Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation process

From	Dave Hansen <dave.hansen@intel.com>
Newsgroups	linux.kernel
Subject	Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation process
Date	2017-01-31 19:10 +0100
Message-ID	<t5L2p-3UK-3@gated-at.bofh.it> (permalink)
References	(1 earlier) <t5aYV-7mm-5@gated-at.bofh.it> <t5o5P-6Mu-3@gated-at.bofh.it> <t5vAl-2Rg-1@gated-at.bofh.it> <t5vTI-2Xq-9@gated-at.bofh.it> <t5BcJ-6lb-9@gated-at.bofh.it>
Organization	linux.* mail to news gateway

Show all headers | View raw

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?

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.

Thread

Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation  process Anshuman Khandual <khandual@linux.vnet.ibm.com> - 2017-01-31 02:40 +0100
  Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation  process Dave Hansen <dave.hansen@intel.com> - 2017-01-31 03:00 +0100
    Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation  process John Hubbard <jhubbard@nvidia.com> - 2017-01-31 08:40 +0100
      Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation  process Dave Hansen <dave.hansen@intel.com> - 2017-01-31 19:10 +0100
        Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation  process David Nellans <dnellans@nvidia.com> - 2017-01-31 20:20 +0100
        Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation  process Anshuman Khandual <khandual@linux.vnet.ibm.com> - 2017-02-01 08:00 +0100
      Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation  process Anshuman Khandual <khandual@linux.vnet.ibm.com> - 2017-02-01 07:50 +0100
    Re: [RFC V2 03/12] mm: Change generic FALLBACK zonelist creation  process Anshuman Khandual <khandual@linux.vnet.ibm.com> - 2017-02-01 07:50 +0100

csiph-web