From: Jeremy Fitzhardinge Subject: Re: [PATCH 02/20] x86/ticketlock: convert spin loop to C Date: Wed, 03 Nov 2010 11:38:59 -0400 Message-ID: <4CD18213.4030105@goop.org> References: <3d9083892c28a7d3ed5a0191b0b4cd013022a186.1288794124.git.jeremy.fitzhardinge@citrix.com> <1288797092.2511.141.camel@edumazet-laptop> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: quoted-printable Return-path: In-Reply-To: <1288797092.2511.141.camel@edumazet-laptop> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Eric Dumazet Cc: Nick Piggin , Xen-devel , Peter Zijlstra , Srivatsa Vaddagiri , Linux Kernel Mailing List , Jan Beulich , Linux Virtualization , Jeremy Fitzhardinge , Avi Kivity , "H. Peter Anvin" List-Id: xen-devel@lists.xenproject.org On 11/03/2010 11:11 AM, Eric Dumazet wrote: > Le mercredi 03 novembre 2010 =C3=A0 10:59 -0400, Jeremy Fitzhardinge a > =C3=A9crit : >> From: Jeremy Fitzhardinge >> >> The inner loop of __ticket_spin_lock isn't doing anything very special= , >> so reimplement it in C. >> >> For the 8 bit ticket lock variant, we use a register union to get dire= ct >> access to the lower and upper bytes in the tickets, but unfortunately = gcc >> won't generate a direct comparison between the two halves of the regis= ter, >> so the generated asm isn't quite as pretty as the hand-coded version. >> However benchmarking shows that this is actually a small improvement i= n >> runtime performance on some benchmarks, and never a slowdown. >> >> We also need to make sure there's a barrier at the end of the lock loo= p >> to make sure that the compiler doesn't move any instructions from with= in >> the locked region into the region where we don't yet own the lock. >> >> Signed-off-by: Jeremy Fitzhardinge >> --- >> arch/x86/include/asm/spinlock.h | 58 +++++++++++++++++++-----------= -------- >> 1 files changed, 29 insertions(+), 29 deletions(-) >> >> diff --git a/arch/x86/include/asm/spinlock.h b/arch/x86/include/asm/sp= inlock.h >> index d6d5784..6711d36 100644 >> --- a/arch/x86/include/asm/spinlock.h >> +++ b/arch/x86/include/asm/spinlock.h >> @@ -58,21 +58,21 @@ >> #if (NR_CPUS < 256) >> static __always_inline void __ticket_spin_lock(arch_spinlock_t *lock) >> { >> - unsigned short inc =3D 1 << TICKET_SHIFT; >> - >> - asm volatile ( >> - LOCK_PREFIX "xaddw %w0, %1\n" >> - "1:\t" >> - "cmpb %h0, %b0\n\t" >> - "je 2f\n\t" >> - "rep ; nop\n\t" >> - "movb %1, %b0\n\t" >> - /* don't need lfence here, because loads are in-order */ >> - "jmp 1b\n" >> - "2:" >> - : "+Q" (inc), "+m" (lock->slock) >> - : >> - : "memory", "cc"); >> + register union { >> + struct __raw_tickets tickets; >> + unsigned short slock; >> + } inc =3D { .slock =3D 1 << TICKET_SHIFT }; >> + >> + asm volatile (LOCK_PREFIX "xaddw %w0, %1\n" >> + : "+Q" (inc), "+m" (lock->slock) : : "memory", "cc"); >> + >> + for (;;) { >> + if (inc.tickets.head =3D=3D inc.tickets.tail) >> + return; >> + cpu_relax(); >> + inc.tickets.head =3D ACCESS_ONCE(lock->tickets.head); >> + } >> + barrier(); /* make sure nothing creeps before the lock is taken */ > Isnt this barrier() never reached ? Sorry, a later patch makes this clearer. I should have folded it in. J