Messages in this thread |  | | From | Nick Piggin <> | | Subject | Re: [PATCH 0 of 9] x86/smp function calls: convert x86 tlb flushes to use function calls [POST 2] | | Date | Tue, 19 Aug 2008 20:49:51 +1000 |
| |
On Tuesday 19 August 2008 20:24, Ingo Molnar wrote: > * Nick Piggin <nickpiggin@yahoo.com.au> wrote: > > > At least we could/should perhaps standardize/generalize all the > > > 'specific' IPI handlers into the smp_function_call() framework: if > > > function address equals to a pre-cooked IPI entry point we could > > > call that function without a kmalloc. As these are all hardwired, > > > __builtin_is_constant_p() could come to the help as well. Hm? > > > > No, it's not just the function call but also payload, list entry for > > queue, scoreboard of CPUs have processed it, a lock, etc etc etc. > > > > smp_call_function is *always* going to be heavier than a hard wired > > special case, no matter how it is implemented. For such low level > > performance critical functionality, I miss the days when people were > > rabid about saving every cycle rather than every line of code ;) > > no, i was thinking about really high level hardwiring, i.e. hardwiring > the _function pointer_ knowledge into smp_function_call(). > > for example for the reschedule IPI, it would be hardwired on x86 to just > call into the special IPI handler, via: > > smp_call_function_mask(target_mask, smp_send_reschedule, NULL, 0); > > Exactly same cost and call sequence as a direct hardwired-to-IPI > function call (and the same underlying mechanism) - just consolidated > around a single cross-call API.
So it would magically use a different IPI vector, magically be callable from interrupt context, have a different function type etc.? Why would we do such a thing? If it doesn't walk like a duck and doesn't quack like a duck, why should we call it a duck? ;)
> Same for all the other special cross-CPU handlers. That way some > architectures would hardwire it, some wouldnt, etc.
I think the right way is just to keep the "special" classes of IPI-ish functions (I guess reschedule and TLB flush importantly). The arch code is still free to implement these using their call function or a special vector or whatever.
|  |