Commit 22767544 authored by Uros Bizjak's avatar Uros Bizjak Committed by Dave Hansen

x86/ACPI/boot: Improve __acpi_acquire_global_lock

Improve __acpi_acquire_global_lock by using a temporary variable.
This enables compiler to perform if-conversion and improves generated
code from:

 ...
 72a:	d1 ea                	shr    %edx
 72c:	83 e1 fc             	and    $0xfffffffc,%ecx
 72f:	83 e2 01             	and    $0x1,%edx
 732:	09 ca                	or     %ecx,%edx
 734:	83 c2 02             	add    $0x2,%edx
 737:	f0 0f b1 17          	lock cmpxchg %edx,(%rdi)
 73b:	75 e9                	jne    726 <__acpi_acquire_global_lock+0x6>
 73d:	83 e2 03             	and    $0x3,%edx
 740:	31 c0                	xor    %eax,%eax
 742:	83 fa 03             	cmp    $0x3,%edx
 745:	0f 95 c0             	setne  %al
 748:	f7 d8                	neg    %eax

to:

 ...
 72a:	d1 e9                	shr    %ecx
 72c:	83 e2 fc             	and    $0xfffffffc,%edx
 72f:	83 e1 01             	and    $0x1,%ecx
 732:	09 ca                	or     %ecx,%edx
 734:	83 c2 02             	add    $0x2,%edx
 737:	f0 0f b1 17          	lock cmpxchg %edx,(%rdi)
 73b:	75 e9                	jne    726 <__acpi_acquire_global_lock+0x6>
 73d:	8d 41 ff             	lea    -0x1(%rcx),%eax

BTW: the compiler could generate:

	lea 0x2(%rcx,%rdx,1),%edx

instead of:

	or     %ecx,%edx
	add    $0x2,%edx

but unwated conversion from add to or when bits are known to be zero
prevents this improvement. This is GCC PR108477.

No functional change intended.
Signed-off-by: default avatarUros Bizjak <ubizjak@gmail.com>
Signed-off-by: default avatarDave Hansen <dave.hansen@linux.intel.com>
Acked-by: default avatarRafael J. Wysocki <rafael.j.wysocki@intel.com>
Link: https://lore.kernel.org/all/20230320212012.12704-1-ubizjak%40gmail.com
parent e8d018dd
......@@ -1853,13 +1853,18 @@ early_param("acpi_sci", setup_acpi_sci);
int __acpi_acquire_global_lock(unsigned int *lock)
{
unsigned int old, new;
unsigned int old, new, val;
old = READ_ONCE(*lock);
do {
new = (((old & ~0x3) + 2) + ((old >> 1) & 0x1));
val = (old >> 1) & 0x1;
new = (old & ~0x3) + 2 + val;
} while (!try_cmpxchg(lock, &old, new));
return ((new & 0x3) < 3) ? -1 : 0;
if (val)
return 0;
return -1;
}
int __acpi_release_global_lock(unsigned int *lock)
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment