Re: [lmi] [lmi-commits] odd/eraseme_long_double_irr cca9eaa5 1/3: Pass floating-point scalar arguments by value

[Greg Chicares]

On 4/30/22 16:55, Vadim Zeitlin wrote:
> On Sat, 30 Apr 2022 12:17:39 -0400 (EDT) Greg Chicares 
> <gchicares@sbcglobal.net> wrote:
> 
> GC> branch: odd/eraseme_long_double_irr
> GC> commit cca9eaa5559d7f7aadefd590e148849b6e0d327b
> GC> Author: Gregory W. Chicares <gchicares@sbcglobal.net>
> GC> Commit: Gregory W. Chicares <gchicares@sbcglobal.net>
> GC> 
> GC>     Pass floating-point scalar arguments by value
> 
>  This is another of those curious-but-not-curious-enough-to-test-myself
> questions: does this have any noticeable effect on performance?

/opt/lmi/src/lmi[0]$git switch --detach HEAD~2                     
HEAD is now at d7c5304dd Don't store unhelpful reciprocals

i686-w64-mingw32

Running math_functions_test:
Speed tests:
  std::pow         5.750e-02 s mean;      55555 us least of  18 runs
  std::expm1       4.807e-02 s mean;      47531 us least of  21 runs
  double      i365 7.562e-03 s mean;       7388 us least of 100 runs
  long double i365 6.902e-03 s mean;       6711 us least of 100 runs
  10^-9 nonstd     5.785e-04 s mean;        564 us least of 100 runs
  10^-9 std        2.731e-03 s mean;       2622 us least of 100 runs

x86_64-pc-linux-gnu

  std::pow         2.314e-01 s mean;     220195 us least of   5 runs
  std::expm1       1.983e-02 s mean;      18702 us least of  51 runs
  double      i365 2.491e-03 s mean;       2214 us least of 100 runs
  long double i365 1.431e-02 s mean;      13835 us least of  70 runs
  10^-9 nonstd     4.645e-04 s mean;        438 us least of 100 runs
  10^-9 std        2.210e-03 s mean;       2125 us least of 100 runs

/opt/lmi/src/lmi[0]$git switch -         
Previous HEAD position was d7c5304dd Don't store unhelpful reciprocals
Switched to branch 'master'
/opt/lmi/src/lmi[0]$git --no-pager log -2 --oneline 
7fcf08e2b (HEAD -> master) Annotate an arcanum
cca9eaa55 Pass floating-point scalar arguments by value

i686-w64-mingw32

  std::pow         5.809e-02 s mean;      57337 us least of  18 runs
  std::expm1       4.789e-02 s mean;      46639 us least of  21 runs
  double      i365 7.512e-03 s mean;       7345 us least of 100 runs
  long double i365 6.868e-03 s mean;       6676 us least of 100 runs
  10^-9 nonstd     5.799e-04 s mean;        564 us least of 100 runs
  10^-9 std        2.535e-03 s mean;       2444 us least of 100 runs

x86_64-pc-linux-gnu

  std::pow         2.194e-01 s mean;     212414 us least of   5 runs
  std::expm1       1.895e-02 s mean;      18572 us least of  53 runs
  double      i365 2.231e-03 s mean;       2214 us least of 100 runs
  long double i365 1.389e-02 s mean;      13782 us least of  72 runs
  10^-9 nonstd     4.427e-04 s mean;        438 us least of 100 runs
  10^-9 std        2.143e-03 s mean;       2125 us least of 100 runs

IOW, TL;DR: "No".