Re: [boost] Re: Technical session at the October C++ meeting?

From: Greg Colvin (gcolvin_at_[hidden])
Date: 2001-08-20 00:38:23

From: Aleksey Gurtovoy <alexy_at_[hidden]>
> John Max Skaller wrote:
> > But such an animal cannot exist. There is no way
> > to allocate n elements of an array on the stack with n being
> > dynamically determined -- without using VLA's. That is the whole
> > point of VLA's.
> 
> You can get pretty close, though:
> 
> fast_buffer<255, char> buf(n);
> char* p = buf.data();
> //..
> 
> Here, if n <= 255, then no dynamic allocation will be performed, and the
> buf.data() will point to a stack space; otherwise, the memory will be
> allocated on the heap (and released in the destructor of the class). Of
> course, in the latter case 255 bytes of the stack space will be "wasted",
> but usually you don't care :). In return, if you know that a typical value
> of n <= 255, but that also there are (relatively rare) cases then the
> condition does not hold (one-line character buffer is a typical example of
> such situation), then the 'fast_buffer' class can save you quite a few
> processor cycles :). Not exactly a VLA, but in many cases close enough.

And such an animal can also be allocated in terms of alloca()
or LocalAlloc() or such on systems on which this is a good idea.