/*
 * $Id: benchmark.dox,v 1.8 2004/04/10 04:58:35 opetzold Exp $
 */

/**

  \page benchmark Benchmark

  Prior to version 0.8.0, tvmet had benchmark results listed here. Benchmarking
  isn't as easy as had been thought. The compiler's optimizer is very tricky -
  even more than I am. 8( To outwit the optimizer takes time - I would rather
  put the time into tvmet, itself.

  Laurent Plagne has created some benchmarks using tvmet and other libraries,
  please have a look at
  <a href="http://www.opencascade.org/upload/87">opencascade</a>.

  The following benchmarks are done on my
  Intel(R) Pentium(R) 4 CPU 2.40GHz
  Linux box:
  - \ref bench_axpy
  - \ref bench_matrix_matrix
  - \ref bench_matrix_vector
  - \ref bench_aat
  - \ref bench_ata

  using a patched version of BTL 20030124. The patch is necessary to be able
  to compile tvmet due to the high inline level used here. The patch is
  distributed with the package.

  \section bench_axpy axpy
  \image html axpy.png "axpy"

  \section bench_matrix_matrix Matrix-Matrix
  \image html matrix_matrix.png "Matrix-Matrix"

  \section bench_matrix_vector Matrix-Vector
  \image html matrix_vector.png "Matrix-Vector"

  \section bench_aat MMt
  \image html aat.png "M M^t"

  \section bench_ata MtM
  \image html ata.png "M^t M"

*/


// Local Variables:
// mode:c++
// End:
//  LocalWords:  BTL GHz Plagne