From: Daniel Stenberg <daniel_at_haxx.se>
Date: Wed, 9 Jun 2010 12:17:50 +0200 (CEST)

On Wed, 9 Jun 2010, Thomas Martitz wrote:

>> version. So the question here is whether malloc in core (even as a loadable
>> component) is an absolute no-do. I'm not going to try and hide the ugly
>> truth - the memory fragmentation is MASSIVE, even with something like tlsf.
>
> Unfortunately it's not only a NODO, it's simply not doable properly. tlsf is
> the best you can get on the devices Rockbox runs on. Like you said,
> fragmentation is massive even with tlsf, and you cannot help it without
> proper hardware support, e.g. MMU (on some targets) and memory paging is not
> available.

malloc has been used in embedded systems with MMU-less processors and RTOSes
for decades. It is perfectly doable.

Yes, fragementation can become an issue.

Do we want to introduce malloc in the Rockbox core? We've fought pretty
strongly against this for many years now, and I've not yet seen a convincing
argument why we would let it in anytime soon. My personal opinion of course.

I would rather turn around and say that if a thing such as a text to speach
engine can't be done without mallocing itself to death, then perhaps that's
telling of an inferior design to start with and perhaps it isn't that suitable
for use in and by Rockbox?

-- 
  / daniel.haxx.se

Received on 2010-06-09