git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH v3 1/2] bloom: replace struct bloom_key * with struct bloom_keyvec

From
Lidong Yan <yldhome2d2@gmail.com>
Date
Jun 28, 2025, 04:21 UTC
Message-ID
<20250628042140.1097910-2-502024330056@smail.nju.edu.cn>
In-Reply-To
<20250627062154.1121530-1-502024330056@smail.nju.edu.cn>

The revision traversal limited by pathspec has optimization when the pathspec has only one element. To support optimization for multiple pathspec items, we need to modify the data structures in struct rev_info.

struct rev_info uses bloom_keys and bloom_nr to store the bloom keys corresponding to a single pathspec item. To allow struct rev_info to store bloom keys for multiple pathspec items, a new data structure `struct bloom_keyvec` is introduced. Each `struct bloom_keyvec` corresponds to a single pathspec item.

In `struct rev_info`, replace bloom_keys and bloom_nr with bloom_keyvecs and bloom_keyvec_nr. This commit still optimize one pathspec item, thus bloom_keyvec_nr can only be 0 or 1.

New *_bloom_keyvec functions are added to create and destroy a keyvec. bloom_filter_contains_vec() is added to check if all key in keyvec is contained in a bloom filter. fill_bloom_keyvec_key() is added to initialize a key in keyvec.

Signed-off-by: Lidong Yan <502024330056@smail.nju.edu.cn>
---
 bloom.c    | 31 +++++++++++++++++++++++++++++++
 bloom.h    | 20 ++++++++++++++++++++
 revision.c | 36 ++++++++++++++++++------------------
 revision.h |  6 +++---
 4 files changed, 72 insertions(+), 21 deletions(-)
diff --git a/bloom.c b/bloom.c
index 0c8d2cebf9..8259cfce51 100644
--- a/bloom.c
+++ b/bloom.c
@@ -280,6 +280,25 @@ void deinit_bloom_filters(void)
 	deep_clear_bloom_filter_slab(&bloom_filters, free_one_bloom_filter);
 }
 
+struct bloom_keyvec *create_bloom_keyvec(size_t count)
+{
+	struct bloom_keyvec *vec;
+	size_t sz = sizeof(struct bloom_keyvec);
+	sz += count * sizeof(struct bloom_key);
+	vec = (struct bloom_keyvec *)xcalloc(1, sz);
+	vec->count = count;
+	return vec;
+}
+
+void destroy_bloom_keyvec(struct bloom_keyvec *vec)
+{
+	if (!vec)
+		return;
+	for (size_t nr = 0; nr < vec->count; nr++)
+		clear_bloom_key(&vec->key[nr]);
+	free(vec);
+}
+
 static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
 		       const struct hashmap_entry *eptr,
 		       const struct hashmap_entry *entry_or_key,
@@ -540,3 +559,15 @@ int bloom_filter_contains(const struct bloom_filter *filter,
 
 	return 1;
 }
+
+int bloom_filter_contains_vec(const struct bloom_filter *filter,
+			      const struct bloom_keyvec *vec,
+			      const struct bloom_filter_settings *settings)
+{
+	int ret = 1;
+
+	for (size_t nr = 0; ret > 0 && nr < vec->count; nr++)
+		ret = bloom_filter_contains(filter, &vec->key[nr], settings);
+
+	return ret;
+}
diff --git a/bloom.h b/bloom.h
index 6e46489a20..9e4e832c8c 100644
--- a/bloom.h
+++ b/bloom.h
@@ -74,6 +74,11 @@ struct bloom_key {
 	uint32_t *hashes;
 };
 
+struct bloom_keyvec {
+	size_t count;
+	struct bloom_key key[FLEX_ARRAY];
+};
+
 int load_bloom_filter_from_graph(struct commit_graph *g,
 				 struct bloom_filter *filter,
 				 uint32_t graph_pos);
@@ -100,6 +105,17 @@ void add_key_to_filter(const struct bloom_key *key,
 void init_bloom_filters(void);
 void deinit_bloom_filters(void);
 
+struct bloom_keyvec *create_bloom_keyvec(size_t count);
+void destroy_bloom_keyvec(struct bloom_keyvec *vec);
+
+static inline void fill_bloom_keyvec_key(const char *data, size_t len,
+					 struct bloom_keyvec *vec, size_t nr,
+					 const struct bloom_filter_settings *settings)
+{
+	assert(nr < vec->count);
+	fill_bloom_key(data, len, &vec->key[nr], settings);
+}
+
 enum bloom_filter_computed {
 	BLOOM_NOT_COMPUTED = (1 << 0),
 	BLOOM_COMPUTED     = (1 << 1),
@@ -137,4 +153,8 @@ int bloom_filter_contains(const struct bloom_filter *filter,
 			  const struct bloom_key *key,
 			  const struct bloom_filter_settings *settings);
 
+int bloom_filter_contains_vec(const struct bloom_filter *filter,
+			      const struct bloom_keyvec *v,
+			      const struct bloom_filter_settings *settings);
+
 #endif
diff --git a/revision.c b/revision.c
index afee111196..3aa544c137 100644
--- a/revision.c
+++ b/revision.c
@@ -688,6 +688,7 @@ static int forbid_bloom_filters(struct pathspec *spec)
 static void prepare_to_use_bloom_filter(struct rev_info *revs)
 {
 	struct pathspec_item *pi;
+	struct bloom_keyvec *bloom_keyvec;
 	char *path_alloc = NULL;
 	const char *path, *p;
 	size_t len;
@@ -736,19 +737,21 @@ static void prepare_to_use_bloom_filter(struct rev_info *revs)
 		p++;
 	}
 
-	revs->bloom_keys_nr = path_component_nr;
-	ALLOC_ARRAY(revs->bloom_keys, revs->bloom_keys_nr);
+	revs->bloom_keyvecs_nr = 1;
+	CALLOC_ARRAY(revs->bloom_keyvecs, 1);
+	bloom_keyvec = create_bloom_keyvec(path_component_nr);
+	revs->bloom_keyvecs[0] = bloom_keyvec;
 
-	fill_bloom_key(path, len, &revs->bloom_keys[0],
-		       revs->bloom_filter_settings);
+	fill_bloom_keyvec_key(path, len, bloom_keyvec, 0,
+			      revs->bloom_filter_settings);
 	path_component_nr = 1;
 
 	p = path + len - 1;
 	while (p > path) {
 		if (*p == '/')
-			fill_bloom_key(path, p - path,
-				       &revs->bloom_keys[path_component_nr++],
-				       revs->bloom_filter_settings);
+			fill_bloom_keyvec_key(path, p - path, bloom_keyvec,
+					      path_component_nr++,
+					      revs->bloom_filter_settings);
 		p--;
 	}
 
@@ -779,11 +782,8 @@ static int check_maybe_different_in_bloom_filter(struct rev_info *revs,
 		return -1;
 	}
 
-	for (j = 0; result && j < revs->bloom_keys_nr; j++) {
-		result = bloom_filter_contains(filter,
-					       &revs->bloom_keys[j],
-					       revs->bloom_filter_settings);
-	}
+	result = bloom_filter_contains_vec(filter, revs->bloom_keyvecs[0],
+					   revs->bloom_filter_settings);
 
 	if (result)
 		count_bloom_filter_maybe++;
@@ -823,7 +823,7 @@ static int rev_compare_tree(struct rev_info *revs,
 			return REV_TREE_SAME;
 	}
 
-	if (revs->bloom_keys_nr && !nth_parent) {
+	if (revs->bloom_keyvecs_nr && !nth_parent) {
 		bloom_ret = check_maybe_different_in_bloom_filter(revs, commit);
 
 		if (bloom_ret == 0)
@@ -850,7 +850,7 @@ static int rev_same_tree_as_empty(struct rev_info *revs, struct commit *commit,
 	if (!t1)
 		return 0;
 
-	if (!nth_parent && revs->bloom_keys_nr) {
+	if (!nth_parent && revs->bloom_keyvecs_nr) {
 		bloom_ret = check_maybe_different_in_bloom_filter(revs, commit);
 		if (!bloom_ret)
 			return 1;
@@ -3230,10 +3230,10 @@ void release_revisions(struct rev_info *revs)
 	line_log_free(revs);
 	oidset_clear(&revs->missing_commits);
 
-	for (int i = 0; i < revs->bloom_keys_nr; i++)
-		clear_bloom_key(&revs->bloom_keys[i]);
-	FREE_AND_NULL(revs->bloom_keys);
-	revs->bloom_keys_nr = 0;
+	for (int i = 0; i < revs->bloom_keyvecs_nr; i++)
+		destroy_bloom_keyvec(revs->bloom_keyvecs[i]);
+	FREE_AND_NULL(revs->bloom_keyvecs);
+	revs->bloom_keyvecs_nr = 0;
 }
 
 static void add_child(struct rev_info *revs, struct commit *parent, struct commit *child)
diff --git a/revision.h b/revision.h
index 6d369cdad6..ac843f58d0 100644
--- a/revision.h
+++ b/revision.h
@@ -62,7 +62,7 @@ struct repository;
 struct rev_info;
 struct string_list;
 struct saved_parents;
-struct bloom_key;
+struct bloom_keyvec;
 struct bloom_filter_settings;
 struct option;
 struct parse_opt_ctx_t;
@@ -360,8 +360,8 @@ struct rev_info {
 
 	/* Commit graph bloom filter fields */
 	/* The bloom filter key(s) for the pathspec */
-	struct bloom_key *bloom_keys;
-	int bloom_keys_nr;
+	struct bloom_keyvec **bloom_keyvecs;
+	int bloom_keyvecs_nr;
 
 	/*
 	 * The bloom filter settings used to generate the key.
-- 
2.50.0.108.g6ae0c543ae
Previous: Junio C HamanoNext: Patrick Steinhardt
Message 65 of 72 in “bloom: use bloom filter given multiple pathspec”
  1. 0/2 bloom: use bloom filter given multiple pathspecLidong Yan, Jun 25, 2025
  2. 1/2 bloom: replace struct bloom_key * with struct bloom_keyvecLidong Yan, Jun 25, 2025
  3. Junio C HamanoJun 25, 2025
  4. Lidong YanJun 26, 2025
  5. 2/2 bloom: enable multiple pathspec bloom keysLidong Yan, Jun 25, 2025
  6. Junio C HamanoJun 27, 2025
  7. Lidong YanJun 27, 2025
  8. Junio C HamanoJun 27, 2025
  9. Lidong YanJul 1, 2025
  10. Junio C HamanoJul 1, 2025
  11. Lidong YanJul 2, 2025
  12. Junio C HamanoJul 2, 2025
  13. Lidong YanJul 3, 2025
  14. Lidong YanJul 4, 2025
  15. SZEDER GáborJul 1, 2025
  16. Lidong YanJul 1, 2025
  17. Junio C HamanoJul 1, 2025
  18. Junio C HamanoJun 27, 2025
  19. Lidong YanJun 28, 2025
  20. Junio C HamanoJun 25, 2025
  21. Lidong YanJun 26, 2025
  22. Junio C HamanoJun 26, 2025
  23. 0/2 bloom: enable bloom filter optimization for multiple pathspec elements in revision traversalLidong Yan, Jun 27, 2025
  24. 0/2 bloom: enable bloom filter optimization for multiple pathspec elements in revision traversalLidong Yan, Jun 28, 2025
  25. 0/4 bloom: enable bloom filter optimization for multiple pathspec elements in revision traversalLidong Yan, Jul 4, 2025
  26. 1/4 bloom: add test helper to return murmur3 hashLidong Yan, Jul 4, 2025
  27. 2/4 bloom: rename function operates on bloom_keyLidong Yan, Jul 4, 2025
  28. 3/4 bloom: replace struct bloom_key * with struct bloom_keyvecLidong Yan, Jul 4, 2025
  29. Derrick StoleeJul 7, 2025
  30. Lidong YanJul 7, 2025
  31. 4/4 bloom: optimize multiple pathspec items in revision traversalLidong Yan, Jul 4, 2025
  32. Derrick StoleeJul 7, 2025
  33. Lidong YanJul 7, 2025
  34. Junio C HamanoJul 7, 2025
  35. 0/4 bloom: enable bloom filter optimization for multiple pathspec elements in revision traversalLidong Yan, Jul 10, 2025
  36. 1/4 bloom: add test helper to return murmur3 hashLidong Yan, Jul 10, 2025
  37. 2/4 bloom: rename function operates on bloom_keyLidong Yan, Jul 10, 2025
  38. 3/4 bloom: replace struct bloom_key * with struct bloom_keyvecLidong Yan, Jul 10, 2025
  39. Junio C HamanoJul 10, 2025
  40. Lidong YanJul 11, 2025
  41. Junio C HamanoJul 11, 2025
  42. 4/4 bloom: optimize multiple pathspec items in revision traversalLidong Yan, Jul 10, 2025
  43. 5/4 revision: make helper for pathspec to bloom keyDerrick Stolee, Jul 10, 2025
  44. Lidong YanJul 10, 2025
  45. 4/4 bloom: optimize multiple pathspec items in revisionDerrick Stolee, Jul 10, 2025
  46. Lidong YanJul 10, 2025
  47. Derrick StoleeJul 10, 2025
  48. 0/5 bloom: enable bloom filter optimization for multiple pathspec elements in revision traversalLidong Yan, Jul 12, 2025
  49. 1/5 bloom: add test helper to return murmur3 hashLidong Yan, Jul 12, 2025
  50. 2/5 bloom: rename function operates on bloom_keyLidong Yan, Jul 12, 2025
  51. 3/5 bloom: replace struct bloom_key * with struct bloom_keyvecLidong Yan, Jul 12, 2025
  52. 4/5 revision: make helper for pathspec to bloom keyvecLidong Yan, Jul 12, 2025
  53. 5/5 To enable optimize multiple pathspec items in revision traversal, return 0 if all pathspec item is literal in forbid_bloom_filters(). Add for loops to initialize and check each pathspec item's bloom_keyvec when optimization is possible.Lidong Yan, Jul 12, 2025
  54. Lidong YanJul 12, 2025
  55. 5/5 bloom: optimize multiple pathspec items in revisionLidong Yan, Jul 12, 2025
  56. Derrick StoleeJul 14, 2025
  57. Junio C HamanoJul 14, 2025
  58. Lidong YanJul 15, 2025
  59. [RESEND][PATCH v6 5/5] bloom: optimize multiple pathspec items in revisionLidong Yan, Jul 15, 2025
  60. Derrick StoleeJul 14, 2025
  61. Junio C HamanoJul 14, 2025
  62. Lidong YanJul 15, 2025
  63. Derrick StoleeJul 15, 2025
  64. Junio C HamanoJul 15, 2025
  65. 1/2 bloom: replace struct bloom_key * with struct bloom_keyvecLidong Yan, Jun 28, 2025
  66. Patrick SteinhardtJul 2, 2025
  67. Lidong YanJul 2, 2025
  68. Junio C HamanoJul 2, 2025
  69. Lidong YanJul 3, 2025
  70. 2/2 bloom: optimize multiple pathspec items in revision traversalLidong Yan, Jun 28, 2025
  71. 1/2 bloom: replace struct bloom_key * with struct bloom_keyvecLidong Yan, Jun 27, 2025
  72. 2/2 bloom: optimize multiple pathspec items in revision traversalLidong Yan, Jun 27, 2025

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.