Boost logo

Boost :

From: Peter Dimov (pdimov_at_[hidden])
Date: 2000-12-04 12:12:35


> > new[] will guarantee alignment for C++ types; unfortunately, SIMD types
do
> > not map to C++ and the default new[] is not guaranteed to work for them.
>
> I thought SIMD was just special (parallel) operations on arrays of doubles
> (or long doubles?)...

True; the point is that these operation require an alignment that is not
needed for any C++ type, so new[] is not guaranteed to be 'SIMD-aware'.

> > Yes, this is an option; but wouldn't it be better to allow user-supplied
> > allocation/deallocation functions?
>
> Maybe. I have to question how often it would be used without overriding
> global new[]/delete[] anyway. In other words: is it really worth the
> trouble?

I may want to use std::vector<float, pool_alloc> in a library; overriding
the global new[] is a per-project decision, not per-component.

--
Peter Dimov
Multi Media Ltd.

Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk