Subject: Re: [boost] [pool] Calling all library designers: The future of Boost.Pool???
From: Christopher Kormanyos (e_float_at_[hidden])
Date: 2012-07-18 17:32:13
> In non-game environment, you may have systems where you don't even
> have a heap or that design ruels prevent use of calls to malloc etc...
> for WCET or stability concerns.
> The only solution is to have a pool of memory on the stack and fake
> ctor/dtor call inside it.
Yeah! That's what I'm talking about. I actually embed very much C++
in microcontrollers down to 1kB of RAM and no heap because I exclude
it in my self-written linker files.
Over the years I wrote all kinds of high-performance allocation stuff:
constructor w/ reinterpret_cast memory mapped to pools, circular allocators,
static allocators with pools. It's fast if you want it to be.
But Boost.Pool, I have not figured out yet.
I would like to say: There really are some of us fitting junk into 64 bytes,
or 1024, or whatever.
Best regards, Chris.