git/bloom.c

362 lines
9.2 KiB
C
Raw Normal View History

#include "git-compat-util.h"
#include "bloom.h"
#include "diff.h"
#include "diffcore.h"
#include "hashmap.h"
#include "commit-graph.h"
#include "commit.h"
#include "commit-slab.h"
define_commit_slab(bloom_filter_slab, struct bloom_filter);
static struct bloom_filter_slab bloom_filters;
struct pathmap_hash_entry {
struct hashmap_entry entry;
const char path[FLEX_ARRAY];
};
static uint32_t rotate_left(uint32_t value, int32_t count)
{
uint32_t mask = 8 * sizeof(uint32_t) - 1;
count &= mask;
return ((value << count) | (value >> ((-count) & mask)));
}
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>
2020-03-30 02:31:25 +02:00
static inline unsigned char get_bitmask(uint32_t pos)
{
return ((unsigned char)1) << (pos & (BITS_PER_WORD - 1));
}
commit-graph: check bounds when accessing BDAT chunk When loading Bloom filters from a commit-graph file, we use the offset values in the BIDX chunk to index into the memory mapped for the BDAT chunk. But since we don't record how big the BDAT chunk is, we just trust that the BIDX offsets won't cause us to read outside of the chunk memory. A corrupted or malicious commit-graph file will cause us to segfault (in practice this isn't a very interesting attack, since commit-graph files are local-only, and the worst case is an out-of-bounds read). We can't fix this by checking the chunk size during parsing, since the data in the BDAT chunk doesn't have a fixed size (that's why we need the BIDX in the first place). So we'll fix it in two parts: 1. Record the BDAT chunk size during parsing, and then later check that the BIDX offsets we look up are within bounds. 2. Because the offsets are relative to the end of the BDAT header, we must also make sure that the BDAT chunk is at least as large as the expected header size. Otherwise, we overflow when trying to move past the header, even for an offset of "0". We can check this early, during the parsing stage. The error messages are rather verbose, but since this is not something you'd expect to see outside of severe bugs or corruption, it makes sense to err on the side of too many details. Sadly we can't mention the filename during the chunk-parsing stage, as we haven't set g->filename at this point, nor passed it down through the stack. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2023-10-09 23:05:50 +02:00
static int check_bloom_offset(struct commit_graph *g, uint32_t pos,
uint32_t offset)
{
/*
* Note that we allow offsets equal to the data size, which would set
* our pointers at one past the end of the chunk memory. This is
* necessary because the on-disk index points to the end of the
* entries (so we can compute size by comparing adjacent ones). And
* naturally the final entry's end is one-past-the-end of the chunk.
*/
if (offset <= g->chunk_bloom_data_size - BLOOMDATA_CHUNK_HEADER_SIZE)
return 0;
warning("ignoring out-of-range offset (%"PRIuMAX") for changed-path"
" filter at pos %"PRIuMAX" of %s (chunk size: %"PRIuMAX")",
(uintmax_t)offset, (uintmax_t)pos,
g->filename, (uintmax_t)g->chunk_bloom_data_size);
return -1;
}
static int load_bloom_filter_from_graph(struct commit_graph *g,
struct bloom_filter *filter,
commit-graph: fix corrupt upgrade from generation v1 to v2 The previous commit demonstrates a bug where a commit-graph using generation v2 could enter a state where one of the GDA2 values has its most-significant bit set (indicating that its value should be read from the extended offset table in the GDO2 chunk) without having a GDO2 chunk to read from. This results in the following error message being displayed to the caller: fatal: commit-graph requires overflow generation data but has none This bug arises in the following scenario: - We decide to write a commit-graph using generation number v2, and decide (correctly) that no GDO2 chunk is necessary (e.g., because all of the commiter date offsets are no larger than 2^31-1). - The v2 generation numbers are stored in the `->generation` member of the commit slab holding `struct commit_graph_data`'s. - Later on, `load_commit_graph_info()` is called, overwriting the v2 generation data in the aforementioned slab with any existing v1 generation data. Then, when the commit-graph code goes to write the GDA2 chunk via `write_graph_chunk_generation_data()`, we use the overwritten generation v1 data in a place where we expect to use a v2 generation number: offset = commit_graph_data_at(c)->generation - c->date; ...because `commit_graph_data_at(c)->generation` used to hold the v2 generation data, but it was overwritten to contain the v1 generation number via `load_commit_graph_info()`. If the `offset` computation above overflows the v2 generation number max, then `write_graph_chunk_generation_data()` will update its count of large offsets and write the marker accordingly: if (offset > GENERATION_NUMBER_V2_OFFSET_MAX) { offset = CORRECTED_COMMIT_DATE_OFFSET_OVERFLOW | num_generation_data_overflows; num_generation_data_overflows++; } and reads will look for the GDO2 chunk containing the overflowing v2 generation number, *after* the commit-graph code decided that no such chunk was necessary. The main problem is that the slab containing `struct commit_graph_data` has a dual purpose. It is used to hold data that we are about to write to disk while generating a commit-graph, as well as hold data that was read from an existing commit-graph. When the two mix, namely when the result of reading the commit-graph has a side-effect that mixes poorly with an in-progress commit-graph write, we end up with corrupt data. A complete fix might be to introduce a new slab that is used exclusively for writing, and gate access between the two slabs based on context provided by the caller (e.g., whether this computation is part of a "read" or "write" operation). But a more minimal fix addresses the only known path which overwrites the slab data, which is `compute_bloom_filters()` -> `get_or_compute_bloom_filter()` -> `load_commit_graph_info()` -> `fill_commit_graph_info()` by avoiding the last call which clobbers the data altogether. This path only needs to learn the graph position of a given commit so that it can be used in `load_bloom_filter_from_graph()`. By replacing the last steps of the above with one that records the graph position into a temporary variable which is then used to load the existing Bloom data, we eliminate the clobbering, removing the corruption. Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-07-13 01:10:33 +02:00
uint32_t graph_pos)
{
uint32_t lex_pos, start_index, end_index;
while (graph_pos < g->num_commits_in_base)
g = g->base_graph;
commit-graph: introduce 'get_bloom_filter_settings()' Many places in the code often need a pointer to the commit-graph's 'struct bloom_filter_settings', in which case they often take the value from the top-most commit-graph. In the non-split case, this works as expected. In the split case, however, things get a little tricky. Not all layers in a chain of incremental commit-graphs are required to themselves have Bloom data, and so whether or not some part of the code uses Bloom filters depends entirely on whether or not the top-most level of the commit-graph chain has Bloom filters. This has been the behavior since Bloom filters were introduced, and has been codified into the tests since a759bfa9ee (t4216: add end to end tests for git log with Bloom filters, 2020-04-06). In fact, t4216.130 requires that Bloom filters are not used in exactly the case described earlier. There is no reason that this needs to be the case, since it is perfectly valid for commits in an earlier layer to have Bloom filters when commits in a newer layer do not. Since Bloom settings are guaranteed in practice to be the same for any layer in a chain that has Bloom data, it is sufficient to traverse the '->base_graph' pointer until either (1) a non-null 'struct bloom_filter_settings *' is found, or (2) until we are at the root of the commit-graph chain. Introduce a 'get_bloom_filter_settings()' function that does just this, and use it instead of purely dereferencing the top-most graph's '->bloom_filter_settings' pointer. While we're at it, add an additional test in t5324 to guard against code in the commit-graph writing machinery that doesn't correctly handle a NULL 'struct bloom_filter *'. Co-authored-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-09-09 17:22:44 +02:00
/* The commit graph commit 'c' lives in doesn't carry Bloom filters. */
if (!g->chunk_bloom_indexes)
return 0;
lex_pos = graph_pos - g->num_commits_in_base;
end_index = get_be32(g->chunk_bloom_indexes + 4 * lex_pos);
if (lex_pos > 0)
start_index = get_be32(g->chunk_bloom_indexes + 4 * (lex_pos - 1));
else
start_index = 0;
commit-graph: check bounds when accessing BDAT chunk When loading Bloom filters from a commit-graph file, we use the offset values in the BIDX chunk to index into the memory mapped for the BDAT chunk. But since we don't record how big the BDAT chunk is, we just trust that the BIDX offsets won't cause us to read outside of the chunk memory. A corrupted or malicious commit-graph file will cause us to segfault (in practice this isn't a very interesting attack, since commit-graph files are local-only, and the worst case is an out-of-bounds read). We can't fix this by checking the chunk size during parsing, since the data in the BDAT chunk doesn't have a fixed size (that's why we need the BIDX in the first place). So we'll fix it in two parts: 1. Record the BDAT chunk size during parsing, and then later check that the BIDX offsets we look up are within bounds. 2. Because the offsets are relative to the end of the BDAT header, we must also make sure that the BDAT chunk is at least as large as the expected header size. Otherwise, we overflow when trying to move past the header, even for an offset of "0". We can check this early, during the parsing stage. The error messages are rather verbose, but since this is not something you'd expect to see outside of severe bugs or corruption, it makes sense to err on the side of too many details. Sadly we can't mention the filename during the chunk-parsing stage, as we haven't set g->filename at this point, nor passed it down through the stack. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2023-10-09 23:05:50 +02:00
if (check_bloom_offset(g, lex_pos, end_index) < 0 ||
check_bloom_offset(g, lex_pos - 1, start_index) < 0)
return 0;
if (end_index < start_index) {
warning("ignoring decreasing changed-path index offsets"
" (%"PRIuMAX" > %"PRIuMAX") for positions"
" %"PRIuMAX" and %"PRIuMAX" of %s",
(uintmax_t)start_index, (uintmax_t)end_index,
(uintmax_t)(lex_pos-1), (uintmax_t)lex_pos,
g->filename);
return 0;
}
filter->len = end_index - start_index;
filter->data = (unsigned char *)(g->chunk_bloom_data +
sizeof(unsigned char) * start_index +
BLOOMDATA_CHUNK_HEADER_SIZE);
return 1;
}
/*
* Calculate the murmur3 32-bit hash value for the given data
* using the given seed.
* Produces a uniformly distributed hash value.
* Not considered to be cryptographically secure.
* Implemented as described in https://en.wikipedia.org/wiki/MurmurHash#Algorithm
*/
uint32_t murmur3_seeded(uint32_t seed, const char *data, size_t len)
{
const uint32_t c1 = 0xcc9e2d51;
const uint32_t c2 = 0x1b873593;
const uint32_t r1 = 15;
const uint32_t r2 = 13;
const uint32_t m = 5;
const uint32_t n = 0xe6546b64;
int i;
uint32_t k1 = 0;
const char *tail;
int len4 = len / sizeof(uint32_t);
uint32_t k;
for (i = 0; i < len4; i++) {
uint32_t byte1 = (uint32_t)data[4*i];
uint32_t byte2 = ((uint32_t)data[4*i + 1]) << 8;
uint32_t byte3 = ((uint32_t)data[4*i + 2]) << 16;
uint32_t byte4 = ((uint32_t)data[4*i + 3]) << 24;
k = byte1 | byte2 | byte3 | byte4;
k *= c1;
k = rotate_left(k, r1);
k *= c2;
seed ^= k;
seed = rotate_left(seed, r2) * m + n;
}
tail = (data + len4 * sizeof(uint32_t));
switch (len & (sizeof(uint32_t) - 1)) {
case 3:
k1 ^= ((uint32_t)tail[2]) << 16;
/*-fallthrough*/
case 2:
k1 ^= ((uint32_t)tail[1]) << 8;
/*-fallthrough*/
case 1:
k1 ^= ((uint32_t)tail[0]) << 0;
k1 *= c1;
k1 = rotate_left(k1, r1);
k1 *= c2;
seed ^= k1;
break;
}
seed ^= (uint32_t)len;
seed ^= (seed >> 16);
seed *= 0x85ebca6b;
seed ^= (seed >> 13);
seed *= 0xc2b2ae35;
seed ^= (seed >> 16);
return seed;
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>
2020-03-30 02:31:25 +02:00
}
void fill_bloom_key(const char *data,
size_t len,
struct bloom_key *key,
const struct bloom_filter_settings *settings)
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>
2020-03-30 02:31:25 +02:00
{
int i;
const uint32_t seed0 = 0x293ae76f;
const uint32_t seed1 = 0x7e646e2c;
const uint32_t hash0 = murmur3_seeded(seed0, data, len);
const uint32_t hash1 = murmur3_seeded(seed1, data, len);
key->hashes = (uint32_t *)xcalloc(settings->num_hashes, sizeof(uint32_t));
for (i = 0; i < settings->num_hashes; i++)
key->hashes[i] = hash0 + i * hash1;
}
line-log: integrate with changed-path Bloom filters The previous changes to the line-log machinery focused on making the first result appear faster. This was achieved by no longer walking the entire commit history before returning the early results. There is still another way to improve the performance: walk most commits much faster. Let's use the changed-path Bloom filters to reduce time spent computing diffs. Since the line-log computation requires opening blobs and checking the content-diff, there is still a lot of necessary computation that cannot be replaced with changed-path Bloom filters. The part that we can reduce is most effective when checking the history of a file that is deep in several directories and those directories are modified frequently. In this case, the computation to check if a commit is TREESAME to its first parent takes a large fraction of the time. That is ripe for improvement with changed-path Bloom filters. We must ensure that prepare_to_use_bloom_filters() is called in revision.c so that the bloom_filter_settings are loaded into the struct rev_info from the commit-graph. Of course, some cases are still forbidden, but in the line-log case the pathspec is provided in a different way than normal. Since multiple paths and segments could be requested, we compute the struct bloom_key data dynamically during the commit walk. This could likely be improved, but adds code complexity that is not valuable at this time. There are two cases to care about: merge commits and "ordinary" commits. Merge commits have multiple parents, but if we are TREESAME to our first parent in every range, then pass the blame for all ranges to the first parent. Ordinary commits have the same condition, but each is done slightly differently in the process_ranges_[merge|ordinary]_commit() methods. By checking if the changed-path Bloom filter can guarantee TREESAME, we can avoid that tree-diff cost. If the filter says "probably changed", then we need to run the tree-diff and then the blob-diff if there was a real edit. The Linux kernel repository is a good testing ground for the performance improvements claimed here. There are two different cases to test. The first is the "entire history" case, where we output the entire history to /dev/null to see how long it would take to compute the full line-log history. The second is the "first result" case, where we find how long it takes to show the first value, which is an indicator of how quickly a user would see responses when waiting at a terminal. To test, I selected the paths that were changed most frequently in the top 10,000 commits using this command (stolen from StackOverflow [1]): git log --pretty=format: --name-only -n 10000 | sort | \ uniq -c | sort -rg | head -10 which results in 121 MAINTAINERS 63 fs/namei.c 60 arch/x86/kvm/cpuid.c 59 fs/io_uring.c 58 arch/x86/kvm/vmx/vmx.c 51 arch/x86/kvm/x86.c 45 arch/x86/kvm/svm.c 42 fs/btrfs/disk-io.c 42 Documentation/scsi/index.rst (along with a bogus first result). It appears that the path arch/x86/kvm/svm.c was renamed, so we ignore that entry. This leaves the following results for the real command time: | | Entire History | First Result | | Path | Before | After | Before | After | |------------------------------|--------|--------|--------|--------| | MAINTAINERS | 4.26 s | 3.87 s | 0.41 s | 0.39 s | | fs/namei.c | 1.99 s | 0.99 s | 0.42 s | 0.21 s | | arch/x86/kvm/cpuid.c | 5.28 s | 1.12 s | 0.16 s | 0.09 s | | fs/io_uring.c | 4.34 s | 0.99 s | 0.94 s | 0.27 s | | arch/x86/kvm/vmx/vmx.c | 5.01 s | 1.34 s | 0.21 s | 0.12 s | | arch/x86/kvm/x86.c | 2.24 s | 1.18 s | 0.21 s | 0.14 s | | fs/btrfs/disk-io.c | 1.82 s | 1.01 s | 0.06 s | 0.05 s | | Documentation/scsi/index.rst | 3.30 s | 0.89 s | 1.46 s | 0.03 s | It is worth noting that the least speedup comes for the MAINTAINERS file which is * edited frequently, * low in the directory heirarchy, and * quite a large file. All of those points lead to spending more time doing the blob diff and less time doing the tree diff. Still, we see some improvement in that case and significant improvement in other cases. A 2-4x speedup is likely the more typical case as opposed to the small 5% change for that file. Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-05-11 13:56:19 +02:00
void clear_bloom_key(struct bloom_key *key)
{
FREE_AND_NULL(key->hashes);
}
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>
2020-03-30 02:31:25 +02:00
void add_key_to_filter(const struct bloom_key *key,
struct bloom_filter *filter,
const struct bloom_filter_settings *settings)
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>
2020-03-30 02:31:25 +02:00
{
int i;
uint64_t mod = filter->len * BITS_PER_WORD;
for (i = 0; i < settings->num_hashes; i++) {
uint64_t hash_mod = key->hashes[i] % mod;
uint64_t block_pos = hash_mod / BITS_PER_WORD;
filter->data[block_pos] |= get_bitmask(hash_mod);
}
}
void init_bloom_filters(void)
{
init_bloom_filter_slab(&bloom_filters);
}
static int pathmap_cmp(const void *hashmap_cmp_fn_data UNUSED,
const struct hashmap_entry *eptr,
const struct hashmap_entry *entry_or_key,
const void *keydata UNUSED)
{
const struct pathmap_hash_entry *e1, *e2;
e1 = container_of(eptr, const struct pathmap_hash_entry, entry);
e2 = container_of(entry_or_key, const struct pathmap_hash_entry, entry);
return strcmp(e1->path, e2->path);
}
bloom: encode out-of-bounds filters as non-empty When a changed-path Bloom filter has either zero, or more than a certain number (commonly 512) of entries, the commit-graph machinery encodes it as "missing". More specifically, it sets the indices adjacent in the BIDX chunk as equal to each other to indicate a "length 0" filter; that is, that the filter occupies zero bytes on disk. This has heretofore been fine, since the commit-graph machinery has no need to care about these filters with too few or too many changed paths. Both cases act like no filter has been generated at all, and so there is no need to store them. In a subsequent commit, however, the commit-graph machinery will learn to only compute Bloom filters for some commits in the current commit-graph layer. This is a change from the current implementation which computes Bloom filters for all commits that are in the layer being written. Critically for this patch, only computing some of the Bloom filters means adding a third state for length 0 Bloom filters: zero entries, too many entries, or "hasn't been computed". It will be important for that future patch to distinguish between "not representable" (i.e., zero or too-many changed paths), and "hasn't been computed". In particular, we don't want to waste time recomputing filters that have already been computed. To that end, change how we store Bloom filters in the "computed but not representable" category: - Bloom filters with no entries are stored as a single byte with all bits low (i.e., all queries to that Bloom filter will return "definitely not") - Bloom filters with too many entries are stored as a single byte with all bits set high (i.e., all queries to that Bloom filter will return "maybe"). These rules are sufficient to not incur a behavior change by changing the on-disk representation of these two classes. Likewise, no specification changes are necessary for the commit-graph format, either: - Filters that were previously empty will be recomputed and stored according to the new rules, and - old clients reading filters generated by new clients will interpret the filters correctly and be none the wiser to how they were generated. Clients will invoke the Bloom machinery in more cases than before, but this can be addressed by returning a NULL filter when all bits are set high. This can be addressed in a future patch. Note that this does increase the size of on-disk commit-graphs, but far less than other proposals. In particular, this is generally more efficient than storing a bitmap for which commits haven't computed their Bloom filters. Storing a bitmap incurs a penalty of one bit per commit, whereas storing explicit filters as above incurs a penalty of one byte per too-large or empty commit. In practice, these boundary commits likely occupy a small proportion of the overall number of commits, and so the size penalty is likely smaller than storing a bitmap for all commits. See, for example, these relative proportions of such boundary commits (collected by SZEDER Gábor): | Percentage of | commit-graph | | | commits modifying | file size | | ├────────┬──────────────┼───────────────────┤ pct. | | 0 path | >= 512 paths | before | after | change | ┌────────────────┼────────┼──────────────┼─────────┼─────────┼───────────┤ | android-base | 13.20% | 0.13% | 37.468M | 37.534M | +0.1741 % | | cmssw | 0.15% | 0.23% | 17.118M | 17.119M | +0.0091 % | | cpython | 3.07% | 0.01% | 7.967M | 7.971M | +0.0423 % | | elasticsearch | 0.70% | 1.00% | 8.833M | 8.835M | +0.0128 % | | gcc | 0.00% | 0.08% | 16.073M | 16.074M | +0.0030 % | | gecko-dev | 0.14% | 0.64% | 59.868M | 59.874M | +0.0105 % | | git | 0.11% | 0.02% | 3.895M | 3.895M | +0.0020 % | | glibc | 0.02% | 0.10% | 3.555M | 3.555M | +0.0021 % | | go | 0.00% | 0.07% | 3.186M | 3.186M | +0.0018 % | | homebrew-cask | 0.40% | 0.02% | 7.035M | 7.035M | +0.0065 % | | homebrew-core | 0.01% | 0.01% | 11.611M | 11.611M | +0.0002 % | | jdk | 0.26% | 5.64% | 5.537M | 5.540M | +0.0590 % | | linux | 0.01% | 0.51% | 63.735M | 63.740M | +0.0073 % | | llvm-project | 0.12% | 0.03% | 25.515M | 25.516M | +0.0050 % | | rails | 0.10% | 0.10% | 6.252M | 6.252M | +0.0027 % | | rust | 0.07% | 0.17% | 9.364M | 9.364M | +0.0033 % | | tensorflow | 0.09% | 1.02% | 7.009M | 7.010M | +0.0158 % | | webkit | 0.05% | 0.31% | 17.405M | 17.406M | +0.0047 % | (where the above increase is determined by computing a non-split commit-graph before and after this patch). Given that these projects are all "large" by commit count, the storage cost by writing these filters explicitly is negligible. In the most extreme example, android-base (which has 494,848 commits at the time of writing) would have its commit-graph increase by a modest 68.4 KB. Finally, a test to exercise filters which contain too many changed path entries will be introduced in a subsequent patch. Suggested-by: SZEDER Gábor <szeder.dev@gmail.com> Suggested-by: Jakub Narębski <jnareb@gmail.com> Helped-by: Derrick Stolee <dstolee@microsoft.com> Helped-by: SZEDER Gábor <szeder.dev@gmail.com> Helped-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-09-18 04:59:44 +02:00
static void init_truncated_large_filter(struct bloom_filter *filter)
{
filter->data = xmalloc(1);
filter->data[0] = 0xFF;
filter->len = 1;
}
struct bloom_filter *get_or_compute_bloom_filter(struct repository *r,
struct commit *c,
int compute_if_not_present,
const struct bloom_filter_settings *settings,
enum bloom_filter_computed *computed)
{
struct bloom_filter *filter;
int i;
struct diff_options diffopt;
if (computed)
*computed = BLOOM_NOT_COMPUTED;
if (!bloom_filters.slab_size)
return NULL;
filter = bloom_filter_slab_at(&bloom_filters, c);
if (!filter->data) {
commit-graph: fix corrupt upgrade from generation v1 to v2 The previous commit demonstrates a bug where a commit-graph using generation v2 could enter a state where one of the GDA2 values has its most-significant bit set (indicating that its value should be read from the extended offset table in the GDO2 chunk) without having a GDO2 chunk to read from. This results in the following error message being displayed to the caller: fatal: commit-graph requires overflow generation data but has none This bug arises in the following scenario: - We decide to write a commit-graph using generation number v2, and decide (correctly) that no GDO2 chunk is necessary (e.g., because all of the commiter date offsets are no larger than 2^31-1). - The v2 generation numbers are stored in the `->generation` member of the commit slab holding `struct commit_graph_data`'s. - Later on, `load_commit_graph_info()` is called, overwriting the v2 generation data in the aforementioned slab with any existing v1 generation data. Then, when the commit-graph code goes to write the GDA2 chunk via `write_graph_chunk_generation_data()`, we use the overwritten generation v1 data in a place where we expect to use a v2 generation number: offset = commit_graph_data_at(c)->generation - c->date; ...because `commit_graph_data_at(c)->generation` used to hold the v2 generation data, but it was overwritten to contain the v1 generation number via `load_commit_graph_info()`. If the `offset` computation above overflows the v2 generation number max, then `write_graph_chunk_generation_data()` will update its count of large offsets and write the marker accordingly: if (offset > GENERATION_NUMBER_V2_OFFSET_MAX) { offset = CORRECTED_COMMIT_DATE_OFFSET_OVERFLOW | num_generation_data_overflows; num_generation_data_overflows++; } and reads will look for the GDO2 chunk containing the overflowing v2 generation number, *after* the commit-graph code decided that no such chunk was necessary. The main problem is that the slab containing `struct commit_graph_data` has a dual purpose. It is used to hold data that we are about to write to disk while generating a commit-graph, as well as hold data that was read from an existing commit-graph. When the two mix, namely when the result of reading the commit-graph has a side-effect that mixes poorly with an in-progress commit-graph write, we end up with corrupt data. A complete fix might be to introduce a new slab that is used exclusively for writing, and gate access between the two slabs based on context provided by the caller (e.g., whether this computation is part of a "read" or "write" operation). But a more minimal fix addresses the only known path which overwrites the slab data, which is `compute_bloom_filters()` -> `get_or_compute_bloom_filter()` -> `load_commit_graph_info()` -> `fill_commit_graph_info()` by avoiding the last call which clobbers the data altogether. This path only needs to learn the graph position of a given commit so that it can be used in `load_bloom_filter_from_graph()`. By replacing the last steps of the above with one that records the graph position into a temporary variable which is then used to load the existing Bloom data, we eliminate the clobbering, removing the corruption. Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2022-07-13 01:10:33 +02:00
uint32_t graph_pos;
if (repo_find_commit_pos_in_graph(r, c, &graph_pos))
load_bloom_filter_from_graph(r->objects->commit_graph,
filter, graph_pos);
}
2020-09-18 15:27:27 +02:00
if (filter->data && filter->len)
return filter;
if (!compute_if_not_present)
return NULL;
repo_diff_setup(r, &diffopt);
diffopt.flags.recursive = 1;
diffopt.detect_rename = 0;
diffopt.max_changes = settings->max_changed_paths;
diff_setup_done(&diffopt);
/* ensure commit is parsed so we have parent information */
repo_parse_commit(r, c);
if (c->parents)
diff_tree_oid(&c->parents->item->object.oid, &c->object.oid, "", &diffopt);
else
diff_tree_oid(NULL, &c->object.oid, "", &diffopt);
diffcore_std(&diffopt);
bloom/diff: properly short-circuit on max_changes Commit e3696980 (diff: halt tree-diff early after max_changes, 2020-03-30) intended to create a mechanism to short-circuit a diff calculation after a certain number of paths were modified. By incrementing a "num_changes" counter throughout the recursive ll_diff_tree_paths(), this was supposed to match the number of changes that would be written into the changed-path Bloom filters. Unfortunately, this was not implemented correctly and instead misses simple cases like file modifications. This then does not stop very large changed-path filters from being written (unless they add or remove many files). To start, change the implementation in ll_diff_tree_paths() to instead use the global diff_queue_diff struct's 'nr' member as the count. This is a way to simplify the logic instead of making more mistakes in the complicated diff code. This has a drawback: the diff_queue_diff struct only lists the paths corresponding to blob changes, not their leading directories. Thus, get_or_compute_bloom_filter() needs an additional check to see if the hashmap with the leading directories becomes too large. One reason why this was not caught by test cases was that the test in t4216-log-bloom.sh that was supposed to check this "too many changes" condition only checked this on the initial commit of a repository. The old logic counted these values correctly. Update this test in a few ways: 1. Use GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS to reduce the limit, allowing smaller commits to engage with this logic. 2. Create several interesting cases of edits, adds, removes, and mode changes (in the second commit). By testing both sides of the inequality with the *_MAX_CHANGED_PATHS variable, we can see that the count is exactly correct, so none of these changes are missed or over-counted. 3. Use the trace2 data value filter_found_large to verify that these commits are on the correct side of the limit. Another way to verify the behavior is correct is through performance tests. By testing on my local copies of the Git repository and the Linux kernel repository, I could measure the effect of these short-circuits when computing a fresh commit-graph file with changed-path Bloom filters using the command GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS=N time \ git commit-graph write --reachable --changed-paths and reporting the wall time and resulting commit-graph size. For Git, the results are | | N=1 | N=10 | N=512 | |--------|----------------|----------------|----------------| | HEAD~1 | 10.90s 9.18MB | 11.11s 9.34MB | 11.31s 9.35MB | | HEAD | 9.21s 8.62MB | 11.11s 9.29MB | 11.29s 9.34MB | For Linux, the results are | | N=1 | N=20 | N=512 | |--------|----------------|---------------|---------------| | HEAD~1 | 61.28s 64.3MB | 76.9s 72.6MB | 77.6s 72.6MB | | HEAD | 49.44s 56.3MB | 68.7s 65.9MB | 69.2s 65.9MB | Naturally, the improvement becomes much less as the limit grows, as fewer commits satisfy the short-circuit. Reported-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-09-16 20:07:52 +02:00
if (diff_queued_diff.nr <= settings->max_changed_paths) {
struct hashmap pathmap = HASHMAP_INIT(pathmap_cmp, NULL);
struct pathmap_hash_entry *e;
struct hashmap_iter iter;
for (i = 0; i < diff_queued_diff.nr; i++) {
const char *path = diff_queued_diff.queue[i]->two->path;
/*
* Add each leading directory of the changed file, i.e. for
* 'dir/subdir/file' add 'dir' and 'dir/subdir' as well, so
* the Bloom filter could be used to speed up commands like
* 'git log dir/subdir', too.
*
* Note that directories are added without the trailing '/'.
*/
do {
char *last_slash = strrchr(path, '/');
FLEX_ALLOC_STR(e, path, path);
hashmap_entry_init(&e->entry, strhash(path));
if (!hashmap_get(&pathmap, &e->entry, NULL))
hashmap_add(&pathmap, &e->entry);
else
free(e);
if (!last_slash)
last_slash = (char*)path;
*last_slash = '\0';
} while (*path);
diff_free_filepair(diff_queued_diff.queue[i]);
}
bloom/diff: properly short-circuit on max_changes Commit e3696980 (diff: halt tree-diff early after max_changes, 2020-03-30) intended to create a mechanism to short-circuit a diff calculation after a certain number of paths were modified. By incrementing a "num_changes" counter throughout the recursive ll_diff_tree_paths(), this was supposed to match the number of changes that would be written into the changed-path Bloom filters. Unfortunately, this was not implemented correctly and instead misses simple cases like file modifications. This then does not stop very large changed-path filters from being written (unless they add or remove many files). To start, change the implementation in ll_diff_tree_paths() to instead use the global diff_queue_diff struct's 'nr' member as the count. This is a way to simplify the logic instead of making more mistakes in the complicated diff code. This has a drawback: the diff_queue_diff struct only lists the paths corresponding to blob changes, not their leading directories. Thus, get_or_compute_bloom_filter() needs an additional check to see if the hashmap with the leading directories becomes too large. One reason why this was not caught by test cases was that the test in t4216-log-bloom.sh that was supposed to check this "too many changes" condition only checked this on the initial commit of a repository. The old logic counted these values correctly. Update this test in a few ways: 1. Use GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS to reduce the limit, allowing smaller commits to engage with this logic. 2. Create several interesting cases of edits, adds, removes, and mode changes (in the second commit). By testing both sides of the inequality with the *_MAX_CHANGED_PATHS variable, we can see that the count is exactly correct, so none of these changes are missed or over-counted. 3. Use the trace2 data value filter_found_large to verify that these commits are on the correct side of the limit. Another way to verify the behavior is correct is through performance tests. By testing on my local copies of the Git repository and the Linux kernel repository, I could measure the effect of these short-circuits when computing a fresh commit-graph file with changed-path Bloom filters using the command GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS=N time \ git commit-graph write --reachable --changed-paths and reporting the wall time and resulting commit-graph size. For Git, the results are | | N=1 | N=10 | N=512 | |--------|----------------|----------------|----------------| | HEAD~1 | 10.90s 9.18MB | 11.11s 9.34MB | 11.31s 9.35MB | | HEAD | 9.21s 8.62MB | 11.11s 9.29MB | 11.29s 9.34MB | For Linux, the results are | | N=1 | N=20 | N=512 | |--------|----------------|---------------|---------------| | HEAD~1 | 61.28s 64.3MB | 76.9s 72.6MB | 77.6s 72.6MB | | HEAD | 49.44s 56.3MB | 68.7s 65.9MB | 69.2s 65.9MB | Naturally, the improvement becomes much less as the limit grows, as fewer commits satisfy the short-circuit. Reported-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-09-16 20:07:52 +02:00
if (hashmap_get_size(&pathmap) > settings->max_changed_paths) {
bloom: encode out-of-bounds filters as non-empty When a changed-path Bloom filter has either zero, or more than a certain number (commonly 512) of entries, the commit-graph machinery encodes it as "missing". More specifically, it sets the indices adjacent in the BIDX chunk as equal to each other to indicate a "length 0" filter; that is, that the filter occupies zero bytes on disk. This has heretofore been fine, since the commit-graph machinery has no need to care about these filters with too few or too many changed paths. Both cases act like no filter has been generated at all, and so there is no need to store them. In a subsequent commit, however, the commit-graph machinery will learn to only compute Bloom filters for some commits in the current commit-graph layer. This is a change from the current implementation which computes Bloom filters for all commits that are in the layer being written. Critically for this patch, only computing some of the Bloom filters means adding a third state for length 0 Bloom filters: zero entries, too many entries, or "hasn't been computed". It will be important for that future patch to distinguish between "not representable" (i.e., zero or too-many changed paths), and "hasn't been computed". In particular, we don't want to waste time recomputing filters that have already been computed. To that end, change how we store Bloom filters in the "computed but not representable" category: - Bloom filters with no entries are stored as a single byte with all bits low (i.e., all queries to that Bloom filter will return "definitely not") - Bloom filters with too many entries are stored as a single byte with all bits set high (i.e., all queries to that Bloom filter will return "maybe"). These rules are sufficient to not incur a behavior change by changing the on-disk representation of these two classes. Likewise, no specification changes are necessary for the commit-graph format, either: - Filters that were previously empty will be recomputed and stored according to the new rules, and - old clients reading filters generated by new clients will interpret the filters correctly and be none the wiser to how they were generated. Clients will invoke the Bloom machinery in more cases than before, but this can be addressed by returning a NULL filter when all bits are set high. This can be addressed in a future patch. Note that this does increase the size of on-disk commit-graphs, but far less than other proposals. In particular, this is generally more efficient than storing a bitmap for which commits haven't computed their Bloom filters. Storing a bitmap incurs a penalty of one bit per commit, whereas storing explicit filters as above incurs a penalty of one byte per too-large or empty commit. In practice, these boundary commits likely occupy a small proportion of the overall number of commits, and so the size penalty is likely smaller than storing a bitmap for all commits. See, for example, these relative proportions of such boundary commits (collected by SZEDER Gábor): | Percentage of | commit-graph | | | commits modifying | file size | | ├────────┬──────────────┼───────────────────┤ pct. | | 0 path | >= 512 paths | before | after | change | ┌────────────────┼────────┼──────────────┼─────────┼─────────┼───────────┤ | android-base | 13.20% | 0.13% | 37.468M | 37.534M | +0.1741 % | | cmssw | 0.15% | 0.23% | 17.118M | 17.119M | +0.0091 % | | cpython | 3.07% | 0.01% | 7.967M | 7.971M | +0.0423 % | | elasticsearch | 0.70% | 1.00% | 8.833M | 8.835M | +0.0128 % | | gcc | 0.00% | 0.08% | 16.073M | 16.074M | +0.0030 % | | gecko-dev | 0.14% | 0.64% | 59.868M | 59.874M | +0.0105 % | | git | 0.11% | 0.02% | 3.895M | 3.895M | +0.0020 % | | glibc | 0.02% | 0.10% | 3.555M | 3.555M | +0.0021 % | | go | 0.00% | 0.07% | 3.186M | 3.186M | +0.0018 % | | homebrew-cask | 0.40% | 0.02% | 7.035M | 7.035M | +0.0065 % | | homebrew-core | 0.01% | 0.01% | 11.611M | 11.611M | +0.0002 % | | jdk | 0.26% | 5.64% | 5.537M | 5.540M | +0.0590 % | | linux | 0.01% | 0.51% | 63.735M | 63.740M | +0.0073 % | | llvm-project | 0.12% | 0.03% | 25.515M | 25.516M | +0.0050 % | | rails | 0.10% | 0.10% | 6.252M | 6.252M | +0.0027 % | | rust | 0.07% | 0.17% | 9.364M | 9.364M | +0.0033 % | | tensorflow | 0.09% | 1.02% | 7.009M | 7.010M | +0.0158 % | | webkit | 0.05% | 0.31% | 17.405M | 17.406M | +0.0047 % | (where the above increase is determined by computing a non-split commit-graph before and after this patch). Given that these projects are all "large" by commit count, the storage cost by writing these filters explicitly is negligible. In the most extreme example, android-base (which has 494,848 commits at the time of writing) would have its commit-graph increase by a modest 68.4 KB. Finally, a test to exercise filters which contain too many changed path entries will be introduced in a subsequent patch. Suggested-by: SZEDER Gábor <szeder.dev@gmail.com> Suggested-by: Jakub Narębski <jnareb@gmail.com> Helped-by: Derrick Stolee <dstolee@microsoft.com> Helped-by: SZEDER Gábor <szeder.dev@gmail.com> Helped-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-09-18 04:59:44 +02:00
init_truncated_large_filter(filter);
bloom/diff: properly short-circuit on max_changes Commit e3696980 (diff: halt tree-diff early after max_changes, 2020-03-30) intended to create a mechanism to short-circuit a diff calculation after a certain number of paths were modified. By incrementing a "num_changes" counter throughout the recursive ll_diff_tree_paths(), this was supposed to match the number of changes that would be written into the changed-path Bloom filters. Unfortunately, this was not implemented correctly and instead misses simple cases like file modifications. This then does not stop very large changed-path filters from being written (unless they add or remove many files). To start, change the implementation in ll_diff_tree_paths() to instead use the global diff_queue_diff struct's 'nr' member as the count. This is a way to simplify the logic instead of making more mistakes in the complicated diff code. This has a drawback: the diff_queue_diff struct only lists the paths corresponding to blob changes, not their leading directories. Thus, get_or_compute_bloom_filter() needs an additional check to see if the hashmap with the leading directories becomes too large. One reason why this was not caught by test cases was that the test in t4216-log-bloom.sh that was supposed to check this "too many changes" condition only checked this on the initial commit of a repository. The old logic counted these values correctly. Update this test in a few ways: 1. Use GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS to reduce the limit, allowing smaller commits to engage with this logic. 2. Create several interesting cases of edits, adds, removes, and mode changes (in the second commit). By testing both sides of the inequality with the *_MAX_CHANGED_PATHS variable, we can see that the count is exactly correct, so none of these changes are missed or over-counted. 3. Use the trace2 data value filter_found_large to verify that these commits are on the correct side of the limit. Another way to verify the behavior is correct is through performance tests. By testing on my local copies of the Git repository and the Linux kernel repository, I could measure the effect of these short-circuits when computing a fresh commit-graph file with changed-path Bloom filters using the command GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS=N time \ git commit-graph write --reachable --changed-paths and reporting the wall time and resulting commit-graph size. For Git, the results are | | N=1 | N=10 | N=512 | |--------|----------------|----------------|----------------| | HEAD~1 | 10.90s 9.18MB | 11.11s 9.34MB | 11.31s 9.35MB | | HEAD | 9.21s 8.62MB | 11.11s 9.29MB | 11.29s 9.34MB | For Linux, the results are | | N=1 | N=20 | N=512 | |--------|----------------|---------------|---------------| | HEAD~1 | 61.28s 64.3MB | 76.9s 72.6MB | 77.6s 72.6MB | | HEAD | 49.44s 56.3MB | 68.7s 65.9MB | 69.2s 65.9MB | Naturally, the improvement becomes much less as the limit grows, as fewer commits satisfy the short-circuit. Reported-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-09-16 20:07:52 +02:00
if (computed)
*computed |= BLOOM_TRUNC_LARGE;
goto cleanup;
}
filter->len = (hashmap_get_size(&pathmap) * settings->bits_per_entry + BITS_PER_WORD - 1) / BITS_PER_WORD;
bloom: encode out-of-bounds filters as non-empty When a changed-path Bloom filter has either zero, or more than a certain number (commonly 512) of entries, the commit-graph machinery encodes it as "missing". More specifically, it sets the indices adjacent in the BIDX chunk as equal to each other to indicate a "length 0" filter; that is, that the filter occupies zero bytes on disk. This has heretofore been fine, since the commit-graph machinery has no need to care about these filters with too few or too many changed paths. Both cases act like no filter has been generated at all, and so there is no need to store them. In a subsequent commit, however, the commit-graph machinery will learn to only compute Bloom filters for some commits in the current commit-graph layer. This is a change from the current implementation which computes Bloom filters for all commits that are in the layer being written. Critically for this patch, only computing some of the Bloom filters means adding a third state for length 0 Bloom filters: zero entries, too many entries, or "hasn't been computed". It will be important for that future patch to distinguish between "not representable" (i.e., zero or too-many changed paths), and "hasn't been computed". In particular, we don't want to waste time recomputing filters that have already been computed. To that end, change how we store Bloom filters in the "computed but not representable" category: - Bloom filters with no entries are stored as a single byte with all bits low (i.e., all queries to that Bloom filter will return "definitely not") - Bloom filters with too many entries are stored as a single byte with all bits set high (i.e., all queries to that Bloom filter will return "maybe"). These rules are sufficient to not incur a behavior change by changing the on-disk representation of these two classes. Likewise, no specification changes are necessary for the commit-graph format, either: - Filters that were previously empty will be recomputed and stored according to the new rules, and - old clients reading filters generated by new clients will interpret the filters correctly and be none the wiser to how they were generated. Clients will invoke the Bloom machinery in more cases than before, but this can be addressed by returning a NULL filter when all bits are set high. This can be addressed in a future patch. Note that this does increase the size of on-disk commit-graphs, but far less than other proposals. In particular, this is generally more efficient than storing a bitmap for which commits haven't computed their Bloom filters. Storing a bitmap incurs a penalty of one bit per commit, whereas storing explicit filters as above incurs a penalty of one byte per too-large or empty commit. In practice, these boundary commits likely occupy a small proportion of the overall number of commits, and so the size penalty is likely smaller than storing a bitmap for all commits. See, for example, these relative proportions of such boundary commits (collected by SZEDER Gábor): | Percentage of | commit-graph | | | commits modifying | file size | | ├────────┬──────────────┼───────────────────┤ pct. | | 0 path | >= 512 paths | before | after | change | ┌────────────────┼────────┼──────────────┼─────────┼─────────┼───────────┤ | android-base | 13.20% | 0.13% | 37.468M | 37.534M | +0.1741 % | | cmssw | 0.15% | 0.23% | 17.118M | 17.119M | +0.0091 % | | cpython | 3.07% | 0.01% | 7.967M | 7.971M | +0.0423 % | | elasticsearch | 0.70% | 1.00% | 8.833M | 8.835M | +0.0128 % | | gcc | 0.00% | 0.08% | 16.073M | 16.074M | +0.0030 % | | gecko-dev | 0.14% | 0.64% | 59.868M | 59.874M | +0.0105 % | | git | 0.11% | 0.02% | 3.895M | 3.895M | +0.0020 % | | glibc | 0.02% | 0.10% | 3.555M | 3.555M | +0.0021 % | | go | 0.00% | 0.07% | 3.186M | 3.186M | +0.0018 % | | homebrew-cask | 0.40% | 0.02% | 7.035M | 7.035M | +0.0065 % | | homebrew-core | 0.01% | 0.01% | 11.611M | 11.611M | +0.0002 % | | jdk | 0.26% | 5.64% | 5.537M | 5.540M | +0.0590 % | | linux | 0.01% | 0.51% | 63.735M | 63.740M | +0.0073 % | | llvm-project | 0.12% | 0.03% | 25.515M | 25.516M | +0.0050 % | | rails | 0.10% | 0.10% | 6.252M | 6.252M | +0.0027 % | | rust | 0.07% | 0.17% | 9.364M | 9.364M | +0.0033 % | | tensorflow | 0.09% | 1.02% | 7.009M | 7.010M | +0.0158 % | | webkit | 0.05% | 0.31% | 17.405M | 17.406M | +0.0047 % | (where the above increase is determined by computing a non-split commit-graph before and after this patch). Given that these projects are all "large" by commit count, the storage cost by writing these filters explicitly is negligible. In the most extreme example, android-base (which has 494,848 commits at the time of writing) would have its commit-graph increase by a modest 68.4 KB. Finally, a test to exercise filters which contain too many changed path entries will be introduced in a subsequent patch. Suggested-by: SZEDER Gábor <szeder.dev@gmail.com> Suggested-by: Jakub Narębski <jnareb@gmail.com> Helped-by: Derrick Stolee <dstolee@microsoft.com> Helped-by: SZEDER Gábor <szeder.dev@gmail.com> Helped-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-09-18 04:59:44 +02:00
if (!filter->len) {
if (computed)
*computed |= BLOOM_TRUNC_EMPTY;
filter->len = 1;
}
CALLOC_ARRAY(filter->data, filter->len);
hashmap_for_each_entry(&pathmap, &iter, e, entry) {
struct bloom_key key;
fill_bloom_key(e->path, strlen(e->path), &key, settings);
add_key_to_filter(&key, filter, settings);
clear_bloom_key(&key);
}
bloom/diff: properly short-circuit on max_changes Commit e3696980 (diff: halt tree-diff early after max_changes, 2020-03-30) intended to create a mechanism to short-circuit a diff calculation after a certain number of paths were modified. By incrementing a "num_changes" counter throughout the recursive ll_diff_tree_paths(), this was supposed to match the number of changes that would be written into the changed-path Bloom filters. Unfortunately, this was not implemented correctly and instead misses simple cases like file modifications. This then does not stop very large changed-path filters from being written (unless they add or remove many files). To start, change the implementation in ll_diff_tree_paths() to instead use the global diff_queue_diff struct's 'nr' member as the count. This is a way to simplify the logic instead of making more mistakes in the complicated diff code. This has a drawback: the diff_queue_diff struct only lists the paths corresponding to blob changes, not their leading directories. Thus, get_or_compute_bloom_filter() needs an additional check to see if the hashmap with the leading directories becomes too large. One reason why this was not caught by test cases was that the test in t4216-log-bloom.sh that was supposed to check this "too many changes" condition only checked this on the initial commit of a repository. The old logic counted these values correctly. Update this test in a few ways: 1. Use GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS to reduce the limit, allowing smaller commits to engage with this logic. 2. Create several interesting cases of edits, adds, removes, and mode changes (in the second commit). By testing both sides of the inequality with the *_MAX_CHANGED_PATHS variable, we can see that the count is exactly correct, so none of these changes are missed or over-counted. 3. Use the trace2 data value filter_found_large to verify that these commits are on the correct side of the limit. Another way to verify the behavior is correct is through performance tests. By testing on my local copies of the Git repository and the Linux kernel repository, I could measure the effect of these short-circuits when computing a fresh commit-graph file with changed-path Bloom filters using the command GIT_TEST_BLOOM_SETTINGS_MAX_CHANGED_PATHS=N time \ git commit-graph write --reachable --changed-paths and reporting the wall time and resulting commit-graph size. For Git, the results are | | N=1 | N=10 | N=512 | |--------|----------------|----------------|----------------| | HEAD~1 | 10.90s 9.18MB | 11.11s 9.34MB | 11.31s 9.35MB | | HEAD | 9.21s 8.62MB | 11.11s 9.29MB | 11.29s 9.34MB | For Linux, the results are | | N=1 | N=20 | N=512 | |--------|----------------|---------------|---------------| | HEAD~1 | 61.28s 64.3MB | 76.9s 72.6MB | 77.6s 72.6MB | | HEAD | 49.44s 56.3MB | 68.7s 65.9MB | 69.2s 65.9MB | Naturally, the improvement becomes much less as the limit grows, as fewer commits satisfy the short-circuit. Reported-by: SZEDER Gábor <szeder.dev@gmail.com> Signed-off-by: Derrick Stolee <dstolee@microsoft.com> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-09-16 20:07:52 +02:00
cleanup:
hashmap_clear_and_free(&pathmap, struct pathmap_hash_entry, entry);
} else {
for (i = 0; i < diff_queued_diff.nr; i++)
diff_free_filepair(diff_queued_diff.queue[i]);
bloom: encode out-of-bounds filters as non-empty When a changed-path Bloom filter has either zero, or more than a certain number (commonly 512) of entries, the commit-graph machinery encodes it as "missing". More specifically, it sets the indices adjacent in the BIDX chunk as equal to each other to indicate a "length 0" filter; that is, that the filter occupies zero bytes on disk. This has heretofore been fine, since the commit-graph machinery has no need to care about these filters with too few or too many changed paths. Both cases act like no filter has been generated at all, and so there is no need to store them. In a subsequent commit, however, the commit-graph machinery will learn to only compute Bloom filters for some commits in the current commit-graph layer. This is a change from the current implementation which computes Bloom filters for all commits that are in the layer being written. Critically for this patch, only computing some of the Bloom filters means adding a third state for length 0 Bloom filters: zero entries, too many entries, or "hasn't been computed". It will be important for that future patch to distinguish between "not representable" (i.e., zero or too-many changed paths), and "hasn't been computed". In particular, we don't want to waste time recomputing filters that have already been computed. To that end, change how we store Bloom filters in the "computed but not representable" category: - Bloom filters with no entries are stored as a single byte with all bits low (i.e., all queries to that Bloom filter will return "definitely not") - Bloom filters with too many entries are stored as a single byte with all bits set high (i.e., all queries to that Bloom filter will return "maybe"). These rules are sufficient to not incur a behavior change by changing the on-disk representation of these two classes. Likewise, no specification changes are necessary for the commit-graph format, either: - Filters that were previously empty will be recomputed and stored according to the new rules, and - old clients reading filters generated by new clients will interpret the filters correctly and be none the wiser to how they were generated. Clients will invoke the Bloom machinery in more cases than before, but this can be addressed by returning a NULL filter when all bits are set high. This can be addressed in a future patch. Note that this does increase the size of on-disk commit-graphs, but far less than other proposals. In particular, this is generally more efficient than storing a bitmap for which commits haven't computed their Bloom filters. Storing a bitmap incurs a penalty of one bit per commit, whereas storing explicit filters as above incurs a penalty of one byte per too-large or empty commit. In practice, these boundary commits likely occupy a small proportion of the overall number of commits, and so the size penalty is likely smaller than storing a bitmap for all commits. See, for example, these relative proportions of such boundary commits (collected by SZEDER Gábor): | Percentage of | commit-graph | | | commits modifying | file size | | ├────────┬──────────────┼───────────────────┤ pct. | | 0 path | >= 512 paths | before | after | change | ┌────────────────┼────────┼──────────────┼─────────┼─────────┼───────────┤ | android-base | 13.20% | 0.13% | 37.468M | 37.534M | +0.1741 % | | cmssw | 0.15% | 0.23% | 17.118M | 17.119M | +0.0091 % | | cpython | 3.07% | 0.01% | 7.967M | 7.971M | +0.0423 % | | elasticsearch | 0.70% | 1.00% | 8.833M | 8.835M | +0.0128 % | | gcc | 0.00% | 0.08% | 16.073M | 16.074M | +0.0030 % | | gecko-dev | 0.14% | 0.64% | 59.868M | 59.874M | +0.0105 % | | git | 0.11% | 0.02% | 3.895M | 3.895M | +0.0020 % | | glibc | 0.02% | 0.10% | 3.555M | 3.555M | +0.0021 % | | go | 0.00% | 0.07% | 3.186M | 3.186M | +0.0018 % | | homebrew-cask | 0.40% | 0.02% | 7.035M | 7.035M | +0.0065 % | | homebrew-core | 0.01% | 0.01% | 11.611M | 11.611M | +0.0002 % | | jdk | 0.26% | 5.64% | 5.537M | 5.540M | +0.0590 % | | linux | 0.01% | 0.51% | 63.735M | 63.740M | +0.0073 % | | llvm-project | 0.12% | 0.03% | 25.515M | 25.516M | +0.0050 % | | rails | 0.10% | 0.10% | 6.252M | 6.252M | +0.0027 % | | rust | 0.07% | 0.17% | 9.364M | 9.364M | +0.0033 % | | tensorflow | 0.09% | 1.02% | 7.009M | 7.010M | +0.0158 % | | webkit | 0.05% | 0.31% | 17.405M | 17.406M | +0.0047 % | (where the above increase is determined by computing a non-split commit-graph before and after this patch). Given that these projects are all "large" by commit count, the storage cost by writing these filters explicitly is negligible. In the most extreme example, android-base (which has 494,848 commits at the time of writing) would have its commit-graph increase by a modest 68.4 KB. Finally, a test to exercise filters which contain too many changed path entries will be introduced in a subsequent patch. Suggested-by: SZEDER Gábor <szeder.dev@gmail.com> Suggested-by: Jakub Narębski <jnareb@gmail.com> Helped-by: Derrick Stolee <dstolee@microsoft.com> Helped-by: SZEDER Gábor <szeder.dev@gmail.com> Helped-by: Junio C Hamano <gitster@pobox.com> Signed-off-by: Taylor Blau <me@ttaylorr.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-09-18 04:59:44 +02:00
init_truncated_large_filter(filter);
if (computed)
*computed |= BLOOM_TRUNC_LARGE;
}
if (computed)
*computed |= BLOOM_COMPUTED;
free(diff_queued_diff.queue);
DIFF_QUEUE_CLEAR(&diff_queued_diff);
return filter;
}
revision.c: use Bloom filters to speed up path based revision walks Revision walk will now use Bloom filters for commits to speed up revision walks for a particular path (for computing history for that path), if they are present in the commit-graph file. We load the Bloom filters during the prepare_revision_walk step, currently only when dealing with a single pathspec. Extending it to work with multiple pathspecs can be explored and built on top of this series in the future. While comparing trees in rev_compare_trees(), if the Bloom filter says that the file is not different between the two trees, we don't need to compute the expensive diff. This is where we get our performance gains. The other response of the Bloom filter is '`:maybe', in which case we fall back to the full diff calculation to determine if the path was changed in the commit. We do not try to use Bloom filters when the '--walk-reflogs' option is specified. The '--walk-reflogs' option does not walk the commit ancestry chain like the rest of the options. Incorporating the performance gains when walking reflog entries would add more complexity, and can be explored in a later series. Performance Gains: We tested the performance of `git log -- <path>` on the git repo, the linux and some internal large repos, with a variety of paths of varying depths. On the git and linux repos: - we observed a 2x to 5x speed up. On a large internal repo with files seated 6-10 levels deep in the tree: - we observed 10x to 20x speed ups, with some paths going up to 28 times faster. Helped-by: Derrick Stolee <dstolee@microsoft.com Helped-by: SZEDER Gábor <szeder.dev@gmail.com> Helped-by: Jonathan Tan <jonathantanmy@google.com> Signed-off-by: Garima Singh <garima.singh@microsoft.com> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2020-04-06 18:59:52 +02:00
int bloom_filter_contains(const struct bloom_filter *filter,
const struct bloom_key *key,
const struct bloom_filter_settings *settings)
{
int i;
uint64_t mod = filter->len * BITS_PER_WORD;
if (!mod)
return -1;
for (i = 0; i < settings->num_hashes; i++) {
uint64_t hash_mod = key->hashes[i] % mod;
uint64_t block_pos = hash_mod / BITS_PER_WORD;
if (!(filter->data[block_pos] & get_bitmask(hash_mod)))
return 0;
}
return 1;
}