Mercurial > hg
view mercurial/ancestor.py @ 42044:bb271ec2fbfb
compression: introduce a `storage.revlog.zstd.level` configuration
This option control the zstd compression level used when compressing revlog
chunk. The usage of zstd for revlog compression has not graduated from
experimental yet, but we intend to fix that soon.
The option name for the compression level is more straight forward to pick, so
this changesets comes first. Having a dedicated option for each compression
engine is useful because they don't support the same range of values.
I ran the same measurement as for the zlib compression level (in the parent
changesets). The variation in repository size is stay mostly in the same (small)
range. The "read/write" performance see smallish variation, but are overall much
better than zlib. Write performance show the same tend of having better write
performance for when reaching high-end compression.
Again, we don't intend to change the default zstd compression level (currently:
3) in this series. However this is worth investigating in the future.
The Performance comparison of zlib vs zstd is quite impressive. The repository
size stay in the same range, but the performance are much better in all
situations.
Comparison summary
==================
We are looking at:
- performance range for zlib
- performance range for zstd
- comparison of default zstd (level-3) to default zlib (level 6)
- comparison of the slowest zstd time to the fastest zlib time
Read performance:
-----------------
| zlib | zstd | cmp | f2s
mercurial | 0.170159 - 0.189219 | 0.144127 - 0.149624 | 80% | 88%
pypy | 2.679217 - 2.768691 | 1.532317 - 1.705044 | 60% | 63%
netbeans | 122.477027 - 141.620281 | 72.996346 - 89.731560 | 58% | 73%
mozilla | 147.867662 - 170.572118 | 91.700995 - 105.853099 | 56% | 71%
Write performance:
------------------
| zlib | zstd | cmp | f2s
mercurial | 53.250304 - 56.2936129 | 40.877025 - 45.677286 | 75% | 86%
pypy | 460.721984 - 476.589918 | 270.545409 - 301.002219 | 63% | 65%
netbeans | 520.560316 - 715.930400 | 370.356311 - 428.329652 | 55% | 82%
mozilla | 739.803002 - 987.056093 | 505.152906 - 591.930683 | 57% | 80%
Raw data
--------
repo alg lvl .hg/store size 00manifest.d read write
mercurial zlib 1 49,402,813 5,963,475 0.170159 53.250304
mercurial zlib 6 47,197,397 5,875,730 0.182820 56.264320
mercurial zlib 9 47,121,596 5,849,781 0.189219 56.293612
mercurial zstd 1 49,737,084 5,966,355 0.144127 40.877025
mercurial zstd 3 48,961,867 5,895,208 0.146376 42.268142
mercurial zstd 5 48,200,592 5,938,676 0.149624 43.162875
mercurial zstd 10 47,833,520 5,913,353 0.145185 44.012489
mercurial zstd 15 47,314,604 5,728,679 0.147686 45.677286
mercurial zstd 20 47,330,502 5,830,539 0.145789 45.025407
mercurial zstd 22 47,330,076 5,830,539 0.143996 44.690460
pypy zlib 1 370,830,572 28,462,425 2.679217 460.721984
pypy zlib 6 340,112,317 27,648,747 2.768691 467.537158
pypy zlib 9 338,360,736 27,639,003 2.763495 476.589918
pypy zstd 1 362,377,479 27,916,214 1.532317 270.545409
pypy zstd 3 354,137,693 27,905,988 1.686718 294.951509
pypy zstd 5 342,640,043 27,655,774 1.705044 301.002219
pypy zstd 10 334,224,327 27,164,493 1.567287 285.186239
pypy zstd 15 329,000,363 26,645,965 1.637729 299.561332
pypy zstd 20 324,534,039 26,199,547 1.526813 302.149827
pypy zstd 22 324,530,595 26,198,932 1.525718 307.821218
netbeans zlib 1 1,281,847,810 165,495,457 122.477027 520.560316
netbeans zlib 6 1,205,284,353 159,161,207 139.876147 715.930400
netbeans zlib 9 1,197,135,671 155,034,586 141.620281 678.297064
netbeans zstd 1 1,259,581,737 160,840,613 72.996346 370.356311
netbeans zstd 3 1,232,978,122 157,691,551 81.622317 396.733087
netbeans zstd 5 1,208,034,075 160,246,880 83.080549 364.342626
netbeans zstd 10 1,188,624,176 156,083,417 79.323935 403.594602
netbeans zstd 15 1,176,973,589 153,859,477 89.731560 428.329652
netbeans zstd 20 1,162,958,258 151,147,535 82.842667 392.335349
netbeans zstd 22 1,162,707,029 151,150,220 82.565695 402.840655
mozilla zlib 1 2,775,497,186 298,527,987 147.867662 751.263721
mozilla zlib 6 2,596,856,420 286,597,671 170.572118 987.056093
mozilla zlib 9 2,587,542,494 287,018,264 163.622338 739.803002
mozilla zstd 1 2,723,159,348 286,617,532 91.700995 570.042751
mozilla zstd 3 2,665,055,001 286,152,013 95.240155 561.412805
mozilla zstd 5 2,607,819,817 288,060,030 101.978048 505.152906
mozilla zstd 10 2,558,761,085 283,967,648 104.113481 497.771202
mozilla zstd 15 2,526,216,060 275,581,300 105.853099 591.930683
mozilla zstd 20 2,485,114,806 266,478,859 95.268795 576.515389
mozilla zstd 22 2,484,869,080 266,456,505 94.429282 572.785537
author | Pierre-Yves David <pierre-yves.david@octobus.net> |
---|---|
date | Wed, 27 Mar 2019 18:35:59 +0100 |
parents | 4856c9b8cbaf |
children | 2372284d9457 |
line wrap: on
line source
# ancestor.py - generic DAG ancestor algorithm for mercurial # # Copyright 2006 Matt Mackall <mpm@selenic.com> # # This software may be used and distributed according to the terms of the # GNU General Public License version 2 or any later version. from __future__ import absolute_import import heapq from .node import nullrev from . import ( dagop, policy, pycompat, ) parsers = policy.importmod(r'parsers') def commonancestorsheads(pfunc, *nodes): """Returns a set with the heads of all common ancestors of all nodes, heads(::nodes[0] and ::nodes[1] and ...) . pfunc must return a list of parent vertices for a given vertex. """ if not isinstance(nodes, set): nodes = set(nodes) if nullrev in nodes: return set() if len(nodes) <= 1: return nodes allseen = (1 << len(nodes)) - 1 seen = [0] * (max(nodes) + 1) for i, n in enumerate(nodes): seen[n] = 1 << i poison = 1 << (i + 1) gca = set() interesting = len(nodes) nv = len(seen) - 1 while nv >= 0 and interesting: v = nv nv -= 1 if not seen[v]: continue sv = seen[v] if sv < poison: interesting -= 1 if sv == allseen: gca.add(v) sv |= poison if v in nodes: # history is linear return {v} if sv < poison: for p in pfunc(v): sp = seen[p] if p == nullrev: continue if sp == 0: seen[p] = sv interesting += 1 elif sp != sv: seen[p] |= sv else: for p in pfunc(v): if p == nullrev: continue sp = seen[p] if sp and sp < poison: interesting -= 1 seen[p] = sv return gca def ancestors(pfunc, *orignodes): """ Returns the common ancestors of a and b that are furthest from a root (as measured by longest path). pfunc must return a list of parent vertices for a given vertex. """ def deepest(nodes): interesting = {} count = max(nodes) + 1 depth = [0] * count seen = [0] * count mapping = [] for (i, n) in enumerate(sorted(nodes)): depth[n] = 1 b = 1 << i seen[n] = b interesting[b] = 1 mapping.append((b, n)) nv = count - 1 while nv >= 0 and len(interesting) > 1: v = nv nv -= 1 dv = depth[v] if dv == 0: continue sv = seen[v] for p in pfunc(v): if p == nullrev: continue dp = depth[p] nsp = sp = seen[p] if dp <= dv: depth[p] = dv + 1 if sp != sv: interesting[sv] += 1 nsp = seen[p] = sv if sp: interesting[sp] -= 1 if interesting[sp] == 0: del interesting[sp] elif dv == dp - 1: nsp = sp | sv if nsp == sp: continue seen[p] = nsp interesting.setdefault(nsp, 0) interesting[nsp] += 1 interesting[sp] -= 1 if interesting[sp] == 0: del interesting[sp] interesting[sv] -= 1 if interesting[sv] == 0: del interesting[sv] if len(interesting) != 1: return [] k = 0 for i in interesting: k |= i return set(n for (i, n) in mapping if k & i) gca = commonancestorsheads(pfunc, *orignodes) if len(gca) <= 1: return gca return deepest(gca) class incrementalmissingancestors(object): '''persistent state used to calculate missing ancestors incrementally Although similar in spirit to lazyancestors below, this is a separate class because trying to support contains and missingancestors operations with the same internal data structures adds needless complexity.''' def __init__(self, pfunc, bases): self.bases = set(bases) if not self.bases: self.bases.add(nullrev) self.pfunc = pfunc def hasbases(self): '''whether the common set has any non-trivial bases''' return self.bases and self.bases != {nullrev} def addbases(self, newbases): '''grow the ancestor set by adding new bases''' self.bases.update(newbases) def basesheads(self): return dagop.headrevs(self.bases, self.pfunc) def removeancestorsfrom(self, revs): '''remove all ancestors of bases from the set revs (in place)''' bases = self.bases pfunc = self.pfunc revs.difference_update(bases) # nullrev is always an ancestor revs.discard(nullrev) if not revs: return # anything in revs > start is definitely not an ancestor of bases # revs <= start needs to be investigated start = max(bases) keepcount = sum(1 for r in revs if r > start) if len(revs) == keepcount: # no revs to consider return for curr in pycompat.xrange(start, min(revs) - 1, -1): if curr not in bases: continue revs.discard(curr) bases.update(pfunc(curr)) if len(revs) == keepcount: # no more potential revs to discard break def missingancestors(self, revs): '''return all the ancestors of revs that are not ancestors of self.bases This may include elements from revs. Equivalent to the revset (::revs - ::self.bases). Revs are returned in revision number order, which is a topological order.''' revsvisit = set(revs) basesvisit = self.bases pfunc = self.pfunc bothvisit = revsvisit.intersection(basesvisit) revsvisit.difference_update(bothvisit) if not revsvisit: return [] start = max(max(revsvisit), max(basesvisit)) # At this point, we hold the invariants that: # - revsvisit is the set of nodes we know are an ancestor of at least # one of the nodes in revs # - basesvisit is the same for bases # - bothvisit is the set of nodes we know are ancestors of at least one # of the nodes in revs and one of the nodes in bases. bothvisit and # revsvisit are mutually exclusive, but bothvisit is a subset of # basesvisit. # Now we walk down in reverse topo order, adding parents of nodes # already visited to the sets while maintaining the invariants. When a # node is found in both revsvisit and basesvisit, it is removed from # revsvisit and added to bothvisit. When revsvisit becomes empty, there # are no more ancestors of revs that aren't also ancestors of bases, so # exit. missing = [] for curr in pycompat.xrange(start, nullrev, -1): if not revsvisit: break if curr in bothvisit: bothvisit.remove(curr) # curr's parents might have made it into revsvisit through # another path for p in pfunc(curr): revsvisit.discard(p) basesvisit.add(p) bothvisit.add(p) continue if curr in revsvisit: missing.append(curr) revsvisit.remove(curr) thisvisit = revsvisit othervisit = basesvisit elif curr in basesvisit: thisvisit = basesvisit othervisit = revsvisit else: # not an ancestor of revs or bases: ignore continue for p in pfunc(curr): if p == nullrev: pass elif p in othervisit or p in bothvisit: # p is implicitly in thisvisit. This means p is or should be # in bothvisit revsvisit.discard(p) basesvisit.add(p) bothvisit.add(p) else: # visit later thisvisit.add(p) missing.reverse() return missing # Extracted from lazyancestors.__iter__ to avoid a reference cycle def _lazyancestorsiter(parentrevs, initrevs, stoprev, inclusive): seen = {nullrev} heappush = heapq.heappush heappop = heapq.heappop heapreplace = heapq.heapreplace see = seen.add if inclusive: visit = [-r for r in initrevs] seen.update(initrevs) heapq.heapify(visit) else: visit = [] heapq.heapify(visit) for r in initrevs: p1, p2 = parentrevs(r) if p1 not in seen: heappush(visit, -p1) see(p1) if p2 not in seen: heappush(visit, -p2) see(p2) while visit: current = -visit[0] if current < stoprev: break yield current # optimize out heapq operation if p1 is known to be the next highest # revision, which is quite common in linear history. p1, p2 = parentrevs(current) if p1 not in seen: if current - p1 == 1: visit[0] = -p1 else: heapreplace(visit, -p1) see(p1) else: heappop(visit) if p2 not in seen: heappush(visit, -p2) see(p2) class lazyancestors(object): def __init__(self, pfunc, revs, stoprev=0, inclusive=False): """Create a new object generating ancestors for the given revs. Does not generate revs lower than stoprev. This is computed lazily starting from revs. The object supports iteration and membership. cl should be a changelog and revs should be an iterable. inclusive is a boolean that indicates whether revs should be included. Revs lower than stoprev will not be generated. Result does not include the null revision.""" self._parentrevs = pfunc self._initrevs = revs = [r for r in revs if r >= stoprev] self._stoprev = stoprev self._inclusive = inclusive self._containsseen = set() self._containsiter = _lazyancestorsiter(self._parentrevs, self._initrevs, self._stoprev, self._inclusive) def __nonzero__(self): """False if the set is empty, True otherwise.""" try: next(iter(self)) return True except StopIteration: return False __bool__ = __nonzero__ def __iter__(self): """Generate the ancestors of _initrevs in reverse topological order. If inclusive is False, yield a sequence of revision numbers starting with the parents of each revision in revs, i.e., each revision is *not* considered an ancestor of itself. Results are emitted in reverse revision number order. That order is also topological: a child is always emitted before its parent. If inclusive is True, the source revisions are also yielded. The reverse revision number order is still enforced.""" return _lazyancestorsiter(self._parentrevs, self._initrevs, self._stoprev, self._inclusive) def __contains__(self, target): """Test whether target is an ancestor of self._initrevs.""" seen = self._containsseen if target in seen: return True iter = self._containsiter if iter is None: # Iterator exhausted return False # Only integer target is valid, but some callers expect 'None in self' # to be False. So we explicitly allow it. if target is None: return False see = seen.add try: while True: rev = next(iter) see(rev) if rev == target: return True if rev < target: return False except StopIteration: # Set to None to indicate fast-path can be used next time, and to # free up memory. self._containsiter = None return False class rustlazyancestors(object): def __init__(self, index, revs, stoprev=0, inclusive=False): self._index = index self._stoprev = stoprev self._inclusive = inclusive # no need to prefilter out init revs that are smaller than stoprev, # it's done by rustlazyancestors constructor. # we need to convert to a list, because our ruslazyancestors # constructor (from C code) doesn't understand anything else yet self._initrevs = initrevs = list(revs) self._containsiter = parsers.rustlazyancestors( index, initrevs, stoprev, inclusive) def __nonzero__(self): """False if the set is empty, True otherwise. It's better to duplicate this essentially trivial method than to subclass lazyancestors """ try: next(iter(self)) return True except StopIteration: return False def __iter__(self): return parsers.rustlazyancestors(self._index, self._initrevs, self._stoprev, self._inclusive) def __contains__(self, target): return target in self._containsiter