Arm Linux user process space full 3G or 32M in practical project
I have never worked on any embedded Linux project. I'm very much interested in knowing what we really do in real arm linux embedded real-time projects.
Do we use full 3G virtual space for user process virtual address space, so that processor's mm is fully swapped out, and TLB and cache both need to be flushed at process context switch, or we can actually take advantage of using arm PC15 processID register (C13) to reduce the mm switch cost? Or may be it doesn't put in practical use at all? Thanks for any insights on this.
|