Perf tests were run with end-user's shell, but it has been
corrected to use the shell specified by $TEST_SHELL_PATH.
* ja/perf-use-specified-shell:
t/perf: do not run tests in user's $SHELL
Correctness and performance update to "diff --color-moved" feature.
* pw/diff-color-moved-fix:
diff --color-moved: intern strings
diff: use designated initializers for emitted_diff_symbol
diff --color-moved-ws=allow-indentation-change: improve hash lookups
diff --color-moved: stop clearing potential moved blocks
diff --color-moved: shrink potential moved blocks as we go
diff --color-moved: unify moved block growth functions
diff --color-moved: call comparison function directly
diff --color-moved-ws=allow-indentation-change: simplify and optimize
diff: simplify allow-indentation-change delta calculation
diff --color-moved: avoid false short line matches and bad zebra coloring
diff --color-moved=zebra: fix alternate coloring
diff --color-moved: rewind when discarding pmb
diff --color-moved: factor out function
diff --color-moved: clear all flags on blocks that are too short
diff --color-moved: add perf tests
Broken &&-chains in the test scripts have been corrected.
* es/test-chain-lint:
t6000-t9999: detect and signal failure within loop
t5000-t5999: detect and signal failure within loop
t4000-t4999: detect and signal failure within loop
t0000-t3999: detect and signal failure within loop
tests: simplify by dropping unnecessary `for` loops
tests: apply modern idiom for exiting loop upon failure
tests: apply modern idiom for signaling test failure
tests: fix broken &&-chains in `{...}` groups
tests: fix broken &&-chains in `$(...)` command substitutions
tests: fix broken &&-chains in compound statements
tests: use test_write_lines() to generate line-oriented output
tests: simplify construction of large blocks of text
t9107: use shell parameter expansion to avoid breaking &&-chain
t6300: make `%(raw:size) --shell` test more robust
t5516: drop unnecessary subshell and command invocation
t4202: clarify intent by creating expected content less cleverly
t1020: avoid aborting entire test script when one test fails
t1010: fix unnoticed failure on Windows
t/lib-pager: use sane_unset() to avoid breaking &&-chain
The environment variable $SHELL is usually set to the user's
interactive shell. Our build and test scripts never use $SHELL because
there are no guarantees about its input language. Instead, we use
/bin/sh which should be a POSIX shell.
For systems with a broken /bin/sh, we allow to override that path via
SHELL_PATH. To run tests in yet another shell we allow to override
SHELL_PATH with TEST_SHELL_PATH.
Perf tests run in $SHELL via a wrapper defined in t/perf/perf-lib.sh,
so they break with e.g. SHELL=python. Use TEST_SHELL_PATH like
in other tests. TEST_SHELL_PATH is always defined because
t/perf/perf-lib.sh includes t/test-lib.sh, which includes
GIT-BUILD-OPTIONS.
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Johannes Altmanninger <aclopte@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach diff and blame to work well with sparse index.
* ld/sparse-diff-blame:
blame: enable and test the sparse index
diff: enable and test the sparse index
diff: replace --staged with --cached in t1092 tests
repo-settings: prepare_repo_settings only in git repos
test-read-cache: set up repo after git directory
commit-graph: return if there is no git directory
git: ensure correct git directory setup with -h
Failures within `for` and `while` loops can go unnoticed if not detected
and signaled manually since the loop itself does not abort when a
contained command fails, nor will a failure necessarily be detected when
the loop finishes since the loop returns the exit code of the last
command it ran on the final iteration, which may not be the command
which failed. Therefore, detect and signal failures manually within
loops using the idiom `|| return 1` (or `|| exit 1` within subshells).
Signed-off-by: Eric Sunshine <sunshine@sunshineco.com>
Reviewed-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The top-level &&-chain checker built into t/test-lib.sh causes tests to
magically exit with code 117 if the &&-chain is broken. However, it has
the shortcoming that the magic does not work within `{...}` groups,
`(...)` subshells, `$(...)` substitutions, or within bodies of compound
statements, such as `if`, `for`, `while`, `case`, etc. `chainlint.sed`
partly fills in the gap by catching broken &&-chains in `(...)`
subshells, but bugs can still lurk behind broken &&-chains in the other
cases.
Fix broken &&-chains in compound statements in order to reduce the
number of possible lurking bugs.
Signed-off-by: Eric Sunshine <sunshine@sunshineco.com>
Reviewed-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Various operating modes of "git reset" have been made to work
better with the sparse index.
* vd/sparse-reset:
unpack-trees: improve performance of next_cache_entry
reset: make --mixed sparse-aware
reset: make sparse-aware (except --mixed)
reset: integrate with sparse index
reset: expand test coverage for sparse checkouts
sparse-index: update command for expand/collapse test
reset: preserve skip-worktree bit in mixed reset
reset: rename is_missing to !is_in_reset_tree
Add some tests so we can monitor changes to the performance of the
move detection code. The tests record the performance --color-moved
and --color-moved-ws=allow-indentation-change for a large diff and a
sequence of smaller diffs. The range of commits used for the large
diff can be customized by exporting TEST_REV_A and TEST_REV_B when
running the test.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Enable the sparse index for the 'git blame' command. The index was already
not expanded with this command, so the most interesting thing to do is to
add tests that verify that 'git blame' behaves correctly when the sparse
index is enabled and that its performance improves. More specifically, these
cases are:
1. The index is not expanded for 'blame' when given paths in the sparse
checkout cone at multiple levels.
2. Performance measurably improves for 'blame' with sparse index when given
paths in the sparse checkout cone at multiple levels.
The `p2000` tests demonstrate a ~60% execution time reduction when running
'blame' for a file two levels deep and and a ~30% execution time reduction
for a file three levels deep.
Test before after
----------------------------------------------------------------
2000.62: git blame f2/f4/a (full-v3) 0.31 0.32 +3.2%
2000.63: git blame f2/f4/a (full-v4) 0.29 0.31 +6.9%
2000.64: git blame f2/f4/a (sparse-v3) 0.55 0.23 -58.2%
2000.65: git blame f2/f4/a (sparse-v4) 0.57 0.23 -59.6%
2000.66: git blame f2/f4/f3/a (full-v3) 0.77 0.85 +10.4%
2000.67: git blame f2/f4/f3/a (full-v4) 0.78 0.81 +3.8%
2000.68: git blame f2/f4/f3/a (sparse-v3) 1.07 0.72 -32.7%
2000.99: git blame f2/f4/f3/a (sparse-v4) 1.05 0.73 -30.5%
We do not include paths outside the sparse checkout cone because blame
does not support blaming files that are not present in the working
directory. This is true in both sparse and full checkouts.
Signed-off-by: Lessley Dennington <lessleydennington@gmail.com>
Reviewed-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Enable the sparse index within the 'git diff' command. Its implementation
already safely integrates with the sparse index because it shares code
with the 'git status' and 'git checkout' commands that were already
integrated. For more details see:
d76723ee53 (status: use sparse-index throughout, 2021-07-14)
1ba5f45132 (checkout: stop expanding sparse indexes, 2021-06-29)
The most interesting thing to do is to add tests that verify that 'git
diff' behaves correctly when the sparse index is enabled. These cases are:
1. The index is not expanded for 'diff' and 'diff --staged'
2. 'diff' and 'diff --staged' behave the same in full checkout, sparse
checkout, and sparse index repositories in the following partially-staged
scenarios (i.e. the index, HEAD, and working directory differ at a given
path):
1. Path is within sparse-checkout cone
2. Path is outside sparse-checkout cone
3. A merge conflict exists for paths outside sparse-checkout cone
The `p2000` tests demonstrate a ~44% execution time reduction for 'git
diff' and a ~86% execution time reduction for 'git diff --staged' using a
sparse index:
Test before after
-------------------------------------------------------------
2000.30: git diff (full-v3) 0.33 0.34 +3.0%
2000.31: git diff (full-v4) 0.33 0.35 +6.1%
2000.32: git diff (sparse-v3) 0.53 0.31 -41.5%
2000.33: git diff (sparse-v4) 0.54 0.29 -46.3%
2000.34: git diff --cached (full-v3) 0.07 0.07 +0.0%
2000.35: git diff --cached (full-v4) 0.07 0.08 +14.3%
2000.36: git diff --cached (sparse-v3) 0.28 0.04 -85.7%
2000.37: git diff --cached (sparse-v4) 0.23 0.03 -87.0%
Co-authored-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Lessley Dennington <lessleydennington@gmail.com>
Reviewed-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* vd/sparse-reset:
unpack-trees: improve performance of next_cache_entry
reset: make --mixed sparse-aware
reset: make sparse-aware (except --mixed)
reset: integrate with sparse index
reset: expand test coverage for sparse checkouts
sparse-index: update command for expand/collapse test
reset: preserve skip-worktree bit in mixed reset
reset: rename is_missing to !is_in_reset_tree
Add new tests for `--merge` and `--keep` modes, as well as mixed reset with
pathspecs. New performance test cases exercise various execution paths for
`reset`.
Co-authored-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Victoria Dye <vdye@github.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Avoid performance measurements from getting ruined by gc and other
housekeeping pauses interfering in the middle.
* rs/disable-gc-during-perf-tests:
perf: disable automatic housekeeping
The mergesort implementation used to sort linked list has been
optimized.
* rs/mergesort:
test-mergesort: use repeatable random numbers
mergesort: use ranks stack
p0071: test performance of llist_mergesort()
p0071: measure sorting of already sorted and reversed files
test-mergesort: add unriffle_skewed mode
test-mergesort: add unriffle mode
test-mergesort: add generate subcommand
test-mergesort: add test subcommand
test-mergesort: add sort subcommand
test-mergesort: use strbuf_getline()
Turn off automatic background maintenance for perf tests by default to
avoid interference with performance measurements. Do that by using the
new file t/perf/config and using it as the system config file for perf
tests. Future tests intended to measure gc performance can override
the setting locally or call "git gc" explicitly.
This fixes a breakage in p2000 caused by gc automatically emptying the
reflog due its fake dates from 2005 being older than 90 days.
Helped-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git multi-pack-index write --bitmap" learns to propagate the
hashcache from original bitmap to resulting bitmap.
* tb/midx-write-propagate-namehash:
t5326: test propagating hashcache values
p5326: generate pack bitmaps before writing the MIDX bitmap
p5326: don't set core.multiPackIndex unnecessarily
p5326: create missing 'perf-tag' tag
midx.c: respect 'pack.writeBitmapHashcache' when writing bitmaps
pack-bitmap.c: propagate namehash values from existing bitmaps
t/helper/test-bitmap.c: add 'dump-hashes' mode
Teach test_perf_() to remove the temporary test_times.* files
at the end of each test.
test_perf_() runs a particular GIT_PERF_REPEAT_COUNT times and creates
./test_times.[123...]. It then uses a perl script to find the minimum
over "./test_times.*" (note the wildcard) and writes that time to
"test-results/<testname>.<testnumber>.result".
If the repeat count is changed during the pXXXX test script, stale
test_times.* files (from previous steps) may be included in the min()
computation. For example:
...
GIT_PERF_REPEAT_COUNT=3 \
test_perf "status" "
git status
"
GIT_PERF_REPEAT_COUNT=1 \
test_perf "checkout other" "
git checkout other
"
...
The time reported in the summary for "XXXX.2 checkout other" would
be "min( checkout[1], status[2], status[3] )".
We prevent that error by removing the test_times.* files at the end of
each test.
Signed-off-by: Jeff Hostetler <jeffhost@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When using `test_size` with `wc -c`, users on certain platforms can run
into issues when `wc` emits leading space characters in its output,
which confuses get_times.
Callers could switch to use test_file_size instead of `wc -c` (the
former never prints leading space characters, so will always work with
test_size regardless of platform), but this is an easy enough spot to
miss that we should teach get_times to be more tolerant of the input it
accepts.
Teach get_times to do just that by stripping any leading space
characters.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
b3dfeebb92 (rebase: avoid computing unnecessary patch IDs, 2016-07-29)
added a perf test that calls tac(1) from GNU core utilities. Support
systems without it by reversing the generated list using sort -nr
instead. sort(1) with options -n and -r is already used in other tests.
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Check if sorting takes advantage of already sorted or reversed content,
or if that corner case actually decreases performance, like it would for
a simplistic quicksort implementation.
Signed-off-by: René Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The GIT_TEST_INSTALLED was moved from perf-lib.sh to run in df0f5021
(perf-lib.sh: remove GIT_TEST_INSTALLED from perf-lib.sh, 2019-05-07)
and that included a change to how it inspected the existence of a
bin-wrappers directory. However, that included a typo that made the
match of bin-wrappers never work. Specifically, the assignment was
mydir_abs_wrappers="$mydir_abs_wrappers/bin-wrappers"
which uses the same variable before it is initialized. By changing it to
mydir_abs_wrappers="$mydir_abs/bin-wrappers"
We can correctly use the bin-wrappers directory.
This is critical to successfully computing performance of commands that
execute subcommands. The bin-wrappers ensure that the --exec-path is set
correctly.
Reported-by: Victoria Dye <vdye@github.com>
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
To help test the performance of permuting the contents of the hash-cache
when generating a MIDX bitmap, we need a bitmap which has its hash-cache
populated.
And since multi-pack bitmaps don't add *new* values to the hash-cache,
we have to rely on a single-pack bitmap to generate those values for us.
Therefore, generate a pack bitmap before the MIDX one in order to ensure
that the MIDX bitmap has entries in its hash-cache. Since we don't want
to time generating the pack bitmap, move that to a non-perf test run
before we try to generate the MIDX bitmap.
Likewise, get rid of the pack bitmap afterwords, to make certain that we
are not accidentally using it in the performance tests run later on.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When this performance test was originally written, `core.multiPackIndex`
was not the default and thus had to be enabled. But now that we have
18e449f86b (midx: enable core.multiPackIndex by default, 2020-09-25), we
no longer need this.
Drop the unnecessary setup (even though it's not hurting anything, it is
unnecessary at best and confusing at worst).
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some of the tests in test_full_bitmap rely on having a tag named
perf-tag in place. We could create it in test_full_bitmap(), but we want
to have it in place before the repack starts.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
These new performance tests demonstrate effectively the same behavior as
p5310, but use a multi-pack bitmap instead of a single-pack one.
Notably, p5326 does not create a MIDX bitmap with multiple packs. This
is so we can measure a direct comparison between it and p5310. Any
difference between the two is measuring just the overhead of using MIDX
bitmaps.
Here are the results of p5310 and p5326 together, measured at the same
time and on the same machine (using a Xenon W-2255 CPU):
Test HEAD
------------------------------------------------------------------------
5310.2: repack to disk 96.78(93.39+11.33)
5310.3: simulated clone 9.98(9.79+0.19)
5310.4: simulated fetch 1.75(4.26+0.19)
5310.5: pack to file (bitmap) 28.20(27.87+8.70)
5310.6: rev-list (commits) 0.41(0.36+0.05)
5310.7: rev-list (objects) 1.61(1.54+0.07)
5310.8: rev-list count with blob:none 0.25(0.21+0.04)
5310.9: rev-list count with blob:limit=1k 2.65(2.54+0.10)
5310.10: rev-list count with tree:0 0.23(0.19+0.04)
5310.11: simulated partial clone 4.34(4.21+0.12)
5310.13: clone (partial bitmap) 11.05(12.21+0.48)
5310.14: pack to file (partial bitmap) 31.25(34.22+3.70)
5310.15: rev-list with tree filter (partial bitmap) 0.26(0.22+0.04)
versus the same tests (this time using a multi-pack index):
Test HEAD
------------------------------------------------------------------------
5326.2: setup multi-pack index 78.99(75.29+11.58)
5326.3: simulated clone 11.78(11.56+0.22)
5326.4: simulated fetch 1.70(4.49+0.13)
5326.5: pack to file (bitmap) 28.02(27.72+8.76)
5326.6: rev-list (commits) 0.42(0.36+0.06)
5326.7: rev-list (objects) 1.65(1.58+0.06)
5326.8: rev-list count with blob:none 0.26(0.21+0.05)
5326.9: rev-list count with blob:limit=1k 2.97(2.86+0.10)
5326.10: rev-list count with tree:0 0.25(0.20+0.04)
5326.11: simulated partial clone 5.65(5.49+0.16)
5326.13: clone (partial bitmap) 12.22(13.43+0.38)
5326.14: pack to file (partial bitmap) 30.05(31.57+7.25)
5326.15: rev-list with tree filter (partial bitmap) 0.24(0.20+0.04)
There is slight overhead in "simulated clone", "simulated partial
clone", and "clone (partial bitmap)". Unsurprisingly, that overhead is
due to using the MIDX's reverse index to map between bit positions and
MIDX positions.
This can be reproduced by running "git repack -adb" along with "git
multi-pack-index write --bitmap" in a large-ish repository. Then run:
$ perf record -o pack.perf git -c core.multiPackIndex=false \
pack-objects --all --stdout >/dev/null </dev/null
$ perf record -o midx.perf git -c core.multiPackIndex=true \
pack-objects --all --stdout >/dev/null </dev/null
and compare the two with "perf diff -c delta -o 1 pack.perf midx.perf".
The most notable results are below (the next largest positive delta is
+0.14%):
# Event 'cycles'
#
# Baseline Delta Shared Object Symbol
# ........ ....... .................. ..........................
#
+5.86% git [.] nth_midxed_offset
+5.24% git [.] nth_midxed_pack_int_id
3.45% +0.97% git [.] offset_to_pack_pos
3.30% +0.57% git [.] pack_pos_to_offset
+0.30% git [.] pack_pos_to_midx
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A new p5326 introduced by the next patch will want these same tests,
interjecting its own setup in between. Move them out so that both perf
tests can reuse them.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
By using shorter names for the test repos, we will get a slightly more
compressed performance summary without comprimising clarity.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
As we increase our list of commands to test in
p2000-sparse-operations.sh, we will want to have a slightly smaller test
repository. Reduce the size by a factor of four by reducing the depth of
the step that creates a big index around a moderately-sized repository.
Also add a step to run 'git checkout -' on repeat. This requires having
a previous location in the reflog, so add that to the initialization
steps.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Rewrite the backend for "diff -G/-S" to use pcre2 engine when
available.
* ab/pickaxe-pcre2: (22 commits)
xdiff-interface: replace discard_hunk_line() with a flag
xdiff users: use designated initializers for out_line
pickaxe -G: don't special-case create/delete
pickaxe -G: terminate early on matching lines
xdiff-interface: allow early return from xdiff_emit_line_fn
xdiff-interface: prepare for allowing early return
pickaxe -S: slightly optimize contains()
pickaxe: rename variables in has_changes() for brevity
pickaxe -S: support content with NULs under --pickaxe-regex
pickaxe: assert that we must have a needle under -G or -S
pickaxe: refactor function selection in diffcore-pickaxe()
perf: add performance test for pickaxe
pickaxe/style: consolidate declarations and assignments
diff.h: move pickaxe fields together again
pickaxe: die when --find-object and --pickaxe-all are combined
pickaxe: die when -G and --pickaxe-regex are combined
pickaxe tests: add missing test for --no-pickaxe-regex being an error
pickaxe tests: test for -G, -S and --find-object incompatibility
pickaxe tests: add test for "log -S" not being a regex
pickaxe tests: add test for diffgrep_consume() internals
...
When the TEST_OUTPUT_DIRECTORY is defined, then all test data will be
written in that directory instead of the default directory located in
"t/". While this works as expected for our normal tests, performance
tests fail to locate and aggregate performance data because they don't
know to handle TEST_OUTPUT_DIRECTORY correctly and always look at the
default location.
Fix the issue by adding a `--results-dir` parameter to "aggregate.perl"
which identifies the directory where results are and by making the "run"
script awake of the TEST_OUTPUT_DIRECTORY variable.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a test for the -G and -S pickaxe options and related options.
This test supports being run with GIT_TEST_LONG=1 to adjust the limit
on the number of commits from 1k to 10k. The 1k limit seems to hit a
good spot on git.git
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git repack -A -d" in a partial clone unnecessarily loosened
objects in promisor pack.
* rs/repack-without-loosening-promised-objects:
repack: avoid loosening promisor objects in partial clones
Builds on top of the sparse-index infrastructure to mark operations
that are not ready to mark with the sparse index, causing them to
fall back on fully-populated index that they always have worked with.
* ds/sparse-index-protections: (47 commits)
name-hash: use expand_to_path()
sparse-index: expand_to_path()
name-hash: don't add directories to name_hash
revision: ensure full index
resolve-undo: ensure full index
read-cache: ensure full index
pathspec: ensure full index
merge-recursive: ensure full index
entry: ensure full index
dir: ensure full index
update-index: ensure full index
stash: ensure full index
rm: ensure full index
merge-index: ensure full index
ls-files: ensure full index
grep: ensure full index
fsck: ensure full index
difftool: ensure full index
commit: ensure full index
checkout: ensure full index
...
When `git repack -A -d` is run in a partial clone, `pack-objects`
is invoked twice: once to repack all promisor objects, and once to
repack all non-promisor objects. The latter `pack-objects` invocation
is with --exclude-promisor-objects and --unpack-unreachable, which
loosens all objects unused during this invocation. Unfortunately,
this includes promisor objects.
Because the -d argument to `git repack` subsequently deletes all loose
objects also in packs, these just-loosened promisor objects will be
immediately deleted. However, this extra disk churn is unnecessary in
the first place. For example, in a newly-cloned partial repo that
filters all blob objects (e.g. `--filter=blob:none`), `repack` ends up
unpacking all trees and commits into the filesystem because every
object, in this particular case, is a promisor object. Depending on
the repo size, this increases the disk usage considerably: In my copy
of the linux.git, the object directory peaked 26GB of more disk usage.
In order to avoid this extra disk churn, pass the names of the promisor
packfiles as --keep-pack arguments to the second invocation of
`pack-objects`. This informs `pack-objects` that the promisor objects
are already in a safe packfile and, therefore, do not need to be
loosened.
For testing, we need to validate whether any object was loosened.
However, the "evidence" (loosened objects) is deleted during the
process which prevents us from inspecting the object directory.
Instead, let's teach `pack-objects` to count loosened objects and
emit via trace2 thus allowing inspecting the debug events after the
process is finished. This new event is used on the added regression
test.
Lastly, add a new perf test to evaluate the performance impact
made by this changes (tested on git.git):
Test HEAD^ HEAD
----------------------------------------------------------
5600.3: gc 134.38(41.93+90.95) 7.80(6.72+1.35) -94.2%
For a bigger repository, such as linux.git, the improvement is
even bigger:
Test HEAD^ HEAD
-------------------------------------------------------------------
5600.3: gc 6833.00(918.07+3162.74) 268.79(227.02+39.18) -96.1%
These improvements are particular big because every object in the
newly-cloned partial repository is a promisor object.
Reported-by: SZEDER Gábor <szeder.dev@gmail.com>
Helped-by: Jeff King <peff@peff.net>
Helped-by: Jonathan Tan <jonathantanmy@google.com>
Signed-off-by: Rafael Silva <rafaeloliveira.cs@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When --exclude-promisor-objects is given, before traversing any objects
we iterate over all of the objects in any promisor packs, marking them
as UNINTERESTING and SEEN. We turn the oid we get from iterating the
pack into an object with parse_object(), but this has two problems:
- it's slow; we are zlib inflating (and reconstructing from deltas)
every byte of every object in the packfile
- it leaves the tree buffers attached to their structs, which means
our heap usage will grow to store every uncompressed tree
simultaneously. This can be gigabytes.
We can obviously fix the second by freeing the tree buffers after we've
parsed them. But we can observe that the function doesn't look at the
object contents at all! The only reason we call parse_object() is that
we need a "struct object" on which to set the flags. There are two
options here:
- we can look up just the object type via oid_object_info(), and then
call the appropriate lookup_foo() function
- we can call lookup_unknown_object(), which gives us an OBJ_NONE
struct (which will get auto-converted later by object_as_type() via
calls to lookup_commit(), etc).
The first one is closer to the current code, but we do pay the price to
look up the type for each object. The latter should be more efficient in
CPU, though it wastes a little bit of memory (the "unknown" object
structs are a union of all object types, so some of the structs are
bigger than they need to be). It also runs the risk of triggering a
latent bug in code that calls lookup_object() directly but isn't ready
to handle OBJ_NONE (such code would already be buggy, but we use
lookup_unknown_object() infrequently enough that it might be hiding).
I went with the second option here. I don't think the risk is high (and
we'd want to find and fix any such bugs anyway), and it should be more
efficient overall.
The new tests in p5600 show off the improvement (this is on git.git):
Test HEAD^ HEAD
-------------------------------------------------------------------------------
5600.5: count commits 0.37(0.37+0.00) 0.38(0.38+0.00) +2.7%
5600.6: count non-promisor commits 11.74(11.37+0.37) 0.04(0.03+0.00) -99.7%
The improvement is particularly big in this script because _every_
object in the newly-cloned partial repo is a promisor object. So after
marking them all, there's nothing left to traverse.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
To get the list of all promisor objects, we not only include all objects
in promisor packs, but also parse each of those objects to see which
objects they reference. After parsing a tree object, the tree->buffer
field will remain populated until we explicitly free it. So in a partial
clone of blob:none, for example, we are essentially reading every tree
in the repository (since they're all in the initial promisor pack), and
keeping all of their uncompressed contents in memory at once.
This patch frees the tree buffers after we've finished marking all of
their reachable objects. We shouldn't need to do this for any other
object type. While we are using some extra memory to store the structs,
no other object type stores the whole contents in its parsed form (we do
sometimes hold on to commit buffers, but less so these days due to
commit graphs, plus most commands which care about promisor objects turn
off the save_commit_buffer global).
Even for a moderate-sized repository like git.git, this patch drops the
peak heap (as measured by massif) for git-fsck from ~1.7GB to ~138MB.
Fsck is a good candidate for measuring here because it doesn't interact
with the promisor code except to call is_promisor_object(), so we can
isolate just this problem.
The added perf test shows only a tiny improvement on my machine for
git.git, since 1.7GB isn't enough to cause any real memory pressure:
Test HEAD^ HEAD
--------------------------------------------------------------------------------
5600.4: fsck 21.26(20.90+0.35) 20.84(20.79+0.04) -2.0%
With linux.git the absolute change is a bit bigger, though still a small
percentage:
Test HEAD^ HEAD
-----------------------------------------------------------------------------
5600.4: fsck 262.26(259.13+3.12) 254.92(254.62+0.29) -2.8%
I didn't have the patience to run it under massif with linux.git, but
it's probably on the order of about 14GB improvement, since that's the
sum of the sizes of all of the uncompressed trees (but still isn't
enough to create memory pressure on this particular machine, which has
64GB of RAM). Smaller machines would probably see a bigger effect on
runtime (and sadly our perf suite does not measure peak heap).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Optimize "rev-list --use-bitmap-index --objects" corner case that
uses negative tags as the stopping points.
* ps/pack-bitmap-optim:
pack-bitmap: avoid traversal of objects referenced by uninteresting tag
p2000-sparse-operations.sh compares different Git commands in
repositories with many files at HEAD but using sparse-checkout to focus
on a small portion of those files.
Add extra copies of the repository that use the sparse-index format so
we can track how that affects the performance of different commands.
At this point in time, the sparse-index is 100% overhead from the CPU
front, and this is measurable in these tests:
Test
---------------------------------------------------------------
2000.2: git status (full-index-v3) 0.59(0.51+0.12)
2000.3: git status (full-index-v4) 0.59(0.52+0.11)
2000.4: git status (sparse-index-v3) 1.40(1.32+0.12)
2000.5: git status (sparse-index-v4) 1.41(1.36+0.08)
2000.6: git add -A (full-index-v3) 2.32(1.97+0.19)
2000.7: git add -A (full-index-v4) 2.17(1.92+0.14)
2000.8: git add -A (sparse-index-v3) 2.31(2.21+0.15)
2000.9: git add -A (sparse-index-v4) 2.30(2.20+0.13)
2000.10: git add . (full-index-v3) 2.39(2.02+0.20)
2000.11: git add . (full-index-v4) 2.20(1.94+0.16)
2000.12: git add . (sparse-index-v3) 2.36(2.27+0.12)
2000.13: git add . (sparse-index-v4) 2.33(2.21+0.16)
2000.14: git commit -a -m A (full-index-v3) 2.47(2.12+0.20)
2000.15: git commit -a -m A (full-index-v4) 2.26(2.00+0.17)
2000.16: git commit -a -m A (sparse-index-v3) 3.01(2.92+0.16)
2000.17: git commit -a -m A (sparse-index-v4) 3.01(2.94+0.15)
Note that there is very little difference between the v3 and v4 index
formats when the sparse-index is enabled. This is primarily due to the
fact that the relative file sizes are the same, and the command time is
mostly taken up by parsing tree objects to expand the sparse index into
a full one.
With the current file layout, the index file sizes are given by this
table:
| full index | sparse index |
+-------------+--------------+
v3 | 108 MiB | 1.6 MiB |
v4 | 80 MiB | 1.2 MiB |
Future updates will improve the performance of Git commands when the
index is sparse.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Create a test script that takes the default performance test (the Git
codebase) and multiplies it by 256 using four layers of duplicated
trees of width four. This results in nearly one million blob entries in
the index. Then, we can clone this repository with sparse-checkout
patterns that demonstrate four copies of the initial repository. Each
clone will use a different index format or mode so peformance can be
tested across the different options.
Note that the initial repo is stripped of submodules before doing the
copies. This preserves the expected data shape of the sparse index,
because directories containing submodules are not collapsed to a sparse
directory entry.
Run a few Git commands on these clones, especially those that use the
index (status, add, commit).
Here are the results on my Linux machine:
Test
--------------------------------------------------------------
2000.2: git status (full-index-v3) 0.37(0.30+0.09)
2000.3: git status (full-index-v4) 0.39(0.32+0.10)
2000.4: git add -A (full-index-v3) 1.42(1.06+0.20)
2000.5: git add -A (full-index-v4) 1.26(0.98+0.16)
2000.6: git add . (full-index-v3) 1.40(1.04+0.18)
2000.7: git add . (full-index-v4) 1.26(0.98+0.17)
2000.8: git commit -a -m A (full-index-v3) 1.42(1.11+0.16)
2000.9: git commit -a -m A (full-index-v4) 1.33(1.08+0.16)
It is perhaps noteworthy that there is an improvement when using index
version 4. This is because the v3 index uses 108 MiB while the v4
index uses 80 MiB. Since the repeated portions of the directories are
very short (f3/f1/f2, for example) this ratio is less pronounced than in
similarly-sized real repositories.
Signed-off-by: Derrick Stolee <dstolee@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git diff-index" codepath has been taught to trust fsmonitor status
to reduce number of lstat() calls.
* nk/diff-index-fsmonitor:
fsmonitor: add perf test for git diff HEAD
fsmonitor: add assertion that fsmonitor is valid to check_removed
fsmonitor: skip lstat deletion check during git diff-index
"git repack" so far has been only capable of repacking everything
under the sun into a single pack (or split by size). A cleverer
strategy to reduce the cost of repacking a repository has been
introduced.
* tb/geometric-repack:
builtin/pack-objects.c: ignore missing links with --stdin-packs
builtin/repack.c: reword comment around pack-objects flags
builtin/repack.c: be more conservative with unsigned overflows
builtin/repack.c: assign pack split later
t7703: test --geometric repack with loose objects
builtin/repack.c: do not repack single packs with --geometric
builtin/repack.c: add '--geometric' option
packfile: add kept-pack cache for find_kept_pack_entry()
builtin/pack-objects.c: rewrite honor-pack-keep logic
p5303: measure time to repack with keep
p5303: add missing &&-chains
builtin/pack-objects.c: add '--stdin-packs' option
revision: learn '--no-kept-objects'
packfile: introduce 'find_kept_pack_entry()'
Perf test update to work better in secondary worktrees.
* jk/perf-in-worktrees:
t/perf: avoid copying worktree files from test repo
t/perf: handle worktrees as test repos
When preparing the bitmap walk, we first establish the set of of have
and want objects by iterating over the set of pending objects: if an
object is marked as uninteresting, it's declared as an object we already
have, otherwise as an object we want. These two sets are then used to
compute which transitively referenced objects we need to obtain.
One special case here are tag objects: when a tag is requested, we
resolve it to its first not-tag object and add both resolved objects as
well as the tag itself into either the have or want set. Given that the
uninteresting-property always propagates to referenced objects, it is
clear that if the tag is uninteresting, so are its children and vice
versa. But we fail to propagate the flag, which effectively means that
referenced objects will always be interesting except for the case where
they have already been marked as uninteresting explicitly.
This mislabeling does not impact correctness: we now have it in our
"wants" set, and given that we later do an `AND NOT` of the bitmaps of
"wants" and "haves" sets it is clear that the result must be the same.
But we now start to needlessly traverse the tag's referenced objects in
case it is uninteresting, even though we know that each referenced
object will be uninteresting anyway. In the worst case, this can lead to
a complete graph walk just to establish that we do not care for any
object.
Fix the issue by propagating the `UNINTERESTING` flag to pointees of tag
objects and add a benchmark with negative revisions to p5310. This shows
some nice performance benefits, tested with linux.git:
Test HEAD~ HEAD
---------------------------------------------------------------------------------------------------------------
5310.3: repack to disk 193.18(181.46+16.42) 194.61(183.41+15.83) +0.7%
5310.4: simulated clone 25.93(24.88+1.05) 25.81(24.73+1.08) -0.5%
5310.5: simulated fetch 2.64(5.30+0.69) 2.59(5.16+0.65) -1.9%
5310.6: pack to file (bitmap) 58.75(57.56+6.30) 58.29(57.61+5.73) -0.8%
5310.7: rev-list (commits) 1.45(1.18+0.26) 1.46(1.22+0.24) +0.7%
5310.8: rev-list (objects) 15.35(14.22+1.13) 15.30(14.23+1.07) -0.3%
5310.9: rev-list with tag negated via --not --all (objects) 22.49(20.93+1.56) 0.11(0.09+0.01) -99.5%
5310.10: rev-list with negative tag (objects) 0.61(0.44+0.16) 0.51(0.35+0.16) -16.4%
5310.11: rev-list count with blob:none 12.15(11.19+0.96) 12.18(11.19+0.99) +0.2%
5310.12: rev-list count with blob:limit=1k 17.77(15.71+2.06) 17.75(15.63+2.12) -0.1%
5310.13: rev-list count with tree:0 1.69(1.31+0.38) 1.68(1.28+0.39) -0.6%
5310.14: simulated partial clone 20.14(19.15+0.98) 19.98(18.93+1.05) -0.8%
5310.16: clone (partial bitmap) 12.78(13.89+1.07) 12.72(13.99+1.01) -0.5%
5310.17: pack to file (partial bitmap) 42.07(45.44+2.72) 41.44(44.66+2.80) -1.5%
5310.18: rev-list with tree filter (partial bitmap) 0.44(0.29+0.15) 0.46(0.32+0.14) +4.5%
While most benchmarks are probably in the range of noise, the newly
added 5310.9 and 5310.10 benchmarks consistenly perform better.
Signed-off-by: Patrick Steinhardt <ps@pks.im>.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
