On 2014/11/27 18:44, Thomas Gleixner wrote: > On Tue, 25 Nov 2014, Jiang Liu wrote: >> Introduce mechanism to support different vector allocation policies, >> so platform or user may choose the best suitable CPU vector allocation > > I've postponed this and the next patch for 3.20. Hi Thomas, That's fine, thanks for merging all other patches for 3.19:) > >> policy. Currently two policies are supported: >> 1) allocate CPU vector from cpumask_of_node(dev_to_node(dev)) >> 2) allocate from apic->target_cpus(), this is the default policy >> >> Platform driver may call set_vector_alloc_policy() to choose the >> preferred policies. >> >> This mechanism may be used to support NumaConnect systems to allocate >> CPU vectors from device local node. > > So these policies are system wide and need to be selected at boot time? Yes, the polices are system wide. It may be change at runtime, but seems no strong requirement yet so haven't implemented the interface to change it at runtime. Basically this idea is inspired by a patch set to allocate CPU vector from device local node for NumaConnect systems. I found it may also help 2-socket or 4-socket systems, so reimplemented in this way. Thanks! Gerry > > Thanks, > > tglx > -- To unsubscribe from this list: send the line "unsubscribe linux-pci" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
