x86: Remove arch-specific low level lock implementation

This patch removes the arch-specific x86 assembly implementation for
low level locking and consolidate both 64 bits and 32 bits in a
single implementation.

Different than other architectures, x86 lll_trylock, lll_lock, and
lll_unlock implements a single-thread optimization to avoid atomic
operation, using cmpxchgl instead.  This patch implements by using
the new single-thread.h definitions in a generic way, although using
the previous semantic.

The lll_cond_trylock, lll_cond_lock, and lll_timedlock just use
atomic operations plus calls to lll_lock_wait*.

For __lll_lock_wait_private and __lll_lock_wait the generic implemtation
there is no indication that assembly implementation is required
performance-wise.

Checked on x86_64-linux-gnu and i686-linux-gnu.

	* sysdeps/nptl/lowlevellock.h (__lll_trylock): New macro.
	(lll_trylock): Call __lll_trylock.
	* sysdeps/unix/sysv/linux/i386/libc-lowlevellock.S: Remove file.
	* sysdeps/unix/sysv/linux/i386/lll_timedlock_wait.c: Likewise.
	* sysdeps/unix/sysv/linux/i386/lowlevellock.S: Likewise.
	* sysdeps/unix/sysv/linux/i386/lowlevellock.h: Likewise.
	* sysdeps/unix/sysv/linux/x86_64/libc-lowlevellock.S: Likewise.
	* sysdeps/unix/sysv/linux/x86_64/lll_timedlock_wait.c: Likewise.
	* sysdeps/unix/sysv/linux/x86_64/lowlevellock.S: Likewise.
	* sysdeps/unix/sysv/linux/x86_64/lowlevellock.h: Likewise.
	* sysdeps/unix/sysv/linux/x86/lowlevellock.h: New file.
	* sysdeps/unix/sysv/linux/x86_64/cancellation.S: Include
	lowlevellock-futex.h.

1 files changed, 3 insertions, 1 deletions

diff --git a/sysdeps/nptl/lowlevellock.h b/sysdeps/nptl/lowlevellock.h
index 6f017afdd5..e905829ee4 100644
--- a/sysdeps/nptl/lowlevellock.h
+++ b/sysdeps/nptl/lowlevellock.h
@@ -63,8 +63,10 @@
 /* If LOCK is 0 (not acquired), set to 1 (acquired with no waiters) and return
    0.  Otherwise leave lock unchanged and return non-zero to indicate that the
    lock was not acquired.  */
+#define __lll_trylock(lock)	\
+  __glibc_unlikely (atomic_compare_and_exchange_bool_acq ((lock), 1, 0))
 #define lll_trylock(lock)	\
-  __glibc_unlikely (atomic_compare_and_exchange_bool_acq (&(lock), 1, 0))
+   __lll_trylock (&(lock))
 
 /* If LOCK is 0 (not acquired), set to 2 (acquired, possibly with waiters) and
    return 0.  Otherwise leave lock unchanged and return non-zero to indicate