Re: [lmi] Contradictory performance measurements

[Greg Chicares]

On 4/7/21 11:28 PM, Greg Chicares wrote:
[...]
> Now I run
>   gwc/speed_test.sh
> which rebuilds every architecture and runs
>   make cli_timing
> for each one. With the patch, speed decreases by several
> percent in general, affecting all architectures. That's
> broadly consistent with this ca. 2000 comment:
>   // Profiling shows that inlining this member function makes a
>   // realistic application that performs a lot of rounding run about
>   // five percent faster with gcc.
> (that "realistic application" was lmi's predecessor).

For the record (as I'm about to cite this thread in lmi's inline
documentation), that was the wrong 2000-era comment; it was
intended instead to cite this one:

  // Division by an exact integer value should have slightly better
  // accuracy in some cases. But profiling shows that multiplication by
  // the reciprocal stored in scale_back_ makes a realistic application
  // that performs a lot of rounding run about four percent faster with
  // all compilers tested.