perf: clear vfs audit_cache before each run
When generating a stream clone, we spend a large amount of time auditing path.
Before this changes, the first run was warming the vfs cache for the other
runs, leading to a large runtime difference and a "faulty" reported timing for
the operation.
We now clear this important cache between run to get a more realistic timing.
Below are some example of median time change when clearing these cases. The
maximum time for a run did not changed significantly.
### data-env-vars.name = mozilla-central-2018-08-01-zstd-sparse-revlog
# benchmark.name = hg.perf.exchange.stream.generate
# bin-env-vars.hg.flavor = default
# bin-env-vars.hg.py-re2-module = default
# benchmark.variants.version = latest
no-clearing: 17.289905
cache-clearing: 21.587965 (+24.86%, +4.30)
## data-env-vars.name = mozilla-central-2024-03-22-zstd-sparse-revlog
no-clearing: 32.670748
cache-clearing: 40.467095 (+23.86%, +7.80)
## data-env-vars.name = mozilla-try-2019-02-18-zstd-sparse-revlog
no-clearing: 37.838858
cache-clearing: 46.072749 (+21.76%, +8.23)
## data-env-vars.name = mozilla-unified-2024-03-22-zstd-sparse-revlog
no-clearing: 32.969395
cache-clearing: 39.646209 (+20.25%, +6.68)
In addition, this significantly reduce the timing difference between the
performance command, from the perf extensions and a `real `hg bundle` call
producing a stream bundle. Some significant differences remain especially on
the "mozilla-try" repositories, but they are now smaller.
Note that some of that difference will actually not be
attributable to the stream generation (like maybe phases or branch map
computation).
Below are some benchmarks done on a currently draft changeset fixing some
unrelated slowness in `hg bundle` (34a78972af409d1ff37c29e60f6ca811ad1a457d)
### data-env-vars.name = mozilla-central-2018-08-01-zstd-sparse-revlog
# bin-env-vars.hg.flavor = default
# bin-env-vars.hg.py-re2-module = default
hg.perf.exchange.stream.generate: 21.587965
hg.command.bundle: 24.301799 (+12.57%, +2.71)
## data-env-vars.name = mozilla-central-2024-03-22-zstd-sparse-revlog
hg.perf.exchange.stream.generate: 40.467095
hg.command.bundle: 44.831317 (+10.78%, +4.36)
## data-env-vars.name = mozilla-unified-2024-03-22-zstd-sparse-revlog
hg.perf.exchange.stream.generate: 39.646209
hg.command.bundle: 45.395258 (+14.50%, +5.75)
## data-env-vars.name = mozilla-try-2019-02-18-zstd-sparse-revlog
hg.perf.exchange.stream.generate: 46.072749
hg.command.bundle: 55.882608 (+21.29%, +9.81)
## data-env-vars.name = mozilla-try-2023-03-22-zlib-general-delta
hg.perf.exchange.stream.generate: 334.716708
hg.command.bundle: 377.856767 (+12.89%, +43.14)
## data-env-vars.name = mozilla-try-2023-03-22-zstd-sparse-revlog
hg.perf.exchange.stream.generate: 302.972301
hg.command.bundle: 326.098755 (+7.63%, +23.13)
import difflib
import random
import unittest
from mercurial import linelog
vecratio = 3 # number of replacelines / number of replacelines_vec
maxlinenum = 0xFFFFFF
maxb1 = 0xFFFFFF
maxdeltaa = 10
maxdeltab = 10
def _genedits(seed, endrev):
lines = []
random.seed(seed)
for rev in range(0, endrev):
n = len(lines)
a1 = random.randint(0, n)
a2 = random.randint(a1, min(n, a1 + maxdeltaa))
b1 = random.randint(0, maxb1)
b2 = random.randint(b1, b1 + maxdeltab)
usevec = not bool(random.randint(0, vecratio))
if usevec:
blines = [
(random.randint(0, rev), random.randint(0, maxlinenum))
for _ in range(b1, b2)
]
else:
blines = [(rev, bidx) for bidx in range(b1, b2)]
lines[a1:a2] = blines
yield lines, rev, a1, a2, b1, b2, blines, usevec
class linelogtests(unittest.TestCase):
def testlinelogencodedecode(self):
program = [
linelog._eof(0, 0),
linelog._jge(41, 42),
linelog._jump(0, 43),
linelog._eof(0, 0),
linelog._jl(44, 45),
linelog._line(46, 47),
]
ll = linelog.linelog(program, maxrev=100)
enc = ll.encode()
# round-trips okay
self.assertEqual(linelog.linelog.fromdata(enc)._program, ll._program)
self.assertEqual(linelog.linelog.fromdata(enc), ll)
# This encoding matches the encoding used by hg-experimental's
# linelog file, or is supposed to if it doesn't.
self.assertEqual(
enc,
(
b'\x00\x00\x01\x90\x00\x00\x00\x06'
b'\x00\x00\x00\xa4\x00\x00\x00*'
b'\x00\x00\x00\x00\x00\x00\x00+'
b'\x00\x00\x00\x00\x00\x00\x00\x00'
b'\x00\x00\x00\xb1\x00\x00\x00-'
b'\x00\x00\x00\xba\x00\x00\x00/'
),
)
def testsimpleedits(self):
ll = linelog.linelog()
# Initial revision: add lines 0, 1, and 2
ll.replacelines(1, 0, 0, 0, 3)
self.assertEqual(
[(l.rev, l.linenum) for l in ll.annotate(1)],
[
(1, 0),
(1, 1),
(1, 2),
],
)
# Replace line 1 with a new line
ll.replacelines(2, 1, 2, 1, 2)
self.assertEqual(
[(l.rev, l.linenum) for l in ll.annotate(2)],
[
(1, 0),
(2, 1),
(1, 2),
],
)
# delete a line out of 2
ll.replacelines(3, 1, 2, 0, 0)
self.assertEqual(
[(l.rev, l.linenum) for l in ll.annotate(3)],
[
(1, 0),
(1, 2),
],
)
# annotation of 1 is unchanged
self.assertEqual(
[(l.rev, l.linenum) for l in ll.annotate(1)],
[
(1, 0),
(1, 1),
(1, 2),
],
)
ll.annotate(3) # set internal state to revision 3
start = ll.getoffset(0)
end = ll.getoffset(1)
self.assertEqual(
ll.getalllines(start, end),
[
(1, 0),
(2, 1),
(1, 1),
],
)
self.assertEqual(
ll.getalllines(),
[
(1, 0),
(2, 1),
(1, 1),
(1, 2),
],
)
def testparseclinelogfile(self):
# This data is what the replacements in testsimpleedits
# produce when fed to the original linelog.c implementation.
data = (
b'\x00\x00\x00\x0c\x00\x00\x00\x0f'
b'\x00\x00\x00\x00\x00\x00\x00\x02'
b'\x00\x00\x00\x05\x00\x00\x00\x06'
b'\x00\x00\x00\x06\x00\x00\x00\x00'
b'\x00\x00\x00\x00\x00\x00\x00\x07'
b'\x00\x00\x00\x06\x00\x00\x00\x02'
b'\x00\x00\x00\x00\x00\x00\x00\x00'
b'\x00\x00\x00\t\x00\x00\x00\t'
b'\x00\x00\x00\x00\x00\x00\x00\x0c'
b'\x00\x00\x00\x08\x00\x00\x00\x05'
b'\x00\x00\x00\x06\x00\x00\x00\x01'
b'\x00\x00\x00\x00\x00\x00\x00\x05'
b'\x00\x00\x00\x0c\x00\x00\x00\x05'
b'\x00\x00\x00\n\x00\x00\x00\x01'
b'\x00\x00\x00\x00\x00\x00\x00\t'
)
llc = linelog.linelog.fromdata(data)
self.assertEqual(
[(l.rev, l.linenum) for l in llc.annotate(1)],
[
(1, 0),
(1, 1),
(1, 2),
],
)
self.assertEqual(
[(l.rev, l.linenum) for l in llc.annotate(2)],
[
(1, 0),
(2, 1),
(1, 2),
],
)
self.assertEqual(
[(l.rev, l.linenum) for l in llc.annotate(3)],
[
(1, 0),
(1, 2),
],
)
# Check we emit the same bytecode.
ll = linelog.linelog()
# Initial revision: add lines 0, 1, and 2
ll.replacelines(1, 0, 0, 0, 3)
# Replace line 1 with a new line
ll.replacelines(2, 1, 2, 1, 2)
# delete a line out of 2
ll.replacelines(3, 1, 2, 0, 0)
diff = '\n ' + '\n '.join(
difflib.unified_diff(
ll.debugstr().splitlines(),
llc.debugstr().splitlines(),
'python',
'c',
lineterm='',
)
)
self.assertEqual(ll._program, llc._program, 'Program mismatch: ' + diff)
# Done as a secondary step so we get a better result if the
# program is where the mismatch is.
self.assertEqual(ll, llc)
self.assertEqual(ll.encode(), data)
def testanothersimplecase(self):
ll = linelog.linelog()
ll.replacelines(3, 0, 0, 0, 2)
ll.replacelines(4, 0, 2, 0, 0)
self.assertEqual([(l.rev, l.linenum) for l in ll.annotate(4)], [])
self.assertEqual(
[(l.rev, l.linenum) for l in ll.annotate(3)], [(3, 0), (3, 1)]
)
# rev 2 is empty because contents were only ever introduced in rev 3
self.assertEqual([(l.rev, l.linenum) for l in ll.annotate(2)], [])
def testrandomedits(self):
# Inspired by original linelog tests.
seed = random.random()
numrevs = 2000
ll = linelog.linelog()
# Populate linelog
for lines, rev, a1, a2, b1, b2, blines, usevec in _genedits(
seed, numrevs
):
if usevec:
ll.replacelines_vec(rev, a1, a2, blines)
else:
ll.replacelines(rev, a1, a2, b1, b2)
ll.annotate(rev)
self.assertEqual(ll.annotateresult, lines)
# Verify we can get back these states by annotating each rev
for lines, rev, a1, a2, b1, b2, blines, usevec in _genedits(
seed, numrevs
):
ar = ll.annotate(rev)
self.assertEqual([(l.rev, l.linenum) for l in ar], lines)
def testinfinitebadprogram(self):
ll = linelog.linelog.fromdata(
b'\x00\x00\x00\x00\x00\x00\x00\x02' # header
b'\x00\x00\x00\x00\x00\x00\x00\x01' # JUMP to self
)
with self.assertRaises(linelog.LineLogError):
# should not be an infinite loop and raise
ll.annotate(1)
if __name__ == '__main__':
import silenttestrunner
silenttestrunner.main(__name__)