Re: [patch 13/14] x86_64: Use common functions in cluster and physflat mode

[Ashok Raj]

On Fri, Sep 09, 2005 at 10:07:28AM -0700, Zwane Mwaikambo wrote:
> On a slightly different topic, how come we're using physflat for hotplug 
> cpu?
> 
> -#ifndef CONFIG_CPU_HOTPLUG
>  		/* In the CPU hotplug case we cannot use broadcast mode
>  		   because that opens a race when a CPU is removed.
> -		   Stay at physflat mode in this case.
> -		   It is bad to do this unconditionally though. Once
> -		   we have ACPI platform support for CPU hotplug
> -		   we should detect hotplug capablity from ACPI tables and
> -		   only do this when really needed. -AK */
> +		   Stay at physflat mode in this case. - AK */
> +#ifdef CONFIG_HOTPLUG_CPU
>  		if (num_cpus <= 8)
>  			genapic = &apic_flat;

What you say was true before this patch, (Although now that you point out i 
realize the ifdef CONFIG_HOTPLUG_CPU is not required). 

Think Andi is fixing this in his next drop to -mm*

When physflat was introduced, it also  switched to use physflat mode for 
#cpus <=8 when hotplug is enabled, since it doesnt use shortcuts and 
so is also safer (although slower). 

http://marc.theaimsgroup.com/?l=linux-kernel&m=112317686712929&w=2

The link above made using genapic_flat safer by using the
flat_send_IPI_mask(), and hence i switched back to using
logical flat mode when #cpus <=8, since that a little more efficient than
the send_IPI_mask_sequence() used in physflat mode.

In general we need

flat_mode - #cpus <= 8 (Hotplug defined or not, so we use mask version 
                       for safety)

physflat or cluster_mode when #cpus >8.

If we choose physflat as default for #cpus <=8 (with hotplug) would make
IPI performance worse, since it would do one cpu at a time, and requires 2 
writes per cpu for each IPI v.s just 2 for a flat mode mask version of the API.

-- 
Cheers,
Ashok Raj
- Open Source Technology Center
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/