def perfrevlogrevisions(ui, repo, file_=None, startrev=0, reverse=False,
**opts):
"""Benchmark reading a series of revisions from a revlog.
By default, we read every ``-d/--dist`` revision from 0 to tip of
the specified revlog.
The start revision can be defined via ``-s/--startrev``.
"""
rl = cmdutil.openrevlog(repo, 'perfrevlogrevisions', file_, opts)
rllen = getlen(ui)(rl)
def d():
rl.clearcaches()
beginrev = startrev
endrev = rllen
dist = opts['dist']
if reverse:
beginrev, endrev = endrev, beginrev
dist = -1 * dist
for x in xrange(beginrev, endrev, dist):
# Old revisions don't support passing int.
n = rl.node(x)
rl.revision(n)
timer, fm = gettimer(ui, opts)
timer(d)
fm.end()
def perfbdiff(ui, repo, file_, rev=None, count=None, **opts):
"""benchmark a bdiff between revisions
By default, benchmark a bdiff between its delta parent and itself.
With ``--count``, benchmark bdiffs between delta parents and self for N
revisions starting at the specified revision.
With ``--alldata``, assume the requested revision is a changeset and
measure bdiffs for all changes related to that changeset (manifest
and filelogs).
"""
if opts['alldata']:
opts['changelog'] = True
if opts.get('changelog') or opts.get('manifest'):
file_, rev = None, file_
elif rev is None:
raise error.CommandError('perfbdiff', 'invalid arguments')
textpairs = []
r = cmdutil.openrevlog(repo, 'perfbdiff', file_, opts)
startrev = r.rev(r.lookup(rev))
for rev in range(startrev, min(startrev + count, len(r) - 1)):
if opts['alldata']:
# Load revisions associated with changeset.
ctx = repo[rev]
mtext = repo.manifestlog._revlog.revision(ctx.manifestnode())
for pctx in ctx.parents():
pman = repo.manifestlog._revlog.revision(pctx.manifestnode())
textpairs.append((pman, mtext))
# Load filelog revisions by iterating manifest delta.
man = ctx.manifest()
pman = ctx.p1().manifest()
for filename, change in pman.diff(man).items():
fctx = repo.file(filename)
f1 = fctx.revision(change[0][0] or -1)
f2 = fctx.revision(change[1][0] or -1)
textpairs.append((f1, f2))
else:
dp = r.deltaparent(rev)
textpairs.append((r.revision(dp), r.revision(rev)))
def d():
for pair in textpairs:
mdiff.textdiff(*pair)
timer, fm = gettimer(ui, opts)
timer(d)
fm.end()
def perfrevlogrevision(ui, repo, file_, rev=None, cache=None, **opts):
"""Benchmark obtaining a revlog revision.
Obtaining a revlog revision consists of roughly the following steps:
1. Compute the delta chain
2. Obtain the raw chunks for that delta chain
3. Decompress each raw chunk
4. Apply binary patches to obtain fulltext
5. Verify hash of fulltext
This command measures the time spent in each of these phases.
"""
if opts.get('changelog') or opts.get('manifest'):
file_, rev = None, file_
elif rev is None:
raise error.CommandError('perfrevlogrevision', 'invalid arguments')
r = cmdutil.openrevlog(repo, 'perfrevlogrevision', file_, opts)
node = r.lookup(rev)
rev = r.rev(node)
def dodeltachain(rev):
if not cache:
r.clearcaches()
r._deltachain(rev)
def doread(chain):
if not cache:
r.clearcaches()
r._chunkraw(chain[0], chain[-1])
def dodecompress(data, chain):
if not cache:
r.clearcaches()
start = r.start
length = r.length
inline = r._inline
iosize = r._io.size
buffer = util.buffer
offset = start(chain[0])
for rev in chain:
chunkstart = start(rev)
if inline:
chunkstart += (rev + 1) * iosize
chunklength = length(rev)
b = buffer(data, chunkstart - offset, chunklength)
revlog.decompress(b)
def dopatch(text, bins):
if not cache:
r.clearcaches()
mdiff.patches(text, bins)
def dohash(text):
if not cache:
r.clearcaches()
r._checkhash(text, node, rev)
def dorevision():
if not cache:
r.clearcaches()
r.revision(node)
chain = r._deltachain(rev)[0]
data = r._chunkraw(chain[0], chain[-1])[1]
bins = r._chunks(chain)
text = str(bins[0])
bins = bins[1:]
text = mdiff.patches(text, bins)
benches = [
(lambda: dorevision(), 'full'),
(lambda: dodeltachain(rev), 'deltachain'),
(lambda: doread(chain), 'read'),
(lambda: dodecompress(data, chain), 'decompress'),
(lambda: dopatch(text, bins), 'patch'),
(lambda: dohash(text), 'hash'),
]
for fn, title in benches:
timer, fm = gettimer(ui, opts)
timer(fn, title=title)
fm.end()
def d():
r = cmdutil.openrevlog(repo, 'perfrevlog', file_, opts)
for x in xrange(startrev, _len(r), dist):
r.revision(r.node(x))
def perfrevlogrevision(ui, repo, file_, rev=None, cache=None, **opts):
"""Benchmark obtaining a revlog revision.
Obtaining a revlog revision consists of roughly the following steps:
1. Compute the delta chain
2. Obtain the raw chunks for that delta chain
3. Decompress each raw chunk
4. Apply binary patches to obtain fulltext
5. Verify hash of fulltext
This command measures the time spent in each of these phases.
"""
if opts.get('changelog') or opts.get('manifest'):
file_, rev = None, file_
elif rev is None:
raise error.CommandError('perfrevlogrevision', 'invalid arguments')
r = cmdutil.openrevlog(repo, 'perfrevlogrevision', file_, opts)
node = r.lookup(rev)
rev = r.rev(node)
def dodeltachain(rev):
if not cache:
r.clearcaches()
r._deltachain(rev)
def doread(chain):
if not cache:
r.clearcaches()
r._chunkraw(chain[0], chain[-1])
def dodecompress(data, chain):
if not cache:
r.clearcaches()
start = r.start
length = r.length
inline = r._inline
iosize = r._io.size
buffer = util.buffer
offset = start(chain[0])
for rev in chain:
chunkstart = start(rev)
if inline:
chunkstart += (rev + 1) * iosize
chunklength = length(rev)
b = buffer(data, chunkstart - offset, chunklength)
revlog.decompress(b)
def dopatch(text, bins):
if not cache:
r.clearcaches()
mdiff.patches(text, bins)
def dohash(text):
if not cache:
r.clearcaches()
r._checkhash(text, node, rev)
def dorevision():
if not cache:
r.clearcaches()
r.revision(node)
chain = r._deltachain(rev)[0]
data = r._chunkraw(chain[0], chain[-1])[1]
bins = r._chunks(chain)
text = str(bins[0])
bins = bins[1:]
text = mdiff.patches(text, bins)
benches = [
(lambda: dorevision(), 'full'),
(lambda: dodeltachain(rev), 'deltachain'),
(lambda: doread(chain), 'read'),
(lambda: dodecompress(data, chain), 'decompress'),
(lambda: dopatch(text, bins), 'patch'),
(lambda: dohash(text), 'hash'),
]
for fn, title in benches:
timer, fm = gettimer(ui, opts)
timer(fn, title=title)
fm.end()
def d():
r = cmdutil.openrevlog(repo, 'perfrevlog', file_, opts)
for x in xrange(startrev, _len(r), dist):
r.revision(r.node(x))
def perfrevlogindex(ui, repo, file_=None, **opts):
"""Benchmark operations against a revlog index.
This tests constructing a revlog instance, reading index data,
parsing index data, and performing various operations related to
index data.
"""
rl = cmdutil.openrevlog(repo, 'perfrevlogindex', file_, opts)
opener = getattr(rl, 'opener') # trick linter
indexfile = rl.indexfile
data = opener.read(indexfile)
header = struct.unpack('>I', data[0:4])[0]
version = header & 0xFFFF
if version == 1:
revlogio = revlog.revlogio()
inline = header & (1 << 16)
else:
raise error.Abort(('unsupported revlog version: %d') % version)
rllen = len(rl)
node0 = rl.node(0)
node25 = rl.node(rllen // 4)
node50 = rl.node(rllen // 2)
node75 = rl.node(rllen // 4 * 3)
node100 = rl.node(rllen - 1)
allrevs = range(rllen)
allrevsrev = list(reversed(allrevs))
allnodes = [rl.node(rev) for rev in range(rllen)]
allnodesrev = list(reversed(allnodes))
def constructor():
revlog.revlog(opener, indexfile)
def read():
with opener(indexfile) as fh:
fh.read()
def parseindex():
revlogio.parseindex(data, inline)
def getentry(revornode):
index = revlogio.parseindex(data, inline)[0]
index[revornode]
def getentries(revs, count=1):
index = revlogio.parseindex(data, inline)[0]
for i in range(count):
for rev in revs:
index[rev]
def resolvenode(node):
nodemap = revlogio.parseindex(data, inline)[1]
# This only works for the C code.
if nodemap is None:
return
try:
nodemap[node]
except error.RevlogError:
pass
def resolvenodes(nodes, count=1):
nodemap = revlogio.parseindex(data, inline)[1]
if nodemap is None:
return
for i in range(count):
for node in nodes:
try:
nodemap[node]
except error.RevlogError:
pass
benches = [
(constructor, 'revlog constructor'),
(read, 'read'),
(parseindex, 'create index object'),
(lambda: getentry(0), 'retrieve index entry for rev 0'),
(lambda: resolvenode('a' * 20), 'look up missing node'),
(lambda: resolvenode(node0), 'look up node at rev 0'),
(lambda: resolvenode(node25), 'look up node at 1/4 len'),
(lambda: resolvenode(node50), 'look up node at 1/2 len'),
(lambda: resolvenode(node75), 'look up node at 3/4 len'),
(lambda: resolvenode(node100), 'look up node at tip'),
# 2x variation is to measure caching impact.
(lambda: resolvenodes(allnodes),
'look up all nodes (forward)'),
(lambda: resolvenodes(allnodes, 2),
'look up all nodes 2x (forward)'),
(lambda: resolvenodes(allnodesrev),
'look up all nodes (reverse)'),
(lambda: resolvenodes(allnodesrev, 2),
'look up all nodes 2x (reverse)'),
(lambda: getentries(allrevs),
'retrieve all index entries (forward)'),
(lambda: getentries(allrevs, 2),
'retrieve all index entries 2x (forward)'),
(lambda: getentries(allrevsrev),
'retrieve all index entries (reverse)'),
(lambda: getentries(allrevsrev, 2),
'retrieve all index entries 2x (reverse)'),
]
for fn, title in benches:
timer, fm = gettimer(ui, opts)
timer(fn, title=title)
fm.end()
def perfrevlogchunks(ui, repo, file_=None, engines=None, startrev=0, **opts):
"""Benchmark operations on revlog chunks.
Logically, each revlog is a collection of fulltext revisions. However,
stored within each revlog are "chunks" of possibly compressed data. This
data needs to be read and decompressed or compressed and written.
This command measures the time it takes to read+decompress and recompress
chunks in a revlog. It effectively isolates I/O and compression performance.
For measurements of higher-level operations like resolving revisions,
see ``perfrevlogrevisions`` and ``perfrevlogrevision``.
"""
rl = cmdutil.openrevlog(repo, 'perfrevlogchunks', file_, opts)
# _chunkraw was renamed to _getsegmentforrevs.
try:
segmentforrevs = rl._getsegmentforrevs
except AttributeError:
segmentforrevs = rl._chunkraw
# Verify engines argument.
if engines:
engines = set(e.strip() for e in engines.split(','))
for engine in engines:
try:
util.compressionengines[engine]
except KeyError:
raise error.Abort('unknown compression engine: %s' % engine)
else:
engines = []
for e in util.compengines:
engine = util.compengines[e]
try:
if engine.available():
engine.revlogcompressor().compress('dummy')
engines.append(e)
except NotImplementedError:
pass
revs = list(rl.revs(startrev, len(rl) - 1))
def rlfh(rl):
if rl._inline:
return getsvfs(repo)(rl.indexfile)
else:
return getsvfs(repo)(rl.datafile)
def doread():
rl.clearcaches()
for rev in revs:
segmentforrevs(rev, rev)
def doreadcachedfh():
rl.clearcaches()
fh = rlfh(rl)
for rev in revs:
segmentforrevs(rev, rev, df=fh)
def doreadbatch():
rl.clearcaches()
segmentforrevs(revs[0], revs[-1])
def doreadbatchcachedfh():
rl.clearcaches()
fh = rlfh(rl)
segmentforrevs(revs[0], revs[-1], df=fh)
def dochunk():
rl.clearcaches()
fh = rlfh(rl)
for rev in revs:
rl._chunk(rev, df=fh)
chunks = [None]
def dochunkbatch():
rl.clearcaches()
fh = rlfh(rl)
# Save chunks as a side-effect.
chunks[0] = rl._chunks(revs, df=fh)
def docompress(compressor):
rl.clearcaches()
try:
# Swap in the requested compression engine.
oldcompressor = rl._compressor
rl._compressor = compressor
for chunk in chunks[0]:
rl.compress(chunk)
finally:
rl._compressor = oldcompressor
benches = [
(lambda: doread(), 'read'),
(lambda: doreadcachedfh(), 'read w/ reused fd'),
(lambda: doreadbatch(), 'read batch'),
(lambda: doreadbatchcachedfh(), 'read batch w/ reused fd'),
(lambda: dochunk(), 'chunk'),
(lambda: dochunkbatch(), 'chunk batch'),
]
for engine in sorted(engines):
compressor = util.compengines[engine].revlogcompressor()
benches.append((functools.partial(docompress, compressor),
'compress w/ %s' % engine))
for fn, title in benches:
timer, fm = gettimer(ui, opts)
timer(fn, title=title)
fm.end()
