nodemap: introduce an option to use mmap to read the nodemap mapping
The performance and memory benefit is much greater if we don't have to copy all
the data in memory for each information. So we introduce an option (on by
default) to read the data using mmap.
This changeset is the last one definition the API for index support nodemap
data. (they have to be able to use the mmaping).
Below are some benchmark comparing the best we currently have in 5.3 with the
final step of this series (using the persistent nodemap implementation in
Rust). The benchmark run `hg perfindex` with various revset and the following
variants:
Before:
* do not use the persistent nodemap
* use the CPython implementation of the index for nodemap
* use mmapping of the changelog index
After:
* use the MixedIndex Rust code, with the NodeTree object for nodemap access
(still in review)
* use the persistent nodemap data from disk
* access the persistent nodemap data through mmap
* use mmapping of the changelog index
The persistent nodemap greatly speed up most operation on very large
repositories. Some of the previously very fast lookup end up a bit slower because
the persistent nodemap has to be setup. However the absolute slowdown is very
small and won't matters in the big picture.
Here are some numbers (in seconds) for the reference copy of mozilla-try:
Revset Before After abs-change speedup
-10000: 0.004622 0.005532 0.000910 × 0.83
-10: 0.000050 0.000132 0.000082 × 0.37
tip 0.000052 0.000085 0.000033 × 0.61
0 + (-10000:) 0.028222 0.005337 -0.022885 × 5.29
0 0.023521 0.000084 -0.023437 × 280.01
(-10000:) + 0 0.235539 0.005308 -0.230231 × 44.37
(-10:) + :9 0.232883 0.000180 -0.232703 ×1293.79
(-10000:) + (:99) 0.238735 0.005358 -0.233377 × 44.55
:99 + (-10000:) 0.317942 0.005593 -0.312349 × 56.84
:9 + (-10:) 0.313372 0.000179 -0.313193 ×1750.68
:9 0.316450 0.000143 -0.316307 ×2212.93
On smaller repositories, the cost of nodemap related operation is not as big, so
the win is much more modest. Yet it helps shaving a handful of millisecond here
and there.
Here are some numbers (in seconds) for the reference copy of mercurial:
Revset Before After abs-change speedup
-10: 0.000065 0.000097 0.000032 × 0.67
tip 0.000063 0.000078 0.000015 × 0.80
0 0.000561 0.000079 -0.000482 × 7.10
-10000: 0.004609 0.003648 -0.000961 × 1.26
0 + (-10000:) 0.005023 0.003715 -0.001307 × 1.35
(-10:) + :9 0.002187 0.000108 -0.002079 ×20.25
(-10000:) + 0 0.006252 0.003716 -0.002536 × 1.68
(-10000:) + (:99) 0.006367 0.003707 -0.002660 × 1.71
:9 + (-10:) 0.003846 0.000110 -0.003736 ×34.96
:9 0.003854 0.000099 -0.003755 ×38.92
:99 + (-10000:) 0.007644 0.003778 -0.003866 × 2.02
Differential Revision: https://phab.mercurial-scm.org/D7894
--- a/mercurial/configitems.py Fri Feb 14 15:03:26 2020 +0100
+++ b/mercurial/configitems.py Tue Feb 11 11:18:52 2020 +0100
@@ -666,6 +666,9 @@
b'experimental', b'exp-persistent-nodemap', default=False,
)
coreconfigitem(
+ b'experimental', b'exp-persistent-nodemap.mmap', default=True,
+)
+coreconfigitem(
b'experimental', b'server.filesdata.recommended-batch-size', default=50000,
)
coreconfigitem(
--- a/mercurial/debugcommands.py Fri Feb 14 15:03:26 2020 +0100
+++ b/mercurial/debugcommands.py Tue Feb 11 11:18:52 2020 +0100
@@ -2122,7 +2122,7 @@
nm_data = nodemap.persisted_data(cl)
if nm_data is not None:
docket, data = nm_data
- ui.write(data)
+ ui.write(data[:])
elif opts['check']:
unfi = repo.unfiltered()
cl = unfi.changelog
--- a/mercurial/localrepo.py Fri Feb 14 15:03:26 2020 +0100
+++ b/mercurial/localrepo.py Tue Feb 11 11:18:52 2020 +0100
@@ -934,6 +934,8 @@
options[b'rust.index'] = True
if ui.configbool(b'experimental', b'exp-persistent-nodemap'):
options[b'exp-persistent-nodemap'] = True
+ if ui.configbool(b'experimental', b'exp-persistent-nodemap.mmap'):
+ options[b'exp-persistent-nodemap.mmap'] = True
if ui.configbool(b'devel', b'persistent-nodemap'):
options[b'devel-force-nodemap'] = True
--- a/mercurial/revlog.py Fri Feb 14 15:03:26 2020 +0100
+++ b/mercurial/revlog.py Tue Feb 11 11:18:52 2020 +0100
@@ -545,9 +545,6 @@
indexdata = b''
self._initempty = True
try:
- nodemap_data = nodemaputil.persisted_data(self)
- if nodemap_data is not None:
- self._nodemap_docket = nodemap_data[0]
with self._indexfp() as f:
if (
mmapindexthreshold is not None
@@ -639,6 +636,7 @@
if use_nodemap:
nodemap_data = nodemaputil.persisted_data(self)
if nodemap_data is not None:
+ self._nodemap_docket = nodemap_data[0]
index.update_nodemap_data(*nodemap_data)
except (ValueError, IndexError):
raise error.RevlogError(
--- a/mercurial/revlogutils/nodemap.py Fri Feb 14 15:03:26 2020 +0100
+++ b/mercurial/revlogutils/nodemap.py Tue Feb 11 11:18:52 2020 +0100
@@ -8,6 +8,7 @@
from __future__ import absolute_import
+import errno
import os
import re
import struct
@@ -45,11 +46,18 @@
docket.data_unused = data_unused
filename = _rawdata_filepath(revlog, docket)
- data = revlog.opener.tryread(filename)
+ use_mmap = revlog.opener.options.get("exp-persistent-nodemap.mmap")
+ try:
+ with revlog.opener(filename) as fd:
+ if use_mmap:
+ data = util.buffer(util.mmapread(fd, data_length))
+ else:
+ data = fd.read(data_length)
+ except OSError as e:
+ if e.errno != errno.ENOENT:
+ raise
if len(data) < data_length:
return None
- elif len(data) > data_length:
- data = data[:data_length]
return docket, data
@@ -81,6 +89,8 @@
can_incremental = util.safehasattr(revlog.index, "nodemap_data_incremental")
ondisk_docket = revlog._nodemap_docket
+ feed_data = util.safehasattr(revlog.index, "update_nodemap_data")
+ use_mmap = revlog.opener.options.get("exp-persistent-nodemap.mmap")
data = None
# first attemp an incremental update of the data
@@ -97,12 +107,18 @@
datafile = _rawdata_filepath(revlog, target_docket)
# EXP-TODO: if this is a cache, this should use a cache vfs, not a
# store vfs
+ new_length = target_docket.data_length + len(data)
with revlog.opener(datafile, b'r+') as fd:
fd.seek(target_docket.data_length)
fd.write(data)
- fd.seek(0)
- new_data = fd.read(target_docket.data_length + len(data))
- target_docket.data_length += len(data)
+ if feed_data:
+ if use_mmap:
+ fd.seek(0)
+ new_data = fd.read(new_length)
+ else:
+ fd.flush()
+ new_data = util.buffer(util.mmapread(fd, new_length))
+ target_docket.data_length = new_length
target_docket.data_unused += data_changed_count
if data is None:
@@ -115,9 +131,14 @@
data = persistent_data(revlog.index)
# EXP-TODO: if this is a cache, this should use a cache vfs, not a
# store vfs
- new_data = data
- with revlog.opener(datafile, b'w') as fd:
+ with revlog.opener(datafile, b'w+') as fd:
fd.write(data)
+ if feed_data:
+ if use_mmap:
+ new_data = data
+ else:
+ fd.flush()
+ new_data = util.buffer(util.mmapread(fd, len(data)))
target_docket.data_length = len(data)
target_docket.tip_rev = revlog.tiprev()
# EXP-TODO: if this is a cache, this should use a cache vfs, not a
@@ -125,7 +146,7 @@
with revlog.opener(revlog.nodemap_file, b'w', atomictemp=True) as fp:
fp.write(target_docket.serialize())
revlog._nodemap_docket = target_docket
- if util.safehasattr(revlog.index, "update_nodemap_data"):
+ if feed_data:
revlog.index.update_nodemap_data(target_docket, new_data)
# EXP-TODO: if the transaction abort, we should remove the new data and
--- a/tests/test-persistent-nodemap.t Fri Feb 14 15:03:26 2020 +0100
+++ b/tests/test-persistent-nodemap.t Tue Feb 11 11:18:52 2020 +0100
@@ -84,3 +84,37 @@
$ hg debugnodemap --check
revision in index: 5002
revision in nodemap: 5002
+
+Test code path without mmap
+---------------------------
+
+ $ echo bar > bar
+ $ hg add bar
+ $ hg ci -m 'bar' --config experimental.exp-persistent-nodemap.mmap=no
+
+ $ hg debugnodemap --check --config experimental.exp-persistent-nodemap.mmap=yes
+ revision in index: 5003
+ revision in nodemap: 5003
+ $ hg debugnodemap --check --config experimental.exp-persistent-nodemap.mmap=no
+ revision in index: 5003
+ revision in nodemap: 5003
+
+
+#if pure
+ $ hg debugnodemap --metadata
+ uid: ???????????????? (glob)
+ tip-rev: 5002
+ data-length: 123328
+ data-unused: 384
+ $ f --sha256 .hg/store/00changelog-*.nd --size
+ .hg/store/00changelog-????????????????.nd: size=123328, sha256=10d26e9776b6596af0f89143a54eba8cc581e929c38242a02a7b0760698c6c70 (glob)
+
+#else
+ $ hg debugnodemap --metadata
+ uid: ???????????????? (glob)
+ tip-rev: 5002
+ data-length: 122944
+ data-unused: 0
+ $ f --sha256 .hg/store/00changelog-*.nd --size
+ .hg/store/00changelog-????????????????.nd: size=122944, sha256=755976b22b64ab680401b45395953504e64e7fa8c31ac570f58dee21e15f9bc0 (glob)
+#endif