Subject: Re: [boost] [XInt] Some after thoughts about SIMD
From: Simonson, Lucanus J (lucanus.j.simonson_at_[hidden])
Date: 2011-03-15 12:50:03
Dave Abrahams wrote:
> At Fri, 11 Mar 2011 12:04:15 -0800,
> Simonson, Lucanus J wrote:
>>
>>>
>>> More seriously, I fail to see an vector operation that is "limited
>>> by the language". I have very eager to see the one you speak about.
>>
>> If you have the same variable in more than one place in an
>> expression
>
> sometimes called "aliasing"
>
>> you can map that expression to different/fewer
>> instructions than if they were all different variables in different
>> places. There is no way to perform that mapping using expression
>> templates because it cannot be detected in template metaprogramming
>> which varaibles are the same, only that they have the same type.
>> Therefor you cannot map expression to the optimal instructions using
>> template metaprogramming.
>
> I beg to differ, at least most of the time. You can detect aliasing
> at runtime and choose an optimal implementation on that basis, and
> when the sameness of the two variables is information available to the
> compiler (as it often is), constant-folding and dead-code-elimination
> will throw away the test and the dead branch, leaving only the optimal
> code.
Clever. I like how you think.
If the two values happened not to be aliases I think the compiler would leave in the compare and branch, which isn't what we would want.
Regards,
Luke