Re: [boost] performance of a linear algebra/matrix library

Subject: Re: [boost] performance of a linear algebra/matrix library
From: Jeffrey Lee Hellrung, Jr. (jhellrung_at_[hidden])
Date: 2010-05-07 11:44:19

On 05/07/2010 07:57 AM, Marius Stoica wrote:
> On Friday 07 May 2010 16:06:41 DE wrote:
>> on 06.05.2010 at 22:48
>>
>>   Kenny Riddile wrote :
>>> If you mean a language-imposed, inherent abstraction penalty, then I
>>> doubt it (certainly not 33%).  You can find out for yourself though:
>>>
>>> http://www.stepanovpapers.com/AbstractionPenaltyBenchmark.cpp
>>
>> i've run the code for my everyday compiler (msvc80) and it greported
>> no abstraction penalty
>>
>> i've also tried icc11 and it showed minor abstraction penalty
>> (mean = 1.17)
>
> i'm trying gcc 4.4.3 here
>
> (optimization/total/penalty )
>
> O0/ 32.15 /5.29
> O1/ 1.02 /0.39
> O2/ 0.86 /0.94
> O3/ 1.09 /0.39
> Os/ 0.96 /0.44
>
> Do i understand that rigth or does gcc produce better code when you abstract
> away ?

[Continuing the slightly OT discussion...]  I noticed this, too, just 
with the default release settings on MSVC9.  I can only think to 
attribute this to iterating through indices in the C-style version, 
rather than through iterators...?  I haven't looked closely at it at 
all, though.

- Jeff