Exponential operator

[Andrei Alexandrescu]

Jimbob wrote:
> "Andrei Alexandrescu" <SeeWebsiteForEmail at erdani.org> wrote in message 
> news:4A7C5313.10105 at erdani.org...
>> Jimbob wrote:
>>> "bearophile" <bearophileHUGS at lycos.com> wrote in message 
>>> news:h5h3uf$23sg$1 at digitalmars.com...
>>>> Lars T. Kyllingstad:
>>>>> He also proposed that the overload be called opPower.
>>>> I want to add to two small things to that post of mine:
>>>> http://www.digitalmars.com/webnews/newsgroups.php?art_group=digitalmars.D&article_id=95123
>>>>
>>>> The name opPow() may be good enough instead of opPower().
>>>>
>>>> And A^^3 may be faster than A*A*A when A isn't a simple number, so 
>>>> always replacing the
>>>> power with mults may be bad.
>>> It wont be on x86. Multiplication has a latency of around 4 cycles 
>>> whether int or float, so x*x*x will clock around 12 cycles.
>> Yeah, but what's the throughput? With multiple ALUs you can get several 
>> multiplications fast, even though getting the first one incurs a latency.
> 
> In this case you incur the latency of every mul because each one needs the 
> result of the previous mul before it can start. Thats the main reason 
> trancendentals take so long to compute, cause they have large dependancy 
> chains which make it difficult, if not imposible for any of it to be done in 
> parallel.


Oh, you're right. At least if there were four multiplies in there, I 
could've had a case :o).

Andrei