Scaleable multi-threaded allocator solutions for embedded systems with NO heap...

0 views
Skip to first unread message

Chris Thomasson

unread,
May 21, 2007, 2:27:39 AM5/21/07
to
You can create a single threaded allocator (e.g., try to do so without using
any static's)... In other words, the invention needs to be able to "create"
an allocator on a strictly per-thread/cpu basis. So, I need an explicit
constructor function-pointer to "your" extern "C" declared
method/static-func in order to render a viable pointer for my library to
cache "and" use.

I need function pointer to your single-threaded static variable free "heap"
object constructor. After that, the rendered impl, (e.g., rendered wrt our
discussions on how to get everything up and running) needs a pointer to a
destructor function. This is required functionality wrt the vzMalloc, and
the Dtor functions...

You can use a threads "extra" stack space for a multi-threaded allocator.
Indeed, most applications are well engineered? ... :^/

Anyway, most application(s) which are designed by professionals will not
have a chance to "blow their stack(s)", so to speak: A recursive function
death spiral! So, I can most likely realize a 100% lock-=free memory
allocator framework which can be solely based on the stacks of a plurality
of threads that were designed by competent programmers. You know!

The programmers that can program single-threaded allocators. No problem! No
threads, no problems mon!

;^)

What do 'yal-ll think!

;^)

--
Chris M. Thomasson
http://appcore.home.comcast.net

Chris Thomasson

unread,
May 21, 2007, 2:32:54 AM5/21/07
to
"Chris Thomasson" <cri...@comcast.net> wrote in message
news:Z6SdnTh1i4iypszb...@comcast.com...

> You can create a single threaded allocator (e.g., try to do so without
> using any static's)... In other words, the invention needs to be able to
> "create" an allocator on a strictly per-thread/cpu basis.

Need to clarify one point!

Even on per-cpu, the user, You, can still stick to single threaded allocator
design... My invention can handle it for sure.

Chris Thomasson

unread,
May 21, 2007, 3:12:05 AM5/21/07
to
Simple talk:

Allocator based on the memory that is made of a plurality of threads "extra"
stack space.

Chris Thomasson

unread,
May 21, 2007, 3:13:38 AM5/21/07
to
This works well with the Cell processor... Its SPU's work well with
per-thread data and DMA...

Reply all
Reply to author
Forward
0 new messages