Re: [PATCH 6/6] x86/ticketlock: make __ticket_spin_trylock common

[Jeremy Fitzhardinge <jeremy@goop.org>]

On 01/24/2011 05:16 PM, Nick Piggin wrote:
> On Tue, Jan 25, 2011 at 10:41 AM, Jeremy Fitzhardinge <jeremy@goop.org> wrote:
>> From: Jeremy Fitzhardinge <jeremy.fitzhardinge@citrix.com>
>>
>> Make trylock code common regardless of ticket size.
> What's the asm for this look like?

Asm:

        movzwl (%rdi), %eax     # lock_1(D)->slock, tmp
        cmpb %ah,%al    # tmp
        leal 0x100(%rax), %edx  # tmp, new
        jne 1f
        lock; cmpxchgw %dx,(%rdi)       # new, lock_1(D)->slock
1:	sete %dl      # new
        movzbl %dl,%eax # new, tmp



C:

        movw    (%rdi), %dx     # lock_2(D)->D.5949.tickets, old
        xorl    %eax, %eax      # D.13954
        movzbl  %dh, %ecx       # old, tmp70
        cmpb    %dl, %cl        # old, tmp70
        jne     .L5     #,
        leal    256(%rdx), %ecx #, D.13956
        movl    %edx, %eax      # old, __ret
        lock; cmpxchgw %cx,(%rdi)       # D.13956,* lock
        cmpw    %dx, %ax        # old, __ret
        sete    %al     #, D.13954
        movzbl  %al, %eax       # D.13954, D.13954
.L5:


The C version can't take advantage of the fact that the cmpxchg directly
sets the flags, so it ends up re-comparing the old and swapped-out
values to set the return.  And it doesn't re-use the same sete to set
the return value in the quick failed-to-acquire path.

It might be worth having a generic cmpxchg() variant which returns a
succeed/fail flag rather than the fetched value, to avoid comparison in
this case - since many (most?) cmpxchg() callers end up doing that
comparison.

How performance critical is trylock?  I guess the ones in fs/dcache.c
are the ones looming large in your mind.

    J