Re: Fortran vs. C++ : Which is best to choose now i
From: Tom Micevski (none_at_none.au)
Date: 02/10/04
- Next message: Giulia Paramour: "Re: Intel ifort compiler problem - part II"
- Previous message: Catherine Rees Lay: "Re: A tale of three Sun's"
- In reply to: E. Robert Tisdale: "Re: Fortran vs. C++ : Which is best to choose now i"
- Next in thread: E. Robert Tisdale: "Re: Fortran vs. C++ : Which is best to choose now i"
- Reply: E. Robert Tisdale: "Re: Fortran vs. C++ : Which is best to choose now i"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Tue, 10 Feb 2004 21:45:34 +1100
E. Robert Tisdale wrote:
> Are these your best or worst times?
> If you want to know whether your results are *significant* or not
> you should make several (about 30) runs
> and calculate the average and standard deviation for both times.
based on 10 runs:
* total time
auto = 7.039 (0.066)
alloc = 7.404 (0.077)
rand = 5.033 (0.040)
i would say that's *significant*.
* time calc'ing sum
auto = 2.005 (0.070)
alloc = 2.371 (0.061)
i would say that's *significant*.
> For completeness, please report the optimization options
> that you chose to use for this benchmark.
standard cvf "release" optimisations (/optimize:4)
redone with (/optimize:5)
12 runs:
* time calc'ing sum
auto = 2.026 (0.047)
alloc = 1.966 (0.060)
i would say that's *not significant* (although alloc seems faster).
> You could make a stronger argument for the validity of your calculation
> if you could actually elide the random number calculations
that's the whole point of me subtracting the time spent in the random_number loop (and doing the "time calc'ing sum" calcs).
> In actual applications, where you actually compute something
> (besides random numbers fro example) this ratio falls into the noise --
> you can't make a meaningful measurement.
>
that's what seems to happen with /optimize:5 (but not with /optimize:4).
- Next message: Giulia Paramour: "Re: Intel ifort compiler problem - part II"
- Previous message: Catherine Rees Lay: "Re: A tale of three Sun's"
- In reply to: E. Robert Tisdale: "Re: Fortran vs. C++ : Which is best to choose now i"
- Next in thread: E. Robert Tisdale: "Re: Fortran vs. C++ : Which is best to choose now i"
- Reply: E. Robert Tisdale: "Re: Fortran vs. C++ : Which is best to choose now i"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|