Re: sup****ting header-less blocks in memory allocators...

"Chris M. Thomasson" <no@[EMAIL PROTECTED]
> wrote in message 
news:6O4Fk.12965$ex3.8007@[EMAIL PROTECTED]
>I was thinking of an idea of expanding a current limitation in some of my

>allocators which sup****t the use of header-less blocks. The way I
currently 
>do it is to simply align chunks on a fairly large boundary, say 16 pages.

>These chunks contain a header at the head. Layout is simple:
>
> [H][CHUNK]
>
> A 16-page chunk would be aligned on a 16-page chunk boundary. If a 
> header-less block needs to access its header, it simple rounds its
address 
> up to a chunk boundary, and subtracts the size of the chunk plus the
size 
> of the header. This has a limitation that the largest memory allocation 
> can't be larger than the size of a chunk. To overcome this, I was
thinking 
> about generating a unique GUID per-computer,

Or, perhaps generate a unique GUID per-process; I think that would be much

safer...


> and using that as a header identifier. The header would store the GUID 
> right after it. So, the layout for a multi-chunk allocation would look 
> like:
>
>
> [H][GUID][CHUNK1][CHUNK2]
>

[...]