@matheusbsbs / matheus-bagliano / commits / f1294eaf7f
Fork 0

bloom.c: introduce core Bloom filter constructs

Introduce the constructs for Bloom filters, Bloom filter keys and Bloom filter settings. For details on what Bloom filters are and how they work, refer to Dr. Derrick Stolee's blog post [1]. It provides a concise explanation of the adoption of Bloom filters as described in [2] and [3]. Implementation specifics: 1. We currently use 7 and 10 for the number of hashes and the size of each entry respectively. They served as great starting values, the mathematical details behind this choice are described in [1] and [4]. The implementation, while not completely open to it at the moment, is flexible enough to allow for tweaking these settings in the future. Note: The performance gains we have observed with these values are significant enough that we did not need to tweak these settings. The performance numbers are included in the cover letter of this series and in the commit message of the subsequent commit where we use Bloom filters to speed up `git log -- path`. 2. As described in [1] and [3], we do not need 7 independent hashing functions. We use the Murmur3 hashing scheme, seed it twice and then combine those to procure an arbitrary number of hash values. 3. The filters will be sized according to the number of changes in each commit, in multiples of 8 bit words. [1] Derrick Stolee "Supercharging the Git Commit Graph IV: Bloom Filters" https://devblogs.microsoft.com/devops/super-charging-the-git-commit-graph-iv-Bloom-filters/ [2] Flavio Bonomi, Michael Mitzenmacher, Rina Panigrahy, Sushil Singh, George Varghese "An Improved Construction for Counting Bloom Filters" http://theory.stanford.edu/~rinap/papers/esa2006b.pdf https://doi.org/10.1007/11841036_61 [3] Peter C. Dillinger and Panagiotis Manolios "Bloom Filters in Probabilistic Verification" http://www.ccs.neu.edu/home/pete/pub/Bloom-filters-verification.pdf https://doi.org/10.1007/978-3-540-30494-4_26 [4] Thomas Mueller Graf, Daniel Lemire "Xor Filters: Faster and Smaller Than Bloom and Cuckoo Filters" https://arxiv.org/abs/1912.08258 Helped-by: Derrick Stolee <dstolee@microsoft.com> Reviewed-by: Jakub Narębski <jnareb@gmail.com> Signed-off-by: Garima Singh <garima.singh@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>

Garima Singh committed Mar 30, 2020 at 00:31 UTC f1294eaf7fbf7673567b698b11e062566b9f1035
4 files changed +188 -1
bloom.c
+37 -1
@@ -8,6 +8,11 @@ static uint32_t rotate_left(uint32_t value, int32_t count)
8 return ((value << count) | (value >> ((-count) & mask)));
9 }
10
11 +static inline unsigned char get_bitmask(uint32_t pos)
12 +{
13 + return ((unsigned char)1) << (pos & (BITS_PER_WORD - 1));
14 +}
15 +
16 /*
17 * Calculate the murmur3 32-bit hash value for the given data
18 * using the given seed.
@@ -70,4 +75,35 @@ uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len)
75 seed ^= (seed >> 16);
76
77 return seed;
73 -}
\ No newline at end of file
78 +}
79 +
80 +void fill_bloom_key(const char *data,
81 + size_t len,
82 + struct bloom_key *key,
83 + const struct bloom_filter_settings *settings)
84 +{
85 + int i;
86 + const uint32_t seed0 = 0x293ae76f;
87 + const uint32_t seed1 = 0x7e646e2c;
88 + const uint32_t hash0 = murmur3_seeded(seed0, data, len);
89 + const uint32_t hash1 = murmur3_seeded(seed1, data, len);
90 +
91 + key->hashes = (uint32_t *)xcalloc(settings->num_hashes, sizeof(uint32_t));
92 + for (i = 0; i < settings->num_hashes; i++)
93 + key->hashes[i] = hash0 + i * hash1;
94 +}
95 +
96 +void add_key_to_filter(const struct bloom_key *key,
97 + struct bloom_filter *filter,
98 + const struct bloom_filter_settings *settings)
99 +{
100 + int i;
101 + uint64_t mod = filter->len * BITS_PER_WORD;
102 +
103 + for (i = 0; i < settings->num_hashes; i++) {
104 + uint64_t hash_mod = key->hashes[i] % mod;
105 + uint64_t block_pos = hash_mod / BITS_PER_WORD;
106 +
107 + filter->data[block_pos] |= get_bitmask(hash_mod);
108 + }
109 +}
bloom.h
+63
@@ -1,6 +1,60 @@
1 #ifndef BLOOM_H
2 #define BLOOM_H
3
4 +struct bloom_filter_settings {
5 + /*
6 + * The version of the hashing technique being used.
7 + * We currently only support version = 1 which is
8 + * the seeded murmur3 hashing technique implemented
9 + * in bloom.c.
10 + */
11 + uint32_t hash_version;
12 +
13 + /*
14 + * The number of times a path is hashed, i.e. the
15 + * number of bit positions tht cumulatively
16 + * determine whether a path is present in the
17 + * Bloom filter.
18 + */
19 + uint32_t num_hashes;
20 +
21 + /*
22 + * The minimum number of bits per entry in the Bloom
23 + * filter. If the filter contains 'n' entries, then
24 + * filter size is the minimum number of 8-bit words
25 + * that contain n*b bits.
26 + */
27 + uint32_t bits_per_entry;
28 +};
29 +
30 +#define DEFAULT_BLOOM_FILTER_SETTINGS { 1, 7, 10 }
31 +#define BITS_PER_WORD 8
32 +
33 +/*
34 + * A bloom_filter struct represents a data segment to
35 + * use when testing hash values. The 'len' member
36 + * dictates how many entries are stored in
37 + * 'data'.
38 + */
39 +struct bloom_filter {
40 + unsigned char *data;
41 + size_t len;
42 +};
43 +
44 +/*
45 + * A bloom_key represents the k hash values for a
46 + * given string. These can be precomputed and
47 + * stored in a bloom_key for re-use when testing
48 + * against a bloom_filter. The number of hashes is
49 + * given by the Bloom filter settings and is the same
50 + * for all Bloom filters and keys interacting with
51 + * the loaded version of the commit graph file and
52 + * the Bloom data chunks.
53 + */
54 +struct bloom_key {
55 + uint32_t *hashes;
56 +};
57 +
58 /*
59 * Calculate the murmur3 32-bit hash value for the given data
60 * using the given seed.
@@ -10,4 +64,13 @@
64 */
65 uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len);
66
67 +void fill_bloom_key(const char *data,
68 + size_t len,
69 + struct bloom_key *key,
70 + const struct bloom_filter_settings *settings);
71 +
72 +void add_key_to_filter(const struct bloom_key *key,
73 + struct bloom_filter *filter,
74 + const struct bloom_filter_settings *settings);
75 +
76 #endif
\ No newline at end of file
t/helper/test-bloom.c
+48
@@ -2,6 +2,36 @@
2 #include "bloom.h"
3 #include "test-tool.h"
4
5 +struct bloom_filter_settings settings = DEFAULT_BLOOM_FILTER_SETTINGS;
6 +
7 +static void add_string_to_filter(const char *data, struct bloom_filter *filter) {
8 + struct bloom_key key;
9 + int i;
10 +
11 + fill_bloom_key(data, strlen(data), &key, &settings);
12 + printf("Hashes:");
13 + for (i = 0; i < settings.num_hashes; i++){
14 + printf("0x%08x|", key.hashes[i]);
15 + }
16 + printf("\n");
17 + add_key_to_filter(&key, filter, &settings);
18 +}
19 +
20 +static void print_bloom_filter(struct bloom_filter *filter) {
21 + int i;
22 +
23 + if (!filter) {
24 + printf("No filter.\n");
25 + return;
26 + }
27 + printf("Filter_Length:%d\n", (int)filter->len);
28 + printf("Filter_Data:");
29 + for (i = 0; i < filter->len; i++){
30 + printf("%02x|", filter->data[i]);
31 + }
32 + printf("\n");
33 +}
34 +
35 int cmd__bloom(int argc, const char **argv)
36 {
37 if (!strcmp(argv[1], "get_murmur3")) {
@@ -9,5 +39,23 @@ int cmd__bloom(int argc, const char **argv)
39 printf("Murmur3 Hash with seed=0:0x%08x\n", hashed);
40 }
41
42 + if (!strcmp(argv[1], "generate_filter")) {
43 + struct bloom_filter filter;
44 + int i = 2;
45 + filter.len = (settings.bits_per_entry + BITS_PER_WORD - 1) / BITS_PER_WORD;
46 + filter.data = xcalloc(filter.len, sizeof(unsigned char));
47 +
48 + if (!argv[2]){
49 + die("at least one input string expected");
50 + }
51 +
52 + while (argv[i]) {
53 + add_string_to_filter(argv[i], &filter);
54 + i++;
55 + }
56 +
57 + print_bloom_filter(&filter);
58 + }
59 +
60 return 0;
61 }
\ No newline at end of file
t/t0095-bloom.sh
+40
@@ -27,4 +27,44 @@ test_expect_success 'compute unseeded murmur3 hash for test string 2' '
27 test_cmp expect actual
28 '
29
30 +test_expect_success 'compute bloom key for empty string' '
31 + cat >expect <<-\EOF &&
32 + Hashes:0x5615800c|0x5b966560|0x61174ab4|0x66983008|0x6c19155c|0x7199fab0|0x771ae004|
33 + Filter_Length:2
34 + Filter_Data:11|11|
35 + EOF
36 + test-tool bloom generate_filter "" >actual &&
37 + test_cmp expect actual
38 +'
39 +
40 +test_expect_success 'compute bloom key for whitespace' '
41 + cat >expect <<-\EOF &&
42 + Hashes:0xf178874c|0x5f3d6eb6|0xcd025620|0x3ac73d8a|0xa88c24f4|0x16510c5e|0x8415f3c8|
43 + Filter_Length:2
44 + Filter_Data:51|55|
45 + EOF
46 + test-tool bloom generate_filter " " >actual &&
47 + test_cmp expect actual
48 +'
49 +
50 +test_expect_success 'compute bloom key for test string 1' '
51 + cat >expect <<-\EOF &&
52 + Hashes:0xb270de9b|0x1bb6f26e|0x84fd0641|0xee431a14|0x57892de7|0xc0cf41ba|0x2a15558d|
53 + Filter_Length:2
54 + Filter_Data:92|6c|
55 + EOF
56 + test-tool bloom generate_filter "Hello world!" >actual &&
57 + test_cmp expect actual
58 +'
59 +
60 +test_expect_success 'compute bloom key for test string 2' '
61 + cat >expect <<-\EOF &&
62 + Hashes:0x20ab385b|0xf5237fe2|0xc99bc769|0x9e140ef0|0x728c5677|0x47049dfe|0x1b7ce585|
63 + Filter_Length:2
64 + Filter_Data:a5|4a|
65 + EOF
66 + test-tool bloom generate_filter "file.txt" >actual &&
67 + test_cmp expect actual
68 +'
69 +
70 test_done
\ No newline at end of file