1
0
mirror of https://github.com/postgres/postgres.git synced 2025-09-06 13:46:51 +03:00
Files
postgres/src/makefiles/meson.build
Nathan Bossart 792752af4e Optimize pg_popcount() with AVX-512 instructions.
Presently, pg_popcount() processes data in 32-bit or 64-bit chunks
when possible.  Newer hardware that supports AVX-512 instructions
can use 512-bit chunks, which provides a nice speedup, especially
for larger buffers.  This commit introduces the infrastructure
required to detect compiler and CPU support for the required
AVX-512 intrinsic functions, and it adds a new pg_popcount()
implementation that uses these functions.  If CPU support for this
optimized implementation is detected at runtime, a function pointer
is updated so that it is used by subsequent calls to pg_popcount().

Most of the existing in-tree calls to pg_popcount() should benefit
from these instructions, and calls with smaller buffers should at
least not regress compared to v16.  The new infrastructure
introduced by this commit can also be used to optimize
visibilitymap_count(), but that is left for a follow-up commit.

Co-authored-by: Paul Amonson, Ants Aasma
Reviewed-by: Matthias van de Meent, Tom Lane, Noah Misch, Akash Shankaran, Alvaro Herrera, Andres Freund, David Rowley
Discussion: https://postgr.es/m/BL1PR11MB5304097DF7EA81D04C33F3D1DCA6A%40BL1PR11MB5304.namprd11.prod.outlook.com
2024-04-06 21:56:23 -05:00

265 lines
6.8 KiB
Meson

# Copyright (c) 2022-2024, PostgreSQL Global Development Group
### Compute pgxs_data, used in src/meson.build to generate Makefile.global
### etc, that's complete enough for PGXS to work.
# Emulation of PGAC_CHECK_STRIP
strip_bin = find_program(get_option('STRIP'), required: false, native: true)
strip_cmd = strip_bin.found() ? [strip_bin.path()] : [':']
working_strip = false
if strip_bin.found()
strip_version = run_command(strip_bin, '-V', check: false)
if strip_version.returncode() == 0 and (
strip_version.stdout().contains('GNU strip') or
strip_version.stderr().contains('GNU strip'))
working_strip = true
strip_static_cmd = strip_cmd + ['--strip-unneeded']
strip_shared_cmd = strip_cmd + ['--strip-unneeded']
elif host_system == 'darwin'
working_strip = true
strip_static_cmd = strip_cmd + ['-x']
strip_shared_cmd = strip_cmd + ['-x']
endif
endif
if not working_strip
strip_cmd = [':']
strip_static_cmd = [':']
strip_shared_cmd = [':']
endif
pgxs_kv = {
'PACKAGE_URL': pg_url,
'PACKAGE_VERSION': pg_version,
'PG_MAJORVERSION': pg_version_major,
'PG_VERSION_NUM': pg_version_num,
'configure_input': 'meson',
'vpath_build': 'yes',
'autodepend': cc.get_argument_syntax() == 'gcc' ? 'yes' : 'no',
'host_cpu': host_cpu,
'host': '@0@-@1@'.format(host_cpu, host_system),
'host_os': host_system,
'build_os': build_machine.system(),
'PORTNAME': portname,
'PG_SYSROOT': pg_sysroot,
'abs_top_builddir': meson.build_root(),
'abs_top_srcdir': meson.source_root(),
'enable_rpath': get_option('rpath') ? 'yes' : 'no',
'enable_nls': libintl.found() ? 'yes' : 'no',
'enable_injection_points': get_option('injection_points') ? 'yes' : 'no',
'enable_tap_tests': tap_tests_enabled ? 'yes' : 'no',
'enable_debug': get_option('debug') ? 'yes' : 'no',
'enable_coverage': 'no',
'enable_dtrace': dtrace.found() ? 'yes' : 'no',
'DLSUFFIX': dlsuffix,
'EXEEXT': exesuffix,
'SUN_STUDIO_CC': 'no', # not supported so far
# want the chosen option, rather than the library
'with_ssl' : ssl_library,
'with_uuid': uuidopt,
'default_port': get_option('pgport'),
'with_system_tzdata': get_option('system_tzdata'),
'with_krb_srvnam': get_option('krb_srvnam'),
'krb_srvtab': krb_srvtab,
'STRIP': ' '.join(strip_cmd),
'STRIP_STATIC_LIB': ' '.join(strip_static_cmd),
'STRIP_SHARED_LIB': ' '.join(strip_shared_cmd),
# these seem to be standard these days
'MKDIR_P': 'mkdir -p',
'LN_S': 'ln -s',
# Just always use the install_sh fallback that autoconf uses. Unlikely to
# matter performance-wise for extensions. If it turns out to do, we can
'install_bin': '$(SHELL) $(top_srcdir)/config/install-sh -c',
'CC': var_cc,
'CPP': var_cpp,
'GCC': cc.get_argument_syntax() == 'gcc' ? 'yes' : 'no',
'CPPFLAGS': var_cppflags,
'CFLAGS': var_cflags,
'CXXFLAGS': var_cxxflags,
'CFLAGS_SL': var_cflags_sl,
'CFLAGS_SL_MODULE': ' '.join(cflags_mod),
'CXXFLAGS_SL_MODULE': ' '.join(cxxflags_mod),
'PERMIT_DECLARATION_AFTER_STATEMENT':
' '.join(cflags_no_decl_after_statement),
'CFLAGS_CRC': ' '.join(cflags_crc),
'CFLAGS_POPCNT': ' '.join(cflags_popcnt),
'CFLAGS_UNROLL_LOOPS': ' '.join(unroll_loops_cflags),
'CFLAGS_VECTORIZE': ' '.join(vectorize_cflags),
'CFLAGS_XSAVE': ' '.join(cflags_xsave),
'LDFLAGS': var_ldflags,
'LDFLAGS_EX': var_ldflags_ex,
'LDFLAGS_EX_BE':
' '.join(cc.get_supported_link_arguments('-Wl,--export-dynamic')),
'LDFLAGS_SL': var_ldflags_sl,
# TODO: requires bitcode generation to be implemented for meson
'BITCODE_CFLAGS': '',
'BITCODE_CXXFLAGS': '',
'BISONFLAGS': ' '.join(bison_flags),
'FLEXFLAGS': ' '.join(flex_flags),
'LIBS': var_libs,
}
if llvm.found()
pgxs_kv += {
'CLANG': clang.path(),
'CXX': ' '.join(cpp.cmd_array()),
'LLVM_BINPATH': llvm_binpath,
}
else
pgxs_kv += {
'CLANG': '',
'CXX': '',
'LLVM_BINPATH': '',
}
endif
pgxs_bins = {
'AR':
find_program(['ar'], native: true, required: false),
'AWK':
find_program(['gawk', 'mawk', 'nawk', 'awk'], native: true, required: false),
'BISON': bison,
'FLEX': flex,
'GZIP': gzip,
'LZ4': program_lz4,
'OPENSSL': openssl,
'PERL': perl,
'PROVE': prove,
'PYTHON': python,
'TAR': tar,
'ZSTD': program_zstd,
'DTRACE': dtrace,
}
pgxs_empty = [
'ICU_CFLAGS', # needs to be added, included by public server headers
# hard to see why we'd need either?
'ZIC',
'TCLSH',
# docs don't seem to be supported by pgxs
'XMLLINT',
'XSLTPROC',
'DBTOEPUB',
'FOP',
# supporting coverage for pgxs-in-meson build doesn't seem worth it
'GENHTML',
'LCOV',
'GCOV',
'MSGFMT_FLAGS',
# translation doesn't appear to be supported by pgxs
'MSGFMT',
'XGETTEXT',
'MSGMERGE',
'WANTED_LANGUAGES',
# Not needed because we don't build the server / PLs with the generated makefile
'LIBOBJS', 'PG_CRC32C_OBJS', 'PG_POPCNT_OBJS', 'TAS',
'DTRACEFLAGS', # only server has dtrace probes
'perl_archlibexp', 'perl_embed_ccflags', 'perl_embed_ldflags', 'perl_includespec', 'perl_privlibexp',
'python_additional_libs', 'python_includespec', 'python_libdir', 'python_libspec', 'python_majorversion', 'python_version',
# possible that some of these are referenced explicitly in pgxs makefiles?
# For now not worth it.
'TCL_INCLUDE_SPEC', 'TCL_LIBS', 'TCL_LIB_SPEC', 'TCL_SHARED_BUILD',
'LLVM_CFLAGS', 'LLVM_CPPFLAGS', 'LLVM_CXXFLAGS', 'LLVM_LIBS',
'LDAP_LIBS_BE', 'LDAP_LIBS_FE',
'UUID_LIBS',
'PTHREAD_CFLAGS', 'PTHREAD_LIBS',
'ICU_LIBS',
]
if host_system == 'windows' and cc.get_argument_syntax() != 'msvc'
pgxs_bins += {'WINDRES': windres}
else
pgxs_empty += 'WINDRES'
endif
pgxs_dirs = {
'prefix': get_option('prefix'),
'bindir': '${exec_prefix}' / get_option('bindir'),
'datarootdir': '${prefix}' / get_option('datadir'),
'datadir': '${datarootdir}',
'docdir': '${prefix}' / dir_doc,
'exec_prefix': '${prefix}',
'htmldir': '${docdir}',
'includedir': '${prefix}' / get_option('includedir'),
'libdir': '${exec_prefix}' / get_option('libdir'),
'localedir': '${prefix}' / get_option('localedir'),
'mandir': '${prefix}' / get_option('mandir'),
'sysconfdir': '${prefix}' / get_option('sysconfdir'),
}
pgxs_deps = {
'bonjour': bonjour,
'bsd_auth': bsd_auth,
'gssapi': gssapi,
'icu': icu,
'ldap': ldap,
'libxml': libxml,
'libxslt': libxslt,
'llvm': llvm,
'lz4': lz4,
'nls': libintl,
'pam': pam,
'perl': perl_dep,
'python': python3_dep,
'readline': readline,
'selinux': selinux,
'systemd': systemd,
'tcl': tcl_dep,
'zlib': zlib,
'zstd': zstd,
}
pgxs_cdata = configuration_data(pgxs_kv)
foreach b, p : pgxs_bins
pgxs_cdata.set(b, p.found() ? p.path() : '')
endforeach
foreach pe : pgxs_empty
pgxs_cdata.set(pe, '')
endforeach
foreach d, p : pgxs_dirs
pgxs_cdata.set(d, p)
endforeach
foreach d, v : pgxs_deps
pgxs_cdata.set('with_@0@'.format(d), v.found() ? 'yes' : 'no')
endforeach