RepoMirrors/haproxy
1
0
Fork 0
mirror of http://git.haproxy.org/git/haproxy.git/ synced 2025-04-11 03:31:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

MINOR: threads/build: atomic: replace the few inlines with macros

Browse Source 
[ plock commit 61e255286ae32e83e1a3174dd7c49eda99880a8b]

There are a few inlines such as pl_barrier() and pl_cpu_relax() which
are used a lot. Unfortunately, while building test code at -O0, inlining
is disabled and these ones are called a lot and show up a lot in any
profile, are traced into when single-stepping with a debugger, etc, thus
they are polluting the landscape. Since they're single-asm statements,
there is no reason for not turning them into macros.

The result becomes fairly visible here at -O0 :

  $ size latency.inline latency.macro
     text    data     bss     dec     hex filename
    11431     692     656   12779    31eb treelock.inline
    10967     692     656   12315    301b treelock.macro

And it was verified that regularly optimized code remains strictly identical.
This commit is contained in:
Willy Tarreau 2017-11-20 20:55:06 +01:00
parent d0d8ba59d3
commit b5f271555e
1 changed files with 29 additions and 24 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

53
include/import/atomic-ops.h
View File

@ -3,35 +3,42 @@
/* compiler-only memory barrier, for use around locks */
static inline void pl_barrier()
{
asm volatile("" ::: "memory");
}
#define pl_barrier() do { \
asm volatile("" ::: "memory"); \
} while (0)
#if defined(__i386__) || defined (__i486__) || defined (__i586__) || defined (__i686__) || defined (__x86_64__)
/* full memory barrier using mfence when SSE2 is supported, falling back to
* "lock add %esp" (gcc uses "lock add" or "lock or").
*/
static inline void pl_mb()
{
#if defined(__SSE2__)
asm volatile("mfence" ::: "memory");
#define pl_mb() do { \
asm volatile("mfence" ::: "memory"); \
} while (0)
#elif defined(__x86_64__)
asm volatile("lock addl $0,0 (%%rsp)" ::: "memory", "cc");
#else
asm volatile("lock addl $0,0 (%%esp)" ::: "memory", "cc");
#endif
}
#define pl_mb() do { \
asm volatile("lock addl $0,0 (%%rsp)" ::: "memory", "cc"); \
} while (0)
#else /* ix86 */
#define pl_mb() do { \
asm volatile("lock addl $0,0 (%%esp)" ::: "memory", "cc"); \
} while (0)
#endif /* end of pl_mb() case for sse2/x86_64/x86 */
/*
* Generic functions common to the x86 family
*/
static inline void pl_cpu_relax()
{
asm volatile("rep;nop\n");
}
#define pl_cpu_relax() do { \
asm volatile("rep;nop\n"); \
} while (0)
/* increment integer value pointed to by pointer <ptr>, and return non-zero if
* result is non-null.
@ -491,16 +498,14 @@ static inline void pl_cpu_relax()
#else
/* generic implementations */
static inline void pl_cpu_relax()
{
asm volatile("");
}
#define pl_cpu_relax() do { \
asm volatile(""); \
} while (0)
/* full memory barrier */
static inline void pl_mb()
{
__sync_synchronize();
}
#define pl_mb() do { \
__sync_synchronize(); \
} while (0)
#define pl_inc_noret(ptr) ({ __sync_add_and_fetch((ptr), 1); })
#define pl_dec_noret(ptr) ({ __sync_sub_and_fetch((ptr), 1); })