Re: [RFC v1 2/2] mm: SLUB Freelist randomization

[Christoph Lameter]

0.On Wed, 18 May 2016, Thomas Garnier wrote:

> slab_test, before:
> 10000 times kmalloc(8) -> 67 cycles kfree -> 101 cycles
> 10000 times kmalloc(16) -> 68 cycles kfree -> 109 cycles
> 10000 times kmalloc(32) -> 76 cycles kfree -> 119 cycles
> 10000 times kmalloc(64) -> 88 cycles kfree -> 114 cycles

> After:
> 10000 times kmalloc(8) -> 60 cycles kfree -> 74 cycles
> 10000 times kmalloc(16) -> 63 cycles kfree -> 78 cycles
> 10000 times kmalloc(32) -> 72 cycles kfree -> 85 cycles
> 10000 times kmalloc(64) -> 91 cycles kfree -> 99 cycles

Erm... The fastpath was not touched and the tests primarily exercise the
fastpath. This is likely some artifact of code placement by the compiler?