Re: [boost] Re: Re: uBLAS and expression templates

From: David Abrahams (david.abrahams_at_[hidden])
Date: 2002-07-02 14:08:32

"Benedikt Weber" <weber_at_[hidden]> wrote in message
news:afssmk$3ms$1_at_main.gmane.org...
> Joerg
>
> I just want to remind you that there are many other expressions...that
should also evaluate efficiently.

Yes.

> like A*( B * (C * v))

That's no more difficult than A*(B*v) is. The ET machinery should
handle it, since this----------^ is just

    Matrix-Expression * Vector-Expression.

Like this is: A*( B * (C * v))
---------------^

> or norm_2(A * (B * v))

By the same reasoning, this should not require any more temporaries than

    norm_2(A)

does.

> Avoiding temporaries in all cases looks like a real challange, if it's
> possible at all.

It's usually possible, but not always desirable.

-Dave