lkml.org 
[lkml]   [2015]   [Aug]   [1]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [tip:x86/mm] x86/mm/mtrr: Clean up mtrr_type_lookup()
On Sat, Aug 01, 2015 at 09:39:07AM -0700, Linus Torvalds wrote:
> Quite the reverse.
>
> It makes no sense to write-combine normal memory (RAM), because caches
> work and sane memory is always cache-coherent. So marking regular
> memory write-combining is a sign of crap hardware (which admittedly
> exists all too much, but hopefully goes away).
>
> In contrast, marking MMIO memory write-combining is not a sign of crap
> hardware - it's just a sign of things like frame buffers on the card
> etc. Which very much wants write combining. So WC for MMIO at least
> makes sense.
>
> Yes, yes, I realize that "crap hardware" may actually be the more
> common case, but still..

Hmm, ok.

My simplistic mental picture while thinking of this is the IO range
where you send the commands to the device and you don't really want to
delay those but they should reach the device as they get issued.

OTOH, your example with frame buffers really wants to WC because sending
down each write separately is plain dumb.

Ok, I see, so it can make sense to have WC IO memory, depending on the
range and what you're going to use it for, I guess...

Thanks.

--
Regards/Gruss,
Boris.

ECO tip #101: Trim your mails when you reply.
--


\
 
 \ /
  Last update: 2015-08-01 19:01    [W:0.113 / U:0.312 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site