On Oct 10, 2005, at 2:25 PM, Frans Englich wrote:
What makes me ask these questions, is that I have code which
(according to my
opinion) severely needs virtual inheritance(VI). The code VI would
affect is
not under my veto power, and the maintainers of that code have so
far given
as comment "VI is out of the question. Period," without giving a
deeper
motivation, except "performance."
Benchmark on a machine with a timestamp counter (rdtsc on x86 as I
recall). Do up two versions of the code, one the right way, and the
other cleverly avoiding VI. Time them, collects all relevant
numbers, then see what the numbers are, then argue from that
position. Actual numbers are sometimes more compelling. The timings
can be for very small things, not whole program execution times.
Like, time to do:
ref->add_refernce (this);
or whatever you're changing. If it says you've added 3 clocks cycles
to a base of 100 clock cycles, and gcov data says that in whole
program, it runs 987234 times, then you can say your code slows it
down by roughly 3*987234 clock cycles, or 329 usecs (for a 3 GHz
processor), you can then express this as a % of the total run time.
Another way, is to find a way to speed up the code in some other
aspect, to pay for the new cost you are introducing, then bargain
with then, say, if you let me slow this down, I will pay for it by
having a net increase in speed of x. It is then harder for them to
say no. It shows that you are serious above performance and willing
to do what it takes to make things nice.
Hopefully they understand the term premature optimization.
