The following examples provide an indication of the raw performance of some common numerical kernels in Java.

Dense Linear Algebra Kernels | |
---|---|

SDK 2.0 on 266MHz Pentium II running Win 95 | |

Kernel | Performance |

JNT right-looking LU decomposition, 100x100 | 42 Mflops |

Linpack Benchmark, 100x100 | 25 Mflops |

JNT DAXPY, n=200 | 76 Mflops |

SDK 2.0 on 200MHz Pentium P6 running Win 95 | |

Kernel | Performance |

JNT QR decomposition, 150x150 | 0.25 sec |

Visual Numerics QR decomposition, 150x150 | 0.39 sec |

Sparse Matrix Vector Multiply in Various Environments | |||
---|---|---|---|

Matrix | Java JNT Pentium II (266 MHz) Java (SDK 2.0) |
Fortran/C Sparse BLAS Sun SPARC 20 f77, cc -O |
Fortran/C Sparse BLAS IBM RS/6000 (SP2 node) xlf, xlc -O |

GRE 115 | 33.4 Mflops | 10.2 Mflops | 12 Mflops |

WEST0156 | 33.7 Mflops | 8.0 Mflops | 11 Mflops |

IMPCOL C | 32.6 Mflops | 9.8 Mflops | 11 Mflops |