[Intel-gfx] [PATCH] mm/vmalloc: Keep a separate lazy-free list

Roman Peniaev r.peniaev at gmail.com
Sat Apr 23 11:21:08 UTC 2016


On Fri, Apr 22, 2016 at 11:49 PM, Andrew Morton
<akpm at linux-foundation.org> wrote:
> On Fri, 15 Apr 2016 12:14:31 +0100 Chris Wilson <chris at chris-wilson.co.uk> wrote:
>
>> > > purge_fragmented_blocks() manages per-cpu lists, so that looks safe
>> > > under its own rcu_read_lock.
>> > >
>> > > Yes, it looks feasible to remove the purge_lock if we can relax sync.
>> >
>> > what is still left is waiting on vmap_area_lock for !sync mode.
>> > but probably is not that bad.
>>
>> Ok, that's bit beyond my comfort zone with a patch to change the free
>> list handling. I'll chicken out for the time being, atm I am more
>> concerned that i915.ko may call set_page_wb() frequently on individual
>> pages.
>
> Nick Piggin's vmap rewrite.  20x (or more) faster.
> https://lwn.net/Articles/285341/
>
> 10 years ago, never finished.

But that's exactly what we are changing making 20.5x faster :)

--
Roman


More information about the Intel-gfx mailing list