[linux-mm-cc] compcache with swap free notify support

[Nitin Gupta]

On Sat, Jun 21, 2008 at 12:43 PM, John McCabe-Dansted <gmatht at gmail.com> wrote:
> On Mon, Jun 16, 2008 at 4:59 PM, Nitin Gupta <nitingupta910 at gmail.com> wrote:
>> Performance numbers (in term of memory saving and performance gain)
>> should be much better with this patch than compcache-0.3.
>
> I would have thought that the free space in the vmalloc'd memory would
> be too fragmented to be effectively reclaimed under normal conditions,
> and so this patch still wouldn't normally result in tlsf releasing
> memory back to the system. Is it intended more that if one page is
> freed and another is allocated, then the memory used by tlsf does not
> increase?
>

All this can become clear from tlsf stats (disabled by default). So,
its worth running some experiments with these enabled.


> It would perhaps be possible to defragment the free space by
> heuristically selecting a maximum pointer value M, and scanning the
> table for a pointers X  > M and then reallocating X below M. Then all
> space higher than M could be released back to the system.
>


There are lot of papers around that deal with memory compaction.
Perhaps we can refer them for further study.

Nitin