Re: aio-core why not using SuS? [Re: [rfc] aio-core for 2.5.29 (Re: async-io API registration for 2.5.29)]

[Andrea Arcangeli]

On Sat, Aug 17, 2002 at 10:02:23AM -0700, Linus Torvalds wrote:
> So to make this work, you'd have to have:
>  - architecture-specific hacks
>  - realize that not all architectures can do it at all, so the places that 
>    depend on this would have to have some abstraction that makes it go 
>    away when not needed.
>  - fix up lazy TLB switching (conditionally on the hack).
> 
> It just sounds really messy to me.

Indeed. Assuming this is an hack under a CONFIG_X86_NUMA_HACK hardwired
for certain config options and certain architecture, the tlb flushing
across threads sounds the worst part in particular because it's an x86.

Andrea