glibc/soft-fp/op-common.h
Ulrich Drepper 11bf311edc [BZ #2510, BZ #2830, BZ #3137, BZ #3313, BZ #3426, BZ #3465, BZ #3480, BZ #3483, BZ #3493, BZ #3514, BZ #3515, BZ #3664, BZ #3673, BZ #3674]
2007-01-11  Jakub Jelinek  <jakub@redhat.com>
	* sysdeps/i386/soft-fp/sfp-machine.h: Remove.
	* sysdeps/x86_64/soft-fp/sfp-machine.h: Likewise.
2007-01-10  Ulrich Drepper  <drepper@redhat.com>
	* io/fts.c: Make sure fts_cur is always valid after return from
	fts_read.
	Patch by Miloslav Trmac <mitr@redhat.com>.
2006-10-27  Richard Sandiford  <richard@codesourcery.com>
	* elf/elf.h (R_MIPS_GLOB_DAT): Define.
	(R_MIPS_NUM): Bump by 1.
2007-01-03  Jakub Jelinek  <jakub@redhat.com>
	* posix/execvp.c: Include alloca.h.
	(allocate_scripts_argv): Renamed to...
	(scripts_argv): ... this.  Don't allocate buffer here nor count
	arguments.
	(execvp): Use alloca if possible.
	* posix/Makefile: Add rules to build and run tst-vfork3 test.
	* posix/tst-vfork3.c: New test.
	* stdlib/Makefile (tst-strtod3-ENV): Define.
2007-01-02  Ulrich Drepper  <drepper@redhat.com>
	* posix/getconf.c: Update copyright year.
	* nss/getent.c: Likewise.
	* iconv/iconvconfig.c: Likewise.
	* iconv/iconv_prog.c: Likewise.
	* elf/ldconfig.c: Likewise.
	* catgets/gencat.c: Likewise.
	* csu/version.c: Likewise.
	* elf/ldd.bash.in: Likewise.
	* elf/sprof.c (print_version): Likewise.
	* locale/programs/locale.c: Likewise.
	* locale/programs/localedef.c: Likewise.
	* nscd/nscd.c (print_version): Likewise.
	* debug/xtrace.sh: Likewise.
	* malloc/memusage.sh: Likewise.
	* malloc/mtrace.pl: Likewise.
	* debug/catchsegv.sh: Likewise.
2006-12-24  Ulrich Drepper  <drepper@redhat.com>
	* malloc/malloc.c (sYSMALLOc): Remove some unnecessary alignment
	attempts.
2006-12-23  Ulrich Drepper  <drepper@redhat.com>
	* posix/wordexp.c: Remove some unnecessary tests.
2006-12-20  SUGIOKA Toshinobu  <sugioka@itonet.co.jp>

	* sysdeps/unix/sysv/linux/sh/bits/shm.h: New file.

	* nss/getXXbyYY_r.c: Include atomic.h.
	(INTERNAL (REENTRANT_NAME)): Write startp after start_fct,
	add atomic_write_barrier () in between.

2006-11-28  Jakub Jelinek  <jakub@redhat.com>
	* elf/dl-support.c: Include dl-procinfo.h.
	* sysdeps/powerpc/dl-procinfo.h (PPC_PLATFORM_POWER4,
	PPC_PLATFORM_PPC970, PPC_PLATFORM_POWER5, PPC_PLATFORM_POWER5_PLUS,
	PPC_PLATFORM_POWER6, PPC_PLATFORM_CELL_BE, PPC_PLATFORM_POWER6X):
	Define.
	(_dl_string_platform): Use PPC_PLATFORM_* macros instead of
	hardcoded constants.
	* sysdeps/powerpc/dl-procinfo.c (_dl_powerpc_platform): Use
	PPC_PLATFORM_* macros for array designators.
2006-11-11  Steven Munroe  <sjmunroe@us.ibm.com>
	* sysdeps/powerpc/dl-procinfo.c (_dl_powerpc_cap_flags): Add 3 new cap
	names to the beginning.
	(_dl_powerpc_platforms): Add "power6x".
	* sysdeps/powerpc/dl-procinfo.h (_DL_HWCAP_FIRST): Decrease.
	(HWCAP_IMPORTANT): Add PPC_FEATURE_HAS_DFP.
	(_DL_PLATFORMS_COUNT): Increase.
	(_dl_string_platform): Handle power6x case.
	* sysdeps/powerpc/sysdep.h (PPC_FEATURE_PA6T, PPC_FEATURE_HAS_DFP,
	PPC_FEATURE_POWER6_EXT): Define.
	(PPC_FEATURE_POWER5, PPC_FEATURE_POWER5_PLUS): Correct Comment.
	[-2^31 .. 2^31) range.
	* sysdeps/unix/sysv/linux/bits/statvfs.h: Define ST_RELATIME.
	* sysdeps/unix/sysv/linux/internal_statvfs.c (__statvfs_getflags):
	Handle relatime mount option.

2006-12-13  Jakub Jelinek  <jakub@redhat.com>
	* sysdeps/unix/sysv/linux/powerpc/powerpc32/setcontext.S: Include
	kernel-features.h.

2006-12-11  Ulrich Drepper  <drepper@redhat.com>

	* stdlib/strtod_l.c (____STRTOF_INTERNAL): Parse thousand
	separators also if no non-zero digits found.
	* stdlib/Makefile (tests): Add tst-strtod3.
	[BZ #3664]
	* stdlib/strtod_l.c (____STRTOF_INTERNAL): Fix test to recognize
	empty parsed strings.
	* stdlib/Makefile (tests): Add tst-strtod2.
	* stdlib/tst-strtod2.c: New file.

	[BZ #3673]
	* stdlib/strtod_l.c (____STRTOF_INTERNAL): Fix exp_limit
	computation.
	* stdlib/Makefile (tests): Add tst-atof2.
	* stdlib/tst-atof2.c: New file.

	[BZ #3674]
	* stdlib/strtod_l.c (____STRTOF_INTERNAL): Adjust exponent value
	correctly if removing trailing zero of hex-float.
	* stdlib/Makefile (tests): Add tst-atof1.
	* stdlib/tst-atof1.c: New file.

	* misc/mntent_r.c (__hasmntopt): Check p[optlen] even when p == rest.
	Start searching for next comma at p rather than rest.
	* misc/Makefile (tests): Add tst-mntent2.
	* misc/tst-mntent2.c: New test.

2006-12-08  Ulrich Drepper  <drepper@redhat.com>
	* malloc/memusage.c: Handle realloc with new size of zero and
	non-NULL pointer correctly.
	(me): Really write first record twice.
	(struct entry): Make format bi-arch safe.
	(dest): Write out more realloc statistics.
	* malloc/memusagestat.c (struct entry): Make format bi-arch safe.
2006-12-05  Jakub Jelinek  <jakub@redhat.com>
	* nis/nis_subr.c (nis_getnames): Revert last change.
2006-12-03  Kaz Kojima  <kkojima@rr.iij4u.or.jp>
	* sysdeps/unix/sysv/linux/sh/sys/io.h: Removed.
2006-11-30  H.J. Lu  <hongjiu.lu@intel.com>
	* sysdeps/i386/i686/memcmp.S: Use jump table as the base of
	jump table entries.

2006-11-30  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* sysdeps/unix/sysv/linux/i386/clone.S: Provide CFI for the outermost
	`clone' function to ensure proper unwinding stop of gdb.
	* sysdeps/unix/sysv/linux/x86_64/clone.S: Likewise.

2006-12-01  Ulrich Drepper  <drepper@redhat.com>

	* nscd/nscd.init: Remove obsolete and commented-out -S option
	handling.

2006-11-23  Jakub Jelinek  <jakub@redhat.com>

	[BZ #3514]
	* manual/string.texi (strncmp): Fix pastos from wcscmp description.

	[BZ #3515]
	* manual/string.texi (strtok): Remove duplicate paragraph.

2006-12-01  Jan Kratochvil  <jan.kratochvil@redhat.com>

	* sysdeps/unix/sysv/linux/x86_64/sigaction.c: Fix compatibility with
	libgcc not supporting `rflags' unwinding (register # >= 17).

2006-11-30  Jakub Jelinek  <jakub@redhat.com>

	* sunrpc/svc_run.c (svc_run): Set my_pollfd to new_pollfd if realloc
	succeeded.

2006-11-29  Daniel Jacobowitz  <dan@codesourcery.com>
	    Jakub Jelinek  <jakub@redhat.com>
	    Jan Kratochvil  <jan.kratochvil@redhat.com>

	* sysdeps/unix/sysv/linux/x86_64/sigaction.c (restore_rt): Add correct
	unwind information.
	* sysdeps/unix/sysv/linux/x86_64/Makefile: Provide symbols for
	'restore_rt' even in the 'signal' directory.
	* sysdeps/unix/sysv/linux/x86_64/ucontext_i.sym: Extend the regs list.
	malloc crashed.  Don't allocate memory unnecessarily in each
	loop.
2006-10-21  Jakub Jelinek  <jakub@redhat.com>
	* resolv/mapv4v6addr.h (map_v4v6_address): Fix last change.
2006-11-20  Ulrich Drepper  <drepper@redhat.com>
	* resolv/mapv4v6addr.h (map_v4v6_address): Optimize a bit.
2006-11-18  Bruno Haible  <bruno@clisp.org>
	* sysdeps/unix/sysv/linux/i386/getgroups.c (__getgroups): Invoke
	__sysconf only after having tried to call getgroups32.
2006-11-19  Ulrich Drepper  <drepper@redhat.com>
	* nss/nss_files/files-hosts.c (LINE_PARSER): Support IPv6-style
	addresses for IPv4 queries if they can be mapped.
2006-11-16  Jakub Jelinek  <jakub@redhat.com>
	* sysdeps/x86_64/fpu/s_copysignf.S (__copysignf): Switch to .text.
	* sysdeps/x86_64/fpu/s_copysign.S (__copysign): Likewise.
	(signmask): Add .size directive.
	(othermask): Add .type directive.
2006-11-14  Ulrich Drepper  <drepper@redhat.com>
	* po/nl.po: Update from translation team.
	* timezone/zdump.c: Redo fix for BZ #3137.
2006-11-14  Jakub Jelinek  <jakub@redhat.com>
	* nss/nss_files/files-alias.c (get_next_alias): Set line back
	to first_unused after parsing :include: file.
	* timezone/africa: Update from tzdata2006o.
	* timezone/antarctica: Likewise.
	* timezone/asia: Likewise.
	* timezone/australasia: Likewise.
	* timezone/backward: Likewise.
	* timezone/europe: Likewise.
	* timezone/iso3166.tab: Likewise.
	* timezone/northamerica: Likewise.
	* timezone/southamerica: Likewise.
	* timezone/zone.tab: Likewise.

	* time/tzfile.c (__tzfile_read): Extend to handle new file format
	on machines with 64-bit time_t.

	* timezone/checktab.awk: Update from tzcode2006o.
	* timezone/ialloc.c: Likewise.
	* timezone/private.h: Likewise.
	* timezone/scheck.c: Likewise.
	* timezone/tzfile.h: Likewise.
	* timezone/tzselect.ksh: Likewise.
	* timezone/zdump.c: Likewise.
	* timezone/zic.c: Likewise.

	[BZ #3483]
	* elf/ldconfig.c (main): Call setlocale and textdomain.
	Patch mostly by Benno Schulenberg <bensberg@justemail.net>.

	[BZ #3480]
	* manual/argp.texi: Fix typos.
	* manual/charset.texi: Likewise.
	* manual/errno.texi: Likewise.
	* manual/filesys.texi: Likewise.
	* manual/lang.texi: Likewise.
	* manual/maint.texi: Likewise.
	* manual/memory.texi: Likewise.
	* manual/message.texi: Likewise.
	* manual/resource.texi: Likewise.
	* manual/search.texi: Likewise.
	* manual/signal.texi: Likewise.
	* manual/startup.texi: Likewise.
	* manual/stdio.texi: Likewise.
	* manual/sysinfo.texi: Likewise.
	* manual/syslog.texi: Likewise.
	* manual/time.texi: Likewise.
	Patch by Ralf Wildenhues <Ralf.Wildenhues@gmx.de>.

	[BZ #3465]
	* sunrpc/clnt_raw.c: Minimal message improvements.
	* sunrpc/pm_getmaps.c: Likewise.
	* nis/nss_nisplus/nisplus-publickey.c: Likewise.
	* nis/nis_print_group_entry.c: Likewise.
	* locale/programs/repertoire.c: Likewise.
	* locale/programs/charmap.c: Likewise.
	* malloc/memusage.sh: Likewise.
	* elf/dl-deps.c: Likewise.
	* locale/programs/ld-collate.c: Likewise.
	* libio/vswprintf.c: Likewise.
	* malloc/memusagestat.c: Likewise.
	* sunrpc/auth_unix.c: Likewise.
	* sunrpc/rpc_main.c: Likewise.
	* nscd/cache.c: Likewise.
	* locale/programs/repertoire.c: Unify output messages.
	* locale/programs/charmap.c: Likewise.
	* locale/programs/ld-ctype.c: Likewise.
	* locale/programs/ld-monetary.c: Likewise.
	* locale/programs/ld-numeric.c: Likewise.
	* locale/programs/ld-time.c: Likewise.
	* elf/ldconfig.c: Likewise.
	* nscd/selinux.c: Likewise.
	* elf/cache.c: Likewise.
	Patch mostly by Benno Schulenberg <bensberg@justemail.net>.

2006-11-10  Jakub Jelinek  <jakub@redhat.com>

	* string/strxfrm_l.c (STRXFRM): Fix trailing \1 optimization
	if N is one bigger than return value.
	* string/tst-strxfrm2.c (do_test): Also test strxfrm with l1 + 1
	and l1 last arguments, if buf is defined, verify the return value
	equals to strlen (buf) and verify no byte beyond passed length
	is modified.

2006-11-10  Ulrich Drepper  <drepper@redhat.com>

	* po/sv.po: Update from translation team.

	* sysdeps/gnu/siglist.c (__old_sys_siglist, __old_sys_sigabbrev):
	Use __new_sys_siglist instead of _sys_siglist_internal as
	second macro argument.
	(_old_sys_siglist): Use declare_symbol_alias macro instead of
	strong_alias.
2006-11-09  Ulrich Drepper  <drepper@redhat.com>

	[BZ #3493]
	* posix/unistd.h (sysconf): Remove const attribute.

	* sysdeps/posix/getaddrinfo.c (getaddrinfo): Fix test for
	temporary or deprecated addresses.
	Patch by Sridhar Samudrala <sri@us.ibm.com>.

	* string/Makefile (tests): Add tst-strxfrm2.
	* string/tst-strxfrm2.c: New file.

2006-10-09  Jakub Jelinek  <jakub@redhat.com>

	* elf/dl-debug.c (_dl_debug_initialize): Check r->r_map for 0
	rather than r->r_brk.
	* string/strxfrm_l.c (STRXFRM): Do the trailing \1 removal
	optimization even if needed > n.

2006-11-07  Jakub Jelinek  <jakub@redhat.com>

	* include/libc-symbols.h (declare_symbol): Rename to...
	(declare_symbol_alias): ... this.  Add ORIGINAL argument, imply
	strong_alias (ORIGINAL, SYMBOL) in asm to make sure it preceedes
	.size directive.
	* sysdeps/gnu/errlist-compat.awk: Adjust for declare_symbol_alias
	changes.
	* sysdeps/gnu/siglist.c: Likewise.

2006-11-03  Steven Munroe  <sjmunroe@us.ibm.com>

	* sysdeps/powerpc/fpu/bits/mathinline.h
	[__LIBC_INTERNAL_MATH_INLINES]: Moved to ...
	* sysdeps/powerpc/fpu/math_private.h: ...here.  New file.

2006-11-05  Ulrich Drepper  <drepper@redhat.com>

	* sysdeps/unix/sysv/linux/i386/sysconf.c (intel_check_word):
	Update handling of cache descriptor 0x49 for new models.
	* sysdeps/unix/sysv/linux/x86_64/sysconf.c (intel_check_word):
	Likewise.

2006-11-02  Ulrich Drepper  <drepper@redhat.com>

	* configure.in: Work around ld --help change and avoid -z relro
	test completely if the architecture doesn't care about security.

2006-11-01  Ulrich Drepper  <drepper@redhat.com>

	* po/sv.po: Update from translation team.

2006-10-31  Ulrich Drepper  <drepper@redhat.com>

	* stdlib/atexit.c (atexit): Don't mark as hidden when used to
	generate compatibility version.

2006-10-29  Ulrich Drepper  <drepper@redhat.com>

	* configure.in: Relax -z relro requirement a bit.

	* po/sv.po: Update from translation team.

2006-10-29  Jakub Jelinek  <jakub@redhat.com>

	* elf/dl-sym.c (do_sym): Use RTLD_SINGLE_THREAD_P.
	* elf/dl-runtime.c (_dl_fixup, _dl_profile_fixup): Likewise.
	* elf/dl-close.c (_dl_close_worker): Likewise.
	* elf/dl-open.c (_dl_open_worker): Likewise.
	* sysdeps/generic/sysdep-cancel.h (RTLD_SINGLE_THREAD_P): Define.

	* configure.in: Require assembler support for visibility, compiler
	support for visibility and aliases, linker support for various -z
	options.
	* Makeconfig: Remove conditional code which now is unnecessary.
	* config.h.in: Likewise.
	* config.make.in: Likewise.
	* dlfcn/Makefile: Likewise.
	* elf/Makefile: Likewise.
	* elf/dl-load.c: Likewise.
	* elf/rtld.c: Likewise.
	* include/libc-symbols.h: Likewise.
	* include/stdio.h: Likewise.
	* io/Makefile: Likewise.
	* io/fstat.c: Likewise.
	* io/fstat64.c: Likewise.
	* io/fstatat.c: Likewise.
	* io/fstatat64.c: Likewise.
	* io/lstat.c: Likewise.
	* io/lstat64.c: Likewise.
	* io/mknod.c: Likewise.
	* io/mknodat.c: Likewise.
	* io/stat.c: Likewise.
	* io/stat64.c: Likewise.
	* libio/stdio.c: Likewise.
	* nscd/Makefile: Likewise.
	* stdlib/Makefile: Likewise.
	* stdlib/atexit.c: Likewise.
	* sysdeps/generic/ldsodefs.h: Likewise.
	* sysdeps/i386/dl-machine.h: Likewise.
	* sysdeps/i386/sysdep.h: Likewise.
	* sysdeps/i386/i686/memcmp.S: Likewise.
	* sysdeps/powerpc/powerpc32/sysdep.h: Likewise.
	* sysdeps/unix/sysv/linux/i386/sigaction.c: Likewise.
	* sysdeps/unix/sysv/linux/x86_64/sigaction.c: Likewise.

	* Makerules: USE_TLS support is now default.
	* tls.make.c: Likewise.
	* csu/Versions: Likewise.
	* csu/libc-start.c: Likewise.
	* csu/libc-tls.c: Likewise.
	* csu/version.c: Likewise.
	* dlfcn/dlinfo.c: Likewise.
	* elf/dl-addr.c: Likewise.
	* elf/dl-cache.c: Likewise.
	* elf/dl-close.c: Likewise.
	* elf/dl-iteratephdr.c: Likewise.
	* elf/dl-load.c: Likewise.
	* elf/dl-lookup.c: Likewise.
	* elf/dl-object.c: Likewise.
	* elf/dl-open.c: Likewise.
	* elf/dl-reloc.c: Likewise.
	* elf/dl-support.c: Likewise.
	* elf/dl-sym.c: Likewise.
	* elf/dl-sysdep.c: Likewise.
	* elf/dl-tls.c: Likewise.
	* elf/ldconfig.c: Likewise.
	* elf/rtld.c: Likewise.
	* elf/tst-tls-dlinfo.c: Likewise.
	* elf/tst-tls1.c: Likewise.
	* elf/tst-tls10.h: Likewise.
	* elf/tst-tls14.c: Likewise.
	* elf/tst-tls2.c: Likewise.
	* elf/tst-tls3.c: Likewise.
	* elf/tst-tls4.c: Likewise.
	* elf/tst-tls5.c: Likewise.
	* elf/tst-tls6.c: Likewise.
	* elf/tst-tls7.c: Likewise.
	* elf/tst-tls8.c: Likewise.
	* elf/tst-tls9.c: Likewise.
	* elf/tst-tlsmod1.c: Likewise.
	* elf/tst-tlsmod13.c: Likewise.
	* elf/tst-tlsmod13a.c: Likewise.
	* elf/tst-tlsmod14a.c: Likewise.
	* elf/tst-tlsmod2.c: Likewise.
	* elf/tst-tlsmod3.c: Likewise.
	* elf/tst-tlsmod4.c: Likewise.
	* elf/tst-tlsmod5.c: Likewise.
	* elf/tst-tlsmod6.c: Likewise.
	* include/errno.h: Likewise.
	* include/link.h: Likewise.
	* include/tls.h: Likewise.
	* locale/global-locale.c: Likewise.
	* locale/localeinfo.h: Likewise.
	* malloc/arena.c: Likewise.
	* malloc/hooks.c: Likewise.
	* malloc/malloc.c: Likewise.
	* resolv/Versions: Likewise.
	* sysdeps/alpha/dl-machine.h: Likewise.
	* sysdeps/alpha/libc-tls.c: Likewise.
	* sysdeps/generic/ldsodefs.h: Likewise.
	* sysdeps/generic/tls.h: Likewise.
	* sysdeps/i386/dl-machine.h: Likewise.
	* sysdeps/ia64/dl-machine.h: Likewise.
	* sysdeps/ia64/libc-tls.c: Likewise.
	* sysdeps/mach/hurd/fork.c: Likewise.
	* sysdeps/mach/hurd/i386/tls.h: Likewise.
	* sysdeps/powerpc/powerpc32/dl-machine.c: Likwise.
	* sysdeps/powerpc/powerpc32/dl-machine.h: Likewise.
	* sysdeps/powerpc/powerpc64/dl-machine.h: Likewise.
	* sysdeps/s390/libc-tls.c: Likewise.
	* sysdeps/s390/s390-32/dl-machine.h: Likewise.
	* sysdeps/s390/s390-64/dl-machine.h: Likewise.
	* sysdeps/sh/dl-machine.h: Likewise.
	* sysdeps/sparc/sparc32/dl-machine.h: Likewise.
	* sysdeps/sparc/sparc64/dl-machine.h: Likewise.
	* sysdeps/x86_64/dl-machine.h: Likewise.

	[BZ #3426]
	* stdlib/stdlib.h: Adjust comment for canonicalize_file_name to
	reality.

2006-10-27  Jakub Jelinek  <jakub@redhat.com>

	* elf/dl-lookup.c (_dl_debug_bindings): Remove unused symbol_scope
	argument.
	(_dl_lookup_symbol_x): Adjust caller.

	* sysdeps/generic/ldsodefs.h (struct link_namespaces): Remove
	_ns_global_scope.
	* elf/rtld.c (dl_main): Don't initialize _ns_global_scope.

	* elf/dl-libc.c: Revert l_scope name changes.
	* elf/dl-load.c: Likewise.
	* elf/dl-object.c: Likewise.
	* elf/rtld.c: Likewise.
	* elf/dl-close.c (_dl_close): Likewise.
	* elf/dl-open.c (dl_open_worker): Likewise.  If not SINGLE_THREAD_P,
	always use __rtld_mrlock_{change,done}.  Always free old scope list
	here if not l_scope_mem.
	* elf/dl-runtime.c (_dl_fixup, _dl_profile_fixup): Revert l_scope name
	change.  Never free scope list here.  Just __rtld_mrlock_lock before
	the lookup and __rtld_mrlock_unlock it after the lookup.
	* elf/dl-sym.c: Likewise.
	* include/link.h (struct r_scoperec): Remove.
	(struct link_map): Replace l_scoperec with l_scope, l_scoperec_mem
	with l_scope_mem and l_scoperec_lock with l_scope_lock.

2006-10-25  Ulrich Drepper  <drepper@redhat.com>

	* sysdeps/gnu/netinet/tcp.h: Define TCP_CONGESTION.

2006-10-18  Ulrich Drepper  <drepper@redhat.com>

	* configure.in: Disable building profile libraries by default.

2006-10-18  Ulrich Drepper  <drepper@redhat.com>

	* elf/dl-lookup.c (_dl_lookup_symbol_x): Add warning to
	_dl_lookup_symbol_x code.

2006-10-17  Jakub Jelinek  <jakub@redhat.com>

	* elf/dl-runtime.c: Include sysdep-cancel.h.
	(_dl_fixup, _dl_profile_fixup): Use __rtld_mrlock_* and
	scoperec->nusers only if !SINGLE_THREAD_P.  Use atomic_*
	instead of catomic_* macros.
	* elf/dl-sym.c: Include sysdep-cancel.h.
	(do_sym): Use __rtld_mrlock_* and scoperec->nusers only
	if !SINGLE_THREAD_P.  Use atomic_* instead of catomic_* macros.
	* elf/dl-close.c: Include sysdep-cancel.h.
	(_dl_close): Use __rtld_mrlock_* and scoperec->nusers only
	if !SINGLE_THREAD_P.  Use atomic_* instead of catomic_* macros.
	* elf/dl-open.c: Include sysdep-cancel.h.
	(dl_open_worker): Use __rtld_mrlock_* and scoperec->nusers only
	if !SINGLE_THREAD_P.  Use atomic_* instead of catomic_* macros.

2006-10-17  Jakub Jelinek  <jakub@redhat.com>

	[BZ #3313]
	* malloc/malloc.c (malloc_consolidate): Set maxfb to address of last
	fastbin rather than end of fastbin array.

2006-10-18  Ulrich Drepper  <drepper@redhat.com>

	* sysdeps/i386/i486/bits/atomic.h (catomic_decrement): Use correct
	body macro.
	* sysdeps/x86_64/bits/atomic.h
	(__arch_c_compare_and_exchange_val_64_acq): Add missing casts.
	(catomic_decrement): Use correct body macro.

2006-10-17  Jakub Jelinek  <jakub@redhat.com>

	* include/atomic.h: Add a unique prefix to all local variables
	in macros.
	* csu/tst-atomic.c (do_test): Test also catomic_* macros.

2006-10-14  Ulrich Drepper  <drepper@redhat.com>

	* resolv/arpa/nameser.h: Document that ns_t_a6 is deprecated.

	[BZ #3313]
	* malloc/malloc.c (malloc_consolidate): Don't use get_fast_max to
	determine highest fast bin to consolidate, always look into all of
	them.
	(do_check_malloc_state): Only require for empty bins for large
	sizes in main arena.

	* libio/stdio.h: Add more __wur attributes.

2006-11-12  Andreas Jaeger  <aj@suse.de>

	[BZ #2510]
	* manual/search.texi (Hash Search Function): Clarify.
	(Array Search Function): Clarify.

2006-11-12  Joseph Myers  <joseph@codesourcery.com>

	[BZ #2830]
	* math/atest-exp.c (main): Cast hex value to mp_limb_t before
	shifting.
	* math/atest-exp2.c (read_mpn_hex): Likewise.
	* math/atest-sincos.c (main): Likewise.

	* sysdeps/unix/sysv/linux/syscalls.list: Add epoll_pwait.
	* sysdeps/unix/sysv/linux/sys/epoll.h: Declare epoll_pwait.
	* sysdeps/unix/sysv/linux/Versions (libc): Add epoll_pwait for
	version GLIBC_2.6.
	* Versions.def: Add GLIBC_2.6 for libc.

	* sysdeps/i386/i486/bits/atomic.h: Add catomic_* support.

2006-10-11  Jakub Jelinek  <jakub@redhat.com>

	* malloc/malloc.c (_int_malloc): Remove unused any_larger variable.

	* nis/nis_defaults.c (__nis_default_access): Don't call getenv twice.

	* nis/nis_subr.c (nis_getnames): Use __secure_getenv instead of getenv.
	* sysdeps/generic/unsecvars.h: Add NIS_PATH.

2006-10-11  Ulrich Drepper  <drepper@redhat.com>

	* include/atomic.c: Define catomic_* operations.
	* sysdeps/x86_64/bits/atomic.h: Likewise.  Fix a few minor problems.
	* stdlib/cxa_finalize.c: Use catomic_* operations instead of atomic_*.
	* malloc/memusage.c: Likewise.
	* gmon/mcount.c: Likewise.
	* elf/dl-close.c: Likewise.
	* elf/dl-open.c: Likewise.
	* elf/dl-profile.c: Likewise.
	* elf/dl-sym.c: Likewise.
	* elf/dl-runtime.c: Likewise.
	* elf/dl-fptr.c: Likewise.
	* resolv/res_libc.c: Likewise.

2006-10-10  Roland McGrath  <roland@frob.com>
	* sysdeps/mach/hurd/utimes.c: Use a union to avoid an improper cast.
	* sysdeps/mach/hurd/futimes.c: Likewise.
	* sysdeps/mach/hurd/lutimes.c: Likewise.

2006-10-09  Ulrich Drepper  <drepper@redhat.com>
	    Jakub Jelinek  <jakub@redhat.com>

	Implement reference counting of scope records.
	* elf/dl-close.c (_dl_close): Remove all scopes from removed objects
	from the list in objects which remain.  Always allocate new scope
	record.
	* elf/dl-open.c (dl_open_worker): When growing array for scopes,
	don't resize, allocate a new one.
	* elf/dl-runtime.c: Update reference counters before using a scope
	array.
	* elf/dl-sym.c: Likewise.
	* elf/dl-libc.c: Adjust for l_scope name change.
	* elf/dl-load.c: Likewise.
	* elf/dl-object.c: Likewise.
	* elf/rtld.c: Likewise.
	* include/link.h: Include <rtld-lowlevel.h>.  Define struct
	r_scoperec.  Replace r_scope with pointer to r_scoperec structure.
	Add l_scoperec_lock.
	* sysdeps/generic/ldsodefs.h: Include <rtld-lowlevel.h>.
	* sysdeps/generic/rtld-lowlevel.h: New file.

	* include/atomic.h: Rename atomic_and to atomic_and_val and
	atomic_or to atomic_or_val.  Define new macros atomic_and and
	atomic_or which do not return values.
	* sysdeps/x86_64/bits/atomic.h: Define atomic_and and atomic_or.
	Various cleanups.
	* sysdeps/i386/i486/bits/atomic.h: Likewise.

	* po/sv.po: Update from translation team.

2006-10-07  Ulrich Drepper  <drepper@redhat.com>

	* Versions.def: Add GLIBC_2.6 to libpthread.

	* include/shlib-compat.h (SHLIB_COMPAT): Expand parameters before use.
	(versioned_symbol): Likewise.
	(compat_symbol): Likewise.

	* po/tr.po: Update from translation team.
	* nis/Banner: Removed.  It's been integral part forever and the
	author info is incomplete anyway.
	* libio/Banner: Likewise.

2006-10-06  Ulrich Drepper  <drepper@redhat.com>

	* version.h (VERSION): Bump to 2.5.90 for new development tree.
2007-01-11 21:51:07 +00:00

1331 lines
41 KiB
C

/* Software floating-point emulation. Common operations.
Copyright (C) 1997,1998,1999,2006 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Contributed by Richard Henderson (rth@cygnus.com),
Jakub Jelinek (jj@ultra.linux.cz),
David S. Miller (davem@redhat.com) and
Peter Maydell (pmaydell@chiark.greenend.org.uk).
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
In addition to the permissions in the GNU Lesser General Public
License, the Free Software Foundation gives you unlimited
permission to link the compiled version of this file into
combinations with other programs, and to distribute those
combinations without any restriction coming from the use of this
file. (The Lesser General Public License restrictions do apply in
other respects; for example, they cover modification of the file,
and distribution when not linked into a combine executable.)
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, write to the Free
Software Foundation, 51 Franklin Street, Fifth Floor, Boston,
MA 02110-1301, USA. */
#define _FP_DECL(wc, X) \
_FP_I_TYPE X##_c __attribute__((unused)), X##_s, X##_e; \
_FP_FRAC_DECL_##wc(X)
/*
* Finish truely unpacking a native fp value by classifying the kind
* of fp value and normalizing both the exponent and the fraction.
*/
#define _FP_UNPACK_CANONICAL(fs, wc, X) \
do { \
switch (X##_e) \
{ \
default: \
_FP_FRAC_HIGH_RAW_##fs(X) |= _FP_IMPLBIT_##fs; \
_FP_FRAC_SLL_##wc(X, _FP_WORKBITS); \
X##_e -= _FP_EXPBIAS_##fs; \
X##_c = FP_CLS_NORMAL; \
break; \
\
case 0: \
if (_FP_FRAC_ZEROP_##wc(X)) \
X##_c = FP_CLS_ZERO; \
else \
{ \
/* a denormalized number */ \
_FP_I_TYPE _shift; \
_FP_FRAC_CLZ_##wc(_shift, X); \
_shift -= _FP_FRACXBITS_##fs; \
_FP_FRAC_SLL_##wc(X, (_shift+_FP_WORKBITS)); \
X##_e -= _FP_EXPBIAS_##fs - 1 + _shift; \
X##_c = FP_CLS_NORMAL; \
FP_SET_EXCEPTION(FP_EX_DENORM); \
} \
break; \
\
case _FP_EXPMAX_##fs: \
if (_FP_FRAC_ZEROP_##wc(X)) \
X##_c = FP_CLS_INF; \
else \
{ \
X##_c = FP_CLS_NAN; \
/* Check for signaling NaN */ \
if (!(_FP_FRAC_HIGH_RAW_##fs(X) & _FP_QNANBIT_##fs)) \
FP_SET_EXCEPTION(FP_EX_INVALID); \
} \
break; \
} \
} while (0)
/* Finish unpacking an fp value in semi-raw mode: the mantissa is
shifted by _FP_WORKBITS but the implicit MSB is not inserted and
other classification is not done. */
#define _FP_UNPACK_SEMIRAW(fs, wc, X) _FP_FRAC_SLL_##wc(X, _FP_WORKBITS)
/* A semi-raw value has overflowed to infinity. Adjust the mantissa
and exponent appropriately. */
#define _FP_OVERFLOW_SEMIRAW(fs, wc, X) \
do { \
if (FP_ROUNDMODE == FP_RND_NEAREST \
|| (FP_ROUNDMODE == FP_RND_PINF && !X##_s) \
|| (FP_ROUNDMODE == FP_RND_MINF && X##_s)) \
{ \
X##_e = _FP_EXPMAX_##fs; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
} \
else \
{ \
X##_e = _FP_EXPMAX_##fs - 1; \
FP_SET_EXCEPTION(FP_EX_OVERFLOW); \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
_FP_FRAC_SET_##wc(X, _FP_MAXFRAC_##wc); \
} \
} while (0)
/* Check for a semi-raw value being a signaling NaN and raise the
invalid exception if so. */
#define _FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X) \
do { \
if (X##_e == _FP_EXPMAX_##fs \
&& !_FP_FRAC_ZEROP_##wc(X) \
&& !(_FP_FRAC_HIGH_##fs(X) & _FP_QNANBIT_SH_##fs)) \
FP_SET_EXCEPTION(FP_EX_INVALID); \
} while (0)
/* Choose a NaN result from an operation on two semi-raw NaN
values. */
#define _FP_CHOOSENAN_SEMIRAW(fs, wc, R, X, Y, OP) \
do { \
/* _FP_CHOOSENAN expects raw values, so shift as required. */ \
_FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
_FP_FRAC_SRL_##wc(Y, _FP_WORKBITS); \
_FP_CHOOSENAN(fs, wc, R, X, Y, OP); \
_FP_FRAC_SLL_##wc(R, _FP_WORKBITS); \
} while (0)
/* Test whether a biased exponent is normal (not zero or maximum). */
#define _FP_EXP_NORMAL(fs, wc, X) (((X##_e + 1) & _FP_EXPMAX_##fs) > 1)
/* Prepare to pack an fp value in semi-raw mode: the mantissa is
rounded and shifted right, with the rounding possibly increasing
the exponent (including changing a finite value to infinity). */
#define _FP_PACK_SEMIRAW(fs, wc, X) \
do { \
_FP_ROUND(wc, X); \
if (_FP_FRAC_HIGH_##fs(X) \
& (_FP_OVERFLOW_##fs >> 1)) \
{ \
_FP_FRAC_HIGH_##fs(X) &= ~(_FP_OVERFLOW_##fs >> 1); \
X##_e++; \
if (X##_e == _FP_EXPMAX_##fs) \
_FP_OVERFLOW_SEMIRAW(fs, wc, X); \
} \
_FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
if (!_FP_EXP_NORMAL(fs, wc, X) && !_FP_FRAC_ZEROP_##wc(X)) \
{ \
if (X##_e == 0) \
FP_SET_EXCEPTION(FP_EX_UNDERFLOW); \
else \
{ \
if (!_FP_KEEPNANFRACP) \
{ \
_FP_FRAC_SET_##wc(X, _FP_NANFRAC_##fs); \
X##_s = _FP_NANSIGN_##fs; \
} \
else \
_FP_FRAC_HIGH_RAW_##fs(X) |= _FP_QNANBIT_##fs; \
} \
} \
} while (0)
/*
* Before packing the bits back into the native fp result, take care
* of such mundane things as rounding and overflow. Also, for some
* kinds of fp values, the original parts may not have been fully
* extracted -- but that is ok, we can regenerate them now.
*/
#define _FP_PACK_CANONICAL(fs, wc, X) \
do { \
switch (X##_c) \
{ \
case FP_CLS_NORMAL: \
X##_e += _FP_EXPBIAS_##fs; \
if (X##_e > 0) \
{ \
_FP_ROUND(wc, X); \
if (_FP_FRAC_OVERP_##wc(fs, X)) \
{ \
_FP_FRAC_CLEAR_OVERP_##wc(fs, X); \
X##_e++; \
} \
_FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
if (X##_e >= _FP_EXPMAX_##fs) \
{ \
/* overflow */ \
switch (FP_ROUNDMODE) \
{ \
case FP_RND_NEAREST: \
X##_c = FP_CLS_INF; \
break; \
case FP_RND_PINF: \
if (!X##_s) X##_c = FP_CLS_INF; \
break; \
case FP_RND_MINF: \
if (X##_s) X##_c = FP_CLS_INF; \
break; \
} \
if (X##_c == FP_CLS_INF) \
{ \
/* Overflow to infinity */ \
X##_e = _FP_EXPMAX_##fs; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
} \
else \
{ \
/* Overflow to maximum normal */ \
X##_e = _FP_EXPMAX_##fs - 1; \
_FP_FRAC_SET_##wc(X, _FP_MAXFRAC_##wc); \
} \
FP_SET_EXCEPTION(FP_EX_OVERFLOW); \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
} \
} \
else \
{ \
/* we've got a denormalized number */ \
X##_e = -X##_e + 1; \
if (X##_e <= _FP_WFRACBITS_##fs) \
{ \
_FP_FRAC_SRS_##wc(X, X##_e, _FP_WFRACBITS_##fs); \
_FP_ROUND(wc, X); \
if (_FP_FRAC_HIGH_##fs(X) \
& (_FP_OVERFLOW_##fs >> 1)) \
{ \
X##_e = 1; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
} \
else \
{ \
X##_e = 0; \
_FP_FRAC_SRL_##wc(X, _FP_WORKBITS); \
FP_SET_EXCEPTION(FP_EX_UNDERFLOW); \
} \
} \
else \
{ \
/* underflow to zero */ \
X##_e = 0; \
if (!_FP_FRAC_ZEROP_##wc(X)) \
{ \
_FP_FRAC_SET_##wc(X, _FP_MINFRAC_##wc); \
_FP_ROUND(wc, X); \
_FP_FRAC_LOW_##wc(X) >>= (_FP_WORKBITS); \
} \
FP_SET_EXCEPTION(FP_EX_UNDERFLOW); \
} \
} \
break; \
\
case FP_CLS_ZERO: \
X##_e = 0; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
break; \
\
case FP_CLS_INF: \
X##_e = _FP_EXPMAX_##fs; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
break; \
\
case FP_CLS_NAN: \
X##_e = _FP_EXPMAX_##fs; \
if (!_FP_KEEPNANFRACP) \
{ \
_FP_FRAC_SET_##wc(X, _FP_NANFRAC_##fs); \
X##_s = _FP_NANSIGN_##fs; \
} \
else \
_FP_FRAC_HIGH_RAW_##fs(X) |= _FP_QNANBIT_##fs; \
break; \
} \
} while (0)
/* This one accepts raw argument and not cooked, returns
* 1 if X is a signaling NaN.
*/
#define _FP_ISSIGNAN(fs, wc, X) \
({ \
int __ret = 0; \
if (X##_e == _FP_EXPMAX_##fs) \
{ \
if (!_FP_FRAC_ZEROP_##wc(X) \
&& !(_FP_FRAC_HIGH_RAW_##fs(X) & _FP_QNANBIT_##fs)) \
__ret = 1; \
} \
__ret; \
})
/* Addition on semi-raw values. */
#define _FP_ADD_INTERNAL(fs, wc, R, X, Y, OP) \
do { \
if (X##_s == Y##_s) \
{ \
/* Addition. */ \
R##_s = X##_s; \
int ediff = X##_e - Y##_e; \
if (ediff > 0) \
{ \
R##_e = X##_e; \
if (Y##_e == 0) \
{ \
/* Y is zero or denormalized. */ \
if (_FP_FRAC_ZEROP_##wc(Y)) \
{ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
_FP_FRAC_COPY_##wc(R, X); \
goto add_done; \
} \
else \
{ \
FP_SET_EXCEPTION(FP_EX_DENORM); \
ediff--; \
if (ediff == 0) \
{ \
_FP_FRAC_ADD_##wc(R, X, Y); \
goto add3; \
} \
if (X##_e == _FP_EXPMAX_##fs) \
{ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
_FP_FRAC_COPY_##wc(R, X); \
goto add_done; \
} \
goto add1; \
} \
} \
else if (X##_e == _FP_EXPMAX_##fs) \
{ \
/* X is NaN or Inf, Y is normal. */ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
_FP_FRAC_COPY_##wc(R, X); \
goto add_done; \
} \
\
/* Insert implicit MSB of Y. */ \
_FP_FRAC_HIGH_##fs(Y) |= _FP_IMPLBIT_SH_##fs; \
\
add1: \
/* Shift the mantissa of Y to the right EDIFF steps; \
remember to account later for the implicit MSB of X. */ \
if (ediff <= _FP_WFRACBITS_##fs) \
_FP_FRAC_SRS_##wc(Y, ediff, _FP_WFRACBITS_##fs); \
else if (!_FP_FRAC_ZEROP_##wc(Y)) \
_FP_FRAC_SET_##wc(Y, _FP_MINFRAC_##wc); \
_FP_FRAC_ADD_##wc(R, X, Y); \
} \
else if (ediff < 0) \
{ \
ediff = -ediff; \
R##_e = Y##_e; \
if (X##_e == 0) \
{ \
/* X is zero or denormalized. */ \
if (_FP_FRAC_ZEROP_##wc(X)) \
{ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
_FP_FRAC_COPY_##wc(R, Y); \
goto add_done; \
} \
else \
{ \
FP_SET_EXCEPTION(FP_EX_DENORM); \
ediff--; \
if (ediff == 0) \
{ \
_FP_FRAC_ADD_##wc(R, Y, X); \
goto add3; \
} \
if (Y##_e == _FP_EXPMAX_##fs) \
{ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
_FP_FRAC_COPY_##wc(R, Y); \
goto add_done; \
} \
goto add2; \
} \
} \
else if (Y##_e == _FP_EXPMAX_##fs) \
{ \
/* Y is NaN or Inf, X is normal. */ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
_FP_FRAC_COPY_##wc(R, Y); \
goto add_done; \
} \
\
/* Insert implicit MSB of X. */ \
_FP_FRAC_HIGH_##fs(X) |= _FP_IMPLBIT_SH_##fs; \
\
add2: \
/* Shift the mantissa of X to the right EDIFF steps; \
remember to account later for the implicit MSB of Y. */ \
if (ediff <= _FP_WFRACBITS_##fs) \
_FP_FRAC_SRS_##wc(X, ediff, _FP_WFRACBITS_##fs); \
else if (!_FP_FRAC_ZEROP_##wc(X)) \
_FP_FRAC_SET_##wc(X, _FP_MINFRAC_##wc); \
_FP_FRAC_ADD_##wc(R, Y, X); \
} \
else \
{ \
/* ediff == 0. */ \
if (!_FP_EXP_NORMAL(fs, wc, X)) \
{ \
if (X##_e == 0) \
{ \
/* X and Y are zero or denormalized. */ \
R##_e = 0; \
if (_FP_FRAC_ZEROP_##wc(X)) \
{ \
if (!_FP_FRAC_ZEROP_##wc(Y)) \
FP_SET_EXCEPTION(FP_EX_DENORM); \
_FP_FRAC_COPY_##wc(R, Y); \
goto add_done; \
} \
else if (_FP_FRAC_ZEROP_##wc(Y)) \
{ \
FP_SET_EXCEPTION(FP_EX_DENORM); \
_FP_FRAC_COPY_##wc(R, X); \
goto add_done; \
} \
else \
{ \
FP_SET_EXCEPTION(FP_EX_DENORM); \
_FP_FRAC_ADD_##wc(R, X, Y); \
if (_FP_FRAC_HIGH_##fs(R) & _FP_IMPLBIT_SH_##fs) \
{ \
/* Normalized result. */ \
_FP_FRAC_HIGH_##fs(R) \
&= ~(_FP_W_TYPE)_FP_IMPLBIT_SH_##fs; \
R##_e = 1; \
} \
goto add_done; \
} \
} \
else \
{ \
/* X and Y are NaN or Inf. */ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
R##_e = _FP_EXPMAX_##fs; \
if (_FP_FRAC_ZEROP_##wc(X)) \
_FP_FRAC_COPY_##wc(R, Y); \
else if (_FP_FRAC_ZEROP_##wc(Y)) \
_FP_FRAC_COPY_##wc(R, X); \
else \
_FP_CHOOSENAN_SEMIRAW(fs, wc, R, X, Y, OP); \
goto add_done; \
} \
} \
/* The exponents of X and Y, both normal, are equal. The \
implicit MSBs will always add to increase the \
exponent. */ \
_FP_FRAC_ADD_##wc(R, X, Y); \
R##_e = X##_e + 1; \
_FP_FRAC_SRS_##wc(R, 1, _FP_WFRACBITS_##fs); \
if (R##_e == _FP_EXPMAX_##fs) \
/* Overflow to infinity (depending on rounding mode). */ \
_FP_OVERFLOW_SEMIRAW(fs, wc, R); \
goto add_done; \
} \
add3: \
if (_FP_FRAC_HIGH_##fs(R) & _FP_IMPLBIT_SH_##fs) \
{ \
/* Overflow. */ \
_FP_FRAC_HIGH_##fs(R) &= ~(_FP_W_TYPE)_FP_IMPLBIT_SH_##fs; \
R##_e++; \
_FP_FRAC_SRS_##wc(R, 1, _FP_WFRACBITS_##fs); \
if (R##_e == _FP_EXPMAX_##fs) \
/* Overflow to infinity (depending on rounding mode). */ \
_FP_OVERFLOW_SEMIRAW(fs, wc, R); \
} \
add_done: ; \
} \
else \
{ \
/* Subtraction. */ \
int ediff = X##_e - Y##_e; \
if (ediff > 0) \
{ \
R##_e = X##_e; \
R##_s = X##_s; \
if (Y##_e == 0) \
{ \
/* Y is zero or denormalized. */ \
if (_FP_FRAC_ZEROP_##wc(Y)) \
{ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
_FP_FRAC_COPY_##wc(R, X); \
goto sub_done; \
} \
else \
{ \
FP_SET_EXCEPTION(FP_EX_DENORM); \
ediff--; \
if (ediff == 0) \
{ \
_FP_FRAC_SUB_##wc(R, X, Y); \
goto sub3; \
} \
if (X##_e == _FP_EXPMAX_##fs) \
{ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
_FP_FRAC_COPY_##wc(R, X); \
goto sub_done; \
} \
goto sub1; \
} \
} \
else if (X##_e == _FP_EXPMAX_##fs) \
{ \
/* X is NaN or Inf, Y is normal. */ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
_FP_FRAC_COPY_##wc(R, X); \
goto sub_done; \
} \
\
/* Insert implicit MSB of Y. */ \
_FP_FRAC_HIGH_##fs(Y) |= _FP_IMPLBIT_SH_##fs; \
\
sub1: \
/* Shift the mantissa of Y to the right EDIFF steps; \
remember to account later for the implicit MSB of X. */ \
if (ediff <= _FP_WFRACBITS_##fs) \
_FP_FRAC_SRS_##wc(Y, ediff, _FP_WFRACBITS_##fs); \
else if (!_FP_FRAC_ZEROP_##wc(Y)) \
_FP_FRAC_SET_##wc(Y, _FP_MINFRAC_##wc); \
_FP_FRAC_SUB_##wc(R, X, Y); \
} \
else if (ediff < 0) \
{ \
ediff = -ediff; \
R##_e = Y##_e; \
R##_s = Y##_s; \
if (X##_e == 0) \
{ \
/* X is zero or denormalized. */ \
if (_FP_FRAC_ZEROP_##wc(X)) \
{ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
_FP_FRAC_COPY_##wc(R, Y); \
goto sub_done; \
} \
else \
{ \
FP_SET_EXCEPTION(FP_EX_DENORM); \
ediff--; \
if (ediff == 0) \
{ \
_FP_FRAC_SUB_##wc(R, Y, X); \
goto sub3; \
} \
if (Y##_e == _FP_EXPMAX_##fs) \
{ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
_FP_FRAC_COPY_##wc(R, Y); \
goto sub_done; \
} \
goto sub2; \
} \
} \
else if (Y##_e == _FP_EXPMAX_##fs) \
{ \
/* Y is NaN or Inf, X is normal. */ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
_FP_FRAC_COPY_##wc(R, Y); \
goto sub_done; \
} \
\
/* Insert implicit MSB of X. */ \
_FP_FRAC_HIGH_##fs(X) |= _FP_IMPLBIT_SH_##fs; \
\
sub2: \
/* Shift the mantissa of X to the right EDIFF steps; \
remember to account later for the implicit MSB of Y. */ \
if (ediff <= _FP_WFRACBITS_##fs) \
_FP_FRAC_SRS_##wc(X, ediff, _FP_WFRACBITS_##fs); \
else if (!_FP_FRAC_ZEROP_##wc(X)) \
_FP_FRAC_SET_##wc(X, _FP_MINFRAC_##wc); \
_FP_FRAC_SUB_##wc(R, Y, X); \
} \
else \
{ \
/* ediff == 0. */ \
if (!_FP_EXP_NORMAL(fs, wc, X)) \
{ \
if (X##_e == 0) \
{ \
/* X and Y are zero or denormalized. */ \
R##_e = 0; \
if (_FP_FRAC_ZEROP_##wc(X)) \
{ \
_FP_FRAC_COPY_##wc(R, Y); \
if (_FP_FRAC_ZEROP_##wc(Y)) \
R##_s = (FP_ROUNDMODE == FP_RND_MINF); \
else \
{ \
FP_SET_EXCEPTION(FP_EX_DENORM); \
R##_s = Y##_s; \
} \
goto sub_done; \
} \
else if (_FP_FRAC_ZEROP_##wc(Y)) \
{ \
FP_SET_EXCEPTION(FP_EX_DENORM); \
_FP_FRAC_COPY_##wc(R, X); \
R##_s = X##_s; \
goto sub_done; \
} \
else \
{ \
FP_SET_EXCEPTION(FP_EX_DENORM); \
_FP_FRAC_SUB_##wc(R, X, Y); \
R##_s = X##_s; \
if (_FP_FRAC_HIGH_##fs(R) & _FP_IMPLBIT_SH_##fs) \
{ \
/* |X| < |Y|, negate result. */ \
_FP_FRAC_SUB_##wc(R, Y, X); \
R##_s = Y##_s; \
} \
else if (_FP_FRAC_ZEROP_##wc(R)) \
R##_s = (FP_ROUNDMODE == FP_RND_MINF); \
goto sub_done; \
} \
} \
else \
{ \
/* X and Y are NaN or Inf, of opposite signs. */ \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, X); \
_FP_CHECK_SIGNAN_SEMIRAW(fs, wc, Y); \
R##_e = _FP_EXPMAX_##fs; \
if (_FP_FRAC_ZEROP_##wc(X)) \
{ \
if (_FP_FRAC_ZEROP_##wc(Y)) \
{ \
/* Inf - Inf. */ \
R##_s = _FP_NANSIGN_##fs; \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
_FP_FRAC_SLL_##wc(R, _FP_WORKBITS); \
FP_SET_EXCEPTION(FP_EX_INVALID); \
} \
else \
{ \
/* Inf - NaN. */ \
R##_s = Y##_s; \
_FP_FRAC_COPY_##wc(R, Y); \
} \
} \
else \
{ \
if (_FP_FRAC_ZEROP_##wc(Y)) \
{ \
/* NaN - Inf. */ \
R##_s = X##_s; \
_FP_FRAC_COPY_##wc(R, X); \
} \
else \
{ \
/* NaN - NaN. */ \
_FP_CHOOSENAN_SEMIRAW(fs, wc, R, X, Y, OP); \
} \
} \
goto sub_done; \
} \
} \
/* The exponents of X and Y, both normal, are equal. The \
implicit MSBs cancel. */ \
R##_e = X##_e; \
_FP_FRAC_SUB_##wc(R, X, Y); \
R##_s = X##_s; \
if (_FP_FRAC_HIGH_##fs(R) & _FP_IMPLBIT_SH_##fs) \
{ \
/* |X| < |Y|, negate result. */ \
_FP_FRAC_SUB_##wc(R, Y, X); \
R##_s = Y##_s; \
} \
else if (_FP_FRAC_ZEROP_##wc(R)) \
{ \
R##_e = 0; \
R##_s = (FP_ROUNDMODE == FP_RND_MINF); \
goto sub_done; \
} \
goto norm; \
} \
sub3: \
if (_FP_FRAC_HIGH_##fs(R) & _FP_IMPLBIT_SH_##fs) \
{ \
int diff; \
/* Carry into most significant bit of larger one of X and Y, \
canceling it; renormalize. */ \
_FP_FRAC_HIGH_##fs(R) &= _FP_IMPLBIT_SH_##fs - 1; \
norm: \
_FP_FRAC_CLZ_##wc(diff, R); \
diff -= _FP_WFRACXBITS_##fs; \
_FP_FRAC_SLL_##wc(R, diff); \
if (R##_e <= diff) \
{ \
/* R is denormalized. */ \
diff = diff - R##_e + 1; \
_FP_FRAC_SRS_##wc(R, diff, _FP_WFRACBITS_##fs); \
R##_e = 0; \
} \
else \
{ \
R##_e -= diff; \
_FP_FRAC_HIGH_##fs(R) &= ~(_FP_W_TYPE)_FP_IMPLBIT_SH_##fs; \
} \
} \
sub_done: ; \
} \
} while (0)
#define _FP_ADD(fs, wc, R, X, Y) _FP_ADD_INTERNAL(fs, wc, R, X, Y, '+')
#define _FP_SUB(fs, wc, R, X, Y) \
do { \
if (!(Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y))) Y##_s ^= 1; \
_FP_ADD_INTERNAL(fs, wc, R, X, Y, '-'); \
} while (0)
/*
* Main negation routine. FIXME -- when we care about setting exception
* bits reliably, this will not do. We should examine all of the fp classes.
*/
#define _FP_NEG(fs, wc, R, X) \
do { \
_FP_FRAC_COPY_##wc(R, X); \
R##_c = X##_c; \
R##_e = X##_e; \
R##_s = 1 ^ X##_s; \
} while (0)
/*
* Main multiplication routine. The input values should be cooked.
*/
#define _FP_MUL(fs, wc, R, X, Y) \
do { \
R##_s = X##_s ^ Y##_s; \
switch (_FP_CLS_COMBINE(X##_c, Y##_c)) \
{ \
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NORMAL): \
R##_c = FP_CLS_NORMAL; \
R##_e = X##_e + Y##_e + 1; \
\
_FP_MUL_MEAT_##fs(R,X,Y); \
\
if (_FP_FRAC_OVERP_##wc(fs, R)) \
_FP_FRAC_SRS_##wc(R, 1, _FP_WFRACBITS_##fs); \
else \
R##_e--; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NAN): \
_FP_CHOOSENAN(fs, wc, R, X, Y, '*'); \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NORMAL): \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_ZERO): \
R##_s = X##_s; \
\
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NORMAL): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NORMAL): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO): \
_FP_FRAC_COPY_##wc(R, X); \
R##_c = X##_c; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NAN): \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NAN): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NAN): \
R##_s = Y##_s; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_ZERO): \
_FP_FRAC_COPY_##wc(R, Y); \
R##_c = Y##_c; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_ZERO): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_INF): \
R##_s = _FP_NANSIGN_##fs; \
R##_c = FP_CLS_NAN; \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
FP_SET_EXCEPTION(FP_EX_INVALID); \
break; \
\
default: \
abort(); \
} \
} while (0)
/*
* Main division routine. The input values should be cooked.
*/
#define _FP_DIV(fs, wc, R, X, Y) \
do { \
R##_s = X##_s ^ Y##_s; \
switch (_FP_CLS_COMBINE(X##_c, Y##_c)) \
{ \
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NORMAL): \
R##_c = FP_CLS_NORMAL; \
R##_e = X##_e - Y##_e; \
\
_FP_DIV_MEAT_##fs(R,X,Y); \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NAN): \
_FP_CHOOSENAN(fs, wc, R, X, Y, '/'); \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_NORMAL): \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_NAN,FP_CLS_ZERO): \
R##_s = X##_s; \
_FP_FRAC_COPY_##wc(R, X); \
R##_c = X##_c; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_NAN): \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NAN): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NAN): \
R##_s = Y##_s; \
_FP_FRAC_COPY_##wc(R, Y); \
R##_c = Y##_c; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_NORMAL): \
R##_c = FP_CLS_ZERO; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_NORMAL,FP_CLS_ZERO): \
FP_SET_EXCEPTION(FP_EX_DIVZERO); \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_ZERO): \
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_NORMAL): \
R##_c = FP_CLS_INF; \
break; \
\
case _FP_CLS_COMBINE(FP_CLS_INF,FP_CLS_INF): \
case _FP_CLS_COMBINE(FP_CLS_ZERO,FP_CLS_ZERO): \
R##_s = _FP_NANSIGN_##fs; \
R##_c = FP_CLS_NAN; \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
FP_SET_EXCEPTION(FP_EX_INVALID); \
break; \
\
default: \
abort(); \
} \
} while (0)
/*
* Main differential comparison routine. The inputs should be raw not
* cooked. The return is -1,0,1 for normal values, 2 otherwise.
*/
#define _FP_CMP(fs, wc, ret, X, Y, un) \
do { \
/* NANs are unordered */ \
if ((X##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(X)) \
|| (Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y))) \
{ \
ret = un; \
} \
else \
{ \
int __is_zero_x; \
int __is_zero_y; \
\
__is_zero_x = (!X##_e && _FP_FRAC_ZEROP_##wc(X)) ? 1 : 0; \
__is_zero_y = (!Y##_e && _FP_FRAC_ZEROP_##wc(Y)) ? 1 : 0; \
\
if (__is_zero_x && __is_zero_y) \
ret = 0; \
else if (__is_zero_x) \
ret = Y##_s ? 1 : -1; \
else if (__is_zero_y) \
ret = X##_s ? -1 : 1; \
else if (X##_s != Y##_s) \
ret = X##_s ? -1 : 1; \
else if (X##_e > Y##_e) \
ret = X##_s ? -1 : 1; \
else if (X##_e < Y##_e) \
ret = X##_s ? 1 : -1; \
else if (_FP_FRAC_GT_##wc(X, Y)) \
ret = X##_s ? -1 : 1; \
else if (_FP_FRAC_GT_##wc(Y, X)) \
ret = X##_s ? 1 : -1; \
else \
ret = 0; \
} \
} while (0)
/* Simplification for strict equality. */
#define _FP_CMP_EQ(fs, wc, ret, X, Y) \
do { \
/* NANs are unordered */ \
if ((X##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(X)) \
|| (Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y))) \
{ \
ret = 1; \
} \
else \
{ \
ret = !(X##_e == Y##_e \
&& _FP_FRAC_EQ_##wc(X, Y) \
&& (X##_s == Y##_s || (!X##_e && _FP_FRAC_ZEROP_##wc(X)))); \
} \
} while (0)
/* Version to test unordered. */
#define _FP_CMP_UNORD(fs, wc, ret, X, Y) \
do { \
ret = ((X##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(X)) \
|| (Y##_e == _FP_EXPMAX_##fs && !_FP_FRAC_ZEROP_##wc(Y))); \
} while (0)
/*
* Main square root routine. The input value should be cooked.
*/
#define _FP_SQRT(fs, wc, R, X) \
do { \
_FP_FRAC_DECL_##wc(T); _FP_FRAC_DECL_##wc(S); \
_FP_W_TYPE q; \
switch (X##_c) \
{ \
case FP_CLS_NAN: \
_FP_FRAC_COPY_##wc(R, X); \
R##_s = X##_s; \
R##_c = FP_CLS_NAN; \
break; \
case FP_CLS_INF: \
if (X##_s) \
{ \
R##_s = _FP_NANSIGN_##fs; \
R##_c = FP_CLS_NAN; /* NAN */ \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
FP_SET_EXCEPTION(FP_EX_INVALID); \
} \
else \
{ \
R##_s = 0; \
R##_c = FP_CLS_INF; /* sqrt(+inf) = +inf */ \
} \
break; \
case FP_CLS_ZERO: \
R##_s = X##_s; \
R##_c = FP_CLS_ZERO; /* sqrt(+-0) = +-0 */ \
break; \
case FP_CLS_NORMAL: \
R##_s = 0; \
if (X##_s) \
{ \
R##_c = FP_CLS_NAN; /* sNAN */ \
R##_s = _FP_NANSIGN_##fs; \
_FP_FRAC_SET_##wc(R, _FP_NANFRAC_##fs); \
FP_SET_EXCEPTION(FP_EX_INVALID); \
break; \
} \
R##_c = FP_CLS_NORMAL; \
if (X##_e & 1) \
_FP_FRAC_SLL_##wc(X, 1); \
R##_e = X##_e >> 1; \
_FP_FRAC_SET_##wc(S, _FP_ZEROFRAC_##wc); \
_FP_FRAC_SET_##wc(R, _FP_ZEROFRAC_##wc); \
q = _FP_OVERFLOW_##fs >> 1; \
_FP_SQRT_MEAT_##wc(R, S, T, X, q); \
} \
} while (0)
/*
* Convert from FP to integer. Input is raw.
*/
/* RSIGNED can have following values:
* 0: the number is required to be 0..(2^rsize)-1, if not, NV is set plus
* the result is either 0 or (2^rsize)-1 depending on the sign in such
* case.
* 1: the number is required to be -(2^(rsize-1))..(2^(rsize-1))-1, if not,
* NV is set plus the result is either -(2^(rsize-1)) or (2^(rsize-1))-1
* depending on the sign in such case.
* -1: the number is required to be -(2^(rsize-1))..(2^rsize)-1, if not, NV is
* set plus the result is either -(2^(rsize-1)) or (2^(rsize-1))-1
* depending on the sign in such case.
*/
#define _FP_TO_INT(fs, wc, r, X, rsize, rsigned) \
do { \
if (X##_e < _FP_EXPBIAS_##fs) \
{ \
r = 0; \
if (X##_e == 0) \
{ \
if (!_FP_FRAC_ZEROP_##wc(X)) \
{ \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
FP_SET_EXCEPTION(FP_EX_DENORM); \
} \
} \
else \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
} \
else if (X##_e >= _FP_EXPBIAS_##fs + rsize - (rsigned > 0 || X##_s) \
|| (!rsigned && X##_s)) \
{ \
/* Overflow or converting to the most negative integer. */ \
if (rsigned) \
{ \
r = 1; \
r <<= rsize - 1; \
r -= 1 - X##_s; \
} else { \
r = 0; \
if (X##_s) \
r = ~r; \
} \
\
if (rsigned && X##_s && X##_e == _FP_EXPBIAS_##fs + rsize - 1) \
{ \
/* Possibly converting to most negative integer; check the \
mantissa. */ \
int inexact = 0; \
(void)((_FP_FRACBITS_##fs > rsize) \
? ({ _FP_FRAC_SRST_##wc(X, inexact, \
_FP_FRACBITS_##fs - rsize, \
_FP_FRACBITS_##fs); 0; }) \
: 0); \
if (!_FP_FRAC_ZEROP_##wc(X)) \
FP_SET_EXCEPTION(FP_EX_INVALID); \
else if (inexact) \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
} \
else \
FP_SET_EXCEPTION(FP_EX_INVALID); \
} \
else \
{ \
_FP_FRAC_HIGH_RAW_##fs(X) |= _FP_IMPLBIT_##fs; \
if (X##_e >= _FP_EXPBIAS_##fs + _FP_FRACBITS_##fs - 1) \
{ \
_FP_FRAC_ASSEMBLE_##wc(r, X, rsize); \
r <<= X##_e - _FP_EXPBIAS_##fs - _FP_FRACBITS_##fs + 1; \
} \
else \
{ \
int inexact; \
_FP_FRAC_SRST_##wc(X, inexact, \
(_FP_FRACBITS_##fs + _FP_EXPBIAS_##fs - 1 \
- X##_e), \
_FP_FRACBITS_##fs); \
if (inexact) \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
_FP_FRAC_ASSEMBLE_##wc(r, X, rsize); \
} \
if (rsigned && X##_s) \
r = -r; \
} \
} while (0)
/* Convert integer to fp. Output is raw. RTYPE is unsigned even if
input is signed. */
#define _FP_FROM_INT(fs, wc, X, r, rsize, rtype) \
do { \
if (r) \
{ \
rtype ur_; \
\
if ((X##_s = (r < 0))) \
r = -(rtype)r; \
\
ur_ = (rtype) r; \
(void)((rsize <= _FP_W_TYPE_SIZE) \
? ({ \
int lz_; \
__FP_CLZ(lz_, (_FP_W_TYPE)ur_); \
X##_e = _FP_EXPBIAS_##fs + _FP_W_TYPE_SIZE - 1 - lz_; \
}) \
: ((rsize <= 2 * _FP_W_TYPE_SIZE) \
? ({ \
int lz_; \
__FP_CLZ_2(lz_, (_FP_W_TYPE)(ur_ >> _FP_W_TYPE_SIZE), \
(_FP_W_TYPE)ur_); \
X##_e = (_FP_EXPBIAS_##fs + 2 * _FP_W_TYPE_SIZE - 1 \
- lz_); \
}) \
: (abort(), 0))); \
\
if (rsize - 1 + _FP_EXPBIAS_##fs >= _FP_EXPMAX_##fs \
&& X##_e >= _FP_EXPMAX_##fs) \
{ \
/* Exponent too big; overflow to infinity. (May also \
happen after rounding below.) */ \
_FP_OVERFLOW_SEMIRAW(fs, wc, X); \
goto pack_semiraw; \
} \
\
if (rsize <= _FP_FRACBITS_##fs \
|| X##_e < _FP_EXPBIAS_##fs + _FP_FRACBITS_##fs) \
{ \
/* Exactly representable; shift left. */ \
_FP_FRAC_DISASSEMBLE_##wc(X, ur_, rsize); \
_FP_FRAC_SLL_##wc(X, (_FP_EXPBIAS_##fs \
+ _FP_FRACBITS_##fs - 1 - X##_e)); \
} \
else \
{ \
/* More bits in integer than in floating type; need to \
round. */ \
if (_FP_EXPBIAS_##fs + _FP_WFRACBITS_##fs - 1 < X##_e) \
ur_ = ((ur_ >> (X##_e - _FP_EXPBIAS_##fs \
- _FP_WFRACBITS_##fs + 1)) \
| ((ur_ << (rsize - (X##_e - _FP_EXPBIAS_##fs \
- _FP_WFRACBITS_##fs + 1))) \
!= 0)); \
_FP_FRAC_DISASSEMBLE_##wc(X, ur_, rsize); \
if ((_FP_EXPBIAS_##fs + _FP_WFRACBITS_##fs - 1 - X##_e) > 0) \
_FP_FRAC_SLL_##wc(X, (_FP_EXPBIAS_##fs \
+ _FP_WFRACBITS_##fs - 1 - X##_e)); \
_FP_FRAC_HIGH_##fs(X) &= ~(_FP_W_TYPE)_FP_IMPLBIT_SH_##fs; \
pack_semiraw: \
_FP_PACK_SEMIRAW(fs, wc, X); \
} \
} \
else \
{ \
X##_s = 0; \
X##_e = 0; \
_FP_FRAC_SET_##wc(X, _FP_ZEROFRAC_##wc); \
} \
} while (0)
/* Extend from a narrower floating-point format to a wider one. Input
and output are raw. */
#define FP_EXTEND(dfs,sfs,dwc,swc,D,S) \
do { \
if (_FP_FRACBITS_##dfs < _FP_FRACBITS_##sfs \
|| (_FP_EXPMAX_##dfs - _FP_EXPBIAS_##dfs \
< _FP_EXPMAX_##sfs - _FP_EXPBIAS_##sfs) \
|| _FP_EXPBIAS_##dfs < _FP_EXPBIAS_##sfs + _FP_FRACBITS_##sfs - 1) \
abort(); \
D##_s = S##_s; \
_FP_FRAC_COPY_##dwc##_##swc(D, S); \
if (_FP_EXP_NORMAL(sfs, swc, S)) \
{ \
D##_e = S##_e + _FP_EXPBIAS_##dfs - _FP_EXPBIAS_##sfs; \
_FP_FRAC_SLL_##dwc(D, (_FP_FRACBITS_##dfs - _FP_FRACBITS_##sfs)); \
} \
else \
{ \
if (S##_e == 0) \
{ \
if (_FP_FRAC_ZEROP_##swc(S)) \
D##_e = 0; \
else \
{ \
int _lz; \
FP_SET_EXCEPTION(FP_EX_DENORM); \
_FP_FRAC_CLZ_##swc(_lz, S); \
_FP_FRAC_SLL_##dwc(D, \
_lz + _FP_FRACBITS_##dfs \
- _FP_FRACTBITS_##sfs); \
D##_e = (_FP_EXPBIAS_##dfs - _FP_EXPBIAS_##sfs + 1 \
+ _FP_FRACXBITS_##sfs - _lz); \
} \
} \
else \
{ \
D##_e = _FP_EXPMAX_##dfs; \
if (!_FP_FRAC_ZEROP_##swc(S)) \
{ \
if (!(_FP_FRAC_HIGH_RAW_##sfs(S) & _FP_QNANBIT_##sfs)) \
FP_SET_EXCEPTION(FP_EX_INVALID); \
_FP_FRAC_SLL_##dwc(D, (_FP_FRACBITS_##dfs \
- _FP_FRACBITS_##sfs)); \
} \
} \
} \
} while (0)
/* Truncate from a wider floating-point format to a narrower one.
Input and output are semi-raw. */
#define FP_TRUNC(dfs,sfs,dwc,swc,D,S) \
do { \
if (_FP_FRACBITS_##sfs < _FP_FRACBITS_##dfs \
|| _FP_EXPBIAS_##sfs < _FP_EXPBIAS_##dfs + _FP_FRACBITS_##dfs - 1) \
abort(); \
D##_s = S##_s; \
if (_FP_EXP_NORMAL(sfs, swc, S)) \
{ \
D##_e = S##_e + _FP_EXPBIAS_##dfs - _FP_EXPBIAS_##sfs; \
if (D##_e >= _FP_EXPMAX_##dfs) \
_FP_OVERFLOW_SEMIRAW(dfs, dwc, D); \
else \
{ \
if (D##_e <= 0) \
{ \
if (D##_e <= 1 - _FP_FRACBITS_##dfs) \
_FP_FRAC_SET_##swc(S, _FP_ZEROFRAC_##swc); \
else \
{ \
_FP_FRAC_HIGH_##sfs(S) |= _FP_IMPLBIT_SH_##sfs; \
_FP_FRAC_SRS_##swc(S, (_FP_WFRACBITS_##sfs \
- _FP_WFRACBITS_##dfs + 1 - D##_e), \
_FP_WFRACBITS_##sfs); \
} \
D##_e = 0; \
} \
else \
_FP_FRAC_SRS_##swc(S, (_FP_WFRACBITS_##sfs \
- _FP_WFRACBITS_##dfs), \
_FP_WFRACBITS_##sfs); \
_FP_FRAC_COPY_##dwc##_##swc(D, S); \
} \
} \
else \
{ \
if (S##_e == 0) \
{ \
D##_e = 0; \
_FP_FRAC_SET_##dwc(D, _FP_ZEROFRAC_##dwc); \
if (!_FP_FRAC_ZEROP_##swc(S)) \
{ \
FP_SET_EXCEPTION(FP_EX_DENORM); \
FP_SET_EXCEPTION(FP_EX_INEXACT); \
} \
} \
else \
{ \
D##_e = _FP_EXPMAX_##dfs; \
if (_FP_FRAC_ZEROP_##swc(S)) \
_FP_FRAC_SET_##dwc(D, _FP_ZEROFRAC_##dwc); \
else \
{ \
_FP_CHECK_SIGNAN_SEMIRAW(sfs, swc, S); \
_FP_FRAC_SRL_##swc(S, (_FP_WFRACBITS_##sfs \
- _FP_WFRACBITS_##dfs)); \
_FP_FRAC_COPY_##dwc##_##swc(D, S); \
_FP_FRAC_HIGH_##dfs(D) |= _FP_QNANBIT_SH_##dfs; \
} \
} \
} \
} while (0)
/*
* Helper primitives.
*/
/* Count leading zeros in a word. */
#ifndef __FP_CLZ
/* GCC 3.4 and later provide the builtins for us. */
#define __FP_CLZ(r, x) \
do { \
if (sizeof (_FP_W_TYPE) == sizeof (unsigned int)) \
r = __builtin_clz (x); \
else if (sizeof (_FP_W_TYPE) == sizeof (unsigned long)) \
r = __builtin_clzl (x); \
else if (sizeof (_FP_W_TYPE) == sizeof (unsigned long long)) \
r = __builtin_clzll (x); \
else \
abort (); \
} while (0)
#endif /* ndef __FP_CLZ */
#define _FP_DIV_HELP_imm(q, r, n, d) \
do { \
q = n / d, r = n % d; \
} while (0)
/* A restoring bit-by-bit division primitive. */
#define _FP_DIV_MEAT_N_loop(fs, wc, R, X, Y) \
do { \
int count = _FP_WFRACBITS_##fs; \
_FP_FRAC_DECL_##wc (u); \
_FP_FRAC_DECL_##wc (v); \
_FP_FRAC_COPY_##wc (u, X); \
_FP_FRAC_COPY_##wc (v, Y); \
_FP_FRAC_SET_##wc (R, _FP_ZEROFRAC_##wc); \
/* Normalize U and V. */ \
_FP_FRAC_SLL_##wc (u, _FP_WFRACXBITS_##fs); \
_FP_FRAC_SLL_##wc (v, _FP_WFRACXBITS_##fs); \
/* First round. Since the operands are normalized, either the \
first or second bit will be set in the fraction. Produce a \
normalized result by checking which and adjusting the loop \
count and exponent accordingly. */ \
if (_FP_FRAC_GE_1 (u, v)) \
{ \
_FP_FRAC_SUB_##wc (u, u, v); \
_FP_FRAC_LOW_##wc (R) |= 1; \
count--; \
} \
else \
R##_e--; \
/* Subsequent rounds. */ \
do { \
int msb = (_FP_WS_TYPE) _FP_FRAC_HIGH_##wc (u) < 0; \
_FP_FRAC_SLL_##wc (u, 1); \
_FP_FRAC_SLL_##wc (R, 1); \
if (msb || _FP_FRAC_GE_1 (u, v)) \
{ \
_FP_FRAC_SUB_##wc (u, u, v); \
_FP_FRAC_LOW_##wc (R) |= 1; \
} \
} while (--count > 0); \
/* If there's anything left in U, the result is inexact. */ \
_FP_FRAC_LOW_##wc (R) |= !_FP_FRAC_ZEROP_##wc (u); \
} while (0)
#define _FP_DIV_MEAT_1_loop(fs, R, X, Y) _FP_DIV_MEAT_N_loop (fs, 1, R, X, Y)
#define _FP_DIV_MEAT_2_loop(fs, R, X, Y) _FP_DIV_MEAT_N_loop (fs, 2, R, X, Y)
#define _FP_DIV_MEAT_4_loop(fs, R, X, Y) _FP_DIV_MEAT_N_loop (fs, 4, R, X, Y)