Re: [driver-core PATCH v6 2/9] async: Add support for queueing on specific NUMA node
From: Bart Van Assche
Date: Thu Nov 08 2018 - 18:36:44 EST
On Thu, 2018-11-08 at 10:06 -0800, Alexander Duyck wrote:
+AD4 Introduce four new variants of the async+AF8-schedule+AF8 functions that allow
+AD4 scheduling on a specific NUMA node.
+AD4
+AD4 The first two functions are async+AF8-schedule+AF8-near and
+AD4 async+AF8-schedule+AF8-near+AF8-domain end up mapping to async+AF8-schedule and
+AD4 async+AF8-schedule+AF8-domain, but provide NUMA node specific functionality. They
+AD4 replace the original functions which were moved to inline function
+AD4 definitions that call the new functions while passing NUMA+AF8-NO+AF8-NODE.
+AD4
+AD4 The second two functions are async+AF8-schedule+AF8-dev and
+AD4 async+AF8-schedule+AF8-dev+AF8-domain which provide NUMA specific functionality when
+AD4 passing a device as the data member and that device has a NUMA node other
+AD4 than NUMA+AF8-NO+AF8-NODE.
+AD4
+AD4 The main motivation behind this is to address the need to be able to
+AD4 schedule device specific init work on specific NUMA nodes in order to
+AD4 improve performance of memory initialization.
Reviewed-by: Bart Van Assche +ADw-bvanassche+AEA-acm.org+AD4