Re: [patch] reduce IPI noise due to /dev/cdrom open/close

[Milton Miller]

On Jul 4, 2006, at 1:39 AM, Andrew Morton wrote:

> I expect raw_smp_processor_id() is used here as a a microoptimisation -
> avoid a might_sleep() which obviously will never trigger.
>
> But I think it'd be better to do just a single raw_smp_processor_id() 
> for this entire function:
>
> static void bh_lru_install(struct buffer_head *bh)
> {
> struct buffer_head *evictee = NULL;
> struct bh_lru *lru;
> + int cpu;
>
> check_irqs_on();
> bh_lru_lock();
> + cpu = raw_smp_processor_id();
> - lru = &__get_cpu_var(bh_lrus);
> + lru = per_cpu(bh_lrus, cpu);

The problem with this style is that it is an disoptimizatoin for 
architectures who hold their per-cpu data offset in a register, put the 
smp_processor_id in the per-cpu data (or similar) and per_cpu data 
offsets in a global lookup.

Do we need a new macro? (what is that gcc macro function syntax?)

#ifdef PER_CPU_IS_SLOW
#define my_cpu_var(bh_lrus, cpu) \
	({ BUG_ON(cpu != smp_processor_id()); 	&__get_cpu_var(bh_lrus); })
#else
#define my_cpu_var(bh_lrus, cpu) \
	({ BUG_ON(cpu != smp_processor_id()); per_cpu(bh_lrus, cpu); })
#endif

(and yes, the BUG_ON would be for debug checking).

milton

-
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/