There has been discussion related to this question here: http://tech.groups.yahoo.com/group/nuttx/message/685link
Extracts: The Linux Journal article that is referred to is here: www.linuxjournal.com/article/10785#mpart4link
I think that the 8052 and the M68HC12 ports are especially bad choices to characterize NuttX because they both have some issues. And, and NuttX is now at version 5.16 with 63 releases.
I did fill out the interview in the "Publisher" tab here: download.famouswhy.com/publisher/gregory_nutt/. There link; there is a review there too: download.famouswhylink .com/nuttx/
Extensive NuttX documentation is available here: www.nuttx.orglink .
8051/80c52:8051/80c52: This architecture is really RTOS hostile. It has a tiny hardware stack (128 bytes on the 8051, 256 on the 80c52) at a dedicated memory location (address 0). To switch tasks, you have to copy the whole stack of the task to be blocked from its dedicated address to some save location, and then copy the whole stack of the task to be started from its save location to the dedicated stack location. YECH!
hcs12:hcs12: This is a project that I work on in my spare time when I am not doing anything else. It is just not finished and not ready for prime time yet.
FreeRTOS. HasFreeRTOS has tons of downloads and a really tiny footprint of about 4Kb. It is the RTOS of choice for the really small MCUs. A FreeRTOS port is bundled by silicon vendors with just about every MCU. So it is the default RTOS choice.
There are dozens of competitors with FreeRTOS out there. ChiBIOSChiBIOS comes to mind immediately. These are all tiny schedulers of varying types.
RTEMS.RTEMS: I've worked with this one. It has been around forever and should be very stable. It is big. Think >100Kbbig; think >100kb. I think it aims a little above the MCU market.
uCOS.uCOS: Never used it, but this the RTOS under several popular bootloaders, isn't it? My impression is that is it similar to RTEMS, but I don't really know what I am talking about.