Caveat and Notes

  1. This work, started around 1978, has benefited from much help and encouragement; many sources have contributed, and still are contributing, numbers of results: acknowledgements are due to each of the contributors but special thanks are owed to, chronologically, Ðô Ngoc Ky , Patrick Lefèvre , Serge Sevastianov , Louis Métayer , David G. Hough , Pierre Meindre , Norbert Juffa and Stéphane Tsacas .
  2. This report is intended to be dense with only raw values and occasional ratios: no data reduction is done, nor any explanation is given, which implies interpretation and/or judgement of value. However and perhaps because of this, comments and remarks are welcome
  3. Throughout succeeding releases of this report, values for R may have changed: normally, it's no error but a proof of the permanent improvement process of every living system. In particular, a same machine may give significantly different results within this report, or display a not-so-good value (compared with the ones seen or reported elsewhere): different runs at different times with different configurations will most likeky give different results, if simply because many differences may and do occur: new compiler, compiler version, new optimization technique, new preprocessed source ..., on the software side, and cache size, bus width ... on the hardware side .
  4. This benchmark now is frequently used by a large segment of concerned parties: you will find it included in many internal or public suites such as the
    Suite Synthétique des Benchmarks de l' AFUU ("SSBA") or SPEC (89 & 92) Benchmark Suites ...
  5. A subset of the best performing executables for each main architecture is available for standalone ready-to-go as-is execution : as well as the different source versions for recompilation :
  6. Version 55 of this report gets a new table, a single (last) page summarizing the TOP FIFTY machines with a new ratio : the R1996 refers to the 1996 reference machine, a Sun SPARCstation 10/40 (no cache, Sun SC3 (f77 O4 fast)) . Within the scope and limits of this benchmark, the R1978 of this new reference machine was 1908 , with respect to the old 1978 reference machine, an IBM 370/168-3 (HSMU, Fortran 4.H.extended) .
  7. The original document was generated using the LaTeX2HTML translator Version 95.1 (Fri Jan 20 1995) Copyright © 1993, 1994, Nikos Drakos, Computer Based Learning Unit, University of Leeds. The command line arguments were: latex2html ach.tex.

Next: Chapter 1 - Ms/Os machines Up: Fortran Execution Time Previous: The report's organization