Beefy Boxes and Bandwidth Generously Provided by pair Networks
Perl: the Markov chain saw
 
PerlMonks  

Re^2: Measuring time intervals using Time::HiRes -- time vs. gettimeofday and tv_interval

by hrr (Monk)
on Jul 18, 2006 at 19:33 UTC ( [id://562109]=note: print w/replies, xml ) Need Help??


in reply to Re: Measuring time intervals using Time::HiRes -- time vs. gettimeofday and tv_interval
in thread Measuring time intervals using Time::HiRes -- time vs. gettimeofday and tv_interval

Thank you for this hint! I obtained similar benchmark results as you did.

Surprsingly, using Time::HiRes::time instead of the standard time, the performance is much better (but still worse than tv_interval):
Rate Hires Time Hires 98201/s -- -55% Time 220149/s 124% --

Replies are listed 'Best First'.
Re^3: Measuring time intervals using Time::HiRes -- time vs. gettimeofday and tv_interval
by Ieronim (Friar) on Jul 18, 2006 at 19:53 UTC
    Vice versa. Look at the Rate column — it shows the speed of the function! The more times a second the function is called, the faster it is :) So in this table (produced by the benchmark involving all three variants)
    Rate interval HiRes::time buitin time() interval 88470/s -- -53% -95% HiRes::time 188363/s 113% -- -89% buitin time() 1753092/s 1882% 831% --
    builtin time() is 9 times faster than Time::HiRes::time, which is nearly 2 times faster than the weird variant with tv_interval!

    Benchmark code is here:


         s;;Just-me-not-h-Ni-m-P-Ni-lm-I-ar-O-Ni;;tr?IerONim-?HAcker ?d;print

      As long as we're all tweaking the heck out of our benchmarks... ;-)

      What I've noticed is the number of calls to these functions are not all the same. In the 'interval' function, we are calling gettimeofday, constructing an anonymous array, calling tv_interval, and destroying the array. Only one of these are we actually interested in. Meanwhile, both of the other two cases, we're calling *time twice. A completely different set of comparisons. I also eliminated the overhead of returning floats, what little that should be.

      So, I thought I'd throw this into the mix. Put all the initialisation up front. Then see what happens.

      And the results?
      Rate interval time hires interval 685148/s -- -79% -80% time 3340426/s 388% -- -2% hires 3398163/s 396% 2% --
      2% bonus to hires's time over the built-in time? I'll chalk that up to statistically insignificant, and call the two a tie. In fact, I ran it a second time and got it reversed:
      Rate interval hires time interval 971246/s -- -77% -78% hires 4192706/s 332% -- -3% time 4327848/s 346% 3% --
      So, I'd say that the difference between interval and time is non-trivial. But even that operates fast enough that I'm not going to worry about it.

Log In?
Username:
Password:

What's my password?
Create A New User
Domain Nodelet?
Node Status?
node history
Node Type: note [id://562109]
help
Chatterbox?
and the web crawler heard nothing...

How do I use this?Last hourOther CB clients
Other Users?
Others cooling their heels in the Monastery: (6)
As of 2024-03-28 14:44 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    No recent polls found