Re: [PATCH 2/5] vmevent: Convert from deferred timer to deferredwork
From: KOSAKI Motohiro
Date: Fri Jun 08 2012 - 03:33:14 EST
(6/8/12 3:28 AM), leonid.moiseichuk@xxxxxxxxx wrote:
-----Original Message-----
From: ext KOSAKI Motohiro [mailto:kosaki.motohiro@xxxxxxxxx]
Sent: 08 June, 2012 10:23
...
If you wakeup only by signal when memory situation changed you can be
not mlocked.
Mlocking uses memory very inefficient way and usually cannot be applied
for apps which wants to be notified due to resources restrictions.
That's your choice. If you don't need to care cache dropping, We don't
enforce it. I only pointed out your explanation was technically incorrect.
My explanation is correct. That is an overhead you have to pay if start to
use API based on polling from user-space and this overhead narrows API
applicability.
Moving all times/tracking to kernel avoid useless wakeups in user-space.
Wrong. CPU don't realized the running code belong to userspace or kernel. Every
code just consume a power. That's why polling timer is wrong from point of power
consumption view.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/