From 163d14704491e56d98909b0180c46913cd601366 Mon Sep 17 00:00:00 2001
From: Bill Currie <bill@taniwha.org>
Date: Tue, 27 Jul 2021 13:54:22 +0900
Subject: [PATCH] [util] Give set_count a >8x speed boost

I knew counting bits individually was slow, but it never really mattered
until now. However, I didn't expect such a dramatic boost just by going
to mapping bytes to bit counts. 16-bit words would be faster still, but
the 64kB lookup table would probably start hurting cache performance,
and 32-bit words (4GB table) definitely would ruin the cache. The
universe isn't big enough for 64-bits :)
---
 libs/util/set.c           | 27 +++++++++++++++++++++++----
 libs/util/test/test-set.c | 12 ++++++++++++
 2 files changed, 35 insertions(+), 4 deletions(-)

diff --git a/libs/util/set.c b/libs/util/set.c
index ffb0e5489..f0215454c 100644
--- a/libs/util/set.c
+++ b/libs/util/set.c
@@ -553,12 +553,31 @@ set_is_member (const set_t *set, unsigned x)
 unsigned
 set_count (const set_t *set)
 {
+	static byte bit_counts[] = {
+		0, 1, 1, 2, 1, 2, 2, 3, 1, 2, 2, 3, 2, 3, 3, 4,
+		1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+		1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+		2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+		1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+		2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+		2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+		3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+		1, 2, 2, 3, 2, 3, 3, 4, 2, 3, 3, 4, 3, 4, 4, 5,
+		2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+		2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+		3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+		2, 3, 3, 4, 3, 4, 4, 5, 3, 4, 4, 5, 4, 5, 5, 6,
+		3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+		3, 4, 4, 5, 4, 5, 5, 6, 4, 5, 5, 6, 5, 6, 6, 7,
+		4, 5, 5, 6, 5, 6, 6, 7, 5, 6, 6, 7, 6, 7, 7, 8,
+	};
 	unsigned    count = 0;
-	unsigned    i;
+	byte       *b = (byte *) set->map;
+	unsigned    i = SET_WORDS (set) * sizeof (set_bits_t);
 
-	for (i = 0; i < set->size; i++)
-		if (_set_is_member (set, i))
-			count++;
+	while (i-- > 0) {
+		count += bit_counts[*b++];
+	}
 	return count;
 }
 
diff --git a/libs/util/test/test-set.c b/libs/util/test/test-set.c
index e1b93d137..6e326268a 100644
--- a/libs/util/test/test-set.c
+++ b/libs/util/test/test-set.c
@@ -83,6 +83,12 @@ check_size (const set_t *set, const set_t *unused)
 	return set->size;
 }
 
+static int
+check_count (const set_t *set, const set_t *unused)
+{
+	return set_count (set);
+}
+
 static set_t *
 make_5 (void)
 {
@@ -198,6 +204,7 @@ struct {
 	{make_55, make_5,  set_union, set_is_equivalent,   0, "{5 55}"},
 	{make_55, make_5,  set_union, set_is_intersecting, 1, "{5 55}"},
 	{make_55, make_5,  set_union, set_is_disjoint,     0, "{5 55}"},
+	{make_55, make_5,  set_union, check_count,         2, "{5 55}"},
 	{make_not_SIZE, make_everything, 0, set_is_equivalent,   0, 0},
 	{make_not_SIZE, make_everything, 0, set_is_intersecting, 1, 0},
 	{make_not_SIZE, make_everything, 0, set_is_disjoint,     0, 0},
@@ -210,6 +217,11 @@ struct {
 	{make_5, make_everything, 0, set_is_equivalent,   0, 0},
 	{make_5, make_everything, 0, set_is_intersecting, 1, 0},
 	{make_5, make_everything, 0, set_is_disjoint,     0, 0},
+	{make_empty, 0, 0, check_count, 0, 0},
+	{make_everything, 0, 0, check_count, 0, 0},
+	{make_5, 0, 0, check_count, 1, 0},
+	{make_not_5, 0, 0, check_count, 1, 0},
+	{make_0_to_SIZEm1, 0, 0, check_size, SIZE, 0},
 };
 #define num_tests (sizeof (tests) / sizeof (tests[0]))