lkml.org 
[lkml]   [2008]   [Feb]   [26]   [last100]   RSS Feed
Views: [more markup]  [less markup]  [headers]  [forward] 
 
Messages in this thread
/
DateTue, 26 Feb 2008 11:34:41 -0600
FromPaul Jackson <>
SubjectRe: [patch 6/6] mempolicy: update NUMA memory policy documentation
+	MPOL_F_RELATIVE_NODES:  This flag specifies that the nodemask passed
+	by the user should remain in the same context as it is for the
+	current task or VMA's set of accessible nodes after the memory
+	policy has been defined.
+
+	    Without this flag (and without MPOL_F_STATIC_NODES), anytime a
+	    mempolicy is rebound because of a change in the set of
+	    accessible nodes, the node (Preferred) or nodemask (Bind,
+	    Interleave) is remapped to the new set of accessible nodes.
+	    With this flag, the remap is done to ensure the context of the
+	    previous nodemask with its set of allowed mems is preserved.

Hmmm ... I've read this several times now ... still can't figure out
what it's saying ;).  And it doesn't really explain key aspects of
MPOL_F_RELATIVE_NODES, such as that it provides cpuset relative
numbering (use nodes 0..N-1, regardless of your current cpuset, to
refer to the first N nodes in whatever is your current cpuset.)

Perhaps we'd be further ahead of the game if you started with the
documentation changes to Documentation/vm/numa_memory_policy.txt,
in my patch:
	Date: Sun, 23 Dec 2007 22:24:02 -0600
	From: Paul Jackson <pj@sgi.com>
	To: David Rientjes <rientjes@google.com>
	Cc: Lee.Schermerhorn@hp.com, clameter@sgi.com
	Subject: Re: [RFC] cpuset relative memory policies - second choice
	Message-Id: <20071223222402.5486bf0a.pj@sgi.com>

Change MPOL_MASK_REL to MPOL_F_RELATIVE_NODES and similar changes.

-- 
                  I won't rest till it's the best ...
                  Programmer, Linux Scalability
                  Paul Jackson <pj@sgi.com> 1.940.382.4214


\
 
 \ /
  Last update: 2008-02-26 18:37    [W:0.620 / U:0.060 seconds]
©2003-2008 Jasper Spaans