LINPACK Benchmark Ax=b

• 100x100: Performance on the unchanged Fortran code. Only compiler optimization allowed.

• 1000x1000: Performance on an arbitrary optimized code. Count operations same as standard Gaussian Elimination.

• NxN: Optimized code for arbitrary problem size.

  • Parameters nmax and n1/2 are listed as well.
  • Used in the Top500

Previous slide

Next slide

Back to first slide

View graphic version