Types. This change in behavior produced warning messages to assist you with diagnosing any resulting issues. In R2006b, these warning messages no longer exist.
The warning messages for mixed-type inputs to the functions max and min are no longer produced. Turning warning messages on will no longer display messages for this behavior, and you will no longer be able to depend on the messages for the diagnosis of problems.
| Desktop Tools and Development Environment, MATLAB Version 7.3 (R2006b) | Data Analysis, MATLAB Version 7.3 (R2006b) | |
| © 1984-2008- The MathWorks, Inc. - Site Help - Patents - Trademarks - Privacy Policy - Preventing Piracy - RSS |