Hi Rik, On Sun, Jul 01, 2012 at 11:33:17AM -0400, Rik van Riel wrote: > On 06/28/2012 08:56 AM, Andrea Arcangeli wrote: > > > diff --git a/kernel/fork.c b/kernel/fork.c > > index 0adbe09..3e5a0d9 100644 > > --- a/kernel/fork.c > > +++ b/kernel/fork.c > > @@ -527,6 +527,8 @@ static void mm_init_aio(struct mm_struct *mm) > > > > static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct > > *p) > > { > > + if (unlikely(alloc_mm_autonuma(mm))) > > + goto out_free_mm; > > atomic_set(&mm->mm_users, 1); > > atomic_set(&mm->mm_count, 1); > > init_rwsem(&mm->mmap_sem); > > I wonder if it would be possible to defer the allocation > of the mm_autonuma struct to knuma_scand, so short lived > processes never have to allocate and free the mm_autonuma > structure. > > That way we only have a function call at exit time, and > the branch inside kfree that checks for a null pointer.
It would be possible to convert them to prepare_mm/task_autonuma (the mm side especially would be a branch once in a while) but it would then become impossible to inherit the mm/task stats across fork/clone. Right now the default is to reset them, but two sysfs switches control that, and I wouldn't drop those until I've the time to experiment how large kernel builds are affected by enabling the stats inheritance. Right now kernel builds are unaffected because of the default stat-resetting behavior and gcc too quick to be measured. -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/