Received-SPF: pass (google.com: domain of linux-kernel+bounces-189478-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
From: Thorsten Blum <thorsten.blum@toblux.com>
To: Yury Norov <yury.norov@gmail.com>,
	Rasmus Villemoes <linux@rasmusvillemoes.dk>,
	Thomas Gleixner <tglx@linutronix.de>,
	Ingo Molnar <mingo@redhat.com>,
	Borislav Petkov <bp@alien8.de>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	x86@kernel.org,
	"H. Peter Anvin" <hpa@zytor.com>
Cc: linux-kernel@vger.kernel.org,
	Thorsten Blum <thorsten.blum@toblux.com>
Subject: [PATCH] x86/asm/bitops: Change function return types from long to int
Date: Sat, 25 May 2024 19:44:49 +0200
Message-ID: <20240525174448.174824-2-thorsten.blum@toblux.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

Change the return types of bitops functions (ffs, ffz, and fls) from
long to int. The expected return values are in the range [0, 64], for
which int is sufficient.

Additionally, int aligns well with the return types of the corresponding
__builtin_* functions, potentially reducing overall type conversions.

Many of the existing bitops functions already return an int and don't
need to be changed.

With GCC 14 and defconfig, these changes reduced the number of x86-64
instructions by 342 (using objdump -d and compared to 0b32d436c015).

Summary of the most significant changes in instruction counts:

	insn	before	after	diff
	----------------------------
	ja	16044	15166	-878
	cmp	171197	170583	-614
	movsxd	16796	16271	-525
	add	94692	94454	-238
	jbe	9053	8815	-238
	xor	97499	97478	 -21
	...
	and	47291	47312	 +21
	nop	1036207	1036233	 +26
	cs	5622	5656	 +34
	int3	2815714	2815842	+128
	lea	118250	118488	+238
	je	190859	191132	+273
	test	204887	205484	+597
	jne	134316	135161	+845

Signed-off-by: Thorsten Blum <thorsten.blum@toblux.com>
---
 arch/x86/include/asm/bitops.h | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/arch/x86/include/asm/bitops.h b/arch/x86/include/asm/bitops.h
index b96d45944c59..4d91517a395d 100644
--- a/arch/x86/include/asm/bitops.h
+++ b/arch/x86/include/asm/bitops.h
@@ -246,7 +246,7 @@ arch_test_bit_acquire(unsigned long nr, const volatile unsigned long *addr)
 					  variable_test_bit(nr, addr);
 }
 
-static __always_inline unsigned long variable__ffs(unsigned long word)
+static __always_inline unsigned int variable__ffs(unsigned long word)
 {
 	asm("rep; bsf %1,%0"
 		: "=r" (word)
@@ -262,10 +262,10 @@ static __always_inline unsigned long variable__ffs(unsigned long word)
  */
 #define __ffs(word)				\
 	(__builtin_constant_p(word) ?		\
-	 (unsigned long)__builtin_ctzl(word) :	\
+	 (unsigned int)__builtin_ctzl(word) :	\
 	 variable__ffs(word))
 
-static __always_inline unsigned long variable_ffz(unsigned long word)
+static __always_inline unsigned int variable_ffz(unsigned long word)
 {
 	asm("rep; bsf %1,%0"
 		: "=r" (word)
@@ -281,7 +281,7 @@ static __always_inline unsigned long variable_ffz(unsigned long word)
  */
 #define ffz(word)				\
 	(__builtin_constant_p(word) ?		\
-	 (unsigned long)__builtin_ctzl(~word) :	\
+	 (unsigned int)__builtin_ctzl(~word) :	\
 	 variable_ffz(word))
 
 /*
@@ -290,7 +290,7 @@ static __always_inline unsigned long variable_ffz(unsigned long word)
  *
  * Undefined if no set bit exists, so code should check against 0 first.
  */
-static __always_inline unsigned long __fls(unsigned long word)
+static __always_inline unsigned int __fls(unsigned long word)
 {
 	if (__builtin_constant_p(word))
 		return BITS_PER_LONG - 1 - __builtin_clzl(word);

base-commit: 0b32d436c015d5a88b3368405e3d8fe82f195a54
-- 
2.45.1