Re: Efficient Sequential Access to Arrays

[Georg Bauhaus <rm.dash-bauhaus@futureapps.de>]

On 30.07.12 08:31, Jacob Sparre Andersen wrote:
> robin.vowels@gmail.com writes:
> 
>> On Monday, 16 July 2012 04:40:08 UTC+10, Keean Schupke  wrote:
> 
>>> The Monte-Carlo simulator I am working on is now performing better in
>>> Ada than in C++ using profile-guided compilation (57k simulations per
>>> second for Ada vs 56k simulations per second for C++).
>>
>> One part in 59 is not a significant difference.
> 
> That depends on the variance.  But reporting performance numbers without
> the corresponding variance/standard deviation is very bad style.

When I see fluctuations, and the system is otherwise idle,
I'll typically assume that there is something random going on
in the program or in its run-time support system, or that
some contemporary CPU lets its cores run at different
speed without me knowing a way to influence that (by, e.g.,
asking for a constant number of ins per sec that would not
create too much heat (that would force fewer ins per sec)).
Because otherwise, how could the sequence of instructions
create fluctuations if it is not self-modifying?

In this case, the inputs seem of fixed size, tailored to
the caches' sizes in particular, so I'll assume that the results
show that it is not universally true that equivalent programs
written in Ada or C-17 cannot be as fast as the other.