From: Peter Zijlstra <peterz@infradead.org>
To: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Ingo Molnar <mingo@kernel.org>, Andi Kleen <ak@linux.intel.com>,
Peter Anvin <hpa@zytor.com>, Mike Galbraith <bitbucket@online.de>,
Thomas Gleixner <tglx@linutronix.de>,
Arjan van de Ven <arjan@linux.intel.com>,
Frederic Weisbecker <fweisbec@gmail.com>,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
"linux-arch@vger.kernel.org" <linux-arch@vger.kernel.org>
Subject: Re: [PATCH 0/7] preempt_count rework -v2
Date: Thu, 12 Sep 2013 04:20:40 +0200 [thread overview]
Message-ID: <20130912022040.GT31370@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <CA+55aFxDbqeFumkVApf-L2pi+FQn06GQqQf4rWkzOC4LF=fn=A@mail.gmail.com>
On Wed, Sep 11, 2013 at 04:02:14PM -0700, Linus Torvalds wrote:
> On Wed, Sep 11, 2013 at 11:59 AM, Peter Zijlstra <peterz@infradead.org> wrote:
> >
> > OK, stripped it down further, I couldn't quite see how to collapse the
> > unary and binary operator variants though :/
>
> Ok, this looks pretty good. I assume it works too? ;)
Only compile tested that one.. the below is kvm boot tested until root
mount -- I'll try on actual hardware when I've gotten some sleep.
I split the thing up into two macros GEN_UNARY_RMWcc and
GEN_BINARY_RMWcc which ends up being more readable as well as smaller
code overall.
I also attempted to convert asm/bitops.h, although I'm not sure it'll
compile right with older GCCs due to the comment near BITOP_ADDR()
It also changes the fallback from sbb %0,%0 to setc %0, which afaict
should be similar but I'm not too familiar with all that.
I might have to add the clobber to the macro arguments so we can do
version without "memory" clobber, although bitops is inconsistent with
that as well, __test_and_clear_bit() doesn't have a memory clobber but
__test_and_change_bit() does.
---
arch/x86/include/asm/atomic.h | 29 +++---------------
arch/x86/include/asm/atomic64_64.h | 28 ++---------------
arch/x86/include/asm/bitops.h | 58 +++++--------------------------------
arch/x86/include/asm/local.h | 28 ++---------------
arch/x86/include/asm/rmwcc.h | 52 +++++++++++++++++++++++++++++++++
5 files changed, 73 insertions(+), 122 deletions(-)
--- a/arch/x86/include/asm/atomic.h
+++ b/arch/x86/include/asm/atomic.h
@@ -6,6 +6,7 @@
#include <asm/processor.h>
#include <asm/alternative.h>
#include <asm/cmpxchg.h>
+#include <asm/rmwcc.h>
/*
* Atomic operations that C can't guarantee us. Useful for
@@ -76,12 +77,7 @@ static inline void atomic_sub(int i, ato
*/
static inline int atomic_sub_and_test(int i, atomic_t *v)
{
- unsigned char c;
-
- asm volatile(LOCK_PREFIX "subl %2,%0; sete %1"
- : "+m" (v->counter), "=qm" (c)
- : "ir" (i) : "memory");
- return c;
+ GEN_BINARY_RMWcc(LOCK_PREFIX "subl", v->counter, i, "%0", "e");
}
/**
@@ -118,12 +114,7 @@ static inline void atomic_dec(atomic_t *
*/
static inline int atomic_dec_and_test(atomic_t *v)
{
- unsigned char c;
-
- asm volatile(LOCK_PREFIX "decl %0; sete %1"
- : "+m" (v->counter), "=qm" (c)
- : : "memory");
- return c != 0;
+ GEN_UNARY_RMWcc(LOCK_PREFIX "decl", v->counter, "%0", "e");
}
/**
@@ -136,12 +127,7 @@ static inline int atomic_dec_and_test(at
*/
static inline int atomic_inc_and_test(atomic_t *v)
{
- unsigned char c;
-
- asm volatile(LOCK_PREFIX "incl %0; sete %1"
- : "+m" (v->counter), "=qm" (c)
- : : "memory");
- return c != 0;
+ GEN_UNARY_RMWcc(LOCK_PREFIX "incl", v->counter, "%0", "e");
}
/**
@@ -155,12 +141,7 @@ static inline int atomic_inc_and_test(at
*/
static inline int atomic_add_negative(int i, atomic_t *v)
{
- unsigned char c;
-
- asm volatile(LOCK_PREFIX "addl %2,%0; sets %1"
- : "+m" (v->counter), "=qm" (c)
- : "ir" (i) : "memory");
- return c;
+ GEN_BINARY_RMWcc(LOCK_PREFIX "addl", v->counter, i, "%0", "s");
}
/**
--- a/arch/x86/include/asm/atomic64_64.h
+++ b/arch/x86/include/asm/atomic64_64.h
@@ -72,12 +72,7 @@ static inline void atomic64_sub(long i,
*/
static inline int atomic64_sub_and_test(long i, atomic64_t *v)
{
- unsigned char c;
-
- asm volatile(LOCK_PREFIX "subq %2,%0; sete %1"
- : "=m" (v->counter), "=qm" (c)
- : "er" (i), "m" (v->counter) : "memory");
- return c;
+ GEN_BINARY_RMWcc(LOCK_PREFIX "subq", v->counter, i, "%0", "e");
}
/**
@@ -116,12 +111,7 @@ static inline void atomic64_dec(atomic64
*/
static inline int atomic64_dec_and_test(atomic64_t *v)
{
- unsigned char c;
-
- asm volatile(LOCK_PREFIX "decq %0; sete %1"
- : "=m" (v->counter), "=qm" (c)
- : "m" (v->counter) : "memory");
- return c != 0;
+ GEN_UNARY_RMWcc(LOCK_PREFIX "decq", v->counter, "%0", "e");
}
/**
@@ -134,12 +124,7 @@ static inline int atomic64_dec_and_test(
*/
static inline int atomic64_inc_and_test(atomic64_t *v)
{
- unsigned char c;
-
- asm volatile(LOCK_PREFIX "incq %0; sete %1"
- : "=m" (v->counter), "=qm" (c)
- : "m" (v->counter) : "memory");
- return c != 0;
+ GEN_UNARY_RMWcc(LOCK_PREFIX "incq", v->counter, "%0", "e");
}
/**
@@ -153,12 +138,7 @@ static inline int atomic64_inc_and_test(
*/
static inline int atomic64_add_negative(long i, atomic64_t *v)
{
- unsigned char c;
-
- asm volatile(LOCK_PREFIX "addq %2,%0; sets %1"
- : "=m" (v->counter), "=qm" (c)
- : "er" (i), "m" (v->counter) : "memory");
- return c;
+ GEN_BINARY_RMWcc(LOCK_PREFIX "addq", v->counter, i, "%0", "s");
}
/**
--- a/arch/x86/include/asm/bitops.h
+++ b/arch/x86/include/asm/bitops.h
@@ -14,6 +14,7 @@
#include <linux/compiler.h>
#include <asm/alternative.h>
+#include <asm/rmwcc.h>
#if BITS_PER_LONG == 32
# define _BITOPS_LONG_SHIFT 5
@@ -204,12 +205,7 @@ static inline void change_bit(long nr, v
*/
static inline int test_and_set_bit(long nr, volatile unsigned long *addr)
{
- int oldbit;
-
- asm volatile(LOCK_PREFIX "bts %2,%1\n\t"
- "sbb %0,%0" : "=r" (oldbit), ADDR : "Ir" (nr) : "memory");
-
- return oldbit;
+ GEN_BINARY_RMWcc(LOCK_PREFIX "bts", *addr, nr, "%0", "c");
}
/**
@@ -236,13 +232,7 @@ test_and_set_bit_lock(long nr, volatile
*/
static inline int __test_and_set_bit(long nr, volatile unsigned long *addr)
{
- int oldbit;
-
- asm("bts %2,%1\n\t"
- "sbb %0,%0"
- : "=r" (oldbit), ADDR
- : "Ir" (nr));
- return oldbit;
+ GEN_BINARY_RMWcc("bts", *addr, nr, "%0", "c");
}
/**
@@ -255,13 +245,7 @@ static inline int __test_and_set_bit(lon
*/
static inline int test_and_clear_bit(long nr, volatile unsigned long *addr)
{
- int oldbit;
-
- asm volatile(LOCK_PREFIX "btr %2,%1\n\t"
- "sbb %0,%0"
- : "=r" (oldbit), ADDR : "Ir" (nr) : "memory");
-
- return oldbit;
+ GEN_BINARY_RMWcc(LOCK_PREFIX "btr", *addr, nr, "%0", "c");
}
/**
@@ -282,26 +266,13 @@ static inline int test_and_clear_bit(lon
*/
static inline int __test_and_clear_bit(long nr, volatile unsigned long *addr)
{
- int oldbit;
-
- asm volatile("btr %2,%1\n\t"
- "sbb %0,%0"
- : "=r" (oldbit), ADDR
- : "Ir" (nr));
- return oldbit;
+ GEN_BINARY_RMWcc("btr", *addr, nr, "%0", "c");
}
/* WARNING: non atomic and it can be reordered! */
static inline int __test_and_change_bit(long nr, volatile unsigned long *addr)
{
- int oldbit;
-
- asm volatile("btc %2,%1\n\t"
- "sbb %0,%0"
- : "=r" (oldbit), ADDR
- : "Ir" (nr) : "memory");
-
- return oldbit;
+ GEN_BINARY_RMWcc("btc", *addr, nr, "%0", "c");
}
/**
@@ -314,13 +285,7 @@ static inline int __test_and_change_bit(
*/
static inline int test_and_change_bit(long nr, volatile unsigned long *addr)
{
- int oldbit;
-
- asm volatile(LOCK_PREFIX "btc %2,%1\n\t"
- "sbb %0,%0"
- : "=r" (oldbit), ADDR : "Ir" (nr) : "memory");
-
- return oldbit;
+ GEN_BINARY_RMWcc(LOCK_PREFIX "btc", *addr, nr, "%0", "c");
}
static __always_inline int constant_test_bit(long nr, const volatile unsigned long *addr)
@@ -331,14 +296,7 @@ static __always_inline int constant_test
static inline int variable_test_bit(long nr, volatile const unsigned long *addr)
{
- int oldbit;
-
- asm volatile("bt %2,%1\n\t"
- "sbb %0,%0"
- : "=r" (oldbit)
- : "m" (*(unsigned long *)addr), "Ir" (nr));
-
- return oldbit;
+ GEN_BINARY_RMWcc("bt", *(volatile unsigned long *)addr, nr, "%0", "c");
}
#if 0 /* Fool kernel-doc since it doesn't do macros yet */
--- a/arch/x86/include/asm/local.h
+++ b/arch/x86/include/asm/local.h
@@ -52,12 +52,7 @@ static inline void local_sub(long i, loc
*/
static inline int local_sub_and_test(long i, local_t *l)
{
- unsigned char c;
-
- asm volatile(_ASM_SUB "%2,%0; sete %1"
- : "+m" (l->a.counter), "=qm" (c)
- : "ir" (i) : "memory");
- return c;
+ GEN_BINARY_RMWcc(_ASM_SUB, l->a.counter, i, "%0", "e");
}
/**
@@ -70,12 +65,7 @@ static inline int local_sub_and_test(lon
*/
static inline int local_dec_and_test(local_t *l)
{
- unsigned char c;
-
- asm volatile(_ASM_DEC "%0; sete %1"
- : "+m" (l->a.counter), "=qm" (c)
- : : "memory");
- return c != 0;
+ GEN_UNARY_RMWcc(_ASM_DEC, l->a.counter, "%0", "e");
}
/**
@@ -88,12 +78,7 @@ static inline int local_dec_and_test(loc
*/
static inline int local_inc_and_test(local_t *l)
{
- unsigned char c;
-
- asm volatile(_ASM_INC "%0; sete %1"
- : "+m" (l->a.counter), "=qm" (c)
- : : "memory");
- return c != 0;
+ GEN_UNARY_RMWcc(_ASM_INC, l->a.counter, "%0", "e");
}
/**
@@ -107,12 +92,7 @@ static inline int local_inc_and_test(loc
*/
static inline int local_add_negative(long i, local_t *l)
{
- unsigned char c;
-
- asm volatile(_ASM_ADD "%2,%0; sets %1"
- : "+m" (l->a.counter), "=qm" (c)
- : "ir" (i) : "memory");
- return c;
+ GEN_BINARY_RMWcc(_ASM_ADD, l->a.counter, i, "%0", "s");
}
/**
--- /dev/null
+++ b/arch/x86/include/asm/rmwcc.h
@@ -0,0 +1,52 @@
+#ifndef _ASM_X86_RMWcc
+#define _ASM_X86_RMWcc
+
+#ifdef CC_HAVE_ASM_GOTO
+
+#define GEN_UNARY_RMWcc(op, var, arg0, cc) \
+do { \
+ asm volatile goto (op " " arg0 ";" \
+ "j" cc " %l[cc_label]" \
+ : : "m" (var) \
+ : "memory" : cc_label); \
+ return 0; \
+cc_label: \
+ return 1; \
+} while (0)
+
+#define GEN_BINARY_RMWcc(op, var, val, arg0, cc) \
+do { \
+ asm volatile goto (op " %1, " arg0 ";" \
+ "j" cc " %l[cc_label]" \
+ : : "m" (var), "er" (val) \
+ : "memory" : cc_label); \
+ return 0; \
+cc_label: \
+ return 1; \
+} while (0)
+
+#else /* !CC_HAVE_ASM_GOTO */
+
+#define GEN_UNARY_RMWcc(op, var, arg0, cc) \
+do { \
+ char c; \
+ asm volatile (op " " arg0 ";" \
+ "set" cc " %1" \
+ : "+m" (var), "=qm" (c) \
+ : : "memory"); \
+ return c != 0; \
+} while (0)
+
+#define GEN_BINARY_RMWcc(op, var, val, arg0, cc) \
+do { \
+ char c; \
+ asm volatile (op " %2, " arg0 ";" \
+ "set" cc " %1" \
+ : "+m" (var), "=qm" (c) \
+ : "er" (val) : "memory"); \
+ return c != 0; \
+} while (0)
+
+#endif /* CC_HAVE_ASM_GOTO */
+
+#endif /* _ASM_X86_RMWcc */
next prev parent reply other threads:[~2013-09-12 2:20 UTC|newest]
Thread overview: 52+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-09-10 13:08 [PATCH 0/7] preempt_count rework -v2 Peter Zijlstra
2013-09-10 13:08 ` [PATCH 1/7] sched: Introduce preempt_count accessor functions Peter Zijlstra
2013-09-10 13:08 ` [PATCH 2/7] sched: Add NEED_RESCHED to the preempt_count Peter Zijlstra
2013-09-11 1:59 ` Andy Lutomirski
2013-09-11 8:25 ` Peter Zijlstra
2013-09-11 11:06 ` Peter Zijlstra
2013-09-11 13:34 ` Mike Galbraith
2013-09-12 6:01 ` Mike Galbraith
2013-09-11 16:35 ` Andy Lutomirski
2013-09-11 18:05 ` Peter Zijlstra
2013-09-11 18:07 ` Andy Lutomirski
2013-09-11 11:14 ` Peter Zijlstra
2013-09-10 13:08 ` [PATCH 3/7] sched, arch: Create asm/preempt.h Peter Zijlstra
2013-09-10 13:08 ` [PATCH 4/7] sched: Create more preempt_count accessors Peter Zijlstra
2013-09-10 13:08 ` [PATCH 5/7] sched: Extract the basic add/sub preempt_count modifiers Peter Zijlstra
2013-09-10 13:08 ` [PATCH 6/7] sched, x86: Provide a per-cpu preempt_count implementation Peter Zijlstra
2013-09-10 13:27 ` Peter Zijlstra
2013-09-10 14:02 ` Eric Dumazet
2013-09-10 15:25 ` Peter Zijlstra
2013-09-10 16:48 ` Peter Zijlstra
2013-09-10 13:08 ` [PATCH 7/7] sched, x86: Optimize the preempt_schedule() call Peter Zijlstra
2013-09-10 13:42 ` Ingo Molnar
2013-09-10 13:55 ` Jan Beulich
2013-09-10 13:55 ` Jan Beulich
2013-09-10 14:25 ` Ingo Molnar
2013-09-10 13:51 ` [PATCH 0/7] preempt_count rework -v2 Ingo Molnar
2013-09-10 13:56 ` Ingo Molnar
2013-09-10 15:14 ` Peter Zijlstra
2013-09-10 15:29 ` Arjan van de Ven
2013-09-10 15:35 ` Peter Zijlstra
2013-09-10 16:24 ` Linus Torvalds
2013-09-11 16:00 ` H. Peter Anvin
2013-09-10 16:34 ` Linus Torvalds
2013-09-10 16:45 ` Peter Zijlstra
2013-09-10 17:06 ` Linus Torvalds
2013-09-10 21:25 ` Peter Zijlstra
2013-09-10 21:43 ` Linus Torvalds
2013-09-10 21:51 ` H. Peter Anvin
2013-09-10 22:02 ` Linus Torvalds
2013-09-10 22:06 ` H. Peter Anvin
2013-09-11 13:13 ` Peter Zijlstra
2013-09-11 13:26 ` Peter Zijlstra
2013-09-11 15:29 ` H. Peter Anvin
2013-09-11 15:33 ` Linus Torvalds
2013-09-11 18:59 ` Peter Zijlstra
2013-09-11 23:02 ` Linus Torvalds
2013-09-12 2:20 ` Peter Zijlstra [this message]
2013-09-12 2:43 ` Linus Torvalds
2013-09-12 11:51 ` Peter Zijlstra
2013-09-12 12:25 ` Ingo Molnar
2013-09-13 7:25 ` Kevin Easton
2013-09-13 8:06 ` Peter Zijlstra
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20130912022040.GT31370@twins.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=ak@linux.intel.com \
--cc=arjan@linux.intel.com \
--cc=bitbucket@online.de \
--cc=fweisbec@gmail.com \
--cc=hpa@zytor.com \
--cc=linux-arch@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@kernel.org \
--cc=tglx@linutronix.de \
--cc=torvalds@linux-foundation.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.