Hello embedded developers,

As part of a CE WG project [1, 2], we're comparing SLAB allocators
focusing on embedded devices.
Our main concern is memory efficiency (fragmentation / wastage).

We've found that kernel defconfigs shipped with the linux kernel set SLAB
as the default allocator [3].

However, some of our measurements show that SLAB may not be the best choice,
since it suffers from much more fragmentation that SLUB or SLOB.

We'd be very interested in getting some feedback on this question:

* Do you know which SLAB allocator your platform uses ?
* Have you compared different SLABs on your platforms ?
* Do you know the trade offs involved in using SLAB vs SLUB vs SLOB ?

Thanks!

Ezequiel

[1] http://elinux.org/Kernel_dynamic_memory_allocation_tracking_and_reduction
[2] http://elinux.org/Kernel_dynamic_memory_analysis
[3] https://lkml.org/lkml/2012/10/11/232