public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
* [patch 0/2] use asm() for atomic_{read|set} (shot 2)
@ 2007-08-15 20:54 Sebastian Siewior
  2007-08-15 20:54 ` [patch 1/2] i386: use asm() like the other atomic operations already do Sebastian Siewior
  2007-08-15 20:54 ` [patch 2/2] x86_64: " Sebastian Siewior
  0 siblings, 2 replies; 3+ messages in thread
From: Sebastian Siewior @ 2007-08-15 20:54 UTC (permalink / raw)
  To: linux-kernel; +Cc: Andi Kleen, Chris Snook

Diff to last post:
- no __*__ functions used.
- no white space fixes.

I converted i386+x86-64 arch. The description of both patches contains the 
file size of four kernel builds:
- "normal" is 28e8351ac22de25034e048c680014ad824323c65 as it
- "inline asm" is with this patch
- "inline volatile" is *(volatile int *)&(v)->counter as a static inline
  function
- "volatile" is *(volatile int *)&(v)->counter as a #define macro

Sebastian
-- 


^ permalink raw reply	[flat|nested] 3+ messages in thread

* [patch 1/2] i386: use asm() like the other atomic operations already do.
  2007-08-15 20:54 [patch 0/2] use asm() for atomic_{read|set} (shot 2) Sebastian Siewior
@ 2007-08-15 20:54 ` Sebastian Siewior
  2007-08-15 20:54 ` [patch 2/2] x86_64: " Sebastian Siewior
  1 sibling, 0 replies; 3+ messages in thread
From: Sebastian Siewior @ 2007-08-15 20:54 UTC (permalink / raw)
  To: linux-kernel; +Cc: Andi Kleen, Chris Snook

[-- Attachment #1: atomic_i386.diff --]
[-- Type: text/plain, Size: 1648 bytes --]

As Segher pointed out, inline asm is better than the volatile casting all over
the place. From the PowerPC patch description:
 Also use inline functions instead of macros; this actually
 improves code generation (some code becomes a little smaller,
 probably because of improved alias information -- just a few
 hundred bytes total on a default kernel build, nothing shocking).

My config with march=pentium-m and gcc (GCC) 4.1.2 (Gentoo 4.1.2):
  text    data     bss     dec     hex filename
  3434150  249176  176128 3859454  3ae3fe atomic_normal/vmlinux
  3435308  249176  176128 3860612  3ae884 atomic_inlineasm/vmlinux
  3436201  249176  176128 3861505  3aec01 atomic_inline_volatile/vmlinux
  3436203  249176  176128 3861507  3aec03 atomic_volatile/vmlinux

Signed-off-by: Sebastian Siewior <sebastian@breakpoint.cc>
Cc: Andi Kleen <ak@suse.de>
Cc: Chris Snook <csnook@redhat.com>
--- a/include/asm-i386/atomic.h
+++ b/include/asm-i386/atomic.h
@@ -25,7 +25,16 @@ typedef struct { int counter; } atomic_t
  * 
  * Atomically reads the value of @v.
  */ 
-#define atomic_read(v)		((v)->counter)
+static inline int atomic_read(const atomic_t *v)
+{
+	int t;
+
+	asm volatile(
+			"movl %1,%0"
+			: "=r"(t)
+			: "m"(v->counter));
+	return t;
+}
 
 /**
  * atomic_set - set atomic variable
@@ -34,7 +43,13 @@ typedef struct { int counter; } atomic_t
  * 
  * Atomically sets the value of @v to @i.
  */ 
-#define atomic_set(v,i)		(((v)->counter) = (i))
+static inline void atomic_set(atomic_t *v, int i)
+{
+	asm volatile(
+			"movl %1,%0"
+			: "=m"(v->counter)
+			: "ir"(i));
+}
 
 /**
  * atomic_add - add integer to atomic variable

-- 


^ permalink raw reply	[flat|nested] 3+ messages in thread

* [patch 2/2] x86_64: use asm() like the other atomic operations already do.
  2007-08-15 20:54 [patch 0/2] use asm() for atomic_{read|set} (shot 2) Sebastian Siewior
  2007-08-15 20:54 ` [patch 1/2] i386: use asm() like the other atomic operations already do Sebastian Siewior
@ 2007-08-15 20:54 ` Sebastian Siewior
  1 sibling, 0 replies; 3+ messages in thread
From: Sebastian Siewior @ 2007-08-15 20:54 UTC (permalink / raw)
  To: linux-kernel; +Cc: Andi Kleen, Chris Snook

[-- Attachment #1: atomic_x86_64.diff --]
[-- Type: text/plain, Size: 2631 bytes --]

As Segher pointed out, inline asm is better than the volatile casting all over
the place. From the PowerPC patch description:
 Also use inline functions instead of macros; this actually
 improves code generation (some code becomes a little smaller,
 probably because of improved alias information -- just a few
 hundred bytes total on a default kernel build, nothing shocking).

My config with march=k8 and gcc (GCC) 4.1.2 (Gentoo 4.1.2):
  text    data     bss     dec     hex filename
  4002473  385936  474440 4862849  4a3381 atomic_normal/vmlinux
  4002587  385936  474440 4862963  4a33f3 atomic_inlineasm/vmlinux
  4003911  385936  474440 4864287  4a391f atomic_volatile/vmlinux
  4003959  385936  474440 4864335  4a394f atomic_volatile_inline/vmlinux

Signed-off-by: Sebastian Siewior <sebastian@breakpoint.cc>
Cc: Andi Kleen <ak@suse.de>
Cc: Chris Snook <csnook@redhat.com>
--- a/include/asm-x86_64/atomic.h
+++ b/include/asm-x86_64/atomic.h
@@ -32,7 +32,16 @@ typedef struct { int counter; } atomic_t
  * 
  * Atomically reads the value of @v.
  */ 
-#define atomic_read(v)		((v)->counter)
+static inline int atomic_read(const atomic_t *v)
+{
+	int t;
+
+	asm volatile(
+			"movl %1, %0"
+			: "=r"(t)
+			: "m"(v->counter));
+	return t;
+}
 
 /**
  * atomic_set - set atomic variable
@@ -41,7 +50,13 @@ typedef struct { int counter; } atomic_t
  * 
  * Atomically sets the value of @v to @i.
  */ 
-#define atomic_set(v,i)		(((v)->counter) = (i))
+static inline void atomic_set(atomic_t *v, int i)
+{
+	asm volatile(
+			"movl %1, %0"
+			: "=m"(v->counter)
+			: "ir"(i));
+}
 
 /**
  * atomic_add - add integer to atomic variable
@@ -206,7 +221,7 @@ static __inline__ int atomic_sub_return(
 
 /* An 64bit atomic type */
 
-typedef struct { volatile long counter; } atomic64_t;
+typedef struct { long counter; } atomic64_t;
 
 #define ATOMIC64_INIT(i)	{ (i) }
 
@@ -217,7 +232,16 @@ typedef struct { volatile long counter; 
  * Atomically reads the value of @v.
  * Doesn't imply a read memory barrier.
  */
-#define atomic64_read(v)		((v)->counter)
+static inline long atomic64_read(const atomic64_t *v)
+{
+	long t;
+
+	asm volatile(
+			"movq %1, %0"
+			: "=r"(t)
+			: "m"(v->counter));
+	return t;
+}
 
 /**
  * atomic64_set - set atomic64 variable
@@ -226,7 +250,13 @@ typedef struct { volatile long counter; 
  *
  * Atomically sets the value of @v to @i.
  */
-#define atomic64_set(v,i)		(((v)->counter) = (i))
+static inline void atomic64_set(atomic64_t *v, long i)
+{
+	asm volatile(
+			"movq %1, %0"
+			: "=m"(v->counter)
+			: "ir"(i));
+}
 
 /**
  * atomic64_add - add integer to atomic64 variable

-- 


^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2007-08-16  7:18 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2007-08-15 20:54 [patch 0/2] use asm() for atomic_{read|set} (shot 2) Sebastian Siewior
2007-08-15 20:54 ` [patch 1/2] i386: use asm() like the other atomic operations already do Sebastian Siewior
2007-08-15 20:54 ` [patch 2/2] x86_64: " Sebastian Siewior

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox