Messages in this thread | | | Date | Wed, 04 Sep 2013 10:52:51 -0400 | From | Waiman Long <> | Subject | Re: [PATCH v7 1/4] spinlock: A new lockref structure for lockless update of refcount |
| |
On 09/03/2013 03:09 PM, Linus Torvalds wrote: > On Tue, Sep 3, 2013 at 8:34 AM, Linus Torvalds > <torvalds@linux-foundation.org> wrote: >> I suspect the tty_ldisc_lock() could be made to go away if we care. > Heh. I just pulled the tty patches from Greg, and the locking has > changed completely. > > It may actually fix your AIM7 test-case, because while the global > spinlock remains (it got renamed to "tty_ldiscs_lock" - there's an > added "s"), the common operations now take the per-tty lock to get the > ldisc for that tty, rather than that global spinlock (which just > protects the actual ldisk array now). > > That said, I don't know what AIM7 really ends up doing, but your > profile seems to have every access through tty_ldisc_[de]ref() that > now uses only the per-tty lock. Of course, how much that helps ends up > depending on whether AIM7 uses lots of tty's or just one shared one. > > Anyway, it might be worth testing my current -git tree. > > Linus
The latest tty patches did work. The tty related spinlock contention is now completely gone. The short workload can now reach over 8M JPM which is the highest I have ever seen.
The perf profile was:
5.85% reaim reaim [.] mul_short 4.87% reaim [kernel.kallsyms] [k] ebitmap_get_bit 4.72% reaim reaim [.] mul_int 4.71% reaim reaim [.] mul_long 2.67% reaim libc-2.12.so [.] __random_r 2.64% reaim [kernel.kallsyms] [k] lockref_get_not_zero 1.58% reaim [kernel.kallsyms] [k] copy_user_generic_string 1.48% reaim [kernel.kallsyms] [k] mls_level_isvalid 1.35% reaim [kernel.kallsyms] [k] find_next_bit 1.23% reaim [kernel.kallsyms] [k] system_call 1.21% reaim libc-2.12.so [.] memcpy 1.19% reaim [kernel.kallsyms] [k] _raw_spin_lock 1.06% reaim [kernel.kallsyms] [k] avc_has_perm_flags 1.04% reaim libc-2.12.so [.] __srandom_r 1.02% reaim reaim [.] newton_raphson 1.01% reaim [kernel.kallsyms] [k] update_cfs_rq_blocked_load 0.98% reaim [kernel.kallsyms] [k] fsnotify 0.94% reaim [kernel.kallsyms] [k] avtab_search_node 0.91% reaim libm-2.12.so [.] __sincos
I have a patch in linux-next that should eliminate ebitmap_get_bit, mls_leve_isvalid and find_next_bit from the list top once it is merged.
Regards, Longman
| |