Yuxin Ren writes:

I am curious memory mapping/unmapping performance. Are there any applications that stress the memory mapping and unmapping utility?

There aren't currently any microbenchmarks for MMU operations in sel4bench[0]. One thing I can note off the top of my head is that each operation will flush the TLB entries associated with that mapping. This will likely be true of shootdowns in multicore as well, though someone else who knows more should chime in. -- cmr +610481782084 http://octayn.net/

On 2 Aug 2016, at 8:32 , Yuxin Ren wrote:

Hi all,

I am curious memory mapping/unmapping performance. Are there any applications that stress the memory mapping and unmapping utility?

I don’t think we’ve come across any yet. This is almost certainly a result of seL4 presently being deployed with fairly static architectures (and memory management operations done inside a VM are invisible to seL4). I’m sure scenarios where memory management performance is important do exist, we just haven’t come across them yet. I suspect you’re asking because you found that these operations are slow – I’d be surprised if they weren’t ;-) No-one bothered to put any effort in optimising the mapping implementation for lack of a driver of high enough significance. We’ll prioritise that once it becomes a problem. Gernot