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

Subject: Re: [boost] performance of a linear algebra/matrix library
From: Stefan Seefeld (seefeld_at_[hidden])
Date: 2010-05-06 17:32:30

On 05/06/2010 01:02 PM, DE wrote:
> so my question is: how fast such a lib should be to make everybody
> happy?
>    

I think this is the wrong question to ask. You can't "make everybody 
happy". Rather than attempt that, you should think very hard about what 
target (users, architectures, use-cases) are most important to you, and 
then design and optimize for those.
(I'm particularly thinking of multi-core and similar platforms, where 
implicitly parallelizable code is key to acceptance.)

> as you can see starting with 8 by 8 matices C++ code is roughly 33%
> slower than C code
>
> is that enough to make you (the reader) happy?
>    

Definitely not.

         Stefan

-- 
       ...ich hab' noch einen Koffer in Berlin...