Threads Vs Forks in Embedded Environment

OS is embedded Linux. The platform is based on a dual core architecture having an ARM11 core (Application Processor, AP) that handles all the application level functionalities and a Star Core DSP processor ( Base band Processor, BP), which handles all the modem functions. OTASU supports Jffs2, cramfs file system. other products supports MCU (MCORE) also

I suppose by "threads" you mean using the "Posix thread library" and by "fork" you mean using standard processes ?

Another alternative is using clone(), or another api call

In Linux, threads are processes. fork() creates processes.

So they are essential the same.

Threads share a common data space. This is not significant in your application, so where is the difference?

Not much. Have you measured?

If you have a MMU, the memory consumption of a process maybe lower than you think because of the copy-on-write semantics.

regards Wolfgang

Not true. (1) Kernel 2.6 does have a kernel representation of user land threads and (2) AFAIK, there are thread libraries that don't use the OS at all to create the threads but do it completely in user land. Here the Kernel does not know about the threads. (I suppose this is what the OP had in mind.)

-Michael