lib/sort: optimize heapsort for equal elements in sift-down path
authorKuan-Wei Chiu <visitorckw@gmail.com>
Sat, 13 Jan 2024 03:13:51 +0000 (11:13 +0800)
committerAndrew Morton <akpm@linux-foundation.org>
Thu, 22 Feb 2024 23:38:52 +0000 (15:38 -0800)
Patch series "lib/sort: Optimize the number of swaps and comparisons".

This patch series aims to optimize the heapsort algorithm, specifically
targeting a reduction in the number of swaps and comparisons required.

This patch (of 2):

Currently, when searching for the sift-down path and encountering equal
elements, the algorithm chooses the left child.  However, considering that
the height of the right subtree may be one less than that of the left
subtree, selecting the right child in such cases can potentially reduce
the number of comparisons and swaps.

For instance, when sorting an array of 10,000 identical elements, the
current implementation requires 247,209 comparisons.  With this patch, the
number of comparisons can be reduced to 227,241.

Link: https://lkml.kernel.org/r/20240113031352.2395118-1-visitorckw@gmail.com
Link: https://lkml.kernel.org/r/20240113031352.2395118-2-visitorckw@gmail.com
Signed-off-by: Kuan-Wei Chiu <visitorckw@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
lib/sort.c

index b399bf10d6759b47c5ed28a1ff0bd3cd4d26b00c..fe4efd4a1410f72a5c41266820b06b2f2cb02161 100644 (file)
@@ -262,7 +262,7 @@ void sort_r(void *base, size_t num, size_t size,
                 * average, 3/4 worst-case.)
                 */
                for (b = a; c = 2*b + size, (d = c + size) < n;)
-                       b = do_cmp(base + c, base + d, cmp_func, priv) >= 0 ? c : d;
+                       b = do_cmp(base + c, base + d, cmp_func, priv) > 0 ? c : d;
                if (d == n)     /* Special case last leaf with no sibling */
                        b = c;