Mercurial > hg
view contrib/python-zstandard/make_cffi.py @ 40021:c537144fdbef
wireprotov2: support response caching
One of the things I've learned from managing VCS servers over the
years is that they are hard to scale. It is well known that some
companies have very beefy (read: very expensive) servers to power
their VCS needs. It is also known that specialized servers for
various VCS exist in order to facilitate scaling servers. (Mercurial
is in this boat.)
One of the aspects that make a VCS server hard to scale is the
high CPU load incurred by constant client clone/pull operations.
To alleviate the scaling pain associated with data retrieval
operations, I want to integrate caching into the Mercurial wire
protocol server as robustly as possible such that servers can
aggressively cache responses and defer as much server load as
possible.
This commit represents the initial implementation of a general
caching layer in wire protocol version 2.
We define a new interface and behavior for a wire protocol cacher
in repository.py. (This is probably where a reviewer should look
first to understand what is going on.)
The bulk of the added code is in wireprotov2server.py, where we
define how a command can opt in to being cached and integrate
caching into command dispatching.
From a very high-level:
* A command can declare itself as cacheable by providing a callable
that can be used to derive a cache key.
* At dispatch time, if a command is cacheable, we attempt to
construct a cacher and use it for serving the request and/or
caching the request.
* The dispatch layer handles the bulk of the business logic for
caching, making cachers mostly "dumb content stores."
* The mechanism for invalidating cached entries (one of the harder
parts about caching in general) is by varying the cache key when
state changes. As such, cachers don't need to be concerned with
cache invalidation.
Initially, we've hooked up support for caching "manifestdata" and
"filedata" commands. These are the simplest to cache, as they should
be immutable over time. Caching of commands related to changeset
data is a bit harder (because cache validation is impacted by
changes to bookmarks, phases, etc). This will be implemented later.
(Strictly speaking, censoring a file should invalidate caches. I've
added an inline TODO to track this edge case.)
To prove it works, this commit implements a test-only extension
providing in-memory caching backed by an lrucachedict. A new test
showing this extension behaving properly is added. FWIW, the
cacher is ~50 lines of code, demonstrating the relative ease with
which a cache can be added to a server.
While the test cacher is not suitable for production workloads, just
for kicks I performed a clone of just the changeset and manifest data
for the mozilla-unified repository. With a fully warmed cache (of just
the manifest data since changeset data is not cached), server-side
CPU usage dropped from ~73s to ~28s. That's pretty significant and
demonstrates the potential that response caching has on server
scalability!
Differential Revision: https://phab.mercurial-scm.org/D4773
author | Gregory Szorc <gregory.szorc@gmail.com> |
---|---|
date | Wed, 26 Sep 2018 17:16:56 -0700 |
parents | b1fb341d8a61 |
children | 73fef626dae3 |
line wrap: on
line source
# Copyright (c) 2016-present, Gregory Szorc # All rights reserved. # # This software may be modified and distributed under the terms # of the BSD license. See the LICENSE file for details. from __future__ import absolute_import import cffi import distutils.ccompiler import os import re import subprocess import tempfile HERE = os.path.abspath(os.path.dirname(__file__)) SOURCES = ['zstd/%s' % p for p in ( 'common/entropy_common.c', 'common/error_private.c', 'common/fse_decompress.c', 'common/pool.c', 'common/threading.c', 'common/xxhash.c', 'common/zstd_common.c', 'compress/fse_compress.c', 'compress/huf_compress.c', 'compress/zstd_compress.c', 'compress/zstd_double_fast.c', 'compress/zstd_fast.c', 'compress/zstd_lazy.c', 'compress/zstd_ldm.c', 'compress/zstd_opt.c', 'compress/zstdmt_compress.c', 'decompress/huf_decompress.c', 'decompress/zstd_decompress.c', 'dictBuilder/cover.c', 'dictBuilder/divsufsort.c', 'dictBuilder/zdict.c', )] # Headers whose preprocessed output will be fed into cdef(). HEADERS = [os.path.join(HERE, 'zstd', *p) for p in ( ('zstd.h',), ('dictBuilder', 'zdict.h'), )] INCLUDE_DIRS = [os.path.join(HERE, d) for d in ( 'zstd', 'zstd/common', 'zstd/compress', 'zstd/decompress', 'zstd/dictBuilder', )] # cffi can't parse some of the primitives in zstd.h. So we invoke the # preprocessor and feed its output into cffi. compiler = distutils.ccompiler.new_compiler() # Needed for MSVC. if hasattr(compiler, 'initialize'): compiler.initialize() # Distutils doesn't set compiler.preprocessor, so invoke the preprocessor # manually. if compiler.compiler_type == 'unix': args = list(compiler.executables['compiler']) args.extend([ '-E', '-DZSTD_STATIC_LINKING_ONLY', '-DZDICT_STATIC_LINKING_ONLY', ]) elif compiler.compiler_type == 'msvc': args = [compiler.cc] args.extend([ '/EP', '/DZSTD_STATIC_LINKING_ONLY', '/DZDICT_STATIC_LINKING_ONLY', ]) else: raise Exception('unsupported compiler type: %s' % compiler.compiler_type) def preprocess(path): with open(path, 'rb') as fh: lines = [] it = iter(fh) for l in it: # zstd.h includes <stddef.h>, which is also included by cffi's # boilerplate. This can lead to duplicate declarations. So we strip # this include from the preprocessor invocation. # # The same things happens for including zstd.h, so give it the same # treatment. # # We define ZSTD_STATIC_LINKING_ONLY, which is redundant with the inline # #define in zstdmt_compress.h and results in a compiler warning. So drop # the inline #define. if l.startswith((b'#include <stddef.h>', b'#include "zstd.h"', b'#define ZSTD_STATIC_LINKING_ONLY')): continue # ZSTDLIB_API may not be defined if we dropped zstd.h. It isn't # important so just filter it out. if l.startswith(b'ZSTDLIB_API'): l = l[len(b'ZSTDLIB_API '):] lines.append(l) fd, input_file = tempfile.mkstemp(suffix='.h') os.write(fd, b''.join(lines)) os.close(fd) try: process = subprocess.Popen(args + [input_file], stdout=subprocess.PIPE) output = process.communicate()[0] ret = process.poll() if ret: raise Exception('preprocessor exited with error') return output finally: os.unlink(input_file) def normalize_output(output): lines = [] for line in output.splitlines(): # CFFI's parser doesn't like __attribute__ on UNIX compilers. if line.startswith(b'__attribute__ ((visibility ("default"))) '): line = line[len(b'__attribute__ ((visibility ("default"))) '):] if line.startswith(b'__attribute__((deprecated('): continue elif b'__declspec(deprecated(' in line: continue lines.append(line) return b'\n'.join(lines) ffi = cffi.FFI() # zstd.h uses a possible undefined MIN(). Define it until # https://github.com/facebook/zstd/issues/976 is fixed. # *_DISABLE_DEPRECATE_WARNINGS prevents the compiler from emitting a warning # when cffi uses the function. Since we statically link against zstd, even # if we use the deprecated functions it shouldn't be a huge problem. ffi.set_source('_zstd_cffi', ''' #define MIN(a,b) ((a)<(b) ? (a) : (b)) #define ZSTD_STATIC_LINKING_ONLY #include <zstd.h> #define ZDICT_STATIC_LINKING_ONLY #define ZDICT_DISABLE_DEPRECATE_WARNINGS #include <zdict.h> ''', sources=SOURCES, include_dirs=INCLUDE_DIRS, extra_compile_args=['-DZSTD_MULTITHREAD']) DEFINE = re.compile(b'^\\#define ([a-zA-Z0-9_]+) ') sources = [] # Feed normalized preprocessor output for headers into the cdef parser. for header in HEADERS: preprocessed = preprocess(header) sources.append(normalize_output(preprocessed)) # #define's are effectively erased as part of going through preprocessor. # So perform a manual pass to re-add those to the cdef source. with open(header, 'rb') as fh: for line in fh: line = line.strip() m = DEFINE.match(line) if not m: continue if m.group(1) == b'ZSTD_STATIC_LINKING_ONLY': continue # The parser doesn't like some constants with complex values. if m.group(1) in (b'ZSTD_LIB_VERSION', b'ZSTD_VERSION_STRING'): continue # The ... is magic syntax by the cdef parser to resolve the # value at compile time. sources.append(m.group(0) + b' ...') cdeflines = b'\n'.join(sources).splitlines() cdeflines = [l for l in cdeflines if l.strip()] ffi.cdef(b'\n'.join(cdeflines).decode('latin1')) if __name__ == '__main__': ffi.compile()