Array Package and BLAS for Java

Array Package and BLAS for Java - Demonstration programs

You can use the Java demonstration programs, contained in arraydemo.zip (8KB) to test your installation and to serve as examples in the use of the package.

Installing the demo programs:

unzip the file arraydemo.zip
this will generate 5 Java source files: cfft.java cholesky.java cmatmul.java shallow.java testblas.java
Set your CLASSPATH to include the file array.zip
compile each program using, for example: javac X.java, where X is one of cfft, cholesky, cmatmul, shallow, or testblas
run any of the demo programs: java X
each program will report execution time and speed (in Mflops) and will perform a verification test

The demonstration programs are:

Cholesky factorization (cholesky.java). This program compute the Cholesky factor U of a symmetric positive definite matrix A such that A = U^TU. U is an upper triangular matrix. The problem size is specified in the source file by the parameter n.
Shallow-water simulation (shallow.java). This program is a kernel of a shallow-water simulation program. The problem size is specified in the source file by the parameter n. You can change n, but then the validation parameter no longer applies.
A test of some BLAS methods (testblas.java). This program computes the matrix product C = A x B using three different approaches: (i) a single call to the BLAS dgemm method, (ii) one call to the BLAS dgemv method for each column of the result matrix C, (iii) one call to the BLAS ddot method for each element of the result matrix C. The program computes performance and checks correctness of each approach. By default, all matrices are of size 128 x 128. This can be controlled through command line arguments: testblas -m m -n n -p p, where A is m x n, B is n x p, and C is m x p.
Complex matrix-multiply (cmatmul.java). This program computes the matrix product C = A x B, where A, B, and C are matrices of complex numbers. The default problem size is 100 x 100 and it is controlled by the source code parameter n.
FFT (cfft.java). This program is another example of computations of complex numbers. It computes the two-dimensional FFT of a two-dimensional complex function. The problem size is controlled by the source code parameter n, which must be a power of 2.

Instructions for AIX HPCJ users:

If you want to use the provided array.jll, you can compile the demo programs with:

hpj -o X X.class -nofollow

The Array package executables will then be dynamically loaded from your array.jll file (remember to add it to your CLASSPATH). Refer to your HPCJ documentation for further details on jlls.