1
0
mirror of https://github.com/git/git.git synced 2024-10-20 01:48:55 +02:00
git/builtin
Taylor Blau 2f4ba2a867 packfile: prepare for the existence of '*.rev' files
Specify the format of the on-disk reverse index 'pack-*.rev' file, as
well as prepare the code for the existence of such files.

The reverse index maps from pack relative positions (i.e., an index into
the array of object which is sorted by their offsets within the
packfile) to their position within the 'pack-*.idx' file. Today, this is
done by building up a list of (off_t, uint32_t) tuples for each object
(the off_t corresponding to that object's offset, and the uint32_t
corresponding to its position in the index). To convert between pack and
index position quickly, this array of tuples is radix sorted based on
its offset.

This has two major drawbacks:

First, the in-memory cost scales linearly with the number of objects in
a pack.  Each 'struct revindex_entry' is sizeof(off_t) +
sizeof(uint32_t) + padding bytes for a total of 16.

To observe this, force Git to load the reverse index by, for e.g.,
running 'git cat-file --batch-check="%(objectsize:disk)"'. When asking
for a single object in a fresh clone of the kernel, Git needs to
allocate 120+ MB of memory in order to hold the reverse index in memory.

Second, the cost to sort also scales with the size of the pack.
Luckily, this is a linear function since 'load_pack_revindex()' uses a
radix sort, but this cost still must be paid once per pack per process.

As an example, it takes ~60x longer to print the _size_ of an object as
it does to print that entire object's _contents_:

  Benchmark #1: git.compile cat-file --batch <obj
    Time (mean ± σ):       3.4 ms ±   0.1 ms    [User: 3.3 ms, System: 2.1 ms]
    Range (min … max):     3.2 ms …   3.7 ms    726 runs

  Benchmark #2: git.compile cat-file --batch-check="%(objectsize:disk)" <obj
    Time (mean ± σ):     210.3 ms ±   8.9 ms    [User: 188.2 ms, System: 23.2 ms]
    Range (min … max):   193.7 ms … 224.4 ms    13 runs

Instead, avoid computing and sorting the revindex once per process by
writing it to a file when the pack itself is generated.

The format is relatively straightforward. It contains an array of
uint32_t's, the length of which is equal to the number of objects in the
pack.  The ith entry in this table contains the index position of the
ith object in the pack, where "ith object in the pack" is determined by
pack offset.

One thing that the on-disk format does _not_ contain is the full (up to)
eight-byte offset corresponding to each object. This is something that
the in-memory revindex contains (it stores an off_t in 'struct
revindex_entry' along with the same uint32_t that the on-disk format
has). Omit it in the on-disk format, since knowing the index position
for some object is sufficient to get a constant-time lookup in the
pack-*.idx file to ask for an object's offset within the pack.

This trades off between the on-disk size of the 'pack-*.rev' file for
runtime to chase down the offset for some object. Even though the lookup
is constant time, the constant is heavier, since it can potentially
involve two pointer walks in v2 indexes (one to access the 4-byte offset
table, and potentially a second to access the double wide offset table).

Consider trying to map an object's pack offset to a relative position
within that pack. In a cold-cache scenario, more page faults occur while
switching between binary searching through the reverse index and
searching through the *.idx file for an object's offset. Sure enough,
with a cold cache (writing '3' into '/proc/sys/vm/drop_caches' after
'sync'ing), printing out the entire object's contents is still
marginally faster than printing its size:

  Benchmark #1: git.compile cat-file --batch-check="%(objectsize:disk)" <obj >/dev/null
    Time (mean ± σ):      22.6 ms ±   0.5 ms    [User: 2.4 ms, System: 7.9 ms]
    Range (min … max):    21.4 ms …  23.5 ms    41 runs

  Benchmark #2: git.compile cat-file --batch <obj >/dev/null
    Time (mean ± σ):      17.2 ms ±   0.7 ms    [User: 2.8 ms, System: 5.5 ms]
    Range (min … max):    15.6 ms …  18.2 ms    45 runs

(Numbers taken in the kernel after cheating and using the next patch to
generate a reverse index). There are a couple of approaches to improve
cold cache performance not pursued here:

  - We could include the object offsets in the reverse index format.
    Predictably, this does result in fewer page faults, but it triples
    the size of the file, while simultaneously duplicating a ton of data
    already available in the .idx file. (This was the original way I
    implemented the format, and it did show
    `--batch-check='%(objectsize:disk)'` winning out against `--batch`.)

    On the other hand, this increase in size also results in a large
    block-cache footprint, which could potentially hurt other workloads.

  - We could store the mapping from pack to index position in more
    cache-friendly way, like constructing a binary search tree from the
    table and writing the values in breadth-first order. This would
    result in much better locality, but the price you pay is trading
    O(1) lookup in 'pack_pos_to_index()' for an O(log n) one (since you
    can no longer directly index the table).

So, neither of these approaches are taken here. (Thankfully, the format
is versioned, so we are free to pursue these in the future.) But, cold
cache performance likely isn't interesting outside of one-off cases like
asking for the size of an object directly. In real-world usage, Git is
often performing many operations in the revindex (i.e., asking about
many objects rather than a single one).

The trade-off is worth it, since we will avoid the vast majority of the
cost of generating the revindex that the extra pointer chase will look
like noise in the following patch's benchmarks.

This patch describes the format and prepares callers (like in
pack-revindex.c) to be able to read *.rev files once they exist. An
implementation of the writer will appear in the next patch, and callers
will gradually begin to start using the writer in the patches that
follow after that.

Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
2021-01-25 18:32:43 -08:00
..
add.c drop unused argc parameters 2020-09-30 12:53:47 -07:00
am.c builtin/*: update usage format 2021-01-06 15:10:49 -08:00
annotate.c
apply.c
archive.c
bisect--helper.c Merge branch 'mr/bisect-in-c-3' 2020-11-09 14:06:25 -08:00
blame.c Merge branch 'ab/mailmap' 2021-01-25 14:19:19 -08:00
branch.c branch: sort detached HEAD based on a flag 2021-01-07 15:13:21 -08:00
bugreport.c builtin/bugreport.c: use thread-safe localtime_r() 2020-12-01 13:05:37 -08:00
bundle.c Merge branch 'bc/sha-256-part-3' 2020-08-11 18:04:11 -07:00
cat-file.c
check-attr.c
check-ignore.c dir: fix problematic API to avoid memory leaks 2020-08-18 17:17:31 -07:00
check-mailmap.c shortlog: remove unused(?) "repo-abbrev" feature 2021-01-12 14:04:42 -08:00
check-ref-format.c
checkout-index.c checkout-index: propagate errors to exit code 2020-10-27 12:41:56 -07:00
checkout.c Merge branch 'dl/checkout-p-merge-base' 2020-12-23 13:59:46 -08:00
clean.c quote_path: give flags parameter to quote_path() 2020-09-10 10:49:19 -07:00
clone.c Merge branch 'js/init-defaultbranch-advice' 2020-12-18 15:15:17 -08:00
column.c
commit-graph.c builtin/*: update usage format 2021-01-06 15:10:49 -08:00
commit-tree.c
commit.c shortlog: remove unused(?) "repo-abbrev" feature 2021-01-12 14:04:42 -08:00
config.c config: implement --fixed-value with --get* 2020-11-25 14:43:48 -08:00
count-objects.c
credential-cache--daemon.c make credential helpers builtins 2020-08-13 11:02:08 -07:00
credential-cache.c Merge branch 'jc/undash-in-tree-git-callers' 2020-09-03 12:37:03 -07:00
credential-store.c crendential-store: use timeout when locking file 2020-11-25 12:30:18 -08:00
credential.c credential: load default config 2020-10-16 12:30:45 -07:00
describe.c
diff-files.c
diff-index.c builtin/diff-index: learn --merge-base 2020-09-20 21:30:26 -07:00
diff-tree.c builtin/diff-tree: learn --merge-base 2020-09-21 13:37:03 -07:00
diff.c Merge branch 'dl/diff-merge-base' 2020-11-02 13:17:39 -08:00
difftool.c Use new HASHMAP_INIT macro to simplify hashmap initialization 2020-11-11 12:55:27 -08:00
env--helper.c assert PARSE_OPT_NONEG in parse-options callbacks 2020-09-30 12:53:47 -07:00
fast-export.c builtin/*: update usage format 2021-01-06 15:10:49 -08:00
fast-import.c Merge branch 'jk/fast-import-marks-cleanup' 2020-11-02 13:17:40 -08:00
fetch-pack.c fetch-pack: remove no_dependents code 2020-08-18 16:46:53 -07:00
fetch.c fetch: implement support for atomic reference updates 2021-01-12 12:06:15 -08:00
fmt-merge-msg.c
for-each-ref.c ref-filter: move ref_sorting flags to a bitfield 2021-01-07 15:13:21 -08:00
for-each-repo.c for-each-repo: do nothing on empty config 2021-01-07 19:12:02 -08:00
fsck.c fsck: make fsck_config() re-usable 2021-01-05 14:58:29 -08:00
gc.c Merge branch 'tb/pack-revindex-api' 2021-01-25 14:19:20 -08:00
get-tar-commit-id.c
grep.c grep: use designated initializers for grep_defaults 2020-11-21 14:50:33 -08:00
hash-object.c
help.c help: drop usage of 'common' and 'useful' for guides 2020-08-04 18:34:01 -07:00
index-pack.c object-file.c: rename from sha1-file.c 2021-01-04 13:01:55 -08:00
init-db.c get_default_branch_name(): prepare for showing some advice 2020-12-13 15:53:50 -08:00
interpret-trailers.c
log.c Merge branch 'ab/mailmap' 2021-01-25 14:19:19 -08:00
ls-files.c dir: fix problematic API to avoid memory leaks 2020-08-18 17:17:31 -07:00
ls-remote.c Merge branch 'jk/unleak-fixes' 2020-08-24 14:54:30 -07:00
ls-tree.c
mailinfo.c
mailsplit.c
merge-base.c
merge-file.c
merge-index.c
merge-ours.c
merge-recursive.c
merge-tree.c merge-base, xdiff: zero out xpparam_t structures 2020-10-20 12:53:26 -07:00
merge.c Merge branch 'en/merge-ort-api-null-impl' 2020-11-18 13:32:53 -08:00
mktag.c mktag: add a --[no-]strict option 2021-01-06 14:22:24 -08:00
mktree.c
multi-pack-index.c
mv.c
name-rev.c hash-lookup: rename from sha1-lookup 2021-01-04 13:01:55 -08:00
notes.c
pack-objects.c check_object(): convert to new revindex API 2021-01-13 21:53:45 -08:00
pack-redundant.c Merge branch 'jc/deprecate-pack-redundant' 2021-01-25 14:19:18 -08:00
pack-refs.c
patch-id.c
prune-packed.c
prune.c
pull.c pull: display default warning only when non-ff 2020-12-15 17:39:42 -08:00
push.c Merge branch 'sk/force-if-includes' 2020-10-27 15:09:49 -07:00
range-diff.c
read-tree.c
rebase.c Merge branch 'rs/rebase-commit-validation' 2021-01-15 15:20:29 -08:00
receive-pack.c Merge branch 'js/trace2-session-id' 2020-12-08 15:11:20 -08:00
reflog.c
remote-ext.c
remote-fd.c
remote.c config: convert multi_replace to flags 2020-11-25 14:43:47 -08:00
repack.c packfile: prepare for the existence of '*.rev' files 2021-01-25 18:32:43 -08:00
replace.c
rerere.c
reset.c wt-status: tolerate dangling marks 2020-09-02 14:39:25 -07:00
rev-list.c bisect: combine args passed to find_bisection() 2020-08-07 15:13:03 -07:00
rev-parse.c rev-parse: add option for absolute or relative path formatting 2020-12-12 23:35:51 -08:00
revert.c Merge branch 'en/merge-ort-api-null-impl' 2020-11-18 13:32:53 -08:00
rm.c
send-pack.c push: parse and set flag for "--force-if-includes" 2020-10-03 09:59:19 -07:00
shortlog.c Merge branch 'ab/mailmap' 2021-01-25 14:19:19 -08:00
show-branch.c Merge branch 'jt/interpret-branch-name-fallback' 2020-09-09 13:53:09 -07:00
show-index.c
show-ref.c
sparse-checkout.c sparse-checkout: fill in some options boilerplate 2020-09-30 12:53:48 -07:00
stash.c Merge branch 'en/stash-apply-sparse-checkout' 2021-01-15 15:20:29 -08:00
stripspace.c
submodule--helper.c builtin/*: update usage format 2021-01-06 15:10:49 -08:00
symbolic-ref.c
tag.c ref-filter: move ref_sorting flags to a bitfield 2021-01-07 15:13:21 -08:00
unpack-file.c
unpack-objects.c
update-index.c
update-ref.c update-ref: disallow "start" for ongoing transactions 2020-11-16 13:44:01 -08:00
update-server-info.c
upload-archive.c
upload-pack.c
var.c
verify-commit.c
verify-pack.c Merge branch 'bc/sha-256-part-3' 2020-08-11 18:04:11 -07:00
verify-tag.c
worktree.c worktree: teach repair to fix multi-directional breakage 2020-12-21 13:44:28 -08:00
write-tree.c