glibc/SHARED-FILES
Adhemerval Zanella baa495f231 math: Use erfcf from CORE-MATH
The CORE-MATH implementation is correctly rounded (for any rounding mode)
and shows better performance to the generic erfcf.

The code was adapted to glibc style and to use the definition of
math_config.h.

Benchtest on x64_64 (Ryzen 9 5900X, gcc 14.2.1), aarch64 (M1,
gcc 13.2.1), and powerpc (POWER10, gcc 13.2.1):

latency                       master       patched  improvement
x86_64                       98.8796       66.2142       33.04%
x86_64v2                     98.9617       67.4221       31.87%
x86_64v3                     87.4161       53.1754       39.17%
aarch64                      33.8336       22.0781       34.75%
power10                      21.1750       13.5864       35.84%
powerpc                      21.4694       13.8149       35.65%

reciprocal-throughput         master       patched  improvement
x86_64                       48.5620       27.6731       43.01%
x86_64v2                     47.9497       28.3804       40.81%
x86_64v3                     42.0255       18.1355       56.85%
aarch64                      24.3938       13.4041       45.05%
power10                      10.4919        6.1881       41.02%
powerpc                       11.763       6.76468       42.49%

Signed-off-by: Alexei Sibidanov <sibid@uvic.ca>
Signed-off-by: Paul Zimmermann <Paul.Zimmermann@inria.fr>
Signed-off-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
Reviewed-by: DJ Delorie <dj@redhat.com>
2024-11-22 10:52:27 -03:00

283 lines
9.2 KiB
Plaintext

# Files shared with other projects. Pass a file path to the
# get_glibc_shared_code() function in the python library
# scripts/glibc_shared_code.py to get a dict object with this information. See
# the library sources for more information.
# The headers on most of these files indicate that glibc is the canonical
# source for these files, although in many cases there seem to be useful
# changes in the gnulib versions that could be merged back in. Not all gnulib
# files contain such a header and it is not always consistent in its format, so
# it would be useful to make sure that all gnulib files that are using glibc as
# upstream have a greppable header.
#
# These files are quite hard to find without a header to grep for and each file
# has to be compared manually so this list is likely incomplete or may contain
# errors.
gnulib:
argp/argp-ba.c
argp/argp-ba.c
argp/argp-eexst.c
argp/argp-fmtstream.c
argp/argp-fmtstream.h
argp/argp-fs-xinl.c
argp/argp-help.c
argp/argp-namefrob.h
argp/argp-parse.c
argp/argp-pv.c
argp/argp-pvh.c
argp/argp-xinl.c
argp/argp.h
dirent/alphasort.c
dirent/scandir.c
# Merged from gnulib 2021-09-21
include/intprops.h
# Merged from gnulib 2021-09-21
include/regex.h
locale/programs/3level.h
# Merged from gnulib 2014-6-23
malloc/obstack.c
# Merged from gnulib 2014-6-23
malloc/obstack.h
# Merged from gnulib 2014-07-10
misc/error.c
misc/error.h
misc/getpass.c
misc/mkdtemp.c
# Merged from gnulib 2021-09-21
misc/sys/cdefs.h
posix/fnmatch_loop.c
# Intended to be the same. Gnulib copy contains glibc changes.
posix/getopt.c
# Intended to be the same. Gnulib copy contains glibc changes.
posix/getopt1.c
# Intended to be the same. Gnulib copy contains glibc changes.
posix/getopt_int.h
posix/glob.c
# Merged from gnulib 2021-09-21
posix/regcomp.c
# Merged from gnulib 2021-09-21
posix/regex.c
# Merged from gnulib 2021-09-21
posix/regex.h
# Merged from gnulib 2021-09-21
posix/regex_internal.c
# Merged from gnulib 2021-09-21
posix/regex_internal.h
# Merged from gnulib 2021-09-21
posix/regexec.c
posix/spawn.c
posix/spawn_faction_addclose.c
posix/spawn_faction_adddup2.c
posix/spawn_faction_addopen.c
posix/spawn_faction_destroy.c
posix/spawn_faction_init.c
posix/spawn_int.h
posix/spawnattr_destroy.c
posix/spawnattr_getdefault.c
posix/spawnattr_getflags.c
posix/spawnattr_getpgroup.c
posix/spawnattr_getschedparam.c
posix/spawnattr_getschedpolicy.c
posix/spawnattr_getsigmask.c
posix/spawnattr_init.c
posix/spawnattr_setdefault.c
posix/spawnattr_setflags.c
posix/spawnattr_setpgroup.c
posix/spawnattr_setschedparam.c
posix/spawnattr_setschedpolicy.c
posix/spawnattr_setsigmask.c
posix/spawnp.c
stdlib/atoll.c
stdlib/getsubopt.c
stdlib/setenv.c
stdlib/strtoll.c
stdlib/strtoul.c
# Merged from gnulib 2014-6-26, needs merge back
string/memchr.c
string/memcmp.c
string/memmem.c
string/mempcpy.c
string/memrchr.c
string/rawmemchr.c
string/stpcpy.c
string/stpncpy.c
string/str-two-way.h
string/strcasestr.c
string/strcspn.c
string/strdup.c
string/strndup.c
string/strpbrk.c
string/strsignal.c
string/strstr.c
string/strtok_r.c
string/strverscmp.c
# Merged from gnulib 2024-04-08 (gnulib commit 3238349628)
stdio-common/tmpdir.c
stdio-common/tmpdir.h
sysdeps/generic/pty-private.h
sysdeps/generic/siglist.h
sysdeps/posix/euidaccess.c
sysdeps/posix/gai_strerror.c
sysdeps/posix/getcwd.c
sysdeps/posix/pwrite.c
sysdeps/posix/spawni.c
# Merged from gnulib 2024-04-08 (gnulib commit 3238349628)
sysdeps/posix/tempname.c
# Merged from gnulib 2014-6-27
time/mktime.c
time/mktime-internal.h
time/strptime.c
time/timegm.c
# The last merge was 2014-12-11 and merged gettext 0.19.3 into glibc with a
# patch submitted to the gettext mailing list for changes that could be merged
# back.
#
# This commit was omitted from the merge as it does not appear to be compatible
# with how glibc expects things to work:
#
# commit 279b57fc367251666f00e8e2b599b83703451afb
# Author: Bruno Haible <bruno@clisp.org>
# Date: Fri Jun 14 12:03:49 2002 +0000
#
# Make absolute pathnames inside $LANGUAGE work.
gettext:
intl/bindtextdom.c
intl/dcgettext.c
intl/dcigettext.c
intl/dcngettext.c
intl/dgettext.c
intl/dngettext.c
intl/explodename.c
intl/finddomain.c
intl/gettext.c
intl/gettextP.h
intl/gmo.h
intl/hash-string.c
intl/hash-string.h
intl/l10nflist.c
intl/loadinfo.h
intl/loadmsgcat.c
intl/locale.alias
intl/localealias.c
intl/ngettext.c
intl/plural-exp.c
intl/plural-exp.h
intl/plural.y
intl/textdomain.c
# The following files are bundled from upstream Linux 6.10 for FUSE testing
# support.
linux:
support/bundled/linux/COPYING
support/bundled/linux/LICENSES
support/bundled/linux/LICENSES/exceptions/Linux-syscall-note
support/bundled/linux/LICENSES/preferred/GPL-2.0
support/bundled/linux/include/uapi/linux/fuse.h
# The following files are shared with the upstream Unicode project and must be
# updated regularly to stay in sync with the upstream unicode releases.
#
# Merged from Unicode 16.0.0 release.
unicode:
localedata/unicode-gen/UnicodeData.txt
localedata/unicode-gen/unicode-license.txt
localedata/unicode-gen/DerivedCoreProperties.txt
localedata/unicode-gen/EastAsianWidth.txt
localedata/unicode-gen/HangulSyllableType.txt
# The following files are shared with the upstream tzcode project and must be
# updated regularly to stay in sync with the upstream releases.
#
# Currently synced to TZDB 2024b, announced and distributed here:
# https://github.com/eggert/tz/releases/tag/2024b
tzcode:
timezone/private.h
timezone/tzfile.h
timezone/tzselect.ksh
timezone/version
timezone/zdump.c
timezone/zic.c
# The following files are shared with the upstream tzdata project but is not
# synchronized regularly. The data files themselves are used only for testing
# purposes and their data is never used to generate any output. We synchronize
# them only to stay on top of newer data that might help with testing.
#
# Currently synced to tzcode 2009i, announced and distributed here:
# https://mm.icann.org/pipermail/tz/2009-June/040697.html
# https://data.iana.org/time-zones/releases/tzdata2009i.tar.gz
tzdata:
timezone/africa
timezone/antarctica
timezone/asia
timezone/australasia
timezone/europe
timezone/northamerica
timezone/southamerica
timezone/pacificnew
timezone/etcetera
timezone/factory
timezone/backward
timezone/systemv
timezone/solar87
timezone/solar88
timezone/solar89
timezone/iso3166.tab
timezone/zone.tab
timezone/leapseconds
# This is yearistype.sh in the parent project
timezone/yearistype
# The following files are shared with the CORE-MATH project
sysdeps/ieee754/flt-32/e_gammaf_r.c
(file src/binary32/tgamma/tgammaf.c in CORE-MATH)
Instructions to merge new versions:
- change the function name from cr_tgammaf to __ieee754_gammaf_r
- add "int *signgamp" as 2nd argument and add at the beginning:
if (signgamp != NULL) *signgamp = 1;
- remove the errno stuff (this is done by the wrapper)
- replace 0x1p127f * 0x1p127f by math_narrow_eval (x * 0x1p127f)
- replace 0x1p-127f * sgn[k&1] by math_narrow_eval (0x1p-127f * sgn[k&1])
- add libm_alias_finite (__ieee754_gammaf_r, __gammaf_r) at the end
sysdeps/ieee754/flt-32/s_exp10m1f.c
(file src/binary32/exp10m1/exp10m1f.c in CORE-MATH)
- The code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.
sysdeps/ieee754/flt-32/s_exp2m1f.c
(file src/binary32/exp2m1/exp2m1f.c in CORE-MATH)
- The code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.
sysdeps/ieee754/flt-32/s_expm1f.c
(file src/binary32/expm1/expm1f.c in CORE-MATH)
- The code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.
sysdeps/ieee754/flt-32/e_log10f.c
(file src/binary32/log10/log10f.c in CORE-MATH)
- The code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.
sysdeps/ieee754/flt-32/s_log2p1f.c
(file src/binary32/log2p1/log2p1f.c in CORE-MATH)
- The code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.
sysdeps/ieee754/flt-32/s_log1pf.c
(file src/binary32/log1p/log1pf.c in CORE-MATH)
- The code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.
sysdeps/ieee754/flt-32/s_log10p1f.c
(file src/binary32/log10p1/log10p1f.c in CORE-MATH)
- The code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.
sysdeps/ieee754/flt-32/s_cbrtf.c
(file src/binary32/cbrt/cbrtf.c in CORE-MATH)
- The code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.
sysdeps/ieee754/flt-32/s_erff.c
(file src/binary32/erf/erff.c in CORE-MATH)
- The code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.
sysdeps/ieee754/flt-32/s_erfcf.c
(file src/binary32/erfc/erfcf.c in CORE-MATH)
- The code was adapted to use glibc code style and internal
functions to handle errno, overflow, and underflow.