Received-SPF: pass (google.com: domain of linux-kernel+bounces-148499-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:4601:e00::3 as permitted sender) client-ip=2604:1380:4601:e00::3;
From: Chin-Chun Chen <chinchunchen2001@gmail.com>
To: yury.norov@gmail.com
Cc: linux@rasmusvillemoes.dk,
	linux-kernel@vger.kernel.org,
	Chin-Chun Chen <chinchunchen2001@gmail.com>
Subject: [PATCH] linux/bitops.h: Refactor function fns to reduce branch usage and eliminate external calls
Date: Wed, 17 Apr 2024 20:57:07 +0800
Message-Id: <20240417125706.19899-1-chinchunchen2001@gmail.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

Rewrote the function fns to use macro wrote in linux/include/asm-generic/bitops/const_hweight.h
Decrease the number of required branches and eliminate the need for external calls.
The number of branches is now constant, unaffected by input variations.

This commit improves code efficiency and clarity.

Signed-off-by: Chin-Chun Chen <chinchunchen2001@gmail.com>
---
 include/linux/bitops.h | 54 +++++++++++++++++++++++++++++++++++++-----
 1 file changed, 48 insertions(+), 6 deletions(-)

diff --git a/include/linux/bitops.h b/include/linux/bitops.h
index 5842d7d03f19..18899f11aaa7 100644
--- a/include/linux/bitops.h
+++ b/include/linux/bitops.h
@@ -247,6 +247,13 @@ static inline unsigned long __ffs64(u64 word)
 	return __ffs((unsigned long)word);
 }
 
+
+#define __const_hweight2(w)                                                 \
+    ((unsigned int) (!!((word) & (1ULL << 0))) + (!!((word) & (1ULL << 1))))
+
+#define __const_hweight4(w)                                                 \
+    ((unsigned int) (!!((word) & (1ULL << 0))) + (!!((word) & (1ULL << 1))) + \
+                     (!!((word) & (1ULL << 2))) + (!!((word) & (1ULL << 3))))
 /**
  * fns - find N'th set bit in a word
  * @word: The word to search
@@ -255,13 +262,48 @@ static inline unsigned long __ffs64(u64 word)
 static inline unsigned long fns(unsigned long word, unsigned int n)
 {
 	unsigned int bit;
-
-	while (word) {
-		bit = __ffs(word);
-		if (n-- == 0)
-			return bit;
-		__clear_bit(bit, &word);
-       }
+	unsigned int sum = 0;
+#if BITS_PER_LONG == 64
+	bits = __const_hweight32(word & 0xffffffff);
+	if (bits <= n) {
+		n -= bits;
+		word >>= 32;
+		sum += 32;
+ 	}
+#endif
+	bits = __const_hweight16(word & 0xffff);
+       if (bits <= n) {
+               n -= bits;
+               word >>= 16;
+               sum += 16;
+       }
+	bits = __const_hweight8(word & 0xff);
+       if (bits <= n) {
+               n -= bits;
+               word >>= 8;
+               sum += 8;
+       }
+	bits = __const_hweight4(word & 0xf);
+       if (bits <= n) {
+               n -= bits;
+               word >>= 4;
+               sum += 4;
+       }
+       bits = __const_hweight2(word & 0x3);
+       if (bits <= n) {
+               n -= bits;
+               word >>= 2;
+               sum += 2;
+       }
+       bits = (unsigned int) (word & 0x1);
+       if (bits <= n) {
+               n -= bits;
+               word >>= 1;
+               sum += 1;
+       }
+	bits = (unsigned int) (word & 0x1);
+       if (bits <= n) {
+               n -= bits;
+               sum += 1;
+       }
+
+	if (!n)
+		return sum;
+
 
 	return BITS_PER_LONG;
 }
-- 
2.40.1