linux-mm.kvack.org archive mirror
 help / color / mirror / Atom feed
From: Anshuman Khandual <khandual@linux.vnet.ibm.com>
To: Dave Hansen <dave.hansen@intel.com>,
	John Hubbard <jhubbard@nvidia.com>,
	Anshuman Khandual <khandual@linux.vnet.ibm.com>,
	linux-kernel@vger.kernel.org, linux-mm@kvack.org
Cc: mhocko@suse.com, vbabka@suse.cz, mgorman@suse.de,
	minchan@kernel.org, aneesh.kumar@linux.vnet.ibm.com,
	bsingharora@gmail.com, srikar@linux.vnet.ibm.com,
	haren@linux.vnet.ibm.com, jglisse@redhat.com,
	dan.j.williams@intel.com
Subject: Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation process
Date: Wed, 1 Feb 2017 12:26:16 +0530	[thread overview]
Message-ID: <1ab159b5-1b67-9dae-4112-3360d8f909fd@linux.vnet.ibm.com> (raw)
In-Reply-To: <217e817e-2f91-91a5-1bef-16fb0cbacb63@intel.com>

On 01/31/2017 11:34 PM, Dave Hansen wrote:
> On 01/30/2017 11:25 PM, John Hubbard wrote:
>> I also don't like having these policies hard-coded, and your 100x
>> example above helps clarify what can go wrong about it. It would be
>> nicer if, instead, we could better express the "distance" between nodes
>> (bandwidth, latency, relative to sysmem, perhaps), and let the NUMA
>> system figure out the Right Thing To Do.
>>
>> I realize that this is not quite possible with NUMA just yet, but I
>> wonder if that's a reasonable direction to go with this?
> 
> In the end, I don't think the kernel can make the "right" decision very
> widely here.
> 
> Intel's Xeon Phis have some high-bandwidth memory (MCDRAM) that
> evidently has a higher latency than DRAM.  Given a plain malloc(), how
> is the kernel to know that the memory will be used for AVX-512
> instructions that need lots of bandwidth vs. some random data structure
> that's latency-sensitive?

CDM has been designed to work with a driver which can take these kind
of appropriate memory placement decisions along the way. But as per
the above example of an generic malloc() allocated buffer.

(1) System RAM gets allocated if there are first CPU faults
(2) CDM memory gets allocated if there are first device access faults
(3) After monitoring the access patterns there after, the driver can
    then take required "right" decisions about its eventual placement
    and migrates memory as required

> 
> In the end, I think all we can do is keep the kernel's existing default
> of "low latency to the CPU that allocated it", and let apps override
> when that policy doesn't fit them.

I think this is almost similar to what we are trying to achieve with
CDM representation and driver based migrations. Dont you agree ?

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

  parent reply	other threads:[~2017-02-01  6:57 UTC|newest]

Thread overview: 58+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-01-30  3:35 [RFC V2 00/12] Define coherent device memory node Anshuman Khandual
2017-01-30  3:35 ` [RFC V2 01/12] mm: Define coherent device memory (CDM) node Anshuman Khandual
2017-01-30  3:35 ` [RFC V2 02/12] mm: Isolate HugeTLB allocations away from CDM nodes Anshuman Khandual
2017-01-30 17:19   ` Dave Hansen
2017-01-31  1:03     ` Anshuman Khandual
2017-01-31  1:37       ` Dave Hansen
2017-02-01 13:59         ` Anshuman Khandual
2017-02-01 19:01           ` Dave Hansen
2017-01-30  3:35 ` [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation process Anshuman Khandual
2017-01-30 17:34   ` Dave Hansen
2017-01-31  1:36     ` Anshuman Khandual
2017-01-31  1:57       ` Dave Hansen
2017-01-31  7:25         ` John Hubbard
2017-01-31 18:04           ` Dave Hansen
2017-01-31 19:14             ` David Nellans
2017-02-01  6:56             ` Anshuman Khandual [this message]
2017-02-01  6:46           ` Anshuman Khandual
2017-02-01  6:40         ` Anshuman Khandual
2017-01-30  3:35 ` [RFC V2 04/12] mm: Change mbind(MPOL_BIND) implementation for CDM nodes Anshuman Khandual
2017-01-30  3:35 ` [RFC V2 05/12] cpuset: Add cpuset_inc() inside cpuset_init() Anshuman Khandual
2017-01-30 17:36   ` Dave Hansen
2017-01-30 20:30   ` Mel Gorman
2017-01-31 14:22     ` [RFC] cpuset: Enable changing of top_cpuset's mems_allowed nodemask Anshuman Khandual
2017-01-31 16:00       ` Mel Gorman
2017-02-01  7:31         ` Anshuman Khandual
2017-02-01  8:53           ` Michal Hocko
2017-02-01  9:18           ` Mel Gorman
2017-01-31 14:36     ` [RFC V2 05/12] cpuset: Add cpuset_inc() inside cpuset_init() Vlastimil Babka
2017-01-31 15:30       ` Anshuman Khandual
2017-01-30  3:35 ` [RFC V2 06/12] mm: Exclude CDM nodes from task->mems_allowed and root cpuset Anshuman Khandual
2017-01-30  3:35 ` [RFC V2 07/12] mm: Ignore cpuset enforcement when allocation flag has __GFP_THISNODE Anshuman Khandual
2017-01-30  3:35 ` [RFC V2 08/12] mm: Add new VMA flag VM_CDM Anshuman Khandual
2017-01-30 18:52   ` Jerome Glisse
2017-01-31  4:22     ` Anshuman Khandual
2017-01-31  6:05       ` Jerome Glisse
2017-01-30  3:35 ` [RFC V2 09/12] mm: Exclude CDM marked VMAs from auto NUMA Anshuman Khandual
2017-01-30  3:35 ` [RFC V2 10/12] mm: Ignore madvise(MADV_MERGEABLE) request for VM_CDM marked VMAs Anshuman Khandual
2017-01-30  3:35 ` [RFC V2 11/12] mm: Tag VMA with VM_CDM flag during page fault Anshuman Khandual
2017-01-30 17:51   ` Dave Hansen
2017-01-31  5:10     ` Anshuman Khandual
2017-01-31 17:54       ` Dave Hansen
2017-01-30  3:35 ` [RFC V2 12/12] mm: Tag VMA with VM_CDM flag explicitly during mbind(MPOL_BIND) Anshuman Khandual
2017-01-30 17:54   ` Dave Hansen
2017-01-31  4:36     ` Anshuman Khandual
2017-02-07 18:07       ` Dave Hansen
2017-02-08 14:13         ` Anshuman Khandual
2017-02-08 15:04         ` Jerome Glisse
2017-01-30  3:35 ` [DEBUG 13/21] powerpc/mm: Identify coherent device memory nodes during platform init Anshuman Khandual
2017-01-30  3:35 ` [DEBUG 14/21] powerpc/mm: Create numa nodes for hotplug memory Anshuman Khandual
2017-01-30  3:35 ` [DEBUG 15/21] powerpc/mm: Enable CONFIG_MOVABLE_NODE for PPC64 platform Anshuman Khandual
2017-01-30  3:35 ` [DEBUG 16/21] mm: Enable CONFIG_MOVABLE_NODE on powerpc Anshuman Khandual
2017-01-30  3:35 ` [DEBUG 17/21] mm: Export definition of 'zone_names' array through mmzone.h Anshuman Khandual
2017-01-30  3:35 ` [DEBUG 18/21] mm: Add debugfs interface to dump each node's zonelist information Anshuman Khandual
2017-01-30  3:36 ` [DEBUG 19/21] mm: Add migrate_virtual_range migration interface Anshuman Khandual
2017-01-30  3:36 ` [DEBUG 20/21] drivers: Add two drivers for coherent device memory tests Anshuman Khandual
2017-01-30  3:36 ` [DEBUG 21/21] selftests/powerpc: Add a script to perform random VMA migrations Anshuman Khandual
2017-01-31  5:48 ` [RFC V2 00/12] Define coherent device memory node Anshuman Khandual
2017-01-31  6:15   ` Jerome Glisse

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1ab159b5-1b67-9dae-4112-3360d8f909fd@linux.vnet.ibm.com \
    --to=khandual@linux.vnet.ibm.com \
    --cc=aneesh.kumar@linux.vnet.ibm.com \
    --cc=bsingharora@gmail.com \
    --cc=dan.j.williams@intel.com \
    --cc=dave.hansen@intel.com \
    --cc=haren@linux.vnet.ibm.com \
    --cc=jglisse@redhat.com \
    --cc=jhubbard@nvidia.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=mgorman@suse.de \
    --cc=mhocko@suse.com \
    --cc=minchan@kernel.org \
    --cc=srikar@linux.vnet.ibm.com \
    --cc=vbabka@suse.cz \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).