Mercurial > hg
view mercurial/dirs.c @ 30442:41a8106789ca
util: implement zstd compression engine
Now that zstd is vendored and being built (in some configurations), we
can implement a compression engine for zstd!
The zstd engine is a little different from existing engines. Because
it may not always be present, we have to defer load the module in case
importing it fails. We facilitate this via a cached property that holds
a reference to the module or None. The "available" method is
implemented to reflect reality.
The zstd engine declares its ability to handle bundles using the
"zstd" human name and the "ZS" internal name. The latter was chosen
because internal names are 2 characters (by only convention I think)
and "ZS" seems reasonable.
The engine, like others, supports specifying the compression level.
However, there are no consumers of this API that yet pass in that
argument. I have plans to change that, so stay tuned.
Since all we need to do to support bundle generation with a new
compression engine is implement and register the compression engine,
bundle generation with zstd "just works!" Tests demonstrating this
have been added.
How does performance of zstd for bundle generation compare? On the
mozilla-unified repo, `hg bundle --all -t <engine>-v2` yields the
following on my i7-6700K on Linux:
engine CPU time bundle size vs orig size throughput
none 97.0s 4,054,405,584 100.0% 41.8 MB/s
bzip2 (l=9) 393.6s 975,343,098 24.0% 10.3 MB/s
gzip (l=6) 184.0s 1,140,533,074 28.1% 22.0 MB/s
zstd (l=1) 108.2s 1,119,434,718 27.6% 37.5 MB/s
zstd (l=2) 111.3s 1,078,328,002 26.6% 36.4 MB/s
zstd (l=3) 113.7s 1,011,823,727 25.0% 35.7 MB/s
zstd (l=4) 116.0s 1,008,965,888 24.9% 35.0 MB/s
zstd (l=5) 121.0s 977,203,148 24.1% 33.5 MB/s
zstd (l=6) 131.7s 927,360,198 22.9% 30.8 MB/s
zstd (l=7) 139.0s 912,808,505 22.5% 29.2 MB/s
zstd (l=12) 198.1s 854,527,714 21.1% 20.5 MB/s
zstd (l=18) 681.6s 789,750,690 19.5% 5.9 MB/s
On compression, zstd for bundle generation delivers:
* better compression than gzip with significantly less CPU utilization
* better than bzip2 compression ratios while still being significantly
faster than gzip
* ability to aggressively tune compression level to achieve
significantly smaller bundles
That last point is important. With clone bundles, a server can
pre-generate a bundle file, upload it to a static file server, and
redirect clients to transparently download it during clone. The server
could choose to produce a zstd bundle with the highest compression
settings possible. This would take a very long time - a magnitude
longer than a typical zstd bundle generation - but the result would
be hundreds of megabytes smaller! For the clone volume we do at
Mozilla, this could translate to petabytes of bandwidth savings
per year and faster clones (due to smaller transfer size).
I don't have detailed numbers to report on decompression. However,
zstd decompression is fast: >1 GB/s output throughput on this machine,
even through the Python bindings. And it can do that regardless of the
compression level of the input. By the time you have enough data to
worry about overhead of decompression, you have plenty of other things
to worry about performance wise.
zstd is wins all around. I can't wait to implement support for it
on the wire protocol and in revlogs.
author | Gregory Szorc <gregory.szorc@gmail.com> |
---|---|
date | Fri, 11 Nov 2016 01:10:07 -0800 |
parents | 1e5ff5ae1d2b |
children |
line wrap: on
line source
/* dirs.c - dynamic directory diddling for dirstates Copyright 2013 Facebook This software may be used and distributed according to the terms of the GNU General Public License, incorporated herein by reference. */ #define PY_SSIZE_T_CLEAN #include <Python.h> #include "util.h" #ifdef IS_PY3K #define PYLONG_VALUE(o) ((PyLongObject *)o)->ob_digit[1] #else #define PYLONG_VALUE(o) PyInt_AS_LONG(o) #endif /* * This is a multiset of directory names, built from the files that * appear in a dirstate or manifest. * * A few implementation notes: * * We modify Python integers for refcounting, but those integers are * never visible to Python code. * * We mutate strings in-place, but leave them immutable once they can * be seen by Python code. */ typedef struct { PyObject_HEAD PyObject *dict; } dirsObject; static inline Py_ssize_t _finddir(const char *path, Py_ssize_t pos) { while (pos != -1) { if (path[pos] == '/') break; pos -= 1; } return pos; } static int _addpath(PyObject *dirs, PyObject *path) { const char *cpath = PyBytes_AS_STRING(path); Py_ssize_t pos = PyBytes_GET_SIZE(path); PyObject *key = NULL; int ret = -1; /* This loop is super critical for performance. That's why we inline * access to Python structs instead of going through a supported API. * The implementation, therefore, is heavily dependent on CPython * implementation details. We also commit violations of the Python * "protocol" such as mutating immutable objects. But since we only * mutate objects created in this function or in other well-defined * locations, the references are known so these violations should go * unnoticed. The code for adjusting the length of a PyBytesObject is * essentially a minimal version of _PyBytes_Resize. */ while ((pos = _finddir(cpath, pos - 1)) != -1) { PyObject *val; /* It's likely that every prefix already has an entry in our dict. Try to avoid allocating and deallocating a string for each prefix we check. */ if (key != NULL) ((PyBytesObject *)key)->ob_shash = -1; else { /* Force Python to not reuse a small shared string. */ key = PyBytes_FromStringAndSize(cpath, pos < 2 ? 2 : pos); if (key == NULL) goto bail; } /* Py_SIZE(o) refers to the ob_size member of the struct. Yes, * assigning to what looks like a function seems wrong. */ Py_SIZE(key) = pos; ((PyBytesObject *)key)->ob_sval[pos] = '\0'; val = PyDict_GetItem(dirs, key); if (val != NULL) { PYLONG_VALUE(val) += 1; break; } /* Force Python to not reuse a small shared int. */ #ifdef IS_PY3K val = PyLong_FromLong(0x1eadbeef); #else val = PyInt_FromLong(0x1eadbeef); #endif if (val == NULL) goto bail; PYLONG_VALUE(val) = 1; ret = PyDict_SetItem(dirs, key, val); Py_DECREF(val); if (ret == -1) goto bail; Py_CLEAR(key); } ret = 0; bail: Py_XDECREF(key); return ret; } static int _delpath(PyObject *dirs, PyObject *path) { char *cpath = PyBytes_AS_STRING(path); Py_ssize_t pos = PyBytes_GET_SIZE(path); PyObject *key = NULL; int ret = -1; while ((pos = _finddir(cpath, pos - 1)) != -1) { PyObject *val; key = PyBytes_FromStringAndSize(cpath, pos); if (key == NULL) goto bail; val = PyDict_GetItem(dirs, key); if (val == NULL) { PyErr_SetString(PyExc_ValueError, "expected a value, found none"); goto bail; } if (--PYLONG_VALUE(val) <= 0) { if (PyDict_DelItem(dirs, key) == -1) goto bail; } else break; Py_CLEAR(key); } ret = 0; bail: Py_XDECREF(key); return ret; } static int dirs_fromdict(PyObject *dirs, PyObject *source, char skipchar) { PyObject *key, *value; Py_ssize_t pos = 0; while (PyDict_Next(source, &pos, &key, &value)) { if (!PyBytes_Check(key)) { PyErr_SetString(PyExc_TypeError, "expected string key"); return -1; } if (skipchar) { if (!dirstate_tuple_check(value)) { PyErr_SetString(PyExc_TypeError, "expected a dirstate tuple"); return -1; } if (((dirstateTupleObject *)value)->state == skipchar) continue; } if (_addpath(dirs, key) == -1) return -1; } return 0; } static int dirs_fromiter(PyObject *dirs, PyObject *source) { PyObject *iter, *item = NULL; int ret; iter = PyObject_GetIter(source); if (iter == NULL) return -1; while ((item = PyIter_Next(iter)) != NULL) { if (!PyBytes_Check(item)) { PyErr_SetString(PyExc_TypeError, "expected string"); break; } if (_addpath(dirs, item) == -1) break; Py_CLEAR(item); } ret = PyErr_Occurred() ? -1 : 0; Py_DECREF(iter); Py_XDECREF(item); return ret; } /* * Calculate a refcounted set of directory names for the files in a * dirstate. */ static int dirs_init(dirsObject *self, PyObject *args) { PyObject *dirs = NULL, *source = NULL; char skipchar = 0; int ret = -1; self->dict = NULL; if (!PyArg_ParseTuple(args, "|Oc:__init__", &source, &skipchar)) return -1; dirs = PyDict_New(); if (dirs == NULL) return -1; if (source == NULL) ret = 0; else if (PyDict_Check(source)) ret = dirs_fromdict(dirs, source, skipchar); else if (skipchar) PyErr_SetString(PyExc_ValueError, "skip character is only supported " "with a dict source"); else ret = dirs_fromiter(dirs, source); if (ret == -1) Py_XDECREF(dirs); else self->dict = dirs; return ret; } PyObject *dirs_addpath(dirsObject *self, PyObject *args) { PyObject *path; if (!PyArg_ParseTuple(args, "O!:addpath", &PyBytes_Type, &path)) return NULL; if (_addpath(self->dict, path) == -1) return NULL; Py_RETURN_NONE; } static PyObject *dirs_delpath(dirsObject *self, PyObject *args) { PyObject *path; if (!PyArg_ParseTuple(args, "O!:delpath", &PyBytes_Type, &path)) return NULL; if (_delpath(self->dict, path) == -1) return NULL; Py_RETURN_NONE; } static int dirs_contains(dirsObject *self, PyObject *value) { return PyBytes_Check(value) ? PyDict_Contains(self->dict, value) : 0; } static void dirs_dealloc(dirsObject *self) { Py_XDECREF(self->dict); PyObject_Del(self); } static PyObject *dirs_iter(dirsObject *self) { return PyObject_GetIter(self->dict); } static PySequenceMethods dirs_sequence_methods; static PyMethodDef dirs_methods[] = { {"addpath", (PyCFunction)dirs_addpath, METH_VARARGS, "add a path"}, {"delpath", (PyCFunction)dirs_delpath, METH_VARARGS, "remove a path"}, {NULL} /* Sentinel */ }; static PyTypeObject dirsType = { PyVarObject_HEAD_INIT(NULL, 0) }; void dirs_module_init(PyObject *mod) { dirs_sequence_methods.sq_contains = (objobjproc)dirs_contains; dirsType.tp_name = "parsers.dirs"; dirsType.tp_new = PyType_GenericNew; dirsType.tp_basicsize = sizeof(dirsObject); dirsType.tp_dealloc = (destructor)dirs_dealloc; dirsType.tp_as_sequence = &dirs_sequence_methods; dirsType.tp_flags = Py_TPFLAGS_DEFAULT; dirsType.tp_doc = "dirs"; dirsType.tp_iter = (getiterfunc)dirs_iter; dirsType.tp_methods = dirs_methods; dirsType.tp_init = (initproc)dirs_init; if (PyType_Ready(&dirsType) < 0) return; Py_INCREF(&dirsType); PyModule_AddObject(mod, "dirs", (PyObject *)&dirsType); }