Thread (27 messages) 27 messages, 4 authors, 2023-10-15

Re: [PATCH 03/14] bitops: let the compiler optimize __assign_bit()

From: Yury Norov <yury.norov@gmail.com>
Date: 2023-10-09 16:18:47
Also in: dm-devel, linux-btrfs, linux-s390, lkml, ntfs3

On Mon, Oct 09, 2023 at 05:10:15PM +0200, Alexander Lobakin wrote:
Since commit b03fc1173c0c ("bitops: let optimize out non-atomic bitops
on compile-time constants"), the compilers are able to expand inline
bitmap operations to compile-time initializers when possible.
However, during the round of replacement if-__set-else-__clear with
__assign_bit() as per Andy's advice, bloat-o-meter showed +1024 bytes
difference in object code size for one module (even one function),
where the pattern:

	DECLARE_BITMAP(foo) = { }; // on the stack, zeroed

	if (a)
		__set_bit(const_bit_num, foo);
	if (b)
		__set_bit(another_const_bit_num, foo);
	...

is heavily used, although there should be no difference: the bitmap is
zeroed, so the second half of __assign_bit() should be compiled-out as
a no-op.
I either missed the fact that __assign_bit() has bitmap pointer marked
as `volatile` (as we usually do for bitmaps) or was hoping that the
No, we usually don't. Atomic ops on individual bits is a notable exception
for bitmaps, as the comment for generic_test_bit() says, for example:
         /*
          * Unlike the bitops with the '__' prefix above, this one *is* atomic,
          * so `volatile` must always stay here with no cast-aways. See
          * `Documentation/atomic_bitops.txt` for the details.
          */

For non-atomic single-bit operations and all multi-bit ops, volatile is
useless, and generic___test_and_set_bit() in the same file casts the 
*addr to a plain 'unsigned long *'.
quoted hunk ↗ jump to hunk
compilers would at least try to look past the `volatile` for
__always_inline functions. Anyhow, due to that attribute, the compilers
were always compiling the whole expression and no mentioned compile-time
optimizations were working.

Convert __assign_bit() to a macro since it's a very simple if-else and
all of the checks are performed inside __set_bit() and __clear_bit(),
thus that wrapper has to be as transparent as possible. After that
change, despite it showing only -20 bytes change for vmlinux (due to
that it's still relatively unpopular), no drastic code size changes
happen when replacing if-set-else-clear for onstack bitmaps with
__assign_bit(), meaning the compiler now expands them to the actual
operations will all the expected optimizations.

Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Alexander Lobakin <aleksander.lobakin@intel.com>
---
 include/linux/bitops.h | 10 ++--------
 1 file changed, 2 insertions(+), 8 deletions(-)
diff --git a/include/linux/bitops.h b/include/linux/bitops.h
index e0cd09eb91cd..f98f4fd1047f 100644
--- a/include/linux/bitops.h
+++ b/include/linux/bitops.h
@@ -284,14 +284,8 @@ static __always_inline void assign_bit(long nr, volatile unsigned long *addr,
 		clear_bit(nr, addr);
 }
 
-static __always_inline void __assign_bit(long nr, volatile unsigned long *addr,
-					 bool value)
-{
-	if (value)
-		__set_bit(nr, addr);
-	else
-		__clear_bit(nr, addr);
-}
+#define __assign_bit(nr, addr, value)				\
+	((value) ? __set_bit(nr, addr) : __clear_bit(nr, addr))
Can you protect nr and addr with braces just as well?
Can you convert the atomic version too, to keep them synchronized ?
 
 /**
  * __ptr_set_bit - Set bit in a pointer's value
-- 
2.41.0
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help