Re: [lmi] Robust timings in unit tests

[Greg Chicares]

On 2017-05-11 14:57, Greg Chicares wrote:
[...]
> 'expression_template_0_test' [...] I had once set great store by its
> results, though that was probably with gcc-3.0 on a one-core 3.5 GHz
> CPU...because now there's no difference to measure on x64_86-linux-gnu
> except for the "STL" methods that are known to be awful:
> 
>   Speed tests: array length 1000
>   C               : 9.877e-07 s mean;          0 us least of 10125 runs
>   STL plain       : 4.006e-06 s mean;          3 us least of 2496 runs
>   STL fancy       : 1.462e-06 s mean;          1 us least of 6840 runs
>   valarray        : 9.952e-07 s mean;          0 us least of 10048 runs
>   uBLAS           : 1.020e-06 s mean;          0 us least of 9801 runs
>   PETE            : 8.727e-07 s mean;          0 us least of 11459 runs
> 
> ..and the corresponding results for array lengths of {1, 10, 100} just
> look ridiculous now: they're all zero.

An upcoming commit will run each of those one thousand times, so that
results become as meaningful again as they were about two decades ago,
e.g.:

  Speed tests: array length 10
  C               : 9.690e-06 s mean;          9 us least of 1032 runs
  STL plain       : 6.873e-05 s mean;         68 us least of 146 runs
  STL fancy       : 1.823e-05 s mean;         18 us least of 549 runs
  valarray        : 1.029e-05 s mean;         10 us least of 972 runs
  uBLAS           : 4.186e-05 s mean;         41 us least of 239 runs
  PETE            : 9.824e-06 s mean;          9 us least of 1018 runs
...
  Speed tests: array length 1000
  C               : 7.056e-04 s mean;        704 us least of 100 runs
  STL plain       : 2.997e-03 s mean;       2967 us least of 100 runs
  STL fancy       : 1.164e-03 s mean;       1155 us least of 100 runs
  valarray        : 7.251e-04 s mean;        722 us least of 100 runs
  uBLAS           : 7.731e-04 s mean;        763 us least of 100 runs
  PETE            : 7.244e-04 s mean;        720 us least of 100 runs