diff options
Diffstat (limited to 'mpi')
125 files changed, 17016 insertions, 0 deletions
diff --git a/mpi/ChangeLog b/mpi/ChangeLog new file mode 100644 index 0000000..a2af9f3 --- /dev/null +++ b/mpi/ChangeLog @@ -0,0 +1,505 @@ +2005-09-01 David Shaw <dshaw@jabberwocky.com> + + * mpicoder.c (mpi_read): Fix minor bug in reading a zero-length + MPI (was failing unnecessarily). + +2005-05-06 Werner Koch <wk@g10code.com> + + * mpi-scan.c (mpi_putbyte, mpi_getbyte): Removed. Not used. + +2005-04-21 Werner Koch <wk@g10code.com> + + * mpicoder.c (mpi_read): Changed error detection to always return + an error while maintaining the actual number of bytes read. + +2005-03-11 Werner Koch <wk@g10code.com> + + * Makefile.am (ASFLAGS): Renamed to AM_CCASFLAGS and added the + variable for non exectubale stack options. Adapted users. + +2004-12-20 Werner Koch <wk@g10code.com> + + * mpicoder.c (mpi_read_from_buffer): Don't abort in case of an + invalid MPI but print a message and return NULL. Use log_info and + not log_error. + +2004-10-26 Werner Koch <wk@g10code.com> + + * config.links: Use HOST instead of TARGET. + +2004-10-12 Werner Koch <wk@g10code.com> + + * Makefile.am (.S.o): Include MPI_SFLAGS. This is our bug 145. + +2004-09-29 David Shaw <dshaw@jabberwocky.com> + + * mpicoder.c (mpi_read): If we must fail due to a oversize + (generally corrupt) MPI, make sure the number of bytes we read is + valid so we can skip the rest of the bad packet (in hopes the + whole stream isn't invalid). + +2004-05-20 David Shaw <dshaw@jabberwocky.com> + + * longlong.h: Typo. + + * mpi-pow.c (mpi_powm): s/exp/exponent/ to shutup a compiler + warning. From Werner on stable branch. + +2004-01-20 David Shaw <dshaw@jabberwocky.com> + + * hppa1.1/udiv-qrnnd.S: Alignment fix from Lamont Jones for + Debian. + +2004-01-11 David Shaw <dshaw@jabberwocky.com> + + * config.links: OpenBSD 3.4 is now ELF, so use the proper + assembler code for that. Use the portable C MPI code for OpenBSD + before 3.4, and remove the special i386-openbsd assembly + directory. + + * Makefile.am: Add the portable C links to DISTCLEANFILES. Noted + by Nelson H. F. Beebe. + + * mpi-mpow.c (build_index): s/index/idx/ to avoid gcc warning. + From Werner on stable branch. + + * longlong.h: Added PowerPC 64 bit code from GPM-4.1.2 but didn't + enable it yet. From Werner on stable branch. + +2003-12-29 David Shaw <dshaw@jabberwocky.com> + + * g10m.c: Dead code. Remove. + + * Makefile.am: Don't compile g10m.c. + +2003-12-17 David Shaw <dshaw@localhost.localdomain> + + * mpiutil.c (mpi_set_opaque, mpi_get_opaque): Make the length of + an opaque MPI unsigned. + +2003-12-04 David Shaw <dshaw@jabberwocky.com> + + * config.links: Show target in asm-syntax.h file and include + targets for K*BSD (GNU userland with BSD kernel). + +2003-11-20 David Shaw <dshaw@jabberwocky.com> + + * config.links: sparc64 assembler doesn't work on FreeBSD or + OpenBSD either. + +2003-05-26 David Shaw <dshaw@jabberwocky.com> + + * Makefile.am: Make use of AM_CFLAGS. (From wk on stable branch) + +2003-05-24 David Shaw <dshaw@jabberwocky.com> + + * mpicoder.c, mpi-inline.h, mpi-inv.c, mpiutil.c, mpih-div.c, + mpi-internal.h, mpi-scan.c: Edit all preprocessor instructions to + remove whitespace before the '#'. This is not required by C89, + but there are some compilers out there that don't like it. + +2003-04-25 David Shaw <dshaw@jabberwocky.com> + + * config.links: Re-disable assembler on Darwin. Darwin 6.5 broke + it again. + +2003-04-15 Werner Koch <wk@gnupg.org> + + * longlong.h (umul_ppmm): Support SH3 and SH4. Thanks to + kazuya.s@jp.yokogawa.com. + +2002-11-16 David Shaw <dshaw@jabberwocky.com> + + * config.links: Use OpenBSD/NetBSD powerpc assembler code for + Darwin. Successfully tested by Gordon Worley. + +2002-10-17 Werner Koch <wk@gnupg.org> + + * config.links (powerpc-apple-darwin6.1): Disable assembler + due to non-working modules/as. Suggested by Gordon Worley. + +2002-10-02 David Shaw <dshaw@jabberwocky.com> + + * longlong.h: Some whitespace changes in HPPA to fix assembler + problems on HP-UX. From David Ellement. + +2002-09-20 Werner Koch <wk@gnupg.org> + + * mpicoder.c (do_get_buffer): Avoid zero length allocation. + Checked that all callers behave properly when NBYTES returns 0 as + the length of the allocated buffer. + +2002-09-10 Werner Koch <wk@gnupg.org> + + * mpi-bit.c (mpi_normalize): Replaced the check for protected by + is_opaque. + (mpi_get_nbits): Removed the special case for protected MPIs. + * mpicoder.c (do_get_buffer): Likewise. + (mpi_print): Removed the nbit_info printing. + +2002-09-03 Werner Koch <wk@gnupg.org> + + * mpicoder.c (mpi_set_buffer): Cast all left operands of a shift + to a larger type so that 16 bit CPUs don't suffer from an + overflow. Suggested by TOGAWA Satoshi. + +2002-08-24 David Shaw <dshaw@jabberwocky.com> + + * longlong.h: Remove space between \ and newline. gcc is + complaining. + +2002-08-13 Werner Koch <wk@gnupg.org> + + * mpicoder.c (do_get_buffer): Don't remove leading zeros if the + MPI is marked as protected. + +2002-08-02 Timo Schulz <ts@winpt.org> + + * mpicoder.c: Add a '\n' to all log_bug functions. + +2002-08-01 Werner Koch <wk@gnupg.org> + + * config.links: Added case for sparc64-netbsd. Suggested by + ww@styx.org. + +2002-07-25 David Shaw <dshaw@jabberwocky.com> + + * config.links: Add special rule for OpenBSD on x86 to use special + i386-openbsd files. OpenBSD (at least until version 3.1) has an + older assembler that won't work with the files in i386. + +2002-07-24 Stefan Bellon <sbellon@sbellon.de> + + * longlong.h [__riscos__]: Removed #pragma which is not needed + anymore since the K&R multiline strings are gone. + +2002-07-24 Werner Koch <wk@gnupg.org> + + * longlong.h: Replaced all K&R multiline strings by ISO ones for + the sake of modern compilers. Suggested by Marco Parrone. + +2002-05-10 Stefan Bellon <sbellon@sbellon.de> + + * mpiutil.c (mpi_alloc_like/mpi_debug_alloc_like): Added code + for M_DEBUG. + +2002-04-18 Werner Koch <wk@gnupg.org> + + * i386/syntax.h (ALIGN): Removed parens from definition + * i386/mpih-add1.S, i386/mpih-sub1.S, i386/mpih-rshift.S, + i386/mpih-lshift.S: Minor syntax changes suggested by Mark Pettit + after comparing the files with those for GMP 4. + +2001-11-08 Werner Koch <wk@gnupg.org> + + * config.links (mpi_sflags): Add extra rule for sparc64-sun-solaris2. + +2001-08-20 Werner Koch <wk@gnupg.org> + + * longlong.h [__riscos__]: Need a special pragma here. + +2001-08-09 Werner Koch <wk@gnupg.org> + + * config.links: Added configuraton for powerpc-openbsd. By Peter + Valchev + +2001-07-09 Werner Koch <wk@gnupg.org> + + * config.links: Changed the way the list of files to be + symlinked is returned. + +2001-05-27 Werner Koch <wk@gnupg.org> + + * hppa/, hppa1.1/, pa7100/ : Use .label command instead of labels + because there syntax changed. By Matthew Wilcox. + +2001-05-06 Werner Koch <wk@gnupg.org> + + * longlong.h: Fixes for ARM by Phil Blundell. + +2001-04-17 Werner Koch <wk@gnupg.org> + + Updated copyright notices. + +2001-03-24 Werner Koch <wk@gnupg.org> + + * mpi-mul.c (mpi_mul): Make sure that secret temporary results are + not stored in w. Suggested by Florian Weimer. + +2001-03-18 Werner Koch <wk@gnupg.org> + + * config.links: Use i386 code for i386. According to tests by + Kevin Ryde the i586 code runs slow on i386 CPUs. Ditto for i786. + +2000-10-24 Werner Koch <wk@gnupg.org> + + * mips3/: Changed a few comments to C-style. By Jeff Long. + +2000-10-13 Werner Koch <wk@gnupg.org> + + * mpi.h: Removed the inclusion of mpi-asm-defs.h because this + makes some trouble when doing a VPATH build. configure now + takes care of it. + +2000-10-12 Werner Koch <wk@gnupg.org> + + * generic/mpi-asm-defs.h: New. + * mips3/mpi-asm-defs.h: New. + * config.links: Create a link to one of the above files. + +Wed Jul 19 11:26:43 CEST 2000 Werner Koch <wk@> + + * config.links: Support for powerpc--netbsd by Gabriel Rosenkoetter. + +Wed Mar 22 13:50:24 CET 2000 Werner Koch <wk@openit.de> + + * config.links: Add support for FreeBSD 5 and made the case stmt + looking nicer. From Jun Kuriyama. + +Fri Mar 17 17:50:25 CET 2000 Werner Koch <wk@openit.de> + + * config.links (sparc64-unknown-linux-gnu): use udic module. + From Adam Mitchell. + +2000-03-14 12:03:56 Werner Koch (wk@habibti.openit.de) + + * Makefile.am: Do not use .s and .S files but a temp names, so that + OSes with caseinsensitive filenames do work. From Frank Donahoe. + +Tue Mar 7 18:45:31 CET 2000 Werner Koch <wk@gnupg.de> + + * mpih-mul.c (mpihelp_mul_karatsuba_case): It seems that the + untested part works fine. Removed the debugging message. + + * longlong.h (umul_ppmm): Fixes for ARM-4. By Sean MacLennan. + + * config.links: Add support for NetBSD. + +Thu Jan 13 19:31:58 CET 2000 Werner Koch <wk@gnupg.de> + + * mpi-internal.h (karatsuba_ctx): New. + * mpih-mul.c (mpihelp_release_karatsuba_ctx): New. + (mpihelp_mul_karatsuba_case): New. + (mpihelp_mul): Splitted to make use of the new functions. + * mpi-pow.c (mpi_powm): Make use of the new splitted function + to avoid multiple allocation of temporary memory during the + karatsuba operations. + + * mpi_mpow.c: Removed the unused Barrett code. + +Sun Dec 19 15:22:26 CET 1999 Werner Koch <wk@gnupg.de> + + * power/ : Converted more comments to C comments because some AS + complain about ' in comments. + +Thu Dec 16 10:07:58 CET 1999 Werner Koch <wk@gnupg.de> + + * Makefile.am: c/SFLAGS/ASFLAGS/. This has only been used by the + powerpc and actually never passed the -Wa,foo to the cc. + +Thu Dec 9 10:31:05 CET 1999 Werner Koch <wk@gnupg.de> + + * power/: Add all files from GMP for this CPU. + + * config.links: Support for BSDI 4.x. By Wayne Chapeskie. + (sparc8): Made the search path the same as sparc9 + + * mpih-div.c (mpihelp_divrem): The MPN_COPY_DECR copied one + elemnat too many. This is gmp2.0.2p9.txt patch. + +Sat Oct 9 20:34:41 CEST 1999 Werner Koch <wk@gnupg.de> + + * Makefile.am: Removed libtool. + +Mon Aug 30 20:38:33 CEST 1999 Werner Koch <wk@isil.d.shuttle.de> + + * config.links: Add case label for DJGPP + +Wed Jul 14 19:42:08 CEST 1999 Werner Koch <wk@isil.d.shuttle.de> + + + * Makefile.am: Use .s files as temporaries, disabled other .S rules. + +Wed Jul 7 13:08:40 CEST 1999 Werner Koch <wk@isil.d.shuttle.de> + + + * mpicoder.c (g10_log_mpidump): New. + + * Makefile.am: Support for libtool. + +Fri Jul 2 11:45:54 CEST 1999 Werner Koch <wk@isil.d.shuttle.de> + + + * mpi-bit.c (mpi_lshift_limbs,mpi_rshift_limbs): New. + * mpi-mpow.c (barrett_mulm): New but diabled. + +Tue Jun 1 16:01:46 CEST 1999 Werner Koch <wk@isil.d.shuttle.de> + + * config.links (i[56]86*-*-freebsdelf*): New. + +Sun May 23 14:20:22 CEST 1999 Werner Koch <wk@isil.d.shuttle.de> + + * config.links (sysdep.h): Not any more conditionally created. + +Tue May 4 15:47:53 CEST 1999 Werner Koch <wk@isil.d.shuttle.de> + + * mpiutil.c (mpi_alloc_like): New. + +Mon Apr 26 17:48:15 CEST 1999 Werner Koch <wk@isil.d.shuttle.de> + + * mpih-add.c, mpih-sub.c: Removed + * mpi-inline.c: New. + * mpi-inline.h: Make it usable by mpi-inline.c. + +Sun Apr 18 10:11:28 CEST 1999 Werner Koch <wk@isil.d.shuttle.de> + + * mpih-mul.c (mpihelp_mul_n): Fixed use of memory region. + (mpihelp_mul): Ditto. + +Wed Apr 7 20:51:39 CEST 1999 Werner Koch <wk@isil.d.shuttle.de> + + * Makefile.am: Explicit rules to invoke cpp on *.S + +Mon Mar 8 20:47:17 CET 1999 Werner Koch <wk@isil.d.shuttle.de> + + * config.links: Take advantage of the with_symbol_underscore macro. + Add support for freebsd 4. + +Wed Feb 24 11:07:27 CET 1999 Werner Koch <wk@isil.d.shuttle.de> + + * mips3/mpih-sub1.S: Removed left over junk in last line. (Should I + blame me or my editor?). + +Sat Feb 13 12:04:43 CET 1999 Werner Koch <wk@isil.d.shuttle.de> + + * Makefile.am: Removed the +=. Add MPI_OPT_FLAGS. + +Sat Jan 9 16:02:23 CET 1999 Werner Koch <wk@isil.d.shuttle.de> + + * mpi-cmp.c (mpi_cmp_ui): Normalized the arg. + +Thu Jan 7 18:00:58 CET 1999 Werner Koch <wk@isil.d.shuttle.de> + + * mpi-bit.c (mpi_normalize): New. + (mpi_get_nbits): Normalize the MPI. + * mpi-bit.c (mpi_cmp): Normalize the MPI before the compare. + + +Tue Dec 8 13:15:16 CET 1998 Werner Koch <wk@isil.d.shuttle.de> + + * config.links: Moved the case for powerpc*linux + * powerpcp32/*.S: Removed some underscores. + +Thu Nov 26 07:27:52 1998 Werner Koch <werner.koch@guug.de> + + * config.links: Support for ppc with ELF + * powerpc32/syntax.h: New. + * powerpc32/*.S: Applied ELF patches (glibc patches) + +Tue Nov 10 19:31:37 1998 Werner Koch (wk@isil.d.shuttle.de) + + * power*/ : Started with stuff for PPC + * config.links: Some stuff for PPC. + * generic/udiv-w-sdiv.c: New but disabled. + +Tue Oct 27 12:37:46 1998 Werner Koch (wk@isil.d.shuttle.de) + + * config.links (freebsd): Fixes for FreeBSD 3.0 + +Wed Oct 14 09:59:30 1998 Werner Koch (wk@isil.d.shuttle.de) + + * config.links (freebsd): ELF patches from Jun Kuriyama. + +Thu Oct 8 13:28:17 1998 Werner Koch (wk@isil.d.shuttle.de) + + * mpi-mpow.c (mpi_mulpowm): Fixed mem leak (m_free/mpi_free). + +Thu Sep 17 18:08:50 1998 Werner Koch (wk@(none)) + + * hppa1.1/udiv-qrnnd.S: Fix from Steffen Zahn for HPUX 10.20 + +Thu Aug 6 16:39:28 1998 Werner Koch,mobil,,, (wk@tobold) + + * mpi-bit.c (mpi_set_bytes): Removed. + +Wed Aug 5 15:11:12 1998 Werner Koch (wk@(none)) + + * mpicoder.c (mpi_read_from_buffer): New. + + * mpiutil.c (mpi_set_opaque): New. + (mpi_get_opaque): New. + (mpi_copy): Changed to support opauqe flag + (mpi_free): Ditto. + +Sat Jul 4 10:11:11 1998 Werner Koch (wk@isil.d.shuttle.de) + + * mpiutil.c (mpi_clear): Reset flags. + (mpi_set): Ditto. + (mpi_alloc_secure): Set flag to 1 and not ored the 1 in, tsss.. + +Fri Jun 26 11:19:06 1998 Werner Koch (wk@isil.d.shuttle.de) + + * mpiutil.c (mpi_alloc): set nbits to 0. + (mpi_alloc_secure): Ditto. + (mpi_clear): Ditto. + +Thu Jun 25 11:50:01 1998 Werner Koch (wk@isil.d.shuttle.de) + + * mips3/*.S: New + +Mon May 18 13:47:06 1998 Werner Koch (wk@isil.d.shuttle.de) + + * config.links: split mpih-shift into mpih-[lr]shift and + changed all implementations. + * mpi/alpha: add some new assembler stuff. + +Wed May 13 11:04:29 1998 Werner Koch (wk@isil.d.shuttle.de) + + * config.links: Add support for MIPS + +Thu Apr 9 11:31:36 1998 Werner Koch (wk@isil.d.shuttle.de) + + * mpicoder.c (mpi_get_secure_buffer): New. + +Wed Apr 8 09:44:33 1998 Werner Koch (wk@isil.d.shuttle.de) + + * config.links: Applied small fix from Ulf Möller. + +Mon Apr 6 12:38:52 1998 Werner Koch (wk@isil.d.shuttle.de) + + * mpicoder.c (mpi_get_buffer): Removed returned leading zeroes + and changed all callers. + +Tue Mar 10 13:40:34 1998 Werner Koch (wk@isil.d.shuttle.de) + + * mpi-bit.c (mpi_clear_highbit): New. + +Mon Mar 2 19:29:00 1998 Werner Koch (wk@isil.d.shuttle.de) + + * Makefile.am (DISTCLEANFILES): New + +Thu Feb 26 06:48:54 1998 Werner Koch (wk@isil.d.shuttle.de) + + * config.links (X86_BROKEN_ALIGN): Added for some systems. + +Mon Feb 23 12:21:40 1998 Werner Koch (wk@isil.d.shuttle.de) + + * mpi/m68k/mpih-shift.S (Lspecial): Changed duplicate symbol. + +Mon Feb 16 13:00:27 1998 Werner Koch (wk@isil.d.shuttle.de) + + * config.links : Add detection of m68k cpus + + + + Copyright 1998, 1999, 2000, 2001 Free Software Foundation, Inc. + + This file is free software; as a special exception the author gives + unlimited permission to copy and/or distribute it, with or without + modifications, as long as this notice is preserved. + + This file is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY, to the extent permitted by law; without even the + implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. + + diff --git a/mpi/Makefile.am b/mpi/Makefile.am new file mode 100644 index 0000000..84d7757 --- /dev/null +++ b/mpi/Makefile.am @@ -0,0 +1,79 @@ +# Copyright (C) 1998, 1999, 2000, 2001, 2003 Free Software Foundation, Inc. +# +# This file is part of GnuPG. +# +# GnuPG is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 2 of the License, or +# (at your option) any later version. +# +# GnuPG 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 General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program; if not, write to the Free Software +# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA + +## Process this file with automake to produce Makefile.in + + +INCLUDES = -I.. -I$(top_srcdir)/include +AM_CFLAGS = $(MPI_OPT_FLAGS) +AM_CCASFLAGS = $(NOEXECSTACK_FLAGS) $(MPI_SFLAGS) + +EXTRA_DIST = config.links +DISTCLEANFILES = mpi-asm-defs.h \ + mpih-add1.S mpih-add1.c mpih-mul1.S mpih-mul1.c \ + mpih-mul2.S mpih-mul2.c mpih-mul3.S mpih-mul3.c \ + mpih-lshift.S mpih-lshift.c mpih-rshift.S mpih-rshift.c \ + mpih-sub1.S mpih-sub1.c asm-syntax.h sysdep.h + +# Note: we only use .S files so we should delete all left over .s +CLEANFILES = _*.s + +noinst_LIBRARIES = libmpi.a + +# libmpi_a_LDFLAGS = +libmpi_a_SOURCES = longlong.h \ + mpi-add.c \ + mpi-bit.c \ + mpi-cmp.c \ + mpi-div.c \ + mpi-gcd.c \ + mpi-internal.h \ + mpi-inline.h \ + mpi-inline.c \ + mpi-inv.c \ + mpi-mul.c \ + mpi-pow.c \ + mpi-mpow.c \ + mpi-scan.c \ + mpicoder.c \ + mpih-cmp.c \ + mpih-div.c \ + mpih-mul.c \ + mpiutil.c + +# Note this objects are actually links, the sourcefiles are +# distributed by special code in dist-hook +common_asm_objects = mpih-mul1.o \ + mpih-mul2.o \ + mpih-mul3.o \ + mpih-add1.o \ + mpih-sub1.o \ + mpih-lshift.o \ + mpih-rshift.o + +libmpi_a_DEPENDENCIES = $(common_asm_objects) @MPI_EXTRA_ASM_OBJS@ +libmpi_a_LIBADD = $(common_asm_objects) @MPI_EXTRA_ASM_OBJS@ + +# cancel the default rules used by libtool which do not really +# work and add one to cpp .S files +.S.o: + $(CPP) $(INCLUDES) $(DEFS) $< | grep -v '^#' > _$*.s + $(COMPILE) $(AM_CCASFLAGS) -c _$*.s + mv -f _$*.o $*.o + +.S.lo: diff --git a/mpi/Makefile.in b/mpi/Makefile.in new file mode 100644 index 0000000..fa4dc64 --- /dev/null +++ b/mpi/Makefile.in @@ -0,0 +1,606 @@ +# Makefile.in generated by automake 1.9.6 from Makefile.am. +# @configure_input@ + +# Copyright (C) 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, +# 2003, 2004, 2005 Free Software Foundation, Inc. +# This Makefile.in is free software; the Free Software Foundation +# gives unlimited permission to copy and/or distribute it, +# with or without modifications, as long as this notice is preserved. + +# This program is distributed in the hope that it will be useful, +# but WITHOUT ANY WARRANTY, to the extent permitted by law; without +# even the implied warranty of MERCHANTABILITY or FITNESS FOR A +# PARTICULAR PURPOSE. + +@SET_MAKE@ + +# Copyright (C) 1998, 1999, 2000, 2001, 2003 Free Software Foundation, Inc. +# +# This file is part of GnuPG. +# +# GnuPG is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 2 of the License, or +# (at your option) any later version. +# +# GnuPG 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 General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program; if not, write to the Free Software +# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA + +srcdir = @srcdir@ +top_srcdir = @top_srcdir@ +VPATH = @srcdir@ +pkgdatadir = $(datadir)/@PACKAGE@ +pkglibdir = $(libdir)/@PACKAGE@ +pkgincludedir = $(includedir)/@PACKAGE@ +top_builddir = .. +am__cd = CDPATH="$${ZSH_VERSION+.}$(PATH_SEPARATOR)" && cd +INSTALL = @INSTALL@ +install_sh_DATA = $(install_sh) -c -m 644 +install_sh_PROGRAM = $(install_sh) -c +install_sh_SCRIPT = $(install_sh) -c +INSTALL_HEADER = $(INSTALL_DATA) +transform = $(program_transform_name) +NORMAL_INSTALL = : +PRE_INSTALL = : +POST_INSTALL = : +NORMAL_UNINSTALL = : +PRE_UNINSTALL = : +POST_UNINSTALL = : +build_triplet = @build@ +host_triplet = @host@ +subdir = mpi +DIST_COMMON = $(srcdir)/Makefile.am $(srcdir)/Makefile.in ChangeLog +ACLOCAL_M4 = $(top_srcdir)/aclocal.m4 +am__aclocal_m4_deps = $(top_srcdir)/m4/autobuild.m4 \ + $(top_srcdir)/m4/codeset.m4 $(top_srcdir)/m4/gettext.m4 \ + $(top_srcdir)/m4/glibc21.m4 $(top_srcdir)/m4/iconv.m4 \ + $(top_srcdir)/m4/intdiv0.m4 $(top_srcdir)/m4/intmax.m4 \ + $(top_srcdir)/m4/inttypes-pri.m4 $(top_srcdir)/m4/inttypes.m4 \ + $(top_srcdir)/m4/inttypes_h.m4 $(top_srcdir)/m4/isc-posix.m4 \ + $(top_srcdir)/m4/lcmessage.m4 $(top_srcdir)/m4/ldap.m4 \ + $(top_srcdir)/m4/lib-ld.m4 $(top_srcdir)/m4/lib-link.m4 \ + $(top_srcdir)/m4/lib-prefix.m4 $(top_srcdir)/m4/libcurl.m4 \ + $(top_srcdir)/m4/libusb.m4 $(top_srcdir)/m4/longdouble.m4 \ + $(top_srcdir)/m4/longlong.m4 $(top_srcdir)/m4/nls.m4 \ + $(top_srcdir)/m4/noexecstack.m4 $(top_srcdir)/m4/po.m4 \ + $(top_srcdir)/m4/printf-posix.m4 $(top_srcdir)/m4/progtest.m4 \ + $(top_srcdir)/m4/readline.m4 $(top_srcdir)/m4/signed.m4 \ + $(top_srcdir)/m4/size_max.m4 $(top_srcdir)/m4/stdint_h.m4 \ + $(top_srcdir)/m4/tar-ustar.m4 $(top_srcdir)/m4/uintmax_t.m4 \ + $(top_srcdir)/m4/ulonglong.m4 $(top_srcdir)/m4/wchar_t.m4 \ + $(top_srcdir)/m4/wint_t.m4 $(top_srcdir)/m4/xsize.m4 \ + $(top_srcdir)/acinclude.m4 $(top_srcdir)/configure.ac +am__configure_deps = $(am__aclocal_m4_deps) $(CONFIGURE_DEPENDENCIES) \ + $(ACLOCAL_M4) +mkinstalldirs = $(SHELL) $(top_srcdir)/scripts/mkinstalldirs +CONFIG_HEADER = $(top_builddir)/config.h +CONFIG_CLEAN_FILES = +LIBRARIES = $(noinst_LIBRARIES) +ARFLAGS = cru +libmpi_a_AR = $(AR) $(ARFLAGS) +am__DEPENDENCIES_1 = mpih-mul1.o mpih-mul2.o mpih-mul3.o mpih-add1.o \ + mpih-sub1.o mpih-lshift.o mpih-rshift.o +am_libmpi_a_OBJECTS = mpi-add.$(OBJEXT) mpi-bit.$(OBJEXT) \ + mpi-cmp.$(OBJEXT) mpi-div.$(OBJEXT) mpi-gcd.$(OBJEXT) \ + mpi-inline.$(OBJEXT) mpi-inv.$(OBJEXT) mpi-mul.$(OBJEXT) \ + mpi-pow.$(OBJEXT) mpi-mpow.$(OBJEXT) mpi-scan.$(OBJEXT) \ + mpicoder.$(OBJEXT) mpih-cmp.$(OBJEXT) mpih-div.$(OBJEXT) \ + mpih-mul.$(OBJEXT) mpiutil.$(OBJEXT) +libmpi_a_OBJECTS = $(am_libmpi_a_OBJECTS) +DEFAULT_INCLUDES = -I. -I$(srcdir) -I$(top_builddir) +depcomp = $(SHELL) $(top_srcdir)/scripts/depcomp +am__depfiles_maybe = depfiles +COMPILE = $(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) \ + $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS) +CCLD = $(CC) +LINK = $(CCLD) $(AM_CFLAGS) $(CFLAGS) $(AM_LDFLAGS) $(LDFLAGS) -o $@ +SOURCES = $(libmpi_a_SOURCES) +DIST_SOURCES = $(libmpi_a_SOURCES) +ETAGS = etags +CTAGS = ctags +DISTFILES = $(DIST_COMMON) $(DIST_SOURCES) $(TEXINFOS) $(EXTRA_DIST) +ACLOCAL = @ACLOCAL@ +ALLOCA = @ALLOCA@ +AMDEP_FALSE = @AMDEP_FALSE@ +AMDEP_TRUE = @AMDEP_TRUE@ +AMTAR = @AMTAR@ +AR = @AR@ +AUTOCONF = @AUTOCONF@ +AUTOHEADER = @AUTOHEADER@ +AUTOMAKE = @AUTOMAKE@ +AWK = @AWK@ +BUILD_INCLUDED_LIBINTL = @BUILD_INCLUDED_LIBINTL@ +CAPLIBS = @CAPLIBS@ +CATOBJEXT = @CATOBJEXT@ +CC = @CC@ +CCAS = @CCAS@ +CCASFLAGS = @CCASFLAGS@ +CCDEPMODE = @CCDEPMODE@ +CC_FOR_BUILD = @CC_FOR_BUILD@ +CFLAGS = @CFLAGS@ +CPP = @CPP@ +CPPFLAGS = @CPPFLAGS@ +CROSS_COMPILING_FALSE = @CROSS_COMPILING_FALSE@ +CROSS_COMPILING_TRUE = @CROSS_COMPILING_TRUE@ +CYGPATH_W = @CYGPATH_W@ +DATADIRNAME = @DATADIRNAME@ +DEFS = @DEFS@ +DEPDIR = @DEPDIR@ +DLLIBS = @DLLIBS@ +DNSLIBS = @DNSLIBS@ +DOCBOOK_TO_MAN = @DOCBOOK_TO_MAN@ +DOCBOOK_TO_TEXI = @DOCBOOK_TO_TEXI@ +ECHO_C = @ECHO_C@ +ECHO_N = @ECHO_N@ +ECHO_T = @ECHO_T@ +EGREP = @EGREP@ +ENABLE_AGENT_SUPPORT_FALSE = @ENABLE_AGENT_SUPPORT_FALSE@ +ENABLE_AGENT_SUPPORT_TRUE = @ENABLE_AGENT_SUPPORT_TRUE@ +ENABLE_BZIP2_SUPPORT_FALSE = @ENABLE_BZIP2_SUPPORT_FALSE@ +ENABLE_BZIP2_SUPPORT_TRUE = @ENABLE_BZIP2_SUPPORT_TRUE@ +ENABLE_CARD_SUPPORT_FALSE = @ENABLE_CARD_SUPPORT_FALSE@ +ENABLE_CARD_SUPPORT_TRUE = @ENABLE_CARD_SUPPORT_TRUE@ +ENABLE_LOCAL_ZLIB_FALSE = @ENABLE_LOCAL_ZLIB_FALSE@ +ENABLE_LOCAL_ZLIB_TRUE = @ENABLE_LOCAL_ZLIB_TRUE@ +EXEEXT = @EXEEXT@ +FAKE_CURL_FALSE = @FAKE_CURL_FALSE@ +FAKE_CURL_TRUE = @FAKE_CURL_TRUE@ +FAQPROG = @FAQPROG@ +GENCAT = @GENCAT@ +GETOPT = @GETOPT@ +GLIBC21 = @GLIBC21@ +GMSGFMT = @GMSGFMT@ +GPGKEYS_CURL = @GPGKEYS_CURL@ +GPGKEYS_FINGER = @GPGKEYS_FINGER@ +GPGKEYS_HKP = @GPGKEYS_HKP@ +GPGKEYS_LDAP = @GPGKEYS_LDAP@ +GPGKEYS_MAILTO = @GPGKEYS_MAILTO@ +GREP = @GREP@ +HAVE_ASPRINTF = @HAVE_ASPRINTF@ +HAVE_DOCBOOK_TO_MAN_FALSE = @HAVE_DOCBOOK_TO_MAN_FALSE@ +HAVE_DOCBOOK_TO_MAN_TRUE = @HAVE_DOCBOOK_TO_MAN_TRUE@ +HAVE_DOCBOOK_TO_TEXI_FALSE = @HAVE_DOCBOOK_TO_TEXI_FALSE@ +HAVE_DOCBOOK_TO_TEXI_TRUE = @HAVE_DOCBOOK_TO_TEXI_TRUE@ +HAVE_DOSISH_SYSTEM_FALSE = @HAVE_DOSISH_SYSTEM_FALSE@ +HAVE_DOSISH_SYSTEM_TRUE = @HAVE_DOSISH_SYSTEM_TRUE@ +HAVE_POSIX_PRINTF = @HAVE_POSIX_PRINTF@ +HAVE_SNPRINTF = @HAVE_SNPRINTF@ +HAVE_USTAR_FALSE = @HAVE_USTAR_FALSE@ +HAVE_USTAR_TRUE = @HAVE_USTAR_TRUE@ +HAVE_W32_SYSTEM_FALSE = @HAVE_W32_SYSTEM_FALSE@ +HAVE_W32_SYSTEM_TRUE = @HAVE_W32_SYSTEM_TRUE@ +HAVE_WPRINTF = @HAVE_WPRINTF@ +IDEA_O = @IDEA_O@ +INSTALL_DATA = @INSTALL_DATA@ +INSTALL_PROGRAM = @INSTALL_PROGRAM@ +INSTALL_SCRIPT = @INSTALL_SCRIPT@ +INSTALL_STRIP_PROGRAM = @INSTALL_STRIP_PROGRAM@ +INSTOBJEXT = @INSTOBJEXT@ +INTLBISON = @INTLBISON@ +INTLLIBS = @INTLLIBS@ +INTLOBJS = @INTLOBJS@ +INTL_LIBTOOL_SUFFIX_PREFIX = @INTL_LIBTOOL_SUFFIX_PREFIX@ +LDAPLIBS = @LDAPLIBS@ +LDAP_CPPFLAGS = @LDAP_CPPFLAGS@ +LDFLAGS = @LDFLAGS@ +LIBCURL = @LIBCURL@ +LIBCURL_CPPFLAGS = @LIBCURL_CPPFLAGS@ +LIBICONV = @LIBICONV@ +LIBINTL = @LIBINTL@ +LIBOBJS = @LIBOBJS@ +LIBREADLINE = @LIBREADLINE@ +LIBS = @LIBS@ +LIBUSB = @LIBUSB@ +LIBUSB_CPPFLAGS = @LIBUSB_CPPFLAGS@ +LTLIBICONV = @LTLIBICONV@ +LTLIBINTL = @LTLIBINTL@ +LTLIBOBJS = @LTLIBOBJS@ +MAINT = @MAINT@ +MAINTAINER_MODE_FALSE = @MAINTAINER_MODE_FALSE@ +MAINTAINER_MODE_TRUE = @MAINTAINER_MODE_TRUE@ +MAKEINFO = @MAKEINFO@ +MKINSTALLDIRS = @MKINSTALLDIRS@ +MPI_EXTRA_ASM_OBJS = @MPI_EXTRA_ASM_OBJS@ +MPI_OPT_FLAGS = @MPI_OPT_FLAGS@ +MPI_SFLAGS = @MPI_SFLAGS@ +MSGFMT = @MSGFMT@ +MSGMERGE = @MSGMERGE@ +NETLIBS = @NETLIBS@ +NM = @NM@ +NOEXECSTACK_FLAGS = @NOEXECSTACK_FLAGS@ +OBJEXT = @OBJEXT@ +PACKAGE = @PACKAGE@ +PACKAGE_BUGREPORT = @PACKAGE_BUGREPORT@ +PACKAGE_NAME = @PACKAGE_NAME@ +PACKAGE_STRING = @PACKAGE_STRING@ +PACKAGE_TARNAME = @PACKAGE_TARNAME@ +PACKAGE_VERSION = @PACKAGE_VERSION@ +PATH_SEPARATOR = @PATH_SEPARATOR@ +PERL = @PERL@ +POSUB = @POSUB@ +RANLIB = @RANLIB@ +SENDMAIL = @SENDMAIL@ +SET_MAKE = @SET_MAKE@ +SHELL = @SHELL@ +STRIP = @STRIP@ +TAR = @TAR@ +USE_DNS_SRV_FALSE = @USE_DNS_SRV_FALSE@ +USE_DNS_SRV_TRUE = @USE_DNS_SRV_TRUE@ +USE_INCLUDED_LIBINTL = @USE_INCLUDED_LIBINTL@ +USE_INTERNAL_REGEX_FALSE = @USE_INTERNAL_REGEX_FALSE@ +USE_INTERNAL_REGEX_TRUE = @USE_INTERNAL_REGEX_TRUE@ +USE_NLS = @USE_NLS@ +USE_RNDEGD_FALSE = @USE_RNDEGD_FALSE@ +USE_RNDEGD_TRUE = @USE_RNDEGD_TRUE@ +USE_RNDLINUX_FALSE = @USE_RNDLINUX_FALSE@ +USE_RNDLINUX_TRUE = @USE_RNDLINUX_TRUE@ +USE_RNDUNIX_FALSE = @USE_RNDUNIX_FALSE@ +USE_RNDUNIX_TRUE = @USE_RNDUNIX_TRUE@ +USE_RNDW32_FALSE = @USE_RNDW32_FALSE@ +USE_RNDW32_TRUE = @USE_RNDW32_TRUE@ +USE_SHA512_FALSE = @USE_SHA512_FALSE@ +USE_SHA512_TRUE = @USE_SHA512_TRUE@ +USE_SIMPLE_GETTEXT_FALSE = @USE_SIMPLE_GETTEXT_FALSE@ +USE_SIMPLE_GETTEXT_TRUE = @USE_SIMPLE_GETTEXT_TRUE@ +VERSION = @VERSION@ +W32LIBS = @W32LIBS@ +WORKING_FAQPROG_FALSE = @WORKING_FAQPROG_FALSE@ +WORKING_FAQPROG_TRUE = @WORKING_FAQPROG_TRUE@ +XGETTEXT = @XGETTEXT@ +ZLIBS = @ZLIBS@ +_libcurl_config = @_libcurl_config@ +_usb_config = @_usb_config@ +ac_ct_CC = @ac_ct_CC@ +am__fastdepCC_FALSE = @am__fastdepCC_FALSE@ +am__fastdepCC_TRUE = @am__fastdepCC_TRUE@ +am__include = @am__include@ +am__leading_dot = @am__leading_dot@ +am__quote = @am__quote@ +am__tar = @am__tar@ +am__untar = @am__untar@ +bindir = @bindir@ +build = @build@ +build_alias = @build_alias@ +build_cpu = @build_cpu@ +build_os = @build_os@ +build_vendor = @build_vendor@ +datadir = @datadir@ +datarootdir = @datarootdir@ +docdir = @docdir@ +dvidir = @dvidir@ +exec_prefix = @exec_prefix@ +host = @host@ +host_alias = @host_alias@ +host_cpu = @host_cpu@ +host_os = @host_os@ +host_vendor = @host_vendor@ +htmldir = @htmldir@ +includedir = @includedir@ +infodir = @infodir@ +install_sh = @install_sh@ +libdir = @libdir@ +libexecdir = @libexecdir@ +localedir = @localedir@ +localstatedir = @localstatedir@ +mandir = @mandir@ +mkdir_p = @mkdir_p@ +oldincludedir = @oldincludedir@ +pdfdir = @pdfdir@ +prefix = @prefix@ +program_transform_name = @program_transform_name@ +psdir = @psdir@ +sbindir = @sbindir@ +sharedstatedir = @sharedstatedir@ +sysconfdir = @sysconfdir@ +target_alias = @target_alias@ +INCLUDES = -I.. -I$(top_srcdir)/include +AM_CFLAGS = $(MPI_OPT_FLAGS) +AM_CCASFLAGS = $(NOEXECSTACK_FLAGS) $(MPI_SFLAGS) +EXTRA_DIST = config.links +DISTCLEANFILES = mpi-asm-defs.h \ + mpih-add1.S mpih-add1.c mpih-mul1.S mpih-mul1.c \ + mpih-mul2.S mpih-mul2.c mpih-mul3.S mpih-mul3.c \ + mpih-lshift.S mpih-lshift.c mpih-rshift.S mpih-rshift.c \ + mpih-sub1.S mpih-sub1.c asm-syntax.h sysdep.h + + +# Note: we only use .S files so we should delete all left over .s +CLEANFILES = _*.s +noinst_LIBRARIES = libmpi.a + +# libmpi_a_LDFLAGS = +libmpi_a_SOURCES = longlong.h \ + mpi-add.c \ + mpi-bit.c \ + mpi-cmp.c \ + mpi-div.c \ + mpi-gcd.c \ + mpi-internal.h \ + mpi-inline.h \ + mpi-inline.c \ + mpi-inv.c \ + mpi-mul.c \ + mpi-pow.c \ + mpi-mpow.c \ + mpi-scan.c \ + mpicoder.c \ + mpih-cmp.c \ + mpih-div.c \ + mpih-mul.c \ + mpiutil.c + + +# Note this objects are actually links, the sourcefiles are +# distributed by special code in dist-hook +common_asm_objects = mpih-mul1.o \ + mpih-mul2.o \ + mpih-mul3.o \ + mpih-add1.o \ + mpih-sub1.o \ + mpih-lshift.o \ + mpih-rshift.o + +libmpi_a_DEPENDENCIES = $(common_asm_objects) @MPI_EXTRA_ASM_OBJS@ +libmpi_a_LIBADD = $(common_asm_objects) @MPI_EXTRA_ASM_OBJS@ +all: all-am + +.SUFFIXES: +.SUFFIXES: .S .c .lo .o .obj +$(srcdir)/Makefile.in: @MAINTAINER_MODE_TRUE@ $(srcdir)/Makefile.am $(am__configure_deps) + @for dep in $?; do \ + case '$(am__configure_deps)' in \ + *$$dep*) \ + cd $(top_builddir) && $(MAKE) $(AM_MAKEFLAGS) am--refresh \ + && exit 0; \ + exit 1;; \ + esac; \ + done; \ + echo ' cd $(top_srcdir) && $(AUTOMAKE) --gnu mpi/Makefile'; \ + cd $(top_srcdir) && \ + $(AUTOMAKE) --gnu mpi/Makefile +.PRECIOUS: Makefile +Makefile: $(srcdir)/Makefile.in $(top_builddir)/config.status + @case '$?' in \ + *config.status*) \ + cd $(top_builddir) && $(MAKE) $(AM_MAKEFLAGS) am--refresh;; \ + *) \ + echo ' cd $(top_builddir) && $(SHELL) ./config.status $(subdir)/$@ $(am__depfiles_maybe)'; \ + cd $(top_builddir) && $(SHELL) ./config.status $(subdir)/$@ $(am__depfiles_maybe);; \ + esac; + +$(top_builddir)/config.status: $(top_srcdir)/configure $(CONFIG_STATUS_DEPENDENCIES) + cd $(top_builddir) && $(MAKE) $(AM_MAKEFLAGS) am--refresh + +$(top_srcdir)/configure: @MAINTAINER_MODE_TRUE@ $(am__configure_deps) + cd $(top_builddir) && $(MAKE) $(AM_MAKEFLAGS) am--refresh +$(ACLOCAL_M4): @MAINTAINER_MODE_TRUE@ $(am__aclocal_m4_deps) + cd $(top_builddir) && $(MAKE) $(AM_MAKEFLAGS) am--refresh + +clean-noinstLIBRARIES: + -test -z "$(noinst_LIBRARIES)" || rm -f $(noinst_LIBRARIES) +libmpi.a: $(libmpi_a_OBJECTS) $(libmpi_a_DEPENDENCIES) + -rm -f libmpi.a + $(libmpi_a_AR) libmpi.a $(libmpi_a_OBJECTS) $(libmpi_a_LIBADD) + $(RANLIB) libmpi.a + +mostlyclean-compile: + -rm -f *.$(OBJEXT) + +distclean-compile: + -rm -f *.tab.c + +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-add.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-bit.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-cmp.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-div.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-gcd.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-inline.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-inv.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-mpow.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-mul.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-pow.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpi-scan.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpicoder.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpih-cmp.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpih-div.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpih-mul.Po@am__quote@ +@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/mpiutil.Po@am__quote@ + +.c.o: +@am__fastdepCC_TRUE@ if $(COMPILE) -MT $@ -MD -MP -MF "$(DEPDIR)/$*.Tpo" -c -o $@ $<; \ +@am__fastdepCC_TRUE@ then mv -f "$(DEPDIR)/$*.Tpo" "$(DEPDIR)/$*.Po"; else rm -f "$(DEPDIR)/$*.Tpo"; exit 1; fi +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$<' object='$@' libtool=no @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(COMPILE) -c $< + +.c.obj: +@am__fastdepCC_TRUE@ if $(COMPILE) -MT $@ -MD -MP -MF "$(DEPDIR)/$*.Tpo" -c -o $@ `$(CYGPATH_W) '$<'`; \ +@am__fastdepCC_TRUE@ then mv -f "$(DEPDIR)/$*.Tpo" "$(DEPDIR)/$*.Po"; else rm -f "$(DEPDIR)/$*.Tpo"; exit 1; fi +@AMDEP_TRUE@@am__fastdepCC_FALSE@ source='$<' object='$@' libtool=no @AMDEPBACKSLASH@ +@AMDEP_TRUE@@am__fastdepCC_FALSE@ DEPDIR=$(DEPDIR) $(CCDEPMODE) $(depcomp) @AMDEPBACKSLASH@ +@am__fastdepCC_FALSE@ $(COMPILE) -c `$(CYGPATH_W) '$<'` +uninstall-info-am: + +ID: $(HEADERS) $(SOURCES) $(LISP) $(TAGS_FILES) + list='$(SOURCES) $(HEADERS) $(LISP) $(TAGS_FILES)'; \ + unique=`for i in $$list; do \ + if test -f "$$i"; then echo $$i; else echo $(srcdir)/$$i; fi; \ + done | \ + $(AWK) ' { files[$$0] = 1; } \ + END { for (i in files) print i; }'`; \ + mkid -fID $$unique +tags: TAGS + +TAGS: $(HEADERS) $(SOURCES) $(TAGS_DEPENDENCIES) \ + $(TAGS_FILES) $(LISP) + tags=; \ + here=`pwd`; \ + list='$(SOURCES) $(HEADERS) $(LISP) $(TAGS_FILES)'; \ + unique=`for i in $$list; do \ + if test -f "$$i"; then echo $$i; else echo $(srcdir)/$$i; fi; \ + done | \ + $(AWK) ' { files[$$0] = 1; } \ + END { for (i in files) print i; }'`; \ + if test -z "$(ETAGS_ARGS)$$tags$$unique"; then :; else \ + test -n "$$unique" || unique=$$empty_fix; \ + $(ETAGS) $(ETAGSFLAGS) $(AM_ETAGSFLAGS) $(ETAGS_ARGS) \ + $$tags $$unique; \ + fi +ctags: CTAGS +CTAGS: $(HEADERS) $(SOURCES) $(TAGS_DEPENDENCIES) \ + $(TAGS_FILES) $(LISP) + tags=; \ + here=`pwd`; \ + list='$(SOURCES) $(HEADERS) $(LISP) $(TAGS_FILES)'; \ + unique=`for i in $$list; do \ + if test -f "$$i"; then echo $$i; else echo $(srcdir)/$$i; fi; \ + done | \ + $(AWK) ' { files[$$0] = 1; } \ + END { for (i in files) print i; }'`; \ + test -z "$(CTAGS_ARGS)$$tags$$unique" \ + || $(CTAGS) $(CTAGSFLAGS) $(AM_CTAGSFLAGS) $(CTAGS_ARGS) \ + $$tags $$unique + +GTAGS: + here=`$(am__cd) $(top_builddir) && pwd` \ + && cd $(top_srcdir) \ + && gtags -i $(GTAGS_ARGS) $$here + +distclean-tags: + -rm -f TAGS ID GTAGS GRTAGS GSYMS GPATH tags + +distdir: $(DISTFILES) + @srcdirstrip=`echo "$(srcdir)" | sed 's|.|.|g'`; \ + topsrcdirstrip=`echo "$(top_srcdir)" | sed 's|.|.|g'`; \ + list='$(DISTFILES)'; for file in $$list; do \ + case $$file in \ + $(srcdir)/*) file=`echo "$$file" | sed "s|^$$srcdirstrip/||"`;; \ + $(top_srcdir)/*) file=`echo "$$file" | sed "s|^$$topsrcdirstrip/|$(top_builddir)/|"`;; \ + esac; \ + if test -f $$file || test -d $$file; then d=.; else d=$(srcdir); fi; \ + dir=`echo "$$file" | sed -e 's,/[^/]*$$,,'`; \ + if test "$$dir" != "$$file" && test "$$dir" != "."; then \ + dir="/$$dir"; \ + $(mkdir_p) "$(distdir)$$dir"; \ + else \ + dir=''; \ + fi; \ + if test -d $$d/$$file; then \ + if test -d $(srcdir)/$$file && test $$d != $(srcdir); then \ + cp -pR $(srcdir)/$$file $(distdir)$$dir || exit 1; \ + fi; \ + cp -pR $$d/$$file $(distdir)$$dir || exit 1; \ + else \ + test -f $(distdir)/$$file \ + || cp -p $$d/$$file $(distdir)/$$file \ + || exit 1; \ + fi; \ + done +check-am: all-am +check: check-am +all-am: Makefile $(LIBRARIES) +installdirs: +install: install-am +install-exec: install-exec-am +install-data: install-data-am +uninstall: uninstall-am + +install-am: all-am + @$(MAKE) $(AM_MAKEFLAGS) install-exec-am install-data-am + +installcheck: installcheck-am +install-strip: + $(MAKE) $(AM_MAKEFLAGS) INSTALL_PROGRAM="$(INSTALL_STRIP_PROGRAM)" \ + install_sh_PROGRAM="$(INSTALL_STRIP_PROGRAM)" INSTALL_STRIP_FLAG=-s \ + `test -z '$(STRIP)' || \ + echo "INSTALL_PROGRAM_ENV=STRIPPROG='$(STRIP)'"` install +mostlyclean-generic: + +clean-generic: + -test -z "$(CLEANFILES)" || rm -f $(CLEANFILES) + +distclean-generic: + -test -z "$(CONFIG_CLEAN_FILES)" || rm -f $(CONFIG_CLEAN_FILES) + -test -z "$(DISTCLEANFILES)" || rm -f $(DISTCLEANFILES) + +maintainer-clean-generic: + @echo "This command is intended for maintainers to use" + @echo "it deletes files that may require special tools to rebuild." +clean: clean-am + +clean-am: clean-generic clean-noinstLIBRARIES mostlyclean-am + +distclean: distclean-am + -rm -rf ./$(DEPDIR) + -rm -f Makefile +distclean-am: clean-am distclean-compile distclean-generic \ + distclean-tags + +dvi: dvi-am + +dvi-am: + +html: html-am + +info: info-am + +info-am: + +install-data-am: + +install-exec-am: + +install-info: install-info-am + +install-man: + +installcheck-am: + +maintainer-clean: maintainer-clean-am + -rm -rf ./$(DEPDIR) + -rm -f Makefile +maintainer-clean-am: distclean-am maintainer-clean-generic + +mostlyclean: mostlyclean-am + +mostlyclean-am: mostlyclean-compile mostlyclean-generic + +pdf: pdf-am + +pdf-am: + +ps: ps-am + +ps-am: + +uninstall-am: uninstall-info-am + +.PHONY: CTAGS GTAGS all all-am check check-am clean clean-generic \ + clean-noinstLIBRARIES ctags distclean distclean-compile \ + distclean-generic distclean-tags distdir dvi dvi-am html \ + html-am info info-am install install-am install-data \ + install-data-am install-exec install-exec-am install-info \ + install-info-am install-man install-strip installcheck \ + installcheck-am installdirs maintainer-clean \ + maintainer-clean-generic mostlyclean mostlyclean-compile \ + mostlyclean-generic pdf pdf-am ps ps-am tags uninstall \ + uninstall-am uninstall-info-am + + +# cancel the default rules used by libtool which do not really +# work and add one to cpp .S files +.S.o: + $(CPP) $(INCLUDES) $(DEFS) $< | grep -v '^#' > _$*.s + $(COMPILE) $(AM_CCASFLAGS) -c _$*.s + mv -f _$*.o $*.o + +.S.lo: +# Tell versions [3.59,3.63) of GNU make to not export all variables. +# Otherwise a system limit (for SysV at least) may be exceeded. +.NOEXPORT: diff --git a/mpi/alpha/README b/mpi/alpha/README new file mode 100644 index 0000000..55c0a29 --- /dev/null +++ b/mpi/alpha/README @@ -0,0 +1,53 @@ +This directory contains mpn functions optimized for DEC Alpha processors. + +RELEVANT OPTIMIZATION ISSUES + +EV4 + +1. This chip has very limited store bandwidth. The on-chip L1 cache is +write-through, and a cache line is transfered from the store buffer to the +off-chip L2 in as much 15 cycles on most systems. This delay hurts +mpn_add_n, mpn_sub_n, mpn_lshift, and mpn_rshift. + +2. Pairing is possible between memory instructions and integer arithmetic +instructions. + +3. mulq and umulh is documented to have a latency of 23 cycles, but 2 of +these cycles are pipelined. Thus, multiply instructions can be issued at a +rate of one each 21nd cycle. + +EV5 + +1. The memory bandwidth of this chip seems excellent, both for loads and +stores. Even when the working set is larger than the on-chip L1 and L2 +caches, the perfromance remain almost unaffected. + +2. mulq has a measured latency of 13 cycles and an issue rate of 1 each 8th +cycle. umulh has a measured latency of 15 cycles and an issue rate of 1 +each 10th cycle. But the exact timing is somewhat confusing. + +3. mpn_add_n. With 4-fold unrolling, we need 37 instructions, whereof 12 + are memory operations. This will take at least + ceil(37/2) [dual issue] + 1 [taken branch] = 20 cycles + We have 12 memory cycles, plus 4 after-store conflict cycles, or 16 data + cache cycles, which should be completely hidden in the 20 issue cycles. + The computation is inherently serial, with these dependencies: + addq + / \ + addq cmpult + | | + cmpult | + \ / + or + I.e., there is a 4 cycle path for each limb, making 16 cycles the absolute + minimum. We could replace the `or' with a cmoveq/cmovne, which would save + a cycle on EV5, but that might waste a cycle on EV4. Also, cmov takes 2 + cycles. + addq + / \ + addq cmpult + | \ + cmpult -> cmovne + +STATUS + diff --git a/mpi/alpha/distfiles b/mpi/alpha/distfiles new file mode 100644 index 0000000..f2ab9fc --- /dev/null +++ b/mpi/alpha/distfiles @@ -0,0 +1,11 @@ +README +mpih-add1.S +mpih-sub1.S +mpih-mul1.S +mpih-mul2.S +mpih-mul3.S +mpih-lshift.S +mpih-rshift.S + +udiv-qrnnd.S + diff --git a/mpi/alpha/mpih-add1.S b/mpi/alpha/mpih-add1.S new file mode 100644 index 0000000..6ea8e38 --- /dev/null +++ b/mpi/alpha/mpih-add1.S @@ -0,0 +1,125 @@ +/* alpha add_n -- Add two limb vectors of the same length > 0 and store + * sum in a third limb vector. + * Copyright (C) 1995, 1998, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + + +/******************* + * mpi_limb_t + * mpihelp_add_n( mpi_ptr_t res_ptr, ($16) + * mpi_ptr_t s1_ptr, ($17) + * mpi_ptr_t s2_ptr, ($18) + * mpi_size_t size) ($19) + */ + + + .set noreorder + .set noat +.text + .align 3 + .globl mpihelp_add_n + .ent mpihelp_add_n +mpihelp_add_n: + .frame $30,0,$26,0 + + ldq $3,0($17) + ldq $4,0($18) + + subq $19,1,$19 + and $19,4-1,$2 # number of limbs in first loop + bis $31,$31,$0 + beq $2,.L0 # if multiple of 4 limbs, skip first loop + + subq $19,$2,$19 + +.Loop0: subq $2,1,$2 + ldq $5,8($17) + addq $4,$0,$4 + ldq $6,8($18) + cmpult $4,$0,$1 + addq $3,$4,$4 + cmpult $4,$3,$0 + stq $4,0($16) + or $0,$1,$0 + + addq $17,8,$17 + addq $18,8,$18 + bis $5,$5,$3 + bis $6,$6,$4 + addq $16,8,$16 + bne $2,.Loop0 + +.L0: beq $19,.Lend + + .align 3 +.Loop: subq $19,4,$19 + + ldq $5,8($17) + addq $4,$0,$4 + ldq $6,8($18) + cmpult $4,$0,$1 + addq $3,$4,$4 + cmpult $4,$3,$0 + stq $4,0($16) + or $0,$1,$0 + + ldq $3,16($17) + addq $6,$0,$6 + ldq $4,16($18) + cmpult $6,$0,$1 + addq $5,$6,$6 + cmpult $6,$5,$0 + stq $6,8($16) + or $0,$1,$0 + + ldq $5,24($17) + addq $4,$0,$4 + ldq $6,24($18) + cmpult $4,$0,$1 + addq $3,$4,$4 + cmpult $4,$3,$0 + stq $4,16($16) + or $0,$1,$0 + + ldq $3,32($17) + addq $6,$0,$6 + ldq $4,32($18) + cmpult $6,$0,$1 + addq $5,$6,$6 + cmpult $6,$5,$0 + stq $6,24($16) + or $0,$1,$0 + + addq $17,32,$17 + addq $18,32,$18 + addq $16,32,$16 + bne $19,.Loop + +.Lend: addq $4,$0,$4 + cmpult $4,$0,$1 + addq $3,$4,$4 + cmpult $4,$3,$0 + stq $4,0($16) + or $0,$1,$0 + ret $31,($26),1 + + .end mpihelp_add_n + diff --git a/mpi/alpha/mpih-lshift.S b/mpi/alpha/mpih-lshift.S new file mode 100644 index 0000000..45919f5 --- /dev/null +++ b/mpi/alpha/mpih-lshift.S @@ -0,0 +1,124 @@ +/* alpha - left shift + * + * Copyright (C) 1994, 1995, 1998, 2001 Free Software Foundation, Inc. + * + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + +/******************* + * mpi_limb_t + * mpihelp_lshift( mpi_ptr_t wp, (r16) + * mpi_ptr_t up, (r17) + * mpi_size_t usize, (r18) + * unsigned cnt) (r19) + * + * This code runs at 4.8 cycles/limb on the 21064. With infinite unrolling, + * it would take 4 cycles/limb. It should be possible to get down to 3 + * cycles/limb since both ldq and stq can be paired with the other used + * instructions. But there are many restrictions in the 21064 pipeline that + * makes it hard, if not impossible, to get down to 3 cycles/limb: + * + * 1. ldq has a 3 cycle delay, srl and sll have a 2 cycle delay. + * 2. Only aligned instruction pairs can be paired. + * 3. The store buffer or silo might not be able to deal with the bandwidth. + */ + + .set noreorder + .set noat +.text + .align 3 + .globl mpihelp_lshift + .ent mpihelp_lshift +mpihelp_lshift: + .frame $30,0,$26,0 + + s8addq $18,$17,$17 # make r17 point at end of s1 + ldq $4,-8($17) # load first limb + subq $17,8,$17 + subq $31,$19,$7 + s8addq $18,$16,$16 # make r16 point at end of RES + subq $18,1,$18 + and $18,4-1,$20 # number of limbs in first loop + srl $4,$7,$0 # compute function result + + beq $20,.L0 + subq $18,$20,$18 + + .align 3 +.Loop0: + ldq $3,-8($17) + subq $16,8,$16 + subq $17,8,$17 + subq $20,1,$20 + sll $4,$19,$5 + srl $3,$7,$6 + bis $3,$3,$4 + bis $5,$6,$8 + stq $8,0($16) + bne $20,.Loop0 + +.L0: beq $18,.Lend + + .align 3 +.Loop: ldq $3,-8($17) + subq $16,32,$16 + subq $18,4,$18 + sll $4,$19,$5 + srl $3,$7,$6 + + ldq $4,-16($17) + sll $3,$19,$1 + bis $5,$6,$8 + stq $8,24($16) + srl $4,$7,$2 + + ldq $3,-24($17) + sll $4,$19,$5 + bis $1,$2,$8 + stq $8,16($16) + srl $3,$7,$6 + + ldq $4,-32($17) + sll $3,$19,$1 + bis $5,$6,$8 + stq $8,8($16) + srl $4,$7,$2 + + subq $17,32,$17 + bis $1,$2,$8 + stq $8,0($16) + + bgt $18,.Loop + +.Lend: sll $4,$19,$8 + stq $8,-8($16) + ret $31,($26),1 + .end mpihelp_lshift + + diff --git a/mpi/alpha/mpih-mul1.S b/mpi/alpha/mpih-mul1.S new file mode 100644 index 0000000..65b2d5c --- /dev/null +++ b/mpi/alpha/mpih-mul1.S @@ -0,0 +1,91 @@ +/* Alpha 21064 mpih-mul1.S -- Multiply a limb vector with a limb and store + * the result in a second limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1998, + 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +/******************* + * mpi_limb_t + * mpihelp_mul_1( mpi_ptr_t res_ptr, (r16) + * mpi_ptr_t s1_ptr, (r17) + * mpi_size_t s1_size, (r18) + * mpi_limb_t s2_limb) (r19) + * + * This code runs at 42 cycles/limb on the EV4 and 18 cycles/limb on the EV5. + * + * To improve performance for long multiplications, we would use + * 'fetch' for S1 and 'fetch_m' for RES. It's not obvious how to use + * these instructions without slowing down the general code: 1. We can + * only have two prefetches in operation at any time in the Alpha + * architecture. 2. There will seldom be any special alignment + * between RES_PTR and S1_PTR. Maybe we can simply divide the current + * loop into an inner and outer loop, having the inner loop handle + * exactly one prefetch block? + */ + + .set noreorder + .set noat +.text + .align 3 + .globl mpihelp_mul_1 + .ent mpihelp_mul_1 2 +mpihelp_mul_1: + .frame $30,0,$26 + + ldq $2,0($17) # $2 = s1_limb + subq $18,1,$18 # size-- + mulq $2,$19,$3 # $3 = prod_low + bic $31,$31,$4 # clear cy_limb + umulh $2,$19,$0 # $0 = prod_high + beq $18,Lend1 # jump if size was == 1 + ldq $2,8($17) # $2 = s1_limb + subq $18,1,$18 # size-- + stq $3,0($16) + beq $18,Lend2 # jump if size was == 2 + + .align 3 +Loop: mulq $2,$19,$3 # $3 = prod_low + addq $4,$0,$0 # cy_limb = cy_limb + 'cy' + subq $18,1,$18 # size-- + umulh $2,$19,$4 # $4 = cy_limb + ldq $2,16($17) # $2 = s1_limb + addq $17,8,$17 # s1_ptr++ + addq $3,$0,$3 # $3 = cy_limb + prod_low + stq $3,8($16) + cmpult $3,$0,$0 # $0 = carry from (cy_limb + prod_low) + addq $16,8,$16 # res_ptr++ + bne $18,Loop + +Lend2: mulq $2,$19,$3 # $3 = prod_low + addq $4,$0,$0 # cy_limb = cy_limb + 'cy' + umulh $2,$19,$4 # $4 = cy_limb + addq $3,$0,$3 # $3 = cy_limb + prod_low + cmpult $3,$0,$0 # $0 = carry from (cy_limb + prod_low) + stq $3,8($16) + addq $4,$0,$0 # cy_limb = prod_high + cy + ret $31,($26),1 +Lend1: stq $3,0($16) + ret $31,($26),1 + + .end mpihelp_mul_1 + + diff --git a/mpi/alpha/mpih-mul2.S b/mpi/alpha/mpih-mul2.S new file mode 100644 index 0000000..e62ceb9 --- /dev/null +++ b/mpi/alpha/mpih-mul2.S @@ -0,0 +1,98 @@ +/* Alpha 21064 addmul_1 -- Multiply a limb vector with a limb and add + * the result to a second limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +/******************* + * mpi_limb_t + * mpihelp_addmul_1( mpi_ptr_t res_ptr, (r16) + * mpi_ptr_t s1_ptr, (r17) + * mpi_size_t s1_size, (r18) + * mpi_limb_t s2_limb) (r19) + * + * This code runs at 42 cycles/limb on EV4 and 18 cycles/limb on EV5. + */ + + + .set noreorder + .set noat +.text + .align 3 + .globl mpihelp_addmul_1 + .ent mpihelp_addmul_1 2 +mpihelp_addmul_1: + .frame $30,0,$26 + + ldq $2,0($17) # $2 = s1_limb + addq $17,8,$17 # s1_ptr++ + subq $18,1,$18 # size-- + mulq $2,$19,$3 # $3 = prod_low + ldq $5,0($16) # $5 = *res_ptr + umulh $2,$19,$0 # $0 = prod_high + beq $18,.Lend1 # jump if size was == 1 + ldq $2,0($17) # $2 = s1_limb + addq $17,8,$17 # s1_ptr++ + subq $18,1,$18 # size-- + addq $5,$3,$3 + cmpult $3,$5,$4 + stq $3,0($16) + addq $16,8,$16 # res_ptr++ + beq $18,.Lend2 # jump if size was == 2 + + .align 3 +.Loop: mulq $2,$19,$3 # $3 = prod_low + ldq $5,0($16) # $5 = *res_ptr + addq $4,$0,$0 # cy_limb = cy_limb + 'cy' + subq $18,1,$18 # size-- + umulh $2,$19,$4 # $4 = cy_limb + ldq $2,0($17) # $2 = s1_limb + addq $17,8,$17 # s1_ptr++ + addq $3,$0,$3 # $3 = cy_limb + prod_low + cmpult $3,$0,$0 # $0 = carry from (cy_limb + prod_low) + addq $5,$3,$3 + cmpult $3,$5,$5 + stq $3,0($16) + addq $16,8,$16 # res_ptr++ + addq $5,$0,$0 # combine carries + bne $18,.Loop + +.Lend2: mulq $2,$19,$3 # $3 = prod_low + ldq $5,0($16) # $5 = *res_ptr + addq $4,$0,$0 # cy_limb = cy_limb + 'cy' + umulh $2,$19,$4 # $4 = cy_limb + addq $3,$0,$3 # $3 = cy_limb + prod_low + cmpult $3,$0,$0 # $0 = carry from (cy_limb + prod_low) + addq $5,$3,$3 + cmpult $3,$5,$5 + stq $3,0($16) + addq $5,$0,$0 # combine carries + addq $4,$0,$0 # cy_limb = prod_high + cy + ret $31,($26),1 +.Lend1: addq $5,$3,$3 + cmpult $3,$5,$5 + stq $3,0($16) + addq $0,$5,$0 + ret $31,($26),1 + + .end mpihelp_addmul_1 + diff --git a/mpi/alpha/mpih-mul3.S b/mpi/alpha/mpih-mul3.S new file mode 100644 index 0000000..de8b083 --- /dev/null +++ b/mpi/alpha/mpih-mul3.S @@ -0,0 +1,96 @@ +/* Alpha 21064 submul_1 -- Multiply a limb vector with a limb and + * subtract the result from a second limb vector. + * Copyright (C) 1992, 1994, 1995, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +/******************* + * mpi_limb_t + * mpihelp_submul_1( mpi_ptr_t res_ptr, (r16 ) + * mpi_ptr_t s1_ptr, (r17 ) + * mpi_size_t s1_size, (r18 ) + * mpi_limb_t s2_limb) (r19 ) + * + * This code runs at 42 cycles/limb on EV4 and 18 cycles/limb on EV5. + */ + + .set noreorder + .set noat +.text + .align 3 + .globl mpihelp_submul_1 + .ent mpihelp_submul_1 2 +mpihelp_submul_1: + .frame $30,0,$26 + + ldq $2,0($17) # $2 = s1_limb + addq $17,8,$17 # s1_ptr++ + subq $18,1,$18 # size-- + mulq $2,$19,$3 # $3 = prod_low + ldq $5,0($16) # $5 = *res_ptr + umulh $2,$19,$0 # $0 = prod_high + beq $18,.Lend1 # jump if size was == 1 + ldq $2,0($17) # $2 = s1_limb + addq $17,8,$17 # s1_ptr++ + subq $18,1,$18 # size-- + subq $5,$3,$3 + cmpult $5,$3,$4 + stq $3,0($16) + addq $16,8,$16 # res_ptr++ + beq $18,.Lend2 # jump if size was == 2 + + .align 3 +.Loop: mulq $2,$19,$3 # $3 = prod_low + ldq $5,0($16) # $5 = *res_ptr + addq $4,$0,$0 # cy_limb = cy_limb + 'cy' + subq $18,1,$18 # size-- + umulh $2,$19,$4 # $4 = cy_limb + ldq $2,0($17) # $2 = s1_limb + addq $17,8,$17 # s1_ptr++ + addq $3,$0,$3 # $3 = cy_limb + prod_low + cmpult $3,$0,$0 # $0 = carry from (cy_limb + prod_low) + subq $5,$3,$3 + cmpult $5,$3,$5 + stq $3,0($16) + addq $16,8,$16 # res_ptr++ + addq $5,$0,$0 # combine carries + bne $18,.Loop + +.Lend2: mulq $2,$19,$3 # $3 = prod_low + ldq $5,0($16) # $5 = *res_ptr + addq $4,$0,$0 # cy_limb = cy_limb + 'cy' + umulh $2,$19,$4 # $4 = cy_limb + addq $3,$0,$3 # $3 = cy_limb + prod_low + cmpult $3,$0,$0 # $0 = carry from (cy_limb + prod_low) + subq $5,$3,$3 + cmpult $5,$3,$5 + stq $3,0($16) + addq $5,$0,$0 # combine carries + addq $4,$0,$0 # cy_limb = prod_high + cy + ret $31,($26),1 +.Lend1: subq $5,$3,$3 + cmpult $5,$3,$5 + stq $3,0($16) + addq $0,$5,$0 + ret $31,($26),1 + + .end mpihelp_submul_1 + diff --git a/mpi/alpha/mpih-rshift.S b/mpi/alpha/mpih-rshift.S new file mode 100644 index 0000000..242d2b6 --- /dev/null +++ b/mpi/alpha/mpih-rshift.S @@ -0,0 +1,121 @@ +/* alpha rshift + * Copyright (C) 1994, 1995, 1998, 1999, + * 2000, 2001 Free Software Foundation, Inc. + * + * This file is part of GNUPG. + * + * GNUPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GNUPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + + +/******************* + * mpi_limb_t + * mpihelp_rshift( mpi_ptr_t wp, (r16) + * mpi_ptr_t up, (r17) + * mpi_size_t usize, (r18) + * unsigned cnt) (r19) + * + * This code runs at 4.8 cycles/limb on the 21064. With infinite unrolling, + * it would take 4 cycles/limb. It should be possible to get down to 3 + * cycles/limb since both ldq and stq can be paired with the other used + * instructions. But there are many restrictions in the 21064 pipeline that + * makes it hard, if not impossible, to get down to 3 cycles/limb: + * + * 1. ldq has a 3 cycle delay, srl and sll have a 2 cycle delay. + * 2. Only aligned instruction pairs can be paired. + * 3. The store buffer or silo might not be able to deal with the bandwidth. + */ + + .set noreorder + .set noat +.text + .align 3 + .globl mpihelp_rshift + .ent mpihelp_rshift +mpihelp_rshift: + .frame $30,0,$26,0 + + ldq $4,0($17) # load first limb + addq $17,8,$17 + subq $31,$19,$7 + subq $18,1,$18 + and $18,4-1,$20 # number of limbs in first loop + sll $4,$7,$0 # compute function result + + beq $20,.R0 + subq $18,$20,$18 + + .align 3 +.Roop0: + ldq $3,0($17) + addq $16,8,$16 + addq $17,8,$17 + subq $20,1,$20 + srl $4,$19,$5 + sll $3,$7,$6 + bis $3,$3,$4 + bis $5,$6,$8 + stq $8,-8($16) + bne $20,.Roop0 + +.R0: beq $18,.Rend + + .align 3 +.Roop: ldq $3,0($17) + addq $16,32,$16 + subq $18,4,$18 + srl $4,$19,$5 + sll $3,$7,$6 + + ldq $4,8($17) + srl $3,$19,$1 + bis $5,$6,$8 + stq $8,-32($16) + sll $4,$7,$2 + + ldq $3,16($17) + srl $4,$19,$5 + bis $1,$2,$8 + stq $8,-24($16) + sll $3,$7,$6 + + ldq $4,24($17) + srl $3,$19,$1 + bis $5,$6,$8 + stq $8,-16($16) + sll $4,$7,$2 + + addq $17,32,$17 + bis $1,$2,$8 + stq $8,-8($16) + + bgt $18,.Roop + +.Rend: srl $4,$19,$8 + stq $8,0($16) + ret $31,($26),1 + .end mpihelp_rshift + diff --git a/mpi/alpha/mpih-sub1.S b/mpi/alpha/mpih-sub1.S new file mode 100644 index 0000000..2dea14f --- /dev/null +++ b/mpi/alpha/mpih-sub1.S @@ -0,0 +1,125 @@ +/* Alpha sub_n -- Subtract two limb vectors of the same length > 0 and + * store difference in a third limb vector. + * Copyright (C) 1995, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +/******************* + * mpi_limb_t + * mpihelp_sub_n( mpi_ptr_t res_ptr, (r16) + * mpi_ptr_t s1_ptr, (r17) + * mpi_ptr_t s2_ptr, (r18) + * mpi_size_t size) (r19) + */ + + .set noreorder + .set noat +.text + .align 3 + .globl mpihelp_sub_n + .ent mpihelp_sub_n +mpihelp_sub_n: + .frame $30,0,$26,0 + + ldq $3,0($17) + ldq $4,0($18) + + subq $19,1,$19 + and $19,4-1,$2 # number of limbs in first loop + bis $31,$31,$0 + beq $2,.L0 # if multiple of 4 limbs, skip first loop + + subq $19,$2,$19 + +.Loop0: subq $2,1,$2 + ldq $5,8($17) + addq $4,$0,$4 + ldq $6,8($18) + cmpult $4,$0,$1 + subq $3,$4,$4 + cmpult $3,$4,$0 + stq $4,0($16) + or $0,$1,$0 + + addq $17,8,$17 + addq $18,8,$18 + bis $5,$5,$3 + bis $6,$6,$4 + addq $16,8,$16 + bne $2,.Loop0 + +.L0: beq $19,.Lend + + .align 3 +.Loop: subq $19,4,$19 + + ldq $5,8($17) + addq $4,$0,$4 + ldq $6,8($18) + cmpult $4,$0,$1 + subq $3,$4,$4 + cmpult $3,$4,$0 + stq $4,0($16) + or $0,$1,$0 + + ldq $3,16($17) + addq $6,$0,$6 + ldq $4,16($18) + cmpult $6,$0,$1 + subq $5,$6,$6 + cmpult $5,$6,$0 + stq $6,8($16) + or $0,$1,$0 + + ldq $5,24($17) + addq $4,$0,$4 + ldq $6,24($18) + cmpult $4,$0,$1 + subq $3,$4,$4 + cmpult $3,$4,$0 + stq $4,16($16) + or $0,$1,$0 + + ldq $3,32($17) + addq $6,$0,$6 + ldq $4,32($18) + cmpult $6,$0,$1 + subq $5,$6,$6 + cmpult $5,$6,$0 + stq $6,24($16) + or $0,$1,$0 + + addq $17,32,$17 + addq $18,32,$18 + addq $16,32,$16 + bne $19,.Loop + +.Lend: addq $4,$0,$4 + cmpult $4,$0,$1 + subq $3,$4,$4 + cmpult $3,$4,$0 + stq $4,0($16) + or $0,$1,$0 + ret $31,($26),1 + + .end mpihelp_sub_n + + diff --git a/mpi/alpha/udiv-qrnnd.S b/mpi/alpha/udiv-qrnnd.S new file mode 100644 index 0000000..d715908 --- /dev/null +++ b/mpi/alpha/udiv-qrnnd.S @@ -0,0 +1,162 @@ +/* Alpha 21064 __udiv_qrnnd + * + * Copyright (C) 1992, 1994, 1995, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + + .set noreorder + .set noat +.text + .align 3 + .globl __udiv_qrnnd + .ent __udiv_qrnnd +__udiv_qrnnd: + .frame $30,0,$26,0 + .prologue 0 +#define cnt $2 +#define tmp $3 +#define rem_ptr $16 +#define n1 $17 +#define n0 $18 +#define d $19 +#define qb $20 + + ldiq cnt,16 + blt d,.Largedivisor + +.Loop1: cmplt n0,0,tmp + addq n1,n1,n1 + bis n1,tmp,n1 + addq n0,n0,n0 + cmpule d,n1,qb + subq n1,d,tmp + cmovne qb,tmp,n1 + bis n0,qb,n0 + cmplt n0,0,tmp + addq n1,n1,n1 + bis n1,tmp,n1 + addq n0,n0,n0 + cmpule d,n1,qb + subq n1,d,tmp + cmovne qb,tmp,n1 + bis n0,qb,n0 + cmplt n0,0,tmp + addq n1,n1,n1 + bis n1,tmp,n1 + addq n0,n0,n0 + cmpule d,n1,qb + subq n1,d,tmp + cmovne qb,tmp,n1 + bis n0,qb,n0 + cmplt n0,0,tmp + addq n1,n1,n1 + bis n1,tmp,n1 + addq n0,n0,n0 + cmpule d,n1,qb + subq n1,d,tmp + cmovne qb,tmp,n1 + bis n0,qb,n0 + subq cnt,1,cnt + bgt cnt,.Loop1 + stq n1,0(rem_ptr) + bis $31,n0,$0 + ret $31,($26),1 + +.Largedivisor: + and n0,1,$4 + + srl n0,1,n0 + sll n1,63,tmp + or tmp,n0,n0 + srl n1,1,n1 + + and d,1,$6 + srl d,1,$5 + addq $5,$6,$5 + +.Loop2: cmplt n0,0,tmp + addq n1,n1,n1 + bis n1,tmp,n1 + addq n0,n0,n0 + cmpule $5,n1,qb + subq n1,$5,tmp + cmovne qb,tmp,n1 + bis n0,qb,n0 + cmplt n0,0,tmp + addq n1,n1,n1 + bis n1,tmp,n1 + addq n0,n0,n0 + cmpule $5,n1,qb + subq n1,$5,tmp + cmovne qb,tmp,n1 + bis n0,qb,n0 + cmplt n0,0,tmp + addq n1,n1,n1 + bis n1,tmp,n1 + addq n0,n0,n0 + cmpule $5,n1,qb + subq n1,$5,tmp + cmovne qb,tmp,n1 + bis n0,qb,n0 + cmplt n0,0,tmp + addq n1,n1,n1 + bis n1,tmp,n1 + addq n0,n0,n0 + cmpule $5,n1,qb + subq n1,$5,tmp + cmovne qb,tmp,n1 + bis n0,qb,n0 + subq cnt,1,cnt + bgt cnt,.Loop2 + + addq n1,n1,n1 + addq $4,n1,n1 + bne $6,.LOdd + stq n1,0(rem_ptr) + bis $31,n0,$0 + ret $31,($26),1 + +.LOdd: + /* q' in n0. r' in n1 */ + addq n1,n0,n1 + cmpult n1,n0,tmp # tmp := carry from addq + beq tmp,.LLp6 + addq n0,1,n0 + subq n1,d,n1 +.LLp6: cmpult n1,d,tmp + bne tmp,.LLp7 + addq n0,1,n0 + subq n1,d,n1 +.LLp7: + stq n1,0(rem_ptr) + bis $31,n0,$0 + ret $31,($26),1 + + .end __udiv_qrnnd diff --git a/mpi/config.links b/mpi/config.links new file mode 100644 index 0000000..451a5bf --- /dev/null +++ b/mpi/config.links @@ -0,0 +1,328 @@ +# config.links - helper for ../configure +# Copyright (C) 1998, 1999, 2000, 2001, 2002, 2003 +# 2004 Free Software Foundation, Inc. +# +# This file is part of GnuPG. +# +# GnuPG is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 2 of the License, or +# (at your option) any later version. +# +# GnuPG 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 General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program; if not, write to the Free Software +# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA + +# sourced by ../configure to get the list of files to link +# this should set $mpi_ln_list. +# Note: this is called from the above directory. + + +mpi_extra_modules= +mpi_sflags= + +test -d ./mpi || mkdir ./mpi + +echo '/* created by config.links - do not edit */' >./mpi/asm-syntax.h +echo "/* Host: ${host} */" >>./mpi/asm-syntax.h + +if test "$try_asm_modules" = "yes" ; then +case "${host}" in + powerpc-apple-darwin* | \ + i[34567]86*-*-openbsd[12]* | \ + i[34567]86*-*-openbsd3.[0123]*) + echo '/* No working assembler modules available */' >>./mpi/asm-syntax.h + path="" + ;; + i[3467]86*-*-openbsd* | \ + i[3467]86*-*-freebsd*-elf | \ + i[3467]86*-*-freebsd[3-9]* | \ + i[3467]86*-*-freebsdelf* | \ + i[3467]86*-*-netbsd* | \ + i[3467]86*-*-k*bsd*) + echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h + path="i386" + ;; + i586*-*-openbsd* | \ + i586*-*-freebsd*-elf | \ + i586*-*-freebsd[3-9]* | \ + i586*-*-freebsdelf* | \ + i586*-*-netbsd* | \ + i586*-*-k*bsd* | \ + pentium-*-netbsd* | \ + pentiumpro-*-netbsd*) + echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h + path="i586 i386" + ;; + i[34]86*-*-bsdi4*) + echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h + path="i386" + ;; + i[3467]86*-*-linuxaout* | \ + i[3467]86*-*-linuxoldld* | \ + i[3467]86*-*-*bsd*) + echo '#define BSD_SYNTAX' >>./mpi/asm-syntax.h + echo '#define X86_BROKEN_ALIGN' >>./mpi/asm-syntax.h + cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h + path="i386" + ;; + i586*-*-linuxaout* | \ + i586*-*-linuxoldld* | \ + i586*-*-*bsd*) + echo '#define BSD_SYNTAX' >>./mpi/asm-syntax.h + echo '#define X86_BROKEN_ALIGN' >>./mpi/asm-syntax.h + cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h + path="i586 i386" + ;; + i[3467]86*-msdosdjgpp*) + echo '#define BSD_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h + path="i386" + ;; + i586*-msdosdjgpp*) + echo '#define BSD_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h + path="i586 i386" + ;; + i[3467]86*-*-*) + echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h + path="i386" + ;; + i586*-*-* | \ + pentium-*-* | \ + pentiumpro-*-*) + echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h + path="i586 i386" + ;; + alpha*-*-*) + echo '/* configured for alpha */' >>./mpi/asm-syntax.h + path="alpha" + mpi_extra_modules="udiv-qrnnd" + ;; + hppa7000*-*-*) + echo '/* configured for HPPA (pa7000) */' >>./mpi/asm-syntax.h + path="hppa1.1 hppa" + mpi_extra_modules="udiv-qrnnd" + ;; + hppa1.0*-*-*) + echo '/* configured for HPPA 1.0 */' >>./mpi/asm-syntax.h + path="hppa" + mpi_extra_modules="udiv-qrnnd" + ;; + hppa*-*-*) # assume pa7100 + echo '/* configured for HPPA (pa7100) */' >>./mpi/asm-syntax.h + path="pa7100 hppa1.1 hppa" + mpi_extra_modules="udiv-qrnnd" + ;; + sparc64-*-linux-gnu) + # An extra rule because we have an report for this one only. + # Should be compared against the next GMP version + echo '/* configured for sparc64-*-linux-gnu */' >>./mpi/asm-syntax.h + path="sparc32v8 sparc32" + mpi_extra_modules="udiv" + ;; + sparc64-sun-solaris2*) + # Got a report that udiv is missing, so we try this one + echo '/* configured for sparc64-sun-solaris2 */' >>./mpi/asm-syntax.h + path="sparc32v8 sparc32" + mpi_extra_modules="udiv" + ;; + sparc64-*-netbsd* | sparc64-*-freebsd* | sparc64-*-openbsd*) + # There are no sparc64 assembler modules that work on the + # *BSDs, so use the generic C functions. + echo '/* No working assembler modules available */' >>./mpi/asm-syntax.h + path="" + ;; + sparc9*-*-* | \ + sparc64*-*-* | \ + ultrasparc*-*-* ) + echo '/* configured for sparc9 or higher */' >>./mpi/asm-syntax.h + path="sparc32v8 sparc32" + ;; + sparc8*-*-* | \ + microsparc*-*-*) + echo '/* configured for sparc8 */' >>./mpi/asm-syntax.h + path="sparc32v8 sparc32" + ;; + supersparc*-*-*) + echo '/* configured for supersparc */' >>./mpi/asm-syntax.h + path="supersparc sparc32v8 sparc32" + mpi_extra_modules="udiv" + ;; + sparc*-*-*) + echo '/* configured for sparc */' >>./mpi/asm-syntax.h + path="sparc32" + mpi_extra_modules="udiv" + ;; + mips[34]*-*-* | \ + mips*-*-irix6*) + echo '/* configured for MIPS3 */' >>./mpi/asm-syntax.h + path="mips3" + ;; + mips*-*-*) + echo '/* configured for MIPS2 */' >>./mpi/asm-syntax.h + path="mips2" + ;; + + # Motorola 68k configurations. Let m68k mean 68020-68040. + # mc68000 or mc68060 configurations need to be specified explicitly + m680[234]0*-*-linuxaout* | \ + m68k*-*-linuxaout*) + echo '#define MIT_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/m68k/syntax.h >>./mpi/asm-syntax.h + path="m68k/mc68020 m68k" + ;; + m68060*-*-linuxaout*) + echo '#define MIT_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/m68k/syntax.h >>./mpi/asm-syntax.h + path="m68k" + ;; + m680[234]0*-*-linux* | \ + m68k*-*-linux*) + echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/m68k/syntax.h >>./mpi/asm-syntax.h + ;; + m68060*-*-linux*) + echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/m68k/syntax.h >>./mpi/asm-syntax.h + path="m68k" + ;; + m68k-atari-mint) + echo '#define MIT_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/m68k/syntax.h >>./mpi/asm-syntax.h + path="m68k/mc68020 m68k" + ;; + m68000*-*-* | \ + m68060*-*-*) + echo '#define MIT_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/m68k/syntax.h >>./mpi/asm-syntax.h + path="m68k/mc68000" + ;; + m680[234]0*-*-* | \ + m68k*-*-*) + echo '#define MIT_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/m68k/syntax.h >>./mpi/asm-syntax.h + path="m68k/mc68020 m68k" + ;; + + powerpc*-*-linux*) + echo '/* configured for powerpc/ELF */' >>./mpi/asm-syntax.h + echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/powerpc32/syntax.h >>./mpi/asm-syntax.h + path="powerpc32" + ;; + + powerpc*-*-netbsd* | powerpc*-*-openbsd*) + echo '/* configured for {Open,Net}BSD on powerpc */' >>./mpi/asm-syntax.h + echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h + cat $srcdir/mpi/powerpc32/syntax.h >>./mpi/asm-syntax.h + mpi_sflags="-Wa,-mppc" + path="powerpc32" + ;; + + rs6000-*-aix[456789]* | \ + rs6000-*-aix3.2.[456789]) + mpi_sflags="-Wa,-mpwr" + path="power" + mpi_extra_modules="udiv-w-sdiv" + ;; + rs6000-*-* | \ + power-*-* | \ + power2-*-*) + mpi_sflags="-Wa,-mppc" + path="power" + mpi_extra_modules="udiv-w-sdiv" + ;; + powerpc-ibm-aix4.2.* ) + # I am not sure about this one but a machine identified by + # powerpc-ibm-aix4.2.1.0 cannot use the powerpc32 code. + mpi_sflags="-Wa,-mpwr" + path="power" + mpi_extra_modules="udiv-w-sdiv" + ;; + ppc601-*-*) + mpi_sflags="-Wa,-mppc" + path="power powerpc32" + ;; + ppc60[234]*-*-* | \ + powerpc*-*-*) + mpi_sflags="-Wa,-mppc" + path="powerpc32" + ;; + ppc620-*-* | \ + powerpc64*-*-*) + mpi_sflags="-Wa,-mppc" + path="powerpc64" + ;; + + *) + echo '/* No assembler modules configured */' >>./mpi/asm-syntax.h + path="" + ;; +esac +else + echo '/* Assembler modules disabled on request */' >>./mpi/asm-syntax.h + path="" +fi + + +# Make sysdep.h +echo '/* created by config.links - do not edit */' >./mpi/sysdep.h +if test x$ac_cv_sys_symbol_underscore = xyes; then + cat <<EOF >>./mpi/sysdep.h +#if __STDC__ +#define C_SYMBOL_NAME(name) _##name +#else +#define C_SYMBOL_NAME(name) _/**/name +#endif +EOF +else + cat <<EOF >>./mpi/sysdep.h +#define C_SYMBOL_NAME(name) name +EOF +fi + + +# fixme: grep these modules from Makefile.in +mpi_ln_modules="${mpi_extra_modules} mpih-add1 mpih-mul1 mpih-mul2 mpih-mul3 \ + mpih-lshift mpih-rshift mpih-sub1" + +mpi_ln_objects= +mpi_ln_list= + +# try to get file to link from the assembler subdirectory and +# if this fails get it from the generic subdirectory. +path="$path generic" +for fn in $mpi_ln_modules ; do + mpi_ln_objects="$mpi_ln_objects $fn.o" + for dir in $path ; do + rm -f $srcdir/mpi/$fn.[Sc] + if test -f $srcdir/mpi/$dir/$fn.S ; then + mpi_ln_list="$mpi_ln_list mpi/$fn.S:mpi/$dir/$fn.S" + break; + elif test -f $srcdir/mpi/$dir/$fn.c ; then + mpi_ln_list="$mpi_ln_list mpi/$fn.c:mpi/$dir/$fn.c" + break; + fi + done +done + +# Same thing for the file which defines the limb size +path="$path generic" +for dir in $path ; do + rm -f $srcdir/mpi/mpi-asm-defs.h + if test -f $srcdir/mpi/$dir/mpi-asm-defs.h ; then + mpi_ln_list="$mpi_ln_list mpi/mpi-asm-defs.h:mpi/$dir/mpi-asm-defs.h" + break; + fi +done diff --git a/mpi/generic/distfiles b/mpi/generic/distfiles new file mode 100644 index 0000000..649e829 --- /dev/null +++ b/mpi/generic/distfiles @@ -0,0 +1,10 @@ +mpih-add1.c +mpih-mul1.c +mpih-mul2.c +mpih-mul3.c +mpih-lshift.c +mpih-rshift.c +mpih-sub1.c +udiv-w-sdiv.c +mpi-asm-defs.h + diff --git a/mpi/generic/mpi-asm-defs.h b/mpi/generic/mpi-asm-defs.h new file mode 100644 index 0000000..13424e2 --- /dev/null +++ b/mpi/generic/mpi-asm-defs.h @@ -0,0 +1,10 @@ +/* This file defines some basic constants for the MPI machinery. We + * need to define the types on a per-CPU basis, so it is done with + * this file here. */ +#define BYTES_PER_MPI_LIMB (SIZEOF_UNSIGNED_LONG) + + + + + + diff --git a/mpi/generic/mpih-add1.c b/mpi/generic/mpih-add1.c new file mode 100644 index 0000000..ebcd132 --- /dev/null +++ b/mpi/generic/mpih-add1.c @@ -0,0 +1,66 @@ +/* mpihelp-add_1.c - MPI helper functions + * Copyright (C) 1994, 1996, 1997, 1998, + * 2000 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" +#include "longlong.h" + +mpi_limb_t +mpihelp_add_n( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_ptr_t s2_ptr, mpi_size_t size) +{ + mpi_limb_t x, y, cy; + mpi_size_t j; + + /* The loop counter and index J goes from -SIZE to -1. This way + the loop becomes faster. */ + j = -size; + + /* Offset the base pointers to compensate for the negative indices. */ + s1_ptr -= j; + s2_ptr -= j; + res_ptr -= j; + + cy = 0; + do { + y = s2_ptr[j]; + x = s1_ptr[j]; + y += cy; /* add previous carry to one addend */ + cy = y < cy; /* get out carry from that addition */ + y += x; /* add other addend */ + cy += y < x; /* get out carry from that add, combine */ + res_ptr[j] = y; + } while( ++j ); + + return cy; +} + diff --git a/mpi/generic/mpih-lshift.c b/mpi/generic/mpih-lshift.c new file mode 100644 index 0000000..e55cb49 --- /dev/null +++ b/mpi/generic/mpih-lshift.c @@ -0,0 +1,70 @@ +/* mpihelp-lshift.c - MPI helper functions + * Copyright (C) 1994, 1996, 1998, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" + +/* Shift U (pointed to by UP and USIZE digits long) CNT bits to the left + * and store the USIZE least significant digits of the result at WP. + * Return the bits shifted out from the most significant digit. + * + * Argument constraints: + * 1. 0 < CNT < BITS_PER_MP_LIMB + * 2. If the result is to be written over the input, WP must be >= UP. + */ + +mpi_limb_t +mpihelp_lshift( mpi_ptr_t wp, mpi_ptr_t up, mpi_size_t usize, + unsigned int cnt) +{ + mpi_limb_t high_limb, low_limb; + unsigned sh_1, sh_2; + mpi_size_t i; + mpi_limb_t retval; + + sh_1 = cnt; + wp += 1; + sh_2 = BITS_PER_MPI_LIMB - sh_1; + i = usize - 1; + low_limb = up[i]; + retval = low_limb >> sh_2; + high_limb = low_limb; + while( --i >= 0 ) { + low_limb = up[i]; + wp[i] = (high_limb << sh_1) | (low_limb >> sh_2); + high_limb = low_limb; + } + wp[i] = high_limb << sh_1; + + return retval; +} + + diff --git a/mpi/generic/mpih-mul1.c b/mpi/generic/mpih-mul1.c new file mode 100644 index 0000000..2c5c4ae --- /dev/null +++ b/mpi/generic/mpih-mul1.c @@ -0,0 +1,62 @@ +/* mpihelp-mul_1.c - MPI helper functions + * Copyright (C) 1994, 1996, 1997, 1998, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" +#include "longlong.h" + +mpi_limb_t +mpihelp_mul_1( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, mpi_size_t s1_size, + mpi_limb_t s2_limb) +{ + mpi_limb_t cy_limb; + mpi_size_t j; + mpi_limb_t prod_high, prod_low; + + /* The loop counter and index J goes from -S1_SIZE to -1. This way + * the loop becomes faster. */ + j = -s1_size; + + /* Offset the base pointers to compensate for the negative indices. */ + s1_ptr -= j; + res_ptr -= j; + + cy_limb = 0; + do { + umul_ppmm( prod_high, prod_low, s1_ptr[j], s2_limb ); + prod_low += cy_limb; + cy_limb = (prod_low < cy_limb?1:0) + prod_high; + res_ptr[j] = prod_low; + } while( ++j ); + + return cy_limb; +} + diff --git a/mpi/generic/mpih-mul2.c b/mpi/generic/mpih-mul2.c new file mode 100644 index 0000000..89ef6b9 --- /dev/null +++ b/mpi/generic/mpih-mul2.c @@ -0,0 +1,67 @@ +/* mpihelp-mul_2.c - MPI helper functions + * Copyright (C) 1994, 1996, 1997, 1998, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" +#include "longlong.h" + + +mpi_limb_t +mpihelp_addmul_1( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_size_t s1_size, mpi_limb_t s2_limb) +{ + mpi_limb_t cy_limb; + mpi_size_t j; + mpi_limb_t prod_high, prod_low; + mpi_limb_t x; + + /* The loop counter and index J goes from -SIZE to -1. This way + * the loop becomes faster. */ + j = -s1_size; + res_ptr -= j; + s1_ptr -= j; + + cy_limb = 0; + do { + umul_ppmm( prod_high, prod_low, s1_ptr[j], s2_limb ); + + prod_low += cy_limb; + cy_limb = (prod_low < cy_limb?1:0) + prod_high; + + x = res_ptr[j]; + prod_low = x + prod_low; + cy_limb += prod_low < x?1:0; + res_ptr[j] = prod_low; + } while ( ++j ); + return cy_limb; +} + + diff --git a/mpi/generic/mpih-mul3.c b/mpi/generic/mpih-mul3.c new file mode 100644 index 0000000..ac473a0 --- /dev/null +++ b/mpi/generic/mpih-mul3.c @@ -0,0 +1,68 @@ +/* mpihelp-mul_3.c - MPI helper functions + * Copyright (C) 1994, 1996, 1997, 1998, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" +#include "longlong.h" + + +mpi_limb_t +mpihelp_submul_1( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_size_t s1_size, mpi_limb_t s2_limb) +{ + mpi_limb_t cy_limb; + mpi_size_t j; + mpi_limb_t prod_high, prod_low; + mpi_limb_t x; + + /* The loop counter and index J goes from -SIZE to -1. This way + * the loop becomes faster. */ + j = -s1_size; + res_ptr -= j; + s1_ptr -= j; + + cy_limb = 0; + do { + umul_ppmm( prod_high, prod_low, s1_ptr[j], s2_limb); + + prod_low += cy_limb; + cy_limb = (prod_low < cy_limb?1:0) + prod_high; + + x = res_ptr[j]; + prod_low = x - prod_low; + cy_limb += prod_low > x?1:0; + res_ptr[j] = prod_low; + } while( ++j ); + + return cy_limb; +} + + diff --git a/mpi/generic/mpih-rshift.c b/mpi/generic/mpih-rshift.c new file mode 100644 index 0000000..7aefc59 --- /dev/null +++ b/mpi/generic/mpih-rshift.c @@ -0,0 +1,69 @@ +/* mpih-rshift.c - MPI helper functions + * Copyright (C) 1994, 1996, 1998, 1999, + * 2000, 2001 Free Software Foundation, Inc. + * + * This file is part of GNUPG + * + * GNUPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GNUPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" + + +/* Shift U (pointed to by UP and USIZE limbs long) CNT bits to the right + * and store the USIZE least significant limbs of the result at WP. + * The bits shifted out to the right are returned. + * + * Argument constraints: + * 1. 0 < CNT < BITS_PER_MP_LIMB + * 2. If the result is to be written over the input, WP must be <= UP. + */ + +mpi_limb_t +mpihelp_rshift( mpi_ptr_t wp, mpi_ptr_t up, mpi_size_t usize, unsigned cnt) +{ + mpi_limb_t high_limb, low_limb; + unsigned sh_1, sh_2; + mpi_size_t i; + mpi_limb_t retval; + + sh_1 = cnt; + wp -= 1; + sh_2 = BITS_PER_MPI_LIMB - sh_1; + high_limb = up[0]; + retval = high_limb << sh_2; + low_limb = high_limb; + for( i=1; i < usize; i++) { + high_limb = up[i]; + wp[i] = (low_limb >> sh_1) | (high_limb << sh_2); + low_limb = high_limb; + } + wp[i] = low_limb >> sh_1; + + return retval; +} + diff --git a/mpi/generic/mpih-sub1.c b/mpi/generic/mpih-sub1.c new file mode 100644 index 0000000..2230457 --- /dev/null +++ b/mpi/generic/mpih-sub1.c @@ -0,0 +1,66 @@ +/* mpihelp-add_2.c - MPI helper functions + * Copyright (C) 1994, 1996, 1997, 1998, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" +#include "longlong.h" + +mpi_limb_t +mpihelp_sub_n( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_ptr_t s2_ptr, mpi_size_t size) +{ + mpi_limb_t x, y, cy; + mpi_size_t j; + + /* The loop counter and index J goes from -SIZE to -1. This way + the loop becomes faster. */ + j = -size; + + /* Offset the base pointers to compensate for the negative indices. */ + s1_ptr -= j; + s2_ptr -= j; + res_ptr -= j; + + cy = 0; + do { + y = s2_ptr[j]; + x = s1_ptr[j]; + y += cy; /* add previous carry to subtrahend */ + cy = y < cy; /* get out carry from that addition */ + y = x - y; /* main subtract */ + cy += y > x; /* get out carry from the subtract, combine */ + res_ptr[j] = y; + } while( ++j ); + + return cy; +} + + diff --git a/mpi/generic/udiv-w-sdiv.c b/mpi/generic/udiv-w-sdiv.c new file mode 100644 index 0000000..de71011 --- /dev/null +++ b/mpi/generic/udiv-w-sdiv.c @@ -0,0 +1,134 @@ +/* mpihelp_udiv_w_sdiv -- implement udiv_qrnnd on machines with only signed + * division. + * Copyright (C) 1992, 1994, 1996, 1998 Free Software Foundation, Inc. + * Contributed by Peter L. Montgomery. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" +#include "longlong.h" + + +#if 0 /* not yet ported to MPI */ + +mpi_limb_t +mpihelp_udiv_w_sdiv( mpi_limp_t *rp, + mpi_limp_t *a1, + mpi_limp_t *a0, + mpi_limp_t *d ) +{ + mp_limb_t q, r; + mp_limb_t c0, c1, b1; + + if ((mpi_limb_signed_t) d >= 0) + { + if (a1 < d - a1 - (a0 >> (BITS_PER_MP_LIMB - 1))) + { + /* dividend, divisor, and quotient are nonnegative */ + sdiv_qrnnd (q, r, a1, a0, d); + } + else + { + /* Compute c1*2^32 + c0 = a1*2^32 + a0 - 2^31*d */ + sub_ddmmss (c1, c0, a1, a0, d >> 1, d << (BITS_PER_MP_LIMB - 1)); + /* Divide (c1*2^32 + c0) by d */ + sdiv_qrnnd (q, r, c1, c0, d); + /* Add 2^31 to quotient */ + q += (mp_limb_t) 1 << (BITS_PER_MP_LIMB - 1); + } + } + else + { + b1 = d >> 1; /* d/2, between 2^30 and 2^31 - 1 */ + c1 = a1 >> 1; /* A/2 */ + c0 = (a1 << (BITS_PER_MP_LIMB - 1)) + (a0 >> 1); + + if (a1 < b1) /* A < 2^32*b1, so A/2 < 2^31*b1 */ + { + sdiv_qrnnd (q, r, c1, c0, b1); /* (A/2) / (d/2) */ + + r = 2*r + (a0 & 1); /* Remainder from A/(2*b1) */ + if ((d & 1) != 0) + { + if (r >= q) + r = r - q; + else if (q - r <= d) + { + r = r - q + d; + q--; + } + else + { + r = r - q + 2*d; + q -= 2; + } + } + } + else if (c1 < b1) /* So 2^31 <= (A/2)/b1 < 2^32 */ + { + c1 = (b1 - 1) - c1; + c0 = ~c0; /* logical NOT */ + + sdiv_qrnnd (q, r, c1, c0, b1); /* (A/2) / (d/2) */ + + q = ~q; /* (A/2)/b1 */ + r = (b1 - 1) - r; + + r = 2*r + (a0 & 1); /* A/(2*b1) */ + + if ((d & 1) != 0) + { + if (r >= q) + r = r - q; + else if (q - r <= d) + { + r = r - q + d; + q--; + } + else + { + r = r - q + 2*d; + q -= 2; + } + } + } + else /* Implies c1 = b1 */ + { /* Hence a1 = d - 1 = 2*b1 - 1 */ + if (a0 >= -d) + { + q = -1; + r = a0 + d; + } + else + { + q = -2; + r = a0 + 2*d; + } + } + } + + *rp = r; + return q; +} + +#endif + diff --git a/mpi/hppa/README b/mpi/hppa/README new file mode 100644 index 0000000..5a2d5fd --- /dev/null +++ b/mpi/hppa/README @@ -0,0 +1,84 @@ +This directory contains mpn functions for various HP PA-RISC chips. Code +that runs faster on the PA7100 and later implementations, is in the pa7100 +directory. + +RELEVANT OPTIMIZATION ISSUES + + Load and Store timing + +On the PA7000 no memory instructions can issue the two cycles after a store. +For the PA7100, this is reduced to one cycle. + +The PA7100 has a lookup-free cache, so it helps to schedule loads and the +dependent instruction really far from each other. + +STATUS + +1. mpn_mul_1 could be improved to 6.5 cycles/limb on the PA7100, using the + instructions bwlow (but some sw pipelining is needed to avoid the + xmpyu-fstds delay): + + fldds s1_ptr + + xmpyu + fstds N(%r30) + xmpyu + fstds N(%r30) + + ldws N(%r30) + ldws N(%r30) + ldws N(%r30) + ldws N(%r30) + + addc + stws res_ptr + addc + stws res_ptr + + addib Loop + +2. mpn_addmul_1 could be improved from the current 10 to 7.5 cycles/limb + (asymptotically) on the PA7100, using the instructions below. With proper + sw pipelining and the unrolling level below, the speed becomes 8 + cycles/limb. + + fldds s1_ptr + fldds s1_ptr + + xmpyu + fstds N(%r30) + xmpyu + fstds N(%r30) + xmpyu + fstds N(%r30) + xmpyu + fstds N(%r30) + + ldws N(%r30) + ldws N(%r30) + ldws N(%r30) + ldws N(%r30) + ldws N(%r30) + ldws N(%r30) + ldws N(%r30) + ldws N(%r30) + addc + addc + addc + addc + addc %r0,%r0,cy-limb + + ldws res_ptr + ldws res_ptr + ldws res_ptr + ldws res_ptr + add + stws res_ptr + addc + stws res_ptr + addc + stws res_ptr + addc + stws res_ptr + + addib diff --git a/mpi/hppa/distfiles b/mpi/hppa/distfiles new file mode 100644 index 0000000..7f24205 --- /dev/null +++ b/mpi/hppa/distfiles @@ -0,0 +1,7 @@ +README +udiv-qrnnd.S +mpih-add1.S +mpih-sub1.S +mpih-lshift.S +mpih-rshift.S + diff --git a/mpi/hppa/mpih-add1.S b/mpi/hppa/mpih-add1.S new file mode 100644 index 0000000..fbe93ed --- /dev/null +++ b/mpi/hppa/mpih-add1.S @@ -0,0 +1,73 @@ +/* hppa add_n -- Add two limb vectors of the same length > 0 and store + * sum in a third limb vector. + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Fee Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + +/******************* + * mpi_limb_t + * mpihelp_add_n( mpi_ptr_t res_ptr, (gr26) + * mpi_ptr_t s1_ptr, (gr25) + * mpi_ptr_t s2_ptr, (gr24) + * mpi_size_t size) (gr23) + * + * One might want to unroll this as for other processors, but it turns + * out that the data cache contention after a store makes such + * unrolling useless. We can't come under 5 cycles/limb anyway. + */ + + .code + .export mpihelp_add_n + .label mpihelp_add_n + .proc + .callinfo frame=0,no_calls + .entry + + ldws,ma 4(0,%r25),%r20 + ldws,ma 4(0,%r24),%r19 + + addib,= -1,%r23,L$end ; check for (SIZE == 1) + add %r20,%r19,%r28 ; add first limbs ignoring cy + + .label L$loop + ldws,ma 4(0,%r25),%r20 + ldws,ma 4(0,%r24),%r19 + stws,ma %r28,4(0,%r26) + addib,<> -1,%r23,L$loop + addc %r20,%r19,%r28 + + .label L$end + stws %r28,0(0,%r26) + bv 0(%r2) + addc %r0,%r0,%r28 + + .exit + .procend diff --git a/mpi/hppa/mpih-lshift.S b/mpi/hppa/mpih-lshift.S new file mode 100644 index 0000000..0c4de14 --- /dev/null +++ b/mpi/hppa/mpih-lshift.S @@ -0,0 +1,78 @@ +/* hppa lshift + * + * Copyright (C) 1992, 1994, 1998 + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + + +/******************* + * mpi_limb_t + * mpihelp_lshift( mpi_ptr_t wp, (gr26) + * mpi_ptr_t up, (gr25) + * mpi_size_t usize, (gr24) + * unsigned cnt) (gr23) + */ + + .code + .export mpihelp_lshift + .label mpihelp_lshift + .proc + .callinfo frame=64,no_calls + .entry + + sh2add %r24,%r25,%r25 + sh2add %r24,%r26,%r26 + ldws,mb -4(0,%r25),%r22 + subi 32,%r23,%r1 + mtsar %r1 + addib,= -1,%r24,L$0004 + vshd %r0,%r22,%r28 ; compute carry out limb + ldws,mb -4(0,%r25),%r29 + addib,= -1,%r24,L$0002 + vshd %r22,%r29,%r20 + + .label L$loop + ldws,mb -4(0,%r25),%r22 + stws,mb %r20,-4(0,%r26) + addib,= -1,%r24,L$0003 + vshd %r29,%r22,%r20 + ldws,mb -4(0,%r25),%r29 + stws,mb %r20,-4(0,%r26) + addib,<> -1,%r24,L$loop + vshd %r22,%r29,%r20 + + .label L$0002 + stws,mb %r20,-4(0,%r26) + vshd %r29,%r0,%r20 + bv 0(%r2) + stw %r20,-4(0,%r26) + .label L$0003 + stws,mb %r20,-4(0,%r26) + .label L$0004 + vshd %r22,%r0,%r20 + bv 0(%r2) + stw %r20,-4(0,%r26) + + .exit + .procend + + + diff --git a/mpi/hppa/mpih-rshift.S b/mpi/hppa/mpih-rshift.S new file mode 100644 index 0000000..d7e691b --- /dev/null +++ b/mpi/hppa/mpih-rshift.S @@ -0,0 +1,74 @@ +/* hppa rshift + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + + + +/******************* + * mpi_limb_t + * mpihelp_rshift( mpi_ptr_t wp, (gr26) + * mpi_ptr_t up, (gr25) + * mpi_size_t usize, (gr24) + * unsigned cnt) (gr23) + */ + + .code + .export mpihelp_rshift + .label mpihelp_rshift + .proc + .callinfo frame=64,no_calls + .entry + + ldws,ma 4(0,%r25),%r22 + mtsar %r23 + addib,= -1,%r24,L$r004 + vshd %r22,%r0,%r28 ; compute carry out limb + ldws,ma 4(0,%r25),%r29 + addib,= -1,%r24,L$r002 + vshd %r29,%r22,%r20 + + .label L$roop + ldws,ma 4(0,%r25),%r22 + stws,ma %r20,4(0,%r26) + addib,= -1,%r24,L$r003 + vshd %r22,%r29,%r20 + ldws,ma 4(0,%r25),%r29 + stws,ma %r20,4(0,%r26) + addib,<> -1,%r24,L$roop + vshd %r29,%r22,%r20 + + .label L$r002 + stws,ma %r20,4(0,%r26) + vshd %r0,%r29,%r20 + bv 0(%r2) + stw %r20,0(0,%r26) + .label L$r003 + stws,ma %r20,4(0,%r26) + .label L$r004 + vshd %r0,%r22,%r20 + bv 0(%r2) + stw %r20,0(0,%r26) + + .exit + .procend + diff --git a/mpi/hppa/mpih-sub1.S b/mpi/hppa/mpih-sub1.S new file mode 100644 index 0000000..c1e20d5 --- /dev/null +++ b/mpi/hppa/mpih-sub1.S @@ -0,0 +1,81 @@ +/* hppa sub_n -- Sub two limb vectors of the same length > 0 and store + * sum in a third limb vector. + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_sub_n( mpi_ptr_t res_ptr, (gr26) + * mpi_ptr_t s1_ptr, (gr25) + * mpi_ptr_t s2_ptr, (gr24) + * mpi_size_t size) (gr23) + * + * One might want to unroll this as for other processors, but it turns + * out that the data cache contention after a store makes such + * unrolling useless. We can't come under 5 cycles/limb anyway. + */ + + + .code + .export mpihelp_sub_n + .label mpihelp_sub_n + .proc + .callinfo frame=0,no_calls + .entry + + ldws,ma 4(0,%r25),%r20 + ldws,ma 4(0,%r24),%r19 + + addib,= -1,%r23,L$end ; check for (SIZE == 1) + sub %r20,%r19,%r28 ; subtract first limbs ignoring cy + + .label L$loop + ldws,ma 4(0,%r25),%r20 + ldws,ma 4(0,%r24),%r19 + stws,ma %r28,4(0,%r26) + addib,<> -1,%r23,L$loop + subb %r20,%r19,%r28 + + .label L$end + stws %r28,0(0,%r26) + addc %r0,%r0,%r28 + bv 0(%r2) + subi 1,%r28,%r28 + + .exit + .procend + + + diff --git a/mpi/hppa/udiv-qrnnd.S b/mpi/hppa/udiv-qrnnd.S new file mode 100644 index 0000000..a2024ba --- /dev/null +++ b/mpi/hppa/udiv-qrnnd.S @@ -0,0 +1,299 @@ +/* HP-PA __udiv_qrnnd division support, used from longlong.h. + * This version runs fast on pre-PA7000 CPUs. + * + * Copyright (C) 1993, 1994, 1998, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + +/* INPUT PARAMETERS + * rem_ptr gr26 + * n1 gr25 + * n0 gr24 + * d gr23 + * + * The code size is a bit excessive. We could merge the last two ds;addc + * sequences by simply moving the "bb,< Odd" instruction down. The only + * trouble is the FFFFFFFF code that would need some hacking. + */ + + .code + .export __udiv_qrnnd + .label __udiv_qrnnd + .proc + .callinfo frame=0,no_calls + .entry + + comb,< %r23,0,L$largedivisor + sub %r0,%r23,%r1 ; clear cy as side-effect + ds %r0,%r1,%r0 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r24 + ds %r25,%r23,%r25 + addc %r24,%r24,%r28 + ds %r25,%r23,%r25 + comclr,>= %r25,%r0,%r0 + addl %r25,%r23,%r25 + stws %r25,0(0,%r26) + bv 0(%r2) + addc %r28,%r28,%r28 + + .label L$largedivisor + extru %r24,31,1,%r19 ; r19 = n0 & 1 + bb,< %r23,31,L$odd + extru %r23,30,31,%r22 ; r22 = d >> 1 + shd %r25,%r24,1,%r24 ; r24 = new n0 + extru %r25,30,31,%r25 ; r25 = new n1 + sub %r0,%r22,%r21 + ds %r0,%r21,%r0 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + comclr,>= %r25,%r0,%r0 + addl %r25,%r22,%r25 + sh1addl %r25,%r19,%r25 + stws %r25,0(0,%r26) + bv 0(%r2) + addc %r24,%r24,%r28 + + .label L$odd + addib,sv,n 1,%r22,L$FF.. ; r22 = (d / 2 + 1) + shd %r25,%r24,1,%r24 ; r24 = new n0 + extru %r25,30,31,%r25 ; r25 = new n1 + sub %r0,%r22,%r21 + ds %r0,%r21,%r0 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r24 + ds %r25,%r22,%r25 + addc %r24,%r24,%r28 + comclr,>= %r25,%r0,%r0 + addl %r25,%r22,%r25 + sh1addl %r25,%r19,%r25 +; We have computed (n1,,n0) / (d + 1), q' = r28, r' = r25 + add,nuv %r28,%r25,%r25 + addl %r25,%r1,%r25 + addc %r0,%r28,%r28 + sub,<< %r25,%r23,%r0 + addl %r25,%r1,%r25 + stws %r25,0(0,%r26) + bv 0(%r2) + addc %r0,%r28,%r28 + +; This is just a special case of the code above. +; We come here when d == 0xFFFFFFFF + .label L$FF.. + add,uv %r25,%r24,%r24 + sub,<< %r24,%r23,%r0 + ldo 1(%r24),%r24 + stws %r24,0(0,%r26) + bv 0(%r2) + addc %r0,%r25,%r28 + + .exit + .procend diff --git a/mpi/hppa1.1/distfiles b/mpi/hppa1.1/distfiles new file mode 100644 index 0000000..d68227a --- /dev/null +++ b/mpi/hppa1.1/distfiles @@ -0,0 +1,5 @@ +udiv-qrnnd.S +mpih-mul1.S +mpih-mul2.S +mpih-mul3.S + diff --git a/mpi/hppa1.1/mpih-mul1.S b/mpi/hppa1.1/mpih-mul1.S new file mode 100644 index 0000000..d1272a7 --- /dev/null +++ b/mpi/hppa1.1/mpih-mul1.S @@ -0,0 +1,116 @@ +/* hppa1.1 mul_1 -- Multiply a limb vector with a limb and store + * the result in a second limb vector. + * + * Copyright (C) 1992, 1993, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +/******************* + * mpi_limb_t + * mpihelp_mul_1( mpi_ptr_t res_ptr, (r26) + * mpi_ptr_t s1_ptr, (r25) + * mpi_size_t s1_size, (r24) + * mpi_limb_t s2_limb) (r23) + * + * + * + * This runs at 9 cycles/limb on a PA7000. With the used instructions, it can + * not become faster due to data cache contention after a store. On the + * PA7100 it runs at 7 cycles/limb, and that can not be improved either, since + * only the xmpyu does not need the integer pipeline, so the only dual-issue + * we will get are addc+xmpyu. Unrolling would not help either CPU. + * + * We could use fldds to read two limbs at a time from the S1 array, and that + * could bring down the times to 8.5 and 6.5 cycles/limb for the PA7000 and + * PA7100, respectively. We don't do that since it does not seem worth the + * (alignment) troubles... + * + * At least the PA7100 is rumored to be able to deal with cache-misses + * without stalling instruction issue. If this is true, and the cache is + * actually also lockup-free, we should use a deeper software pipeline, and + * load from S1 very early! (The loads and stores to -12(sp) will surely be + * in the cache.) + */ + + .code + .export mpihelp_mul_1 + .label mpihelp_mul_1 + .proc + .callinfo frame=64,no_calls + .entry + + ldo 64(%r30),%r30 + fldws,ma 4(%r25),%fr5 + stw %r23,-16(%r30) ; move s2_limb ... + addib,= -1,%r24,L$just_one_limb + fldws -16(%r30),%fr4 ; ... into fr4 + add %r0,%r0,%r0 ; clear carry + xmpyu %fr4,%fr5,%fr6 + fldws,ma 4(%r25),%fr7 + fstds %fr6,-16(%r30) + xmpyu %fr4,%fr7,%fr8 + ldw -12(%r30),%r19 ; least significant limb in product + ldw -16(%r30),%r28 + + fstds %fr8,-16(%r30) + addib,= -1,%r24,L$end + ldw -12(%r30),%r1 + +; Main loop + .label L$loop + fldws,ma 4(%r25),%fr5 + stws,ma %r19,4(%r26) + addc %r28,%r1,%r19 + xmpyu %fr4,%fr5,%fr6 + ldw -16(%r30),%r28 + fstds %fr6,-16(%r30) + addib,<> -1,%r24,L$loop + ldw -12(%r30),%r1 + + .label L$end + stws,ma %r19,4(%r26) + addc %r28,%r1,%r19 + ldw -16(%r30),%r28 + stws,ma %r19,4(%r26) + addc %r0,%r28,%r28 + bv 0(%r2) + ldo -64(%r30),%r30 + + .label L$just_one_limb + xmpyu %fr4,%fr5,%fr6 + fstds %fr6,-16(%r30) + ldw -16(%r30),%r28 + ldo -64(%r30),%r30 + bv 0(%r2) + fstws %fr6R,0(%r26) + + .exit + .procend + + diff --git a/mpi/hppa1.1/mpih-mul2.S b/mpi/hppa1.1/mpih-mul2.S new file mode 100644 index 0000000..18c4dfe --- /dev/null +++ b/mpi/hppa1.1/mpih-mul2.S @@ -0,0 +1,118 @@ +/* hppa1.1 addmul_1 -- Multiply a limb vector with a limb and add + * the result to a second limb vector. + * + * Copyright (C) 1992, 1993, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +/******************* + * mpi_limb_t + * mpihelp_addmul_1( mpi_ptr_t res_ptr, (r26) + * mpi_ptr_t s1_ptr, (r25) + * mpi_size_t s1_size, (r24) + * mpi_limb_t s2_limb) (r23) + * + * This runs at 11 cycles/limb on a PA7000. With the used instructions, it + * can not become faster due to data cache contention after a store. On the + * PA7100 it runs at 10 cycles/limb, and that can not be improved either, + * since only the xmpyu does not need the integer pipeline, so the only + * dual-issue we will get are addc+xmpyu. Unrolling could gain a cycle/limb + * on the PA7100. + * + * There are some ideas described in mul1.S that applies to this code too. + */ + + .code + .export mpihelp_addmul_1 + .label mpihelp_addmul_1 + .proc + .callinfo frame=64,no_calls + .entry + + ldo 64(%r30),%r30 + fldws,ma 4(%r25),%fr5 + stw %r23,-16(%r30) ; move s2_limb ... + addib,= -1,%r24,L$just_one_limb + fldws -16(%r30),%fr4 ; ... into fr4 + add %r0,%r0,%r0 ; clear carry + xmpyu %fr4,%fr5,%fr6 + fldws,ma 4(%r25),%fr7 + fstds %fr6,-16(%r30) + xmpyu %fr4,%fr7,%fr8 + ldw -12(%r30),%r19 ; least significant limb in product + ldw -16(%r30),%r28 + + fstds %fr8,-16(%r30) + addib,= -1,%r24,L$end + ldw -12(%r30),%r1 + +; Main loop + .label L$loop + ldws 0(%r26),%r29 + fldws,ma 4(%r25),%fr5 + add %r29,%r19,%r19 + stws,ma %r19,4(%r26) + addc %r28,%r1,%r19 + xmpyu %fr4,%fr5,%fr6 + ldw -16(%r30),%r28 + fstds %fr6,-16(%r30) + addc %r0,%r28,%r28 + addib,<> -1,%r24,L$loop + ldw -12(%r30),%r1 + + .label L$end + ldw 0(%r26),%r29 + add %r29,%r19,%r19 + stws,ma %r19,4(%r26) + addc %r28,%r1,%r19 + ldw -16(%r30),%r28 + ldws 0(%r26),%r29 + addc %r0,%r28,%r28 + add %r29,%r19,%r19 + stws,ma %r19,4(%r26) + addc %r0,%r28,%r28 + bv 0(%r2) + ldo -64(%r30),%r30 + + .label L$just_one_limb + xmpyu %fr4,%fr5,%fr6 + ldw 0(%r26),%r29 + fstds %fr6,-16(%r30) + ldw -12(%r30),%r1 + ldw -16(%r30),%r28 + add %r29,%r1,%r19 + stw %r19,0(%r26) + addc %r0,%r28,%r28 + bv 0(%r2) + ldo -64(%r30),%r30 + + .exit + .procend + + diff --git a/mpi/hppa1.1/mpih-mul3.S b/mpi/hppa1.1/mpih-mul3.S new file mode 100644 index 0000000..c5b3925 --- /dev/null +++ b/mpi/hppa1.1/mpih-mul3.S @@ -0,0 +1,128 @@ +/* hppa1.1 submul_1 -- Multiply a limb vector with a limb and add + * the result to a second limb vector. + * + * Copyright (C) 1992, 1993, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +/******************* + * mpi_limb_t + * mpihelp_submul_1( mpi_ptr_t res_ptr, (r26) + * mpi_ptr_t s1_ptr, (r25) + * mpi_size_t s1_size, (r24) + * mpi_limb_t s2_limb) (r23) + * + * + * This runs at 12 cycles/limb on a PA7000. With the used instructions, it + * can not become faster due to data cache contention after a store. On the + * PA7100 it runs at 11 cycles/limb, and that can not be improved either, + * since only the xmpyu does not need the integer pipeline, so the only + * dual-issue we will get are addc+xmpyu. Unrolling could gain a cycle/limb + * on the PA7100. + * + * There are some ideas described in mul1.S that applies to this code too. + * + * It seems possible to make this run as fast as addmul_1, if we use + * sub,>>= %r29,%r19,%r22 + * addi 1,%r28,%r28 + * but that requires reworking the hairy software pipeline... + */ + + + .code + .export mpihelp_submul_1 + .label mpihelp_submul_1 + .proc + .callinfo frame=64,no_calls + .entry + + ldo 64(%r30),%r30 + fldws,ma 4(%r25),%fr5 + stw %r23,-16(%r30) ; move s2_limb ... + addib,= -1,%r24,L$just_one_limb + fldws -16(%r30),%fr4 ; ... into fr4 + add %r0,%r0,%r0 ; clear carry + xmpyu %fr4,%fr5,%fr6 + fldws,ma 4(%r25),%fr7 + fstds %fr6,-16(%r30) + xmpyu %fr4,%fr7,%fr8 + ldw -12(%r30),%r19 ; least significant limb in product + ldw -16(%r30),%r28 + + fstds %fr8,-16(%r30) + addib,= -1,%r24,L$end + ldw -12(%r30),%r1 + +; Main loop + .label L$loop + ldws 0(%r26),%r29 + fldws,ma 4(%r25),%fr5 + sub %r29,%r19,%r22 + add %r22,%r19,%r0 + stws,ma %r22,4(%r26) + addc %r28,%r1,%r19 + xmpyu %fr4,%fr5,%fr6 + ldw -16(%r30),%r28 + fstds %fr6,-16(%r30) + addc %r0,%r28,%r28 + addib,<> -1,%r24,L$loop + ldw -12(%r30),%r1 + + .label L$end + ldw 0(%r26),%r29 + sub %r29,%r19,%r22 + add %r22,%r19,%r0 + stws,ma %r22,4(%r26) + addc %r28,%r1,%r19 + ldw -16(%r30),%r28 + ldws 0(%r26),%r29 + addc %r0,%r28,%r28 + sub %r29,%r19,%r22 + add %r22,%r19,%r0 + stws,ma %r22,4(%r26) + addc %r0,%r28,%r28 + bv 0(%r2) + ldo -64(%r30),%r30 + + .label L$just_one_limb + xmpyu %fr4,%fr5,%fr6 + ldw 0(%r26),%r29 + fstds %fr6,-16(%r30) + ldw -12(%r30),%r1 + ldw -16(%r30),%r28 + sub %r29,%r1,%r22 + add %r22,%r1,%r0 + stw %r22,0(%r26) + addc %r0,%r28,%r28 + bv 0(%r2) + ldo -64(%r30),%r30 + + .exit + .procend + diff --git a/mpi/hppa1.1/udiv-qrnnd.S b/mpi/hppa1.1/udiv-qrnnd.S new file mode 100644 index 0000000..a086822 --- /dev/null +++ b/mpi/hppa1.1/udiv-qrnnd.S @@ -0,0 +1,92 @@ +/* HP-PA __udiv_qrnnd division support, used from longlong.h. + * This version runs fast on PA 7000 and later. + * + * Copyright (C) 1993, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + +/* INPUT PARAMETERS + * rem_ptr gr26 + * n1 gr25 + * n0 gr24 + * d gr23 + */ + + .code + .align 8 + .label L$0000 + .word 0x43f00000 + .word 0x0 + .export __udiv_qrnnd + .label __udiv_qrnnd + .proc + .callinfo frame=64,no_calls + .entry + ldo 64(%r30),%r30 + + stws %r25,-16(0,%r30) ; n_hi + stws %r24,-12(0,%r30) ; n_lo + ldil L'L$0000,%r19 ; ' + ldo R'L$0000(%r19),%r19 ; ' + fldds -16(0,%r30),%fr5 + stws %r23,-12(0,%r30) + comib,<= 0,%r25,L$1 + fcnvxf,dbl,dbl %fr5,%fr5 + fldds 0(0,%r19),%fr4 + fadd,dbl %fr4,%fr5,%fr5 + .label L$1 + fcpy,sgl %fr0,%fr6L + fldws -12(0,%r30),%fr6R + fcnvxf,dbl,dbl %fr6,%fr4 + + fdiv,dbl %fr5,%fr4,%fr5 + + fcnvfx,dbl,dbl %fr5,%fr4 + fstws %fr4R,-16(%r30) + xmpyu %fr4R,%fr6R,%fr6 + ldws -16(%r30),%r28 + fstds %fr6,-16(0,%r30) + ldws -12(0,%r30),%r21 + ldws -16(0,%r30),%r20 + sub %r24,%r21,%r22 + subb %r25,%r20,%r19 + comib,= 0,%r19,L$2 + ldo -64(%r30),%r30 + + add %r22,%r23,%r22 + ldo -1(%r28),%r28 + .label L$2 + bv 0(%r2) + stws %r22,0(0,%r26) + + .exit + .procend + + diff --git a/mpi/i386/distfiles b/mpi/i386/distfiles new file mode 100644 index 0000000..88d2a30 --- /dev/null +++ b/mpi/i386/distfiles @@ -0,0 +1,9 @@ +mpih-add1.S +mpih-mul1.S +mpih-mul2.S +mpih-mul3.S +mpih-lshift.S +mpih-rshift.S +mpih-sub1.S +syntax.h + diff --git a/mpi/i386/mpih-add1.S b/mpi/i386/mpih-add1.S new file mode 100644 index 0000000..cc96014 --- /dev/null +++ b/mpi/i386/mpih-add1.S @@ -0,0 +1,119 @@ +/* i80386 add_n -- Add two limb vectors of the same length > 0 and store + * sum in a third limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1998, + * 2001, 2002 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_add_n( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_ptr_t s2_ptr, (sp + 12) + * mpi_size_t size) (sp + 16) + */ + +.text + ALIGN (3) + .globl C_SYMBOL_NAME(mpihelp_add_n) +C_SYMBOL_NAME(mpihelp_add_n:) + pushl %edi + pushl %esi + + movl 12(%esp),%edi /* res_ptr */ + movl 16(%esp),%esi /* s1_ptr */ + movl 20(%esp),%edx /* s2_ptr */ + movl 24(%esp),%ecx /* size */ + + movl %ecx,%eax + shrl $3,%ecx /* compute count for unrolled loop */ + negl %eax + andl $7,%eax /* get index where to start loop */ + jz Loop /* necessary special case for 0 */ + incl %ecx /* adjust loop count */ + shll $2,%eax /* adjustment for pointers... */ + subl %eax,%edi /* ... since they are offset ... */ + subl %eax,%esi /* ... by a constant when we ... */ + subl %eax,%edx /* ... enter the loop */ + shrl $2,%eax /* restore previous value */ +#ifdef PIC +/* Calculate start address in loop for PIC. Due to limitations in some + assemblers, Loop-L0-3 cannot be put into the leal */ + call L0 +L0: leal (%eax,%eax,8),%eax + addl (%esp),%eax + addl $(Loop-L0-3),%eax + addl $4,%esp +#else +/* Calculate start address in loop for non-PIC. */ + leal Loop-3(%eax,%eax,8),%eax +#endif + jmp *%eax /* jump into loop */ + ALIGN (3) +Loop: movl (%esi),%eax + adcl (%edx),%eax + movl %eax,(%edi) + movl 4(%esi),%eax + adcl 4(%edx),%eax + movl %eax,4(%edi) + movl 8(%esi),%eax + adcl 8(%edx),%eax + movl %eax,8(%edi) + movl 12(%esi),%eax + adcl 12(%edx),%eax + movl %eax,12(%edi) + movl 16(%esi),%eax + adcl 16(%edx),%eax + movl %eax,16(%edi) + movl 20(%esi),%eax + adcl 20(%edx),%eax + movl %eax,20(%edi) + movl 24(%esi),%eax + adcl 24(%edx),%eax + movl %eax,24(%edi) + movl 28(%esi),%eax + adcl 28(%edx),%eax + movl %eax,28(%edi) + leal 32(%edi),%edi + leal 32(%esi),%esi + leal 32(%edx),%edx + decl %ecx + jnz Loop + + sbbl %eax,%eax + negl %eax + + popl %esi + popl %edi + ret + diff --git a/mpi/i386/mpih-lshift.S b/mpi/i386/mpih-lshift.S new file mode 100644 index 0000000..9233824 --- /dev/null +++ b/mpi/i386/mpih-lshift.S @@ -0,0 +1,97 @@ +/* i80386 lshift + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_lshift( mpi_ptr_t wp, (sp + 4) + * mpi_ptr_t up, (sp + 8) + * mpi_size_t usize, (sp + 12) + * unsigned cnt) (sp + 16) + */ + +.text + ALIGN (3) + .globl C_SYMBOL_NAME(mpihelp_lshift) +C_SYMBOL_NAME(mpihelp_lshift:) + pushl %edi + pushl %esi + pushl %ebx + + movl 16(%esp),%edi /* res_ptr */ + movl 20(%esp),%esi /* s_ptr */ + movl 24(%esp),%edx /* size */ + movl 28(%esp),%ecx /* cnt */ + + subl $4,%esi /* adjust s_ptr */ + + movl (%esi,%edx,4),%ebx /* read most significant limb */ + xorl %eax,%eax + shldl %ebx,%eax /* compute carry limb */ + decl %edx + jz Lend + pushl %eax /* push carry limb onto stack */ + testb $1,%dl + jnz L1 /* enter loop in the middle */ + movl %ebx,%eax + + ALIGN (3) +Loop: movl (%esi,%edx,4),%ebx /* load next lower limb */ + shldl %ebx,%eax /* compute result limb */ + movl %eax,(%edi,%edx,4) /* store it */ + decl %edx +L1: movl (%esi,%edx,4),%eax + shldl %eax,%ebx + movl %ebx,(%edi,%edx,4) + decl %edx + jnz Loop + + shll %cl,%eax /* compute least significant limb */ + movl %eax,(%edi) /* store it */ + + popl %eax /* pop carry limb */ + + popl %ebx + popl %esi + popl %edi + ret + +Lend: shll %cl,%ebx /* compute least significant limb */ + movl %ebx,(%edi) /* store it */ + + popl %ebx + popl %esi + popl %edi + ret + diff --git a/mpi/i386/mpih-mul1.S b/mpi/i386/mpih-mul1.S new file mode 100644 index 0000000..a5f83a5 --- /dev/null +++ b/mpi/i386/mpih-mul1.S @@ -0,0 +1,87 @@ +/* i80386 mul_1 -- Multiply a limb vector with a limb and store + * the result in a second limb vector. + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_mul_1( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_size_t s1_size, (sp + 12) + * mpi_limb_t s2_limb) (sp + 16) + */ + +#define res_ptr edi +#define s1_ptr esi +#define size ecx +#define s2_limb ebp + + TEXT + ALIGN (3) + GLOBL C_SYMBOL_NAME(mpihelp_mul_1) +C_SYMBOL_NAME(mpihelp_mul_1:) + + INSN1(push,l ,R(edi)) + INSN1(push,l ,R(esi)) + INSN1(push,l ,R(ebx)) + INSN1(push,l ,R(ebp)) + + INSN2(mov,l ,R(res_ptr),MEM_DISP(esp,20)) + INSN2(mov,l ,R(s1_ptr),MEM_DISP(esp,24)) + INSN2(mov,l ,R(size),MEM_DISP(esp,28)) + INSN2(mov,l ,R(s2_limb),MEM_DISP(esp,32)) + + INSN2(lea,l ,R(res_ptr),MEM_INDEX(res_ptr,size,4)) + INSN2(lea,l ,R(s1_ptr),MEM_INDEX(s1_ptr,size,4)) + INSN1(neg,l ,R(size)) + INSN2(xor,l ,R(ebx),R(ebx)) + ALIGN (3) +Loop: + INSN2(mov,l ,R(eax),MEM_INDEX(s1_ptr,size,4)) + INSN1(mul,l ,R(s2_limb)) + INSN2(add,l ,R(eax),R(ebx)) + INSN2(mov,l ,MEM_INDEX(res_ptr,size,4),R(eax)) + INSN2(adc,l ,R(edx),$0) + INSN2(mov,l ,R(ebx),R(edx)) + + INSN1(inc,l ,R(size)) + INSN1(jnz, ,Loop) + INSN2(mov,l ,R(eax),R(ebx)) + + INSN1(pop,l ,R(ebp)) + INSN1(pop,l ,R(ebx)) + INSN1(pop,l ,R(esi)) + INSN1(pop,l ,R(edi)) + ret + diff --git a/mpi/i386/mpih-mul2.S b/mpi/i386/mpih-mul2.S new file mode 100644 index 0000000..6b6c806 --- /dev/null +++ b/mpi/i386/mpih-mul2.S @@ -0,0 +1,89 @@ +/* i80386 addmul_1 -- Multiply a limb vector with a limb and add + * the result to a second limb vector. + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_addmul_1( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_size_t s1_size, (sp + 12) + * mpi_limb_t s2_limb) (sp + 16) + */ + +#define res_ptr edi +#define s1_ptr esi +#define size ecx +#define s2_limb ebp + + TEXT + ALIGN (3) + GLOBL C_SYMBOL_NAME(mpihelp_addmul_1) +C_SYMBOL_NAME(mpihelp_addmul_1:) + + INSN1(push,l ,R(edi)) + INSN1(push,l ,R(esi)) + INSN1(push,l ,R(ebx)) + INSN1(push,l ,R(ebp)) + + INSN2(mov,l ,R(res_ptr),MEM_DISP(esp,20)) + INSN2(mov,l ,R(s1_ptr),MEM_DISP(esp,24)) + INSN2(mov,l ,R(size),MEM_DISP(esp,28)) + INSN2(mov,l ,R(s2_limb),MEM_DISP(esp,32)) + + INSN2(lea,l ,R(res_ptr),MEM_INDEX(res_ptr,size,4)) + INSN2(lea,l ,R(s1_ptr),MEM_INDEX(s1_ptr,size,4)) + INSN1(neg,l ,R(size)) + INSN2(xor,l ,R(ebx),R(ebx)) + ALIGN (3) +Loop: + INSN2(mov,l ,R(eax),MEM_INDEX(s1_ptr,size,4)) + INSN1(mul,l ,R(s2_limb)) + INSN2(add,l ,R(eax),R(ebx)) + INSN2(adc,l ,R(edx),$0) + INSN2(add,l ,MEM_INDEX(res_ptr,size,4),R(eax)) + INSN2(adc,l ,R(edx),$0) + INSN2(mov,l ,R(ebx),R(edx)) + + INSN1(inc,l ,R(size)) + INSN1(jnz, ,Loop) + INSN2(mov,l ,R(eax),R(ebx)) + + INSN1(pop,l ,R(ebp)) + INSN1(pop,l ,R(ebx)) + INSN1(pop,l ,R(esi)) + INSN1(pop,l ,R(edi)) + ret + diff --git a/mpi/i386/mpih-mul3.S b/mpi/i386/mpih-mul3.S new file mode 100644 index 0000000..81e1e43 --- /dev/null +++ b/mpi/i386/mpih-mul3.S @@ -0,0 +1,89 @@ +/* i80386 submul_1 -- Multiply a limb vector with a limb and add + * the result to a second limb vector. + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_submul_1( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_size_t s1_size, (sp + 12) + * mpi_limb_t s2_limb) (sp + 16) + */ + +#define res_ptr edi +#define s1_ptr esi +#define size ecx +#define s2_limb ebp + + TEXT + ALIGN (3) + GLOBL C_SYMBOL_NAME(mpihelp_submul_1) +C_SYMBOL_NAME(mpihelp_submul_1:) + + INSN1(push,l ,R(edi)) + INSN1(push,l ,R(esi)) + INSN1(push,l ,R(ebx)) + INSN1(push,l ,R(ebp)) + + INSN2(mov,l ,R(res_ptr),MEM_DISP(esp,20)) + INSN2(mov,l ,R(s1_ptr),MEM_DISP(esp,24)) + INSN2(mov,l ,R(size),MEM_DISP(esp,28)) + INSN2(mov,l ,R(s2_limb),MEM_DISP(esp,32)) + + INSN2(lea,l ,R(res_ptr),MEM_INDEX(res_ptr,size,4)) + INSN2(lea,l ,R(s1_ptr),MEM_INDEX(s1_ptr,size,4)) + INSN1(neg,l ,R(size)) + INSN2(xor,l ,R(ebx),R(ebx)) + ALIGN (3) +Loop: + INSN2(mov,l ,R(eax),MEM_INDEX(s1_ptr,size,4)) + INSN1(mul,l ,R(s2_limb)) + INSN2(add,l ,R(eax),R(ebx)) + INSN2(adc,l ,R(edx),$0) + INSN2(sub,l ,MEM_INDEX(res_ptr,size,4),R(eax)) + INSN2(adc,l ,R(edx),$0) + INSN2(mov,l ,R(ebx),R(edx)) + + INSN1(inc,l ,R(size)) + INSN1(jnz, ,Loop) + INSN2(mov,l ,R(eax),R(ebx)) + + INSN1(pop,l ,R(ebp)) + INSN1(pop,l ,R(ebx)) + INSN1(pop,l ,R(esi)) + INSN1(pop,l ,R(edi)) + ret + diff --git a/mpi/i386/mpih-rshift.S b/mpi/i386/mpih-rshift.S new file mode 100644 index 0000000..43afece --- /dev/null +++ b/mpi/i386/mpih-rshift.S @@ -0,0 +1,100 @@ +/* i80386 rshift + * + * Copyright (C) 1992, 1994, 1998, + * 2001, 2002 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_rshift( mpi_ptr_t wp, (sp + 4) + * mpi_ptr_t up, (sp + 8) + * mpi_size_t usize, (sp + 12) + * unsigned cnt) (sp + 16) + */ + +.text + ALIGN (3) + .globl C_SYMBOL_NAME(mpihelp_rshift) +C_SYMBOL_NAME(mpihelp_rshift:) + pushl %edi + pushl %esi + pushl %ebx + + movl 16(%esp),%edi /* wp */ + movl 20(%esp),%esi /* up */ + movl 24(%esp),%edx /* usize */ + movl 28(%esp),%ecx /* cnt */ + + leal -4(%edi,%edx,4),%edi + leal (%esi,%edx,4),%esi + negl %edx + + movl (%esi,%edx,4),%ebx /* read least significant limb */ + xorl %eax,%eax + shrdl %ebx,%eax /* compute carry limb */ + incl %edx + jz Lend2 + pushl %eax /* push carry limb onto stack */ + testb $1,%dl + jnz L2 /* enter loop in the middle */ + movl %ebx,%eax + + ALIGN (3) +Loop2: movl (%esi,%edx,4),%ebx /* load next higher limb */ + shrdl %ebx,%eax /* compute result limb */ + movl %eax,(%edi,%edx,4) /* store it */ + incl %edx +L2: movl (%esi,%edx,4),%eax + shrdl %eax,%ebx + movl %ebx,(%edi,%edx,4) + incl %edx + jnz Loop2 + + shrl %cl,%eax /* compute most significant limb */ + movl %eax,(%edi) /* store it */ + + popl %eax /* pop carry limb */ + + popl %ebx + popl %esi + popl %edi + ret + +Lend2: shrl %cl,%ebx /* compute most significant limb */ + movl %ebx,(%edi) /* store it */ + + popl %ebx + popl %esi + popl %edi + ret + diff --git a/mpi/i386/mpih-sub1.S b/mpi/i386/mpih-sub1.S new file mode 100644 index 0000000..9f0dbd8 --- /dev/null +++ b/mpi/i386/mpih-sub1.S @@ -0,0 +1,120 @@ +/* i80386 sub_n -- Sub two limb vectors of the same length > 0 and store + * sum in a third limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1998, + * 2001, 2002 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_sub_n( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_ptr_t s2_ptr, (sp + 12) + * mpi_size_t size) (sp + 16) + */ + + +.text + ALIGN (3) + .globl C_SYMBOL_NAME(mpihelp_sub_n) +C_SYMBOL_NAME(mpihelp_sub_n:) + pushl %edi + pushl %esi + + movl 12(%esp),%edi /* res_ptr */ + movl 16(%esp),%esi /* s1_ptr */ + movl 20(%esp),%edx /* s2_ptr */ + movl 24(%esp),%ecx /* size */ + + movl %ecx,%eax + shrl $3,%ecx /* compute count for unrolled loop */ + negl %eax + andl $7,%eax /* get index where to start loop */ + jz Loop /* necessary special case for 0 */ + incl %ecx /* adjust loop count */ + shll $2,%eax /* adjustment for pointers... */ + subl %eax,%edi /* ... since they are offset ... */ + subl %eax,%esi /* ... by a constant when we ... */ + subl %eax,%edx /* ... enter the loop */ + shrl $2,%eax /* restore previous value */ +#ifdef PIC +/* Calculate start address in loop for PIC. Due to limitations in some + assemblers, Loop-L0-3 cannot be put into the leal */ + call L0 +L0: leal (%eax,%eax,8),%eax + addl (%esp),%eax + addl $(Loop-L0-3),%eax + addl $4,%esp +#else +/* Calculate start address in loop for non-PIC. */ + leal Loop-3(%eax,%eax,8),%eax +#endif + jmp *%eax /* jump into loop */ + ALIGN (3) +Loop: movl (%esi),%eax + sbbl (%edx),%eax + movl %eax,(%edi) + movl 4(%esi),%eax + sbbl 4(%edx),%eax + movl %eax,4(%edi) + movl 8(%esi),%eax + sbbl 8(%edx),%eax + movl %eax,8(%edi) + movl 12(%esi),%eax + sbbl 12(%edx),%eax + movl %eax,12(%edi) + movl 16(%esi),%eax + sbbl 16(%edx),%eax + movl %eax,16(%edi) + movl 20(%esi),%eax + sbbl 20(%edx),%eax + movl %eax,20(%edi) + movl 24(%esi),%eax + sbbl 24(%edx),%eax + movl %eax,24(%edi) + movl 28(%esi),%eax + sbbl 28(%edx),%eax + movl %eax,28(%edi) + leal 32(%edi),%edi + leal 32(%esi),%esi + leal 32(%edx),%edx + decl %ecx + jnz Loop + + sbbl %eax,%eax + negl %eax + + popl %esi + popl %edi + ret + diff --git a/mpi/i386/syntax.h b/mpi/i386/syntax.h new file mode 100644 index 0000000..b656166 --- /dev/null +++ b/mpi/i386/syntax.h @@ -0,0 +1,69 @@ +/* syntax.h -- Definitions for x86 syntax variations. + * + * Copyright (C) 1992, 1994, 1995, 1998, + * 2001 Free Software Foundation, Inc. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#undef ALIGN + +#if defined (BSD_SYNTAX) || defined (ELF_SYNTAX) +#define R(r) %r +#define MEM(base)(base) +#define MEM_DISP(base,displacement)displacement(R(base)) +#define MEM_INDEX(base,index,size)(R(base),R(index),size) +#ifdef __STDC__ +#define INSN1(mnemonic,size_suffix,dst)mnemonic##size_suffix dst +#define INSN2(mnemonic,size_suffix,dst,src)mnemonic##size_suffix src,dst +#else +#define INSN1(mnemonic,size_suffix,dst)mnemonic/**/size_suffix dst +#define INSN2(mnemonic,size_suffix,dst,src)mnemonic/**/size_suffix src,dst +#endif +#define TEXT .text +#if defined (BSD_SYNTAX) +#define ALIGN(log) .align log +#endif +#if defined (ELF_SYNTAX) +#define ALIGN(log) .align 1<<log +#endif +#define GLOBL .globl +#endif + +#ifdef INTEL_SYNTAX +#define R(r) r +#define MEM(base)[base] +#define MEM_DISP(base,displacement)[base+(displacement)] +#define MEM_INDEX(base,index,size)[base+index*size] +#define INSN1(mnemonic,size_suffix,dst)mnemonic dst +#define INSN2(mnemonic,size_suffix,dst,src)mnemonic dst,src +#define TEXT .text +#define ALIGN(log) .align log +#define GLOBL .globl +#endif + +#ifdef X86_BROKEN_ALIGN +#undef ALIGN +#define ALIGN(log) .align log,0x90 +#endif diff --git a/mpi/i586/README b/mpi/i586/README new file mode 100644 index 0000000..d73b082 --- /dev/null +++ b/mpi/i586/README @@ -0,0 +1,26 @@ +This directory contains mpn functions optimized for Intel Pentium +processors. + +RELEVANT OPTIMIZATION ISSUES + +1. Pentium doesn't allocate cache lines on writes, unlike most other modern +processors. Since the functions in the mpn class do array writes, we have to +handle allocating the destination cache lines by reading a word from it in the +loops, to achieve the best performance. + +2. Pairing of memory operations requires that the two issued operations refer +to different cache banks. The simplest way to insure this is to read/write +two words from the same object. If we make operations on different objects, +they might or might not be to the same cache bank. + +STATUS + +1. mpn_lshift and mpn_rshift run at about 6 cycles/limb, but the Pentium +documentation indicates that they should take only 43/8 = 5.375 cycles/limb, +or 5 cycles/limb asymptotically. + +2. mpn_add_n and mpn_sub_n run at asymptotically 2 cycles/limb. Due to loop +overhead and other delays (cache refill?), they run at or near 2.5 cycles/limb. + +3. mpn_mul_1, mpn_addmul_1, mpn_submul_1 all run 1 cycle faster than they +should... diff --git a/mpi/i586/distfiles b/mpi/i586/distfiles new file mode 100644 index 0000000..8f821fb --- /dev/null +++ b/mpi/i586/distfiles @@ -0,0 +1,9 @@ +mpih-add1.S +mpih-mul1.S +mpih-mul2.S +mpih-mul3.S +mpih-lshift.S +mpih-rshift.S +mpih-sub1.S +README + diff --git a/mpi/i586/mpih-add1.S b/mpi/i586/mpih-add1.S new file mode 100644 index 0000000..bf9ba27 --- /dev/null +++ b/mpi/i586/mpih-add1.S @@ -0,0 +1,136 @@ +/* i80586 add_n -- Add two limb vectors of the same length > 0 and store + * sum in a third limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_add_n( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_ptr_t s2_ptr, (sp + 12) + * mpi_size_t size) (sp + 16) + */ + +.text + ALIGN (3) + .globl C_SYMBOL_NAME(mpihelp_add_n) +C_SYMBOL_NAME(mpihelp_add_n:) + pushl %edi + pushl %esi + pushl %ebx + pushl %ebp + + movl 20(%esp),%edi /* res_ptr */ + movl 24(%esp),%esi /* s1_ptr */ + movl 28(%esp),%ebp /* s2_ptr */ + movl 32(%esp),%ecx /* size */ + + movl (%ebp),%ebx + + decl %ecx + movl %ecx,%edx + shrl $3,%ecx + andl $7,%edx + testl %ecx,%ecx /* zero carry flag */ + jz Lend + pushl %edx + + ALIGN (3) +Loop: movl 28(%edi),%eax /* fetch destination cache line */ + leal 32(%edi),%edi + +L1: movl (%esi),%eax + movl 4(%esi),%edx + adcl %ebx,%eax + movl 4(%ebp),%ebx + adcl %ebx,%edx + movl 8(%ebp),%ebx + movl %eax,-32(%edi) + movl %edx,-28(%edi) + +L2: movl 8(%esi),%eax + movl 12(%esi),%edx + adcl %ebx,%eax + movl 12(%ebp),%ebx + adcl %ebx,%edx + movl 16(%ebp),%ebx + movl %eax,-24(%edi) + movl %edx,-20(%edi) + +L3: movl 16(%esi),%eax + movl 20(%esi),%edx + adcl %ebx,%eax + movl 20(%ebp),%ebx + adcl %ebx,%edx + movl 24(%ebp),%ebx + movl %eax,-16(%edi) + movl %edx,-12(%edi) + +L4: movl 24(%esi),%eax + movl 28(%esi),%edx + adcl %ebx,%eax + movl 28(%ebp),%ebx + adcl %ebx,%edx + movl 32(%ebp),%ebx + movl %eax,-8(%edi) + movl %edx,-4(%edi) + + leal 32(%esi),%esi + leal 32(%ebp),%ebp + decl %ecx + jnz Loop + + popl %edx +Lend: + decl %edx /* test %edx w/o clobbering carry */ + js Lend2 + incl %edx +Loop2: + leal 4(%edi),%edi + movl (%esi),%eax + adcl %ebx,%eax + movl 4(%ebp),%ebx + movl %eax,-4(%edi) + leal 4(%esi),%esi + leal 4(%ebp),%ebp + decl %edx + jnz Loop2 +Lend2: + movl (%esi),%eax + adcl %ebx,%eax + movl %eax,(%edi) + + sbbl %eax,%eax + negl %eax + + popl %ebp + popl %ebx + popl %esi + popl %edi + ret + + diff --git a/mpi/i586/mpih-lshift.S b/mpi/i586/mpih-lshift.S new file mode 100644 index 0000000..366f929 --- /dev/null +++ b/mpi/i586/mpih-lshift.S @@ -0,0 +1,232 @@ +/* i80586 lshift + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_lshift( mpi_ptr_t wp, (sp + 4) + * mpi_ptr_t up, (sp + 8) + * mpi_size_t usize, (sp + 12) + * unsigned cnt) (sp + 16) + */ + +.text + ALIGN (3) + .globl C_SYMBOL_NAME(mpihelp_lshift) +C_SYMBOL_NAME(mpihelp_lshift:) + + pushl %edi + pushl %esi + pushl %ebx + pushl %ebp + + movl 20(%esp),%edi /* res_ptr */ + movl 24(%esp),%esi /* s_ptr */ + movl 28(%esp),%ebp /* size */ + movl 32(%esp),%ecx /* cnt */ + +/* We can use faster code for shift-by-1 under certain conditions. */ + cmp $1,%ecx + jne Lnormal + leal 4(%esi),%eax + cmpl %edi,%eax + jnc Lspecial /* jump if s_ptr + 1 >= res_ptr */ + leal (%esi,%ebp,4),%eax + cmpl %eax,%edi + jnc Lspecial /* jump if res_ptr >= s_ptr + size */ + +Lnormal: + leal -4(%edi,%ebp,4),%edi + leal -4(%esi,%ebp,4),%esi + + movl (%esi),%edx + subl $4,%esi + xorl %eax,%eax + shldl %cl,%edx,%eax /* compute carry limb */ + pushl %eax /* push carry limb onto stack */ + + decl %ebp + pushl %ebp + shrl $3,%ebp + jz Lend + + movl (%edi),%eax /* fetch destination cache line */ + + ALIGN (2) +Loop: movl -28(%edi),%eax /* fetch destination cache line */ + movl %edx,%ebx + + movl (%esi),%eax + movl -4(%esi),%edx + shldl %cl,%eax,%ebx + shldl %cl,%edx,%eax + movl %ebx,(%edi) + movl %eax,-4(%edi) + + movl -8(%esi),%ebx + movl -12(%esi),%eax + shldl %cl,%ebx,%edx + shldl %cl,%eax,%ebx + movl %edx,-8(%edi) + movl %ebx,-12(%edi) + + movl -16(%esi),%edx + movl -20(%esi),%ebx + shldl %cl,%edx,%eax + shldl %cl,%ebx,%edx + movl %eax,-16(%edi) + movl %edx,-20(%edi) + + movl -24(%esi),%eax + movl -28(%esi),%edx + shldl %cl,%eax,%ebx + shldl %cl,%edx,%eax + movl %ebx,-24(%edi) + movl %eax,-28(%edi) + + subl $32,%esi + subl $32,%edi + decl %ebp + jnz Loop + +Lend: popl %ebp + andl $7,%ebp + jz Lend2 +Loop2: movl (%esi),%eax + shldl %cl,%eax,%edx + movl %edx,(%edi) + movl %eax,%edx + subl $4,%esi + subl $4,%edi + decl %ebp + jnz Loop2 + +Lend2: shll %cl,%edx /* compute least significant limb */ + movl %edx,(%edi) /* store it */ + + popl %eax /* pop carry limb */ + + popl %ebp + popl %ebx + popl %esi + popl %edi + ret + +/* We loop from least significant end of the arrays, which is only + permissable if the source and destination don't overlap, since the + function is documented to work for overlapping source and destination. +*/ + +Lspecial: + movl (%esi),%edx + addl $4,%esi + + decl %ebp + pushl %ebp + shrl $3,%ebp + + addl %edx,%edx + incl %ebp + decl %ebp + jz LLend + + movl (%edi),%eax /* fetch destination cache line */ + + ALIGN (2) +LLoop: movl 28(%edi),%eax /* fetch destination cache line */ + movl %edx,%ebx + + movl (%esi),%eax + movl 4(%esi),%edx + adcl %eax,%eax + movl %ebx,(%edi) + adcl %edx,%edx + movl %eax,4(%edi) + + movl 8(%esi),%ebx + movl 12(%esi),%eax + adcl %ebx,%ebx + movl %edx,8(%edi) + adcl %eax,%eax + movl %ebx,12(%edi) + + movl 16(%esi),%edx + movl 20(%esi),%ebx + adcl %edx,%edx + movl %eax,16(%edi) + adcl %ebx,%ebx + movl %edx,20(%edi) + + movl 24(%esi),%eax + movl 28(%esi),%edx + adcl %eax,%eax + movl %ebx,24(%edi) + adcl %edx,%edx + movl %eax,28(%edi) + + leal 32(%esi),%esi /* use leal not to clobber carry */ + leal 32(%edi),%edi + decl %ebp + jnz LLoop + +LLend: popl %ebp + sbbl %eax,%eax /* save carry in %eax */ + andl $7,%ebp + jz LLend2 + addl %eax,%eax /* restore carry from eax */ +LLoop2: movl %edx,%ebx + movl (%esi),%edx + adcl %edx,%edx + movl %ebx,(%edi) + + leal 4(%esi),%esi /* use leal not to clobber carry */ + leal 4(%edi),%edi + decl %ebp + jnz LLoop2 + + jmp LL1 +LLend2: addl %eax,%eax /* restore carry from eax */ +LL1: movl %edx,(%edi) /* store last limb */ + + sbbl %eax,%eax + negl %eax + + popl %ebp + popl %ebx + popl %esi + popl %edi + ret + + diff --git a/mpi/i586/mpih-mul1.S b/mpi/i586/mpih-mul1.S new file mode 100644 index 0000000..1987825 --- /dev/null +++ b/mpi/i586/mpih-mul1.S @@ -0,0 +1,92 @@ +/* i80586 mul_1 -- Multiply a limb vector with a limb and store + * the result in a second limb vector. + * + * Copyright (C) 1992, 1994, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_mul_1( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_size_t s1_size, (sp + 12) + * mpi_limb_t s2_limb) (sp + 16) + */ + +#define res_ptr edi +#define s1_ptr esi +#define size ecx +#define s2_limb ebp + + TEXT + ALIGN (3) + GLOBL C_SYMBOL_NAME(mpihelp_mul_1) +C_SYMBOL_NAME(mpihelp_mul_1:) + + INSN1(push,l ,R(edi)) + INSN1(push,l ,R(esi)) + INSN1(push,l ,R(ebx)) + INSN1(push,l ,R(ebp)) + + INSN2(mov,l ,R(res_ptr),MEM_DISP(esp,20)) + INSN2(mov,l ,R(s1_ptr),MEM_DISP(esp,24)) + INSN2(mov,l ,R(size),MEM_DISP(esp,28)) + INSN2(mov,l ,R(s2_limb),MEM_DISP(esp,32)) + + INSN2(lea,l ,R(res_ptr),MEM_INDEX(res_ptr,size,4)) + INSN2(lea,l ,R(s1_ptr),MEM_INDEX(s1_ptr,size,4)) + INSN1(neg,l ,R(size)) + INSN2(xor,l ,R(ebx),R(ebx)) + ALIGN (3) + +Loop: INSN2(adc,l ,R(ebx),$0) + INSN2(mov,l ,R(eax),MEM_INDEX(s1_ptr,size,4)) + + INSN1(mul,l ,R(s2_limb)) + + INSN2(add,l ,R(ebx),R(eax)) + + INSN2(mov,l ,MEM_INDEX(res_ptr,size,4),R(ebx)) + INSN1(inc,l ,R(size)) + + INSN2(mov,l ,R(ebx),R(edx)) + INSN1(jnz, ,Loop) + + INSN2(adc,l ,R(ebx),$0) + INSN2(mov,l ,R(eax),R(ebx)) + INSN1(pop,l ,R(ebp)) + INSN1(pop,l ,R(ebx)) + INSN1(pop,l ,R(esi)) + INSN1(pop,l ,R(edi)) + ret + diff --git a/mpi/i586/mpih-mul2.S b/mpi/i586/mpih-mul2.S new file mode 100644 index 0000000..39c0bbd --- /dev/null +++ b/mpi/i586/mpih-mul2.S @@ -0,0 +1,96 @@ +/* i80586 addmul_1 -- Multiply a limb vector with a limb and add + * the result to a second limb vector. + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_addmul_1( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_size_t s1_size, (sp + 12) + * mpi_limb_t s2_limb) (sp + 16) + */ + +#define res_ptr edi +#define s1_ptr esi +#define size ecx +#define s2_limb ebp + + TEXT + ALIGN (3) + GLOBL C_SYMBOL_NAME(mpihelp_addmul_1) +C_SYMBOL_NAME(mpihelp_addmul_1:) + + INSN1(push,l ,R(edi)) + INSN1(push,l ,R(esi)) + INSN1(push,l ,R(ebx)) + INSN1(push,l ,R(ebp)) + + INSN2(mov,l ,R(res_ptr),MEM_DISP(esp,20)) + INSN2(mov,l ,R(s1_ptr),MEM_DISP(esp,24)) + INSN2(mov,l ,R(size),MEM_DISP(esp,28)) + INSN2(mov,l ,R(s2_limb),MEM_DISP(esp,32)) + + INSN2(lea,l ,R(res_ptr),MEM_INDEX(res_ptr,size,4)) + INSN2(lea,l ,R(s1_ptr),MEM_INDEX(s1_ptr,size,4)) + INSN1(neg,l ,R(size)) + INSN2(xor,l ,R(ebx),R(ebx)) + ALIGN (3) + +Loop: INSN2(adc,l ,R(ebx),$0) + INSN2(mov,l ,R(eax),MEM_INDEX(s1_ptr,size,4)) + + INSN1(mul,l ,R(s2_limb)) + + INSN2(add,l ,R(eax),R(ebx)) + INSN2(mov,l ,R(ebx),MEM_INDEX(res_ptr,size,4)) + + INSN2(adc,l ,R(edx),$0) + INSN2(add,l ,R(ebx),R(eax)) + + INSN2(mov,l ,MEM_INDEX(res_ptr,size,4),R(ebx)) + INSN1(inc,l ,R(size)) + + INSN2(mov,l ,R(ebx),R(edx)) + INSN1(jnz, ,Loop) + + INSN2(adc,l ,R(ebx),$0) + INSN2(mov,l ,R(eax),R(ebx)) + INSN1(pop,l ,R(ebp)) + INSN1(pop,l ,R(ebx)) + INSN1(pop,l ,R(esi)) + INSN1(pop,l ,R(edi)) + ret + diff --git a/mpi/i586/mpih-mul3.S b/mpi/i586/mpih-mul3.S new file mode 100644 index 0000000..6c1414f --- /dev/null +++ b/mpi/i586/mpih-mul3.S @@ -0,0 +1,96 @@ +/* i80586 submul_1 -- Multiply a limb vector with a limb and add + * the result to a second limb vector. + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_submul_1( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_size_t s1_size, (sp + 12) + * mpi_limb_t s2_limb) (sp + 16) + */ + +#define res_ptr edi +#define s1_ptr esi +#define size ecx +#define s2_limb ebp + + TEXT + ALIGN (3) + GLOBL C_SYMBOL_NAME(mpihelp_submul_1) +C_SYMBOL_NAME(mpihelp_submul_1:) + + INSN1(push,l ,R(edi)) + INSN1(push,l ,R(esi)) + INSN1(push,l ,R(ebx)) + INSN1(push,l ,R(ebp)) + + INSN2(mov,l ,R(res_ptr),MEM_DISP(esp,20)) + INSN2(mov,l ,R(s1_ptr),MEM_DISP(esp,24)) + INSN2(mov,l ,R(size),MEM_DISP(esp,28)) + INSN2(mov,l ,R(s2_limb),MEM_DISP(esp,32)) + + INSN2(lea,l ,R(res_ptr),MEM_INDEX(res_ptr,size,4)) + INSN2(lea,l ,R(s1_ptr),MEM_INDEX(s1_ptr,size,4)) + INSN1(neg,l ,R(size)) + INSN2(xor,l ,R(ebx),R(ebx)) + ALIGN (3) + +Loop: INSN2(adc,l ,R(ebx),$0) + INSN2(mov,l ,R(eax),MEM_INDEX(s1_ptr,size,4)) + + INSN1(mul,l ,R(s2_limb)) + + INSN2(add,l ,R(eax),R(ebx)) + INSN2(mov,l ,R(ebx),MEM_INDEX(res_ptr,size,4)) + + INSN2(adc,l ,R(edx),$0) + INSN2(sub,l ,R(ebx),R(eax)) + + INSN2(mov,l ,MEM_INDEX(res_ptr,size,4),R(ebx)) + INSN1(inc,l ,R(size)) + + INSN2(mov,l ,R(ebx),R(edx)) + INSN1(jnz, ,Loop) + + INSN2(adc,l ,R(ebx),$0) + INSN2(mov,l ,R(eax),R(ebx)) + INSN1(pop,l ,R(ebp)) + INSN1(pop,l ,R(ebx)) + INSN1(pop,l ,R(esi)) + INSN1(pop,l ,R(edi)) + ret + diff --git a/mpi/i586/mpih-rshift.S b/mpi/i586/mpih-rshift.S new file mode 100644 index 0000000..aeacb68 --- /dev/null +++ b/mpi/i586/mpih-rshift.S @@ -0,0 +1,231 @@ +/* i80586 rshift + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + + +/******************* + * mpi_limb_t + * mpihelp_rshift( mpi_ptr_t wp, (sp + 4) + * mpi_ptr_t up, (sp + 8) + * mpi_size_t usize, (sp + 12) + * unsigned cnt) (sp + 16) + */ + +.text + ALIGN (3) + .globl C_SYMBOL_NAME(mpihelp_rshift) +C_SYMBOL_NAME(mpihelp_rshift:) + pushl %edi + pushl %esi + pushl %ebx + pushl %ebp + + movl 20(%esp),%edi /* res_ptr */ + movl 24(%esp),%esi /* s_ptr */ + movl 28(%esp),%ebp /* size */ + movl 32(%esp),%ecx /* cnt */ + +/* We can use faster code for shift-by-1 under certain conditions. */ + cmp $1,%ecx + jne Rnormal + leal 4(%edi),%eax + cmpl %esi,%eax + jnc Rspecial /* jump if res_ptr + 1 >= s_ptr */ + leal (%edi,%ebp,4),%eax + cmpl %eax,%esi + jnc Rspecial /* jump if s_ptr >= res_ptr + size */ + +Rnormal: + movl (%esi),%edx + addl $4,%esi + xorl %eax,%eax + shrdl %cl,%edx,%eax /* compute carry limb */ + pushl %eax /* push carry limb onto stack */ + + decl %ebp + pushl %ebp + shrl $3,%ebp + jz Rend + + movl (%edi),%eax /* fetch destination cache line */ + + ALIGN (2) +Roop: movl 28(%edi),%eax /* fetch destination cache line */ + movl %edx,%ebx + + movl (%esi),%eax + movl 4(%esi),%edx + shrdl %cl,%eax,%ebx + shrdl %cl,%edx,%eax + movl %ebx,(%edi) + movl %eax,4(%edi) + + movl 8(%esi),%ebx + movl 12(%esi),%eax + shrdl %cl,%ebx,%edx + shrdl %cl,%eax,%ebx + movl %edx,8(%edi) + movl %ebx,12(%edi) + + movl 16(%esi),%edx + movl 20(%esi),%ebx + shrdl %cl,%edx,%eax + shrdl %cl,%ebx,%edx + movl %eax,16(%edi) + movl %edx,20(%edi) + + movl 24(%esi),%eax + movl 28(%esi),%edx + shrdl %cl,%eax,%ebx + shrdl %cl,%edx,%eax + movl %ebx,24(%edi) + movl %eax,28(%edi) + + addl $32,%esi + addl $32,%edi + decl %ebp + jnz Roop + +Rend: popl %ebp + andl $7,%ebp + jz Rend2 +Roop2: movl (%esi),%eax + shrdl %cl,%eax,%edx /* compute result limb */ + movl %edx,(%edi) + movl %eax,%edx + addl $4,%esi + addl $4,%edi + decl %ebp + jnz Roop2 + +Rend2: shrl %cl,%edx /* compute most significant limb */ + movl %edx,(%edi) /* store it */ + + popl %eax /* pop carry limb */ + + popl %ebp + popl %ebx + popl %esi + popl %edi + ret + +/* We loop from least significant end of the arrays, which is only + permissable if the source and destination don't overlap, since the + function is documented to work for overlapping source and destination. +*/ + +Rspecial: + leal -4(%edi,%ebp,4),%edi + leal -4(%esi,%ebp,4),%esi + + movl (%esi),%edx + subl $4,%esi + + decl %ebp + pushl %ebp + shrl $3,%ebp + + shrl $1,%edx + incl %ebp + decl %ebp + jz RLend + + movl (%edi),%eax /* fetch destination cache line */ + + ALIGN (2) +RLoop: movl -28(%edi),%eax /* fetch destination cache line */ + movl %edx,%ebx + + movl (%esi),%eax + movl -4(%esi),%edx + rcrl $1,%eax + movl %ebx,(%edi) + rcrl $1,%edx + movl %eax,-4(%edi) + + movl -8(%esi),%ebx + movl -12(%esi),%eax + rcrl $1,%ebx + movl %edx,-8(%edi) + rcrl $1,%eax + movl %ebx,-12(%edi) + + movl -16(%esi),%edx + movl -20(%esi),%ebx + rcrl $1,%edx + movl %eax,-16(%edi) + rcrl $1,%ebx + movl %edx,-20(%edi) + + movl -24(%esi),%eax + movl -28(%esi),%edx + rcrl $1,%eax + movl %ebx,-24(%edi) + rcrl $1,%edx + movl %eax,-28(%edi) + + leal -32(%esi),%esi /* use leal not to clobber carry */ + leal -32(%edi),%edi + decl %ebp + jnz RLoop + +RLend: popl %ebp + sbbl %eax,%eax /* save carry in %eax */ + andl $7,%ebp + jz RLend2 + addl %eax,%eax /* restore carry from eax */ +RLoop2: movl %edx,%ebx + movl (%esi),%edx + rcrl $1,%edx + movl %ebx,(%edi) + + leal -4(%esi),%esi /* use leal not to clobber carry */ + leal -4(%edi),%edi + decl %ebp + jnz RLoop2 + + jmp RL1 +RLend2: addl %eax,%eax /* restore carry from eax */ +RL1: movl %edx,(%edi) /* store last limb */ + + movl $0,%eax + rcrl $1,%eax + + popl %ebp + popl %ebx + popl %esi + popl %edi + ret + diff --git a/mpi/i586/mpih-sub1.S b/mpi/i586/mpih-sub1.S new file mode 100644 index 0000000..7d3045d --- /dev/null +++ b/mpi/i586/mpih-sub1.S @@ -0,0 +1,145 @@ +/* i80586 sub_n -- Sub two limb vectors of the same length > 0 and store + * sum in a third limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_sub_n( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_ptr_t s2_ptr, (sp + 12) + * mpi_size_t size) (sp + 16) + */ + + +.text + ALIGN (3) + .globl C_SYMBOL_NAME(mpihelp_sub_n) +C_SYMBOL_NAME(mpihelp_sub_n:) + + pushl %edi + pushl %esi + pushl %ebx + pushl %ebp + + movl 20(%esp),%edi /* res_ptr */ + movl 24(%esp),%esi /* s1_ptr */ + movl 28(%esp),%ebp /* s2_ptr */ + movl 32(%esp),%ecx /* size */ + + movl (%ebp),%ebx + + decl %ecx + movl %ecx,%edx + shrl $3,%ecx + andl $7,%edx + testl %ecx,%ecx /* zero carry flag */ + jz Lend + pushl %edx + + ALIGN (3) +Loop: movl 28(%edi),%eax /* fetch destination cache line */ + leal 32(%edi),%edi + +L1: movl (%esi),%eax + movl 4(%esi),%edx + sbbl %ebx,%eax + movl 4(%ebp),%ebx + sbbl %ebx,%edx + movl 8(%ebp),%ebx + movl %eax,-32(%edi) + movl %edx,-28(%edi) + +L2: movl 8(%esi),%eax + movl 12(%esi),%edx + sbbl %ebx,%eax + movl 12(%ebp),%ebx + sbbl %ebx,%edx + movl 16(%ebp),%ebx + movl %eax,-24(%edi) + movl %edx,-20(%edi) + +L3: movl 16(%esi),%eax + movl 20(%esi),%edx + sbbl %ebx,%eax + movl 20(%ebp),%ebx + sbbl %ebx,%edx + movl 24(%ebp),%ebx + movl %eax,-16(%edi) + movl %edx,-12(%edi) + +L4: movl 24(%esi),%eax + movl 28(%esi),%edx + sbbl %ebx,%eax + movl 28(%ebp),%ebx + sbbl %ebx,%edx + movl 32(%ebp),%ebx + movl %eax,-8(%edi) + movl %edx,-4(%edi) + + leal 32(%esi),%esi + leal 32(%ebp),%ebp + decl %ecx + jnz Loop + + popl %edx +Lend: + decl %edx /* test %edx w/o clobbering carry */ + js Lend2 + incl %edx +Loop2: + leal 4(%edi),%edi + movl (%esi),%eax + sbbl %ebx,%eax + movl 4(%ebp),%ebx + movl %eax,-4(%edi) + leal 4(%esi),%esi + leal 4(%ebp),%ebp + decl %edx + jnz Loop2 +Lend2: + movl (%esi),%eax + sbbl %ebx,%eax + movl %eax,(%edi) + + sbbl %eax,%eax + negl %eax + + popl %ebp + popl %ebx + popl %esi + popl %edi + ret + diff --git a/mpi/longlong.h b/mpi/longlong.h new file mode 100644 index 0000000..c0f845c --- /dev/null +++ b/mpi/longlong.h @@ -0,0 +1,1557 @@ +/* longlong.h -- definitions for mixed size 32/64 bit arithmetic. + Note: I added some stuff for use with gnupg + +Copyright (C) 1991, 1992, 1993, 1994, 1996, 1998, + 2000, 2001, 2002, 2003 Free Software Foundation, Inc. + +This file 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. + +This file 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 this file; see the file COPYING.LIB. If not, write to +the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, +MA 02110-1301, USA. */ + +/* You have to define the following before including this file: + + UWtype -- An unsigned type, default type for operations (typically a "word") + UHWtype -- An unsigned type, at least half the size of UWtype. + UDWtype -- An unsigned type, at least twice as large a UWtype + W_TYPE_SIZE -- size in bits of UWtype + + SItype, USItype -- Signed and unsigned 32 bit types. + DItype, UDItype -- Signed and unsigned 64 bit types. + + On a 32 bit machine UWtype should typically be USItype; + on a 64 bit machine, UWtype should typically be UDItype. +*/ + +#define __BITS4 (W_TYPE_SIZE / 4) +#define __ll_B ((UWtype) 1 << (W_TYPE_SIZE / 2)) +#define __ll_lowpart(t) ((UWtype) (t) & (__ll_B - 1)) +#define __ll_highpart(t) ((UWtype) (t) >> (W_TYPE_SIZE / 2)) + +/* This is used to make sure no undesirable sharing between different libraries + that use this file takes place. */ +#ifndef __MPN +#define __MPN(x) __##x +#endif + +/* Define auxiliary asm macros. + + 1) umul_ppmm(high_prod, low_prod, multipler, multiplicand) multiplies two + UWtype integers MULTIPLER and MULTIPLICAND, and generates a two UWtype + word product in HIGH_PROD and LOW_PROD. + + 2) __umulsidi3(a,b) multiplies two UWtype integers A and B, and returns a + UDWtype product. This is just a variant of umul_ppmm. + + 3) udiv_qrnnd(quotient, remainder, high_numerator, low_numerator, + denominator) divides a UDWtype, composed by the UWtype integers + HIGH_NUMERATOR and LOW_NUMERATOR, by DENOMINATOR and places the quotient + in QUOTIENT and the remainder in REMAINDER. HIGH_NUMERATOR must be less + than DENOMINATOR for correct operation. If, in addition, the most + significant bit of DENOMINATOR must be 1, then the pre-processor symbol + UDIV_NEEDS_NORMALIZATION is defined to 1. + + 4) sdiv_qrnnd(quotient, remainder, high_numerator, low_numerator, + denominator). Like udiv_qrnnd but the numbers are signed. The quotient + is rounded towards 0. + + 5) count_leading_zeros(count, x) counts the number of zero-bits from the + msb to the first non-zero bit in the UWtype X. This is the number of + steps X needs to be shifted left to set the msb. Undefined for X == 0, + unless the symbol COUNT_LEADING_ZEROS_0 is defined to some value. + + 6) count_trailing_zeros(count, x) like count_leading_zeros, but counts + from the least significant end. + + 7) add_ssaaaa(high_sum, low_sum, high_addend_1, low_addend_1, + high_addend_2, low_addend_2) adds two UWtype integers, composed by + HIGH_ADDEND_1 and LOW_ADDEND_1, and HIGH_ADDEND_2 and LOW_ADDEND_2 + respectively. The result is placed in HIGH_SUM and LOW_SUM. Overflow + (i.e. carry out) is not stored anywhere, and is lost. + + 8) sub_ddmmss(high_difference, low_difference, high_minuend, low_minuend, + high_subtrahend, low_subtrahend) subtracts two two-word UWtype integers, + composed by HIGH_MINUEND_1 and LOW_MINUEND_1, and HIGH_SUBTRAHEND_2 and + LOW_SUBTRAHEND_2 respectively. The result is placed in HIGH_DIFFERENCE + and LOW_DIFFERENCE. Overflow (i.e. carry out) is not stored anywhere, + and is lost. + + If any of these macros are left undefined for a particular CPU, + C macros are used. */ + +/* The CPUs come in alphabetical order below. + + Please add support for more CPUs here, or improve the current support + for the CPUs below! */ + +#if defined (__GNUC__) && !defined (NO_ASM) + +/* We sometimes need to clobber "cc" with gcc2, but that would not be + understood by gcc1. Use cpp to avoid major code duplication. */ +#if __GNUC__ < 2 +#define __CLOBBER_CC +#define __AND_CLOBBER_CC +#else /* __GNUC__ >= 2 */ +#define __CLOBBER_CC : "cc" +#define __AND_CLOBBER_CC , "cc" +#endif /* __GNUC__ < 2 */ + + +/*************************************** + ************** A29K ***************** + ***************************************/ +#if (defined (__a29k__) || defined (_AM29K)) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("add %1,%4,%5\n" \ + "addc %0,%2,%3" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%r" ((USItype)(ah)), \ + "rI" ((USItype)(bh)), \ + "%r" ((USItype)(al)), \ + "rI" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("sub %1,%4,%5\n" \ + "subc %0,%2,%3" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "r" ((USItype)(ah)), \ + "rI" ((USItype)(bh)), \ + "r" ((USItype)(al)), \ + "rI" ((USItype)(bl))) +#define umul_ppmm(xh, xl, m0, m1) \ + do { \ + USItype __m0 = (m0), __m1 = (m1); \ + __asm__ ("multiplu %0,%1,%2" \ + : "=r" ((USItype)(xl)) \ + : "r" (__m0), \ + "r" (__m1)); \ + __asm__ ("multmu %0,%1,%2" \ + : "=r" ((USItype)(xh)) \ + : "r" (__m0), \ + "r" (__m1)); \ + } while (0) +#define udiv_qrnnd(q, r, n1, n0, d) \ + __asm__ ("dividu %0,%3,%4" \ + : "=r" ((USItype)(q)), \ + "=q" ((USItype)(r)) \ + : "1" ((USItype)(n1)), \ + "r" ((USItype)(n0)), \ + "r" ((USItype)(d))) +#define count_leading_zeros(count, x) \ + __asm__ ("clz %0,%1" \ + : "=r" ((USItype)(count)) \ + : "r" ((USItype)(x))) +#define COUNT_LEADING_ZEROS_0 32 +#endif /* __a29k__ */ + + +#if defined (__alpha) && W_TYPE_SIZE == 64 +#define umul_ppmm(ph, pl, m0, m1) \ + do { \ + UDItype __m0 = (m0), __m1 = (m1); \ + __asm__ ("umulh %r1,%2,%0" \ + : "=r" ((UDItype) ph) \ + : "%rJ" (__m0), \ + "rI" (__m1)); \ + (pl) = __m0 * __m1; \ + } while (0) +#define UMUL_TIME 46 +#ifndef LONGLONG_STANDALONE +#define udiv_qrnnd(q, r, n1, n0, d) \ + do { UDItype __r; \ + (q) = __udiv_qrnnd (&__r, (n1), (n0), (d)); \ + (r) = __r; \ + } while (0) +extern UDItype __udiv_qrnnd (); +#define UDIV_TIME 220 +#endif /* LONGLONG_STANDALONE */ +#endif /* __alpha */ + +/*************************************** + ************** ARM ****************** + ***************************************/ +#if defined (__arm__) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("adds %1, %4, %5\n" \ + "adc %0, %2, %3" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%r" ((USItype)(ah)), \ + "rI" ((USItype)(bh)), \ + "%r" ((USItype)(al)), \ + "rI" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("subs %1, %4, %5\n" \ + "sbc %0, %2, %3" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "r" ((USItype)(ah)), \ + "rI" ((USItype)(bh)), \ + "r" ((USItype)(al)), \ + "rI" ((USItype)(bl))) +#if defined __ARM_ARCH_2__ || defined __ARM_ARCH_3__ +#define umul_ppmm(xh, xl, a, b) \ + __asm__ ("%@ Inlined umul_ppmm\n" \ + "mov %|r0, %2, lsr #16 @ AAAA\n" \ + "mov %|r2, %3, lsr #16 @ BBBB\n" \ + "bic %|r1, %2, %|r0, lsl #16 @ aaaa\n" \ + "bic %0, %3, %|r2, lsl #16 @ bbbb\n" \ + "mul %1, %|r1, %|r2 @ aaaa * BBBB\n" \ + "mul %|r2, %|r0, %|r2 @ AAAA * BBBB\n" \ + "mul %|r1, %0, %|r1 @ aaaa * bbbb\n" \ + "mul %0, %|r0, %0 @ AAAA * bbbb\n" \ + "adds %|r0, %1, %0 @ central sum\n" \ + "addcs %|r2, %|r2, #65536\n" \ + "adds %1, %|r1, %|r0, lsl #16\n" \ + "adc %0, %|r2, %|r0, lsr #16" \ + : "=&r" ((USItype)(xh)), \ + "=r" ((USItype)(xl)) \ + : "r" ((USItype)(a)), \ + "r" ((USItype)(b)) \ + : "r0", "r1", "r2") +#else +#define umul_ppmm(xh, xl, a, b) \ + __asm__ ("%@ Inlined umul_ppmm\n" \ + "umull %r1, %r0, %r2, %r3" \ + : "=&r" ((USItype)(xh)), \ + "=r" ((USItype)(xl)) \ + : "r" ((USItype)(a)), \ + "r" ((USItype)(b)) \ + : "r0", "r1") +#endif +#define UMUL_TIME 20 +#define UDIV_TIME 100 +#endif /* __arm__ */ + +/*************************************** + ************** CLIPPER ************** + ***************************************/ +#if defined (__clipper__) && W_TYPE_SIZE == 32 +#define umul_ppmm(w1, w0, u, v) \ + ({union {UDItype __ll; \ + struct {USItype __l, __h;} __i; \ + } __xx; \ + __asm__ ("mulwux %2,%0" \ + : "=r" (__xx.__ll) \ + : "%0" ((USItype)(u)), \ + "r" ((USItype)(v))); \ + (w1) = __xx.__i.__h; (w0) = __xx.__i.__l;}) +#define smul_ppmm(w1, w0, u, v) \ + ({union {DItype __ll; \ + struct {SItype __l, __h;} __i; \ + } __xx; \ + __asm__ ("mulwx %2,%0" \ + : "=r" (__xx.__ll) \ + : "%0" ((SItype)(u)), \ + "r" ((SItype)(v))); \ + (w1) = __xx.__i.__h; (w0) = __xx.__i.__l;}) +#define __umulsidi3(u, v) \ + ({UDItype __w; \ + __asm__ ("mulwux %2,%0" \ + : "=r" (__w) \ + : "%0" ((USItype)(u)), \ + "r" ((USItype)(v))); \ + __w; }) +#endif /* __clipper__ */ + + +/*************************************** + ************** GMICRO *************** + ***************************************/ +#if defined (__gmicro__) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("add.w %5,%1\n" \ + "addx %3,%0" \ + : "=g" ((USItype)(sh)), \ + "=&g" ((USItype)(sl)) \ + : "%0" ((USItype)(ah)), \ + "g" ((USItype)(bh)), \ + "%1" ((USItype)(al)), \ + "g" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("sub.w %5,%1\n" \ + "subx %3,%0" \ + : "=g" ((USItype)(sh)), \ + "=&g" ((USItype)(sl)) \ + : "0" ((USItype)(ah)), \ + "g" ((USItype)(bh)), \ + "1" ((USItype)(al)), \ + "g" ((USItype)(bl))) +#define umul_ppmm(ph, pl, m0, m1) \ + __asm__ ("mulx %3,%0,%1" \ + : "=g" ((USItype)(ph)), \ + "=r" ((USItype)(pl)) \ + : "%0" ((USItype)(m0)), \ + "g" ((USItype)(m1))) +#define udiv_qrnnd(q, r, nh, nl, d) \ + __asm__ ("divx %4,%0,%1" \ + : "=g" ((USItype)(q)), \ + "=r" ((USItype)(r)) \ + : "1" ((USItype)(nh)), \ + "0" ((USItype)(nl)), \ + "g" ((USItype)(d))) +#define count_leading_zeros(count, x) \ + __asm__ ("bsch/1 %1,%0" \ + : "=g" (count) \ + : "g" ((USItype)(x)), \ + "0" ((USItype)0)) +#endif + + +/*************************************** + ************** HPPA ***************** + ***************************************/ +#if defined (__hppa) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ (" add %4,%5,%1\n" \ + " addc %2,%3,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%rM" ((USItype)(ah)), \ + "rM" ((USItype)(bh)), \ + "%rM" ((USItype)(al)), \ + "rM" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ (" sub %4,%5,%1\n" \ + " subb %2,%3,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "rM" ((USItype)(ah)), \ + "rM" ((USItype)(bh)), \ + "rM" ((USItype)(al)), \ + "rM" ((USItype)(bl))) +#if defined (_PA_RISC1_1) +#define umul_ppmm(wh, wl, u, v) \ + do { \ + union {UDItype __ll; \ + struct {USItype __h, __l;} __i; \ + } __xx; \ + __asm__ (" xmpyu %1,%2,%0" \ + : "=*f" (__xx.__ll) \ + : "*f" ((USItype)(u)), \ + "*f" ((USItype)(v))); \ + (wh) = __xx.__i.__h; \ + (wl) = __xx.__i.__l; \ + } while (0) +#define UMUL_TIME 8 +#define UDIV_TIME 60 +#else +#define UMUL_TIME 40 +#define UDIV_TIME 80 +#endif +#ifndef LONGLONG_STANDALONE +#define udiv_qrnnd(q, r, n1, n0, d) \ + do { USItype __r; \ + (q) = __udiv_qrnnd (&__r, (n1), (n0), (d)); \ + (r) = __r; \ + } while (0) +extern USItype __udiv_qrnnd (); +#endif /* LONGLONG_STANDALONE */ +#define count_leading_zeros(count, x) \ + do { \ + USItype __tmp; \ + __asm__ ( \ + " ldi 1,%0 \n" \ + " extru,= %1,15,16,%%r0 ; Bits 31..16 zero? \n" \ + " extru,tr %1,15,16,%1 ; No. Shift down, skip add.\n" \ + " ldo 16(%0),%0 ; Yes. Perform add. \n" \ + " extru,= %1,23,8,%%r0 ; Bits 15..8 zero? \n" \ + " extru,tr %1,23,8,%1 ; No. Shift down, skip add.\n" \ + " ldo 8(%0),%0 ; Yes. Perform add. \n" \ + " extru,= %1,27,4,%%r0 ; Bits 7..4 zero? \n" \ + " extru,tr %1,27,4,%1 ; No. Shift down, skip add.\n" \ + " ldo 4(%0),%0 ; Yes. Perform add. \n" \ + " extru,= %1,29,2,%%r0 ; Bits 3..2 zero? \n" \ + " extru,tr %1,29,2,%1 ; No. Shift down, skip add.\n" \ + " ldo 2(%0),%0 ; Yes. Perform add. \n" \ + " extru %1,30,1,%1 ; Extract bit 1. \n" \ + " sub %0,%1,%0 ; Subtract it. " \ + : "=r" (count), "=r" (__tmp) : "1" (x)); \ + } while (0) +#endif /* hppa */ + + +/*************************************** + ************** I370 ***************** + ***************************************/ +#if (defined (__i370__) || defined (__mvs__)) && W_TYPE_SIZE == 32 +#define umul_ppmm(xh, xl, m0, m1) \ + do { \ + union {UDItype __ll; \ + struct {USItype __h, __l;} __i; \ + } __xx; \ + USItype __m0 = (m0), __m1 = (m1); \ + __asm__ ("mr %0,%3" \ + : "=r" (__xx.__i.__h), \ + "=r" (__xx.__i.__l) \ + : "%1" (__m0), \ + "r" (__m1)); \ + (xh) = __xx.__i.__h; (xl) = __xx.__i.__l; \ + (xh) += ((((SItype) __m0 >> 31) & __m1) \ + + (((SItype) __m1 >> 31) & __m0)); \ + } while (0) +#define smul_ppmm(xh, xl, m0, m1) \ + do { \ + union {DItype __ll; \ + struct {USItype __h, __l;} __i; \ + } __xx; \ + __asm__ ("mr %0,%3" \ + : "=r" (__xx.__i.__h), \ + "=r" (__xx.__i.__l) \ + : "%1" (m0), \ + "r" (m1)); \ + (xh) = __xx.__i.__h; (xl) = __xx.__i.__l; \ + } while (0) +#define sdiv_qrnnd(q, r, n1, n0, d) \ + do { \ + union {DItype __ll; \ + struct {USItype __h, __l;} __i; \ + } __xx; \ + __xx.__i.__h = n1; __xx.__i.__l = n0; \ + __asm__ ("dr %0,%2" \ + : "=r" (__xx.__ll) \ + : "0" (__xx.__ll), "r" (d)); \ + (q) = __xx.__i.__l; (r) = __xx.__i.__h; \ + } while (0) +#endif + + +/*************************************** + ************** I386 ***************** + ***************************************/ +#if (defined (__i386__) || defined (__i486__)) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("addl %5,%1\n" \ + "adcl %3,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%0" ((USItype)(ah)), \ + "g" ((USItype)(bh)), \ + "%1" ((USItype)(al)), \ + "g" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("subl %5,%1\n" \ + "sbbl %3,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "0" ((USItype)(ah)), \ + "g" ((USItype)(bh)), \ + "1" ((USItype)(al)), \ + "g" ((USItype)(bl))) +#define umul_ppmm(w1, w0, u, v) \ + __asm__ ("mull %3" \ + : "=a" ((USItype)(w0)), \ + "=d" ((USItype)(w1)) \ + : "%0" ((USItype)(u)), \ + "rm" ((USItype)(v))) +#define udiv_qrnnd(q, r, n1, n0, d) \ + __asm__ ("divl %4" \ + : "=a" ((USItype)(q)), \ + "=d" ((USItype)(r)) \ + : "0" ((USItype)(n0)), \ + "1" ((USItype)(n1)), \ + "rm" ((USItype)(d))) +#define count_leading_zeros(count, x) \ + do { \ + USItype __cbtmp; \ + __asm__ ("bsrl %1,%0" \ + : "=r" (__cbtmp) : "rm" ((USItype)(x))); \ + (count) = __cbtmp ^ 31; \ + } while (0) +#define count_trailing_zeros(count, x) \ + __asm__ ("bsfl %1,%0" : "=r" (count) : "rm" ((USItype)(x))) +#ifndef UMUL_TIME +#define UMUL_TIME 40 +#endif +#ifndef UDIV_TIME +#define UDIV_TIME 40 +#endif +#endif /* 80x86 */ + + +/*************************************** + ************** I860 ***************** + ***************************************/ +#if defined (__i860__) && W_TYPE_SIZE == 32 +#define rshift_rhlc(r,h,l,c) \ + __asm__ ("shr %3,r0,r0\n" \ + "shrd %1,%2,%0" \ + "=r" (r) : "r" (h), "r" (l), "rn" (c)) +#endif /* i860 */ + +/*************************************** + ************** I960 ***************** + ***************************************/ +#if defined (__i960__) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("cmpo 1,0\n" \ + "addc %5,%4,%1\n" \ + "addc %3,%2,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%dI" ((USItype)(ah)), \ + "dI" ((USItype)(bh)), \ + "%dI" ((USItype)(al)), \ + "dI" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("cmpo 0,0\n" \ + "subc %5,%4,%1\n" \ + "subc %3,%2,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "dI" ((USItype)(ah)), \ + "dI" ((USItype)(bh)), \ + "dI" ((USItype)(al)), \ + "dI" ((USItype)(bl))) +#define umul_ppmm(w1, w0, u, v) \ + ({union {UDItype __ll; \ + struct {USItype __l, __h;} __i; \ + } __xx; \ + __asm__ ("emul %2,%1,%0" \ + : "=d" (__xx.__ll) \ + : "%dI" ((USItype)(u)), \ + "dI" ((USItype)(v))); \ + (w1) = __xx.__i.__h; (w0) = __xx.__i.__l;}) +#define __umulsidi3(u, v) \ + ({UDItype __w; \ + __asm__ ("emul %2,%1,%0" \ + : "=d" (__w) \ + : "%dI" ((USItype)(u)), \ + "dI" ((USItype)(v))); \ + __w; }) +#define udiv_qrnnd(q, r, nh, nl, d) \ + do { \ + union {UDItype __ll; \ + struct {USItype __l, __h;} __i; \ + } __nn; \ + __nn.__i.__h = (nh); __nn.__i.__l = (nl); \ + __asm__ ("ediv %d,%n,%0" \ + : "=d" (__rq.__ll) \ + : "dI" (__nn.__ll), \ + "dI" ((USItype)(d))); \ + (r) = __rq.__i.__l; (q) = __rq.__i.__h; \ + } while (0) +#define count_leading_zeros(count, x) \ + do { \ + USItype __cbtmp; \ + __asm__ ("scanbit %1,%0" \ + : "=r" (__cbtmp) \ + : "r" ((USItype)(x))); \ + (count) = __cbtmp ^ 31; \ + } while (0) +#define COUNT_LEADING_ZEROS_0 (-32) /* sic */ +#if defined (__i960mx) /* what is the proper symbol to test??? */ +#define rshift_rhlc(r,h,l,c) \ + do { \ + union {UDItype __ll; \ + struct {USItype __l, __h;} __i; \ + } __nn; \ + __nn.__i.__h = (h); __nn.__i.__l = (l); \ + __asm__ ("shre %2,%1,%0" \ + : "=d" (r) : "dI" (__nn.__ll), "dI" (c)); \ + } +#endif /* i960mx */ +#endif /* i960 */ + + +/*************************************** + ************** 68000 **************** + ***************************************/ +#if (defined (__mc68000__) || defined (__mc68020__) || defined (__NeXT__) || defined(mc68020)) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("add%.l %5,%1\n" \ + "addx%.l %3,%0" \ + : "=d" ((USItype)(sh)), \ + "=&d" ((USItype)(sl)) \ + : "%0" ((USItype)(ah)), \ + "d" ((USItype)(bh)), \ + "%1" ((USItype)(al)), \ + "g" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("sub%.l %5,%1\n" \ + "subx%.l %3,%0" \ + : "=d" ((USItype)(sh)), \ + "=&d" ((USItype)(sl)) \ + : "0" ((USItype)(ah)), \ + "d" ((USItype)(bh)), \ + "1" ((USItype)(al)), \ + "g" ((USItype)(bl))) +#if (defined (__mc68020__) || defined (__NeXT__) || defined(mc68020)) +#define umul_ppmm(w1, w0, u, v) \ + __asm__ ("mulu%.l %3,%1:%0" \ + : "=d" ((USItype)(w0)), \ + "=d" ((USItype)(w1)) \ + : "%0" ((USItype)(u)), \ + "dmi" ((USItype)(v))) +#define UMUL_TIME 45 +#define udiv_qrnnd(q, r, n1, n0, d) \ + __asm__ ("divu%.l %4,%1:%0" \ + : "=d" ((USItype)(q)), \ + "=d" ((USItype)(r)) \ + : "0" ((USItype)(n0)), \ + "1" ((USItype)(n1)), \ + "dmi" ((USItype)(d))) +#define UDIV_TIME 90 +#define sdiv_qrnnd(q, r, n1, n0, d) \ + __asm__ ("divs%.l %4,%1:%0" \ + : "=d" ((USItype)(q)), \ + "=d" ((USItype)(r)) \ + : "0" ((USItype)(n0)), \ + "1" ((USItype)(n1)), \ + "dmi" ((USItype)(d))) +#define count_leading_zeros(count, x) \ + __asm__ ("bfffo %1{%b2:%b2},%0" \ + : "=d" ((USItype)(count)) \ + : "od" ((USItype)(x)), "n" (0)) +#define COUNT_LEADING_ZEROS_0 32 +#else /* not mc68020 */ +#define umul_ppmm(xh, xl, a, b) \ + do { USItype __umul_tmp1, __umul_tmp2; \ + __asm__ ("| Inlined umul_ppmm \n" \ + " move%.l %5,%3 \n" \ + " move%.l %2,%0 \n" \ + " move%.w %3,%1 \n" \ + " swap %3 \n" \ + " swap %0 \n" \ + " mulu %2,%1 \n" \ + " mulu %3,%0 \n" \ + " mulu %2,%3 \n" \ + " swap %2 \n" \ + " mulu %5,%2 \n" \ + " add%.l %3,%2 \n" \ + " jcc 1f \n" \ + " add%.l %#0x10000,%0 \n" \ + "1: move%.l %2,%3 \n" \ + " clr%.w %2 \n" \ + " swap %2 \n" \ + " swap %3 \n" \ + " clr%.w %3 \n" \ + " add%.l %3,%1 \n" \ + " addx%.l %2,%0 \n" \ + " | End inlined umul_ppmm" \ + : "=&d" ((USItype)(xh)), "=&d" ((USItype)(xl)), \ + "=d" (__umul_tmp1), "=&d" (__umul_tmp2) \ + : "%2" ((USItype)(a)), "d" ((USItype)(b))); \ + } while (0) +#define UMUL_TIME 100 +#define UDIV_TIME 400 +#endif /* not mc68020 */ +#endif /* mc68000 */ + + +/*************************************** + ************** 88000 **************** + ***************************************/ +#if defined (__m88000__) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("addu.co %1,%r4,%r5\n" \ + "addu.ci %0,%r2,%r3" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%rJ" ((USItype)(ah)), \ + "rJ" ((USItype)(bh)), \ + "%rJ" ((USItype)(al)), \ + "rJ" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("subu.co %1,%r4,%r5\n" \ + "subu.ci %0,%r2,%r3" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "rJ" ((USItype)(ah)), \ + "rJ" ((USItype)(bh)), \ + "rJ" ((USItype)(al)), \ + "rJ" ((USItype)(bl))) +#define count_leading_zeros(count, x) \ + do { \ + USItype __cbtmp; \ + __asm__ ("ff1 %0,%1" \ + : "=r" (__cbtmp) \ + : "r" ((USItype)(x))); \ + (count) = __cbtmp ^ 31; \ + } while (0) +#define COUNT_LEADING_ZEROS_0 63 /* sic */ +#if defined (__m88110__) +#define umul_ppmm(wh, wl, u, v) \ + do { \ + union {UDItype __ll; \ + struct {USItype __h, __l;} __i; \ + } __x; \ + __asm__ ("mulu.d %0,%1,%2" : "=r" (__x.__ll) : "r" (u), "r" (v)); \ + (wh) = __x.__i.__h; \ + (wl) = __x.__i.__l; \ + } while (0) +#define udiv_qrnnd(q, r, n1, n0, d) \ + ({union {UDItype __ll; \ + struct {USItype __h, __l;} __i; \ + } __x, __q; \ + __x.__i.__h = (n1); __x.__i.__l = (n0); \ + __asm__ ("divu.d %0,%1,%2" \ + : "=r" (__q.__ll) : "r" (__x.__ll), "r" (d)); \ + (r) = (n0) - __q.__l * (d); (q) = __q.__l; }) +#define UMUL_TIME 5 +#define UDIV_TIME 25 +#else +#define UMUL_TIME 17 +#define UDIV_TIME 150 +#endif /* __m88110__ */ +#endif /* __m88000__ */ + +/*************************************** + ************** MIPS ***************** + ***************************************/ +#if defined (__mips__) && W_TYPE_SIZE == 32 +#if __GNUC__ > 2 || __GNUC_MINOR__ >= 7 +#define umul_ppmm(w1, w0, u, v) \ + __asm__ ("multu %2,%3" \ + : "=l" ((USItype)(w0)), \ + "=h" ((USItype)(w1)) \ + : "d" ((USItype)(u)), \ + "d" ((USItype)(v))) +#else +#define umul_ppmm(w1, w0, u, v) \ + __asm__ ("multu %2,%3 \n" \ + "mflo %0 \n" \ + "mfhi %1" \ + : "=d" ((USItype)(w0)), \ + "=d" ((USItype)(w1)) \ + : "d" ((USItype)(u)), \ + "d" ((USItype)(v))) +#endif +#define UMUL_TIME 10 +#define UDIV_TIME 100 +#endif /* __mips__ */ + +/*************************************** + ************** MIPS/64 ************** + ***************************************/ +#if (defined (__mips) && __mips >= 3) && W_TYPE_SIZE == 64 +#if __GNUC__ > 2 || __GNUC_MINOR__ >= 7 +#define umul_ppmm(w1, w0, u, v) \ + __asm__ ("dmultu %2,%3" \ + : "=l" ((UDItype)(w0)), \ + "=h" ((UDItype)(w1)) \ + : "d" ((UDItype)(u)), \ + "d" ((UDItype)(v))) +#else +#define umul_ppmm(w1, w0, u, v) \ + __asm__ ("dmultu %2,%3 \n" \ + "mflo %0 \n" \ + "mfhi %1" \ + : "=d" ((UDItype)(w0)), \ + "=d" ((UDItype)(w1)) \ + : "d" ((UDItype)(u)), \ + "d" ((UDItype)(v))) +#endif +#define UMUL_TIME 20 +#define UDIV_TIME 140 +#endif /* __mips__ */ + + +/*************************************** + ************** 32000 **************** + ***************************************/ +#if defined (__ns32000__) && W_TYPE_SIZE == 32 +#define umul_ppmm(w1, w0, u, v) \ + ({union {UDItype __ll; \ + struct {USItype __l, __h;} __i; \ + } __xx; \ + __asm__ ("meid %2,%0" \ + : "=g" (__xx.__ll) \ + : "%0" ((USItype)(u)), \ + "g" ((USItype)(v))); \ + (w1) = __xx.__i.__h; (w0) = __xx.__i.__l;}) +#define __umulsidi3(u, v) \ + ({UDItype __w; \ + __asm__ ("meid %2,%0" \ + : "=g" (__w) \ + : "%0" ((USItype)(u)), \ + "g" ((USItype)(v))); \ + __w; }) +#define udiv_qrnnd(q, r, n1, n0, d) \ + ({union {UDItype __ll; \ + struct {USItype __l, __h;} __i; \ + } __xx; \ + __xx.__i.__h = (n1); __xx.__i.__l = (n0); \ + __asm__ ("deid %2,%0" \ + : "=g" (__xx.__ll) \ + : "0" (__xx.__ll), \ + "g" ((USItype)(d))); \ + (r) = __xx.__i.__l; (q) = __xx.__i.__h; }) +#define count_trailing_zeros(count,x) \ + do { + __asm__ ("ffsd %2,%0" \ + : "=r" ((USItype) (count)) \ + : "0" ((USItype) 0), \ + "r" ((USItype) (x))); \ + } while (0) +#endif /* __ns32000__ */ + + +/*************************************** + ************** PPC ****************** + ***************************************/ +#if (defined (_ARCH_PPC) || defined (_IBMR2)) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + do { \ + if (__builtin_constant_p (bh) && (bh) == 0) \ + __asm__ ("{a%I4|add%I4c} %1,%3,%4\n\t{aze|addze} %0,%2" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%r" ((USItype)(ah)), \ + "%r" ((USItype)(al)), \ + "rI" ((USItype)(bl))); \ + else if (__builtin_constant_p (bh) && (bh) ==~(USItype) 0) \ + __asm__ ("{a%I4|add%I4c} %1,%3,%4\n\t{ame|addme} %0,%2" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%r" ((USItype)(ah)), \ + "%r" ((USItype)(al)), \ + "rI" ((USItype)(bl))); \ + else \ + __asm__ ("{a%I5|add%I5c} %1,%4,%5\n\t{ae|adde} %0,%2,%3" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%r" ((USItype)(ah)), \ + "r" ((USItype)(bh)), \ + "%r" ((USItype)(al)), \ + "rI" ((USItype)(bl))); \ + } while (0) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + do { \ + if (__builtin_constant_p (ah) && (ah) == 0) \ + __asm__ ("{sf%I3|subf%I3c} %1,%4,%3\n\t{sfze|subfze} %0,%2" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "r" ((USItype)(bh)), \ + "rI" ((USItype)(al)), \ + "r" ((USItype)(bl))); \ + else if (__builtin_constant_p (ah) && (ah) ==~(USItype) 0) \ + __asm__ ("{sf%I3|subf%I3c} %1,%4,%3\n\t{sfme|subfme} %0,%2" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "r" ((USItype)(bh)), \ + "rI" ((USItype)(al)), \ + "r" ((USItype)(bl))); \ + else if (__builtin_constant_p (bh) && (bh) == 0) \ + __asm__ ("{sf%I3|subf%I3c} %1,%4,%3\n\t{ame|addme} %0,%2" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "r" ((USItype)(ah)), \ + "rI" ((USItype)(al)), \ + "r" ((USItype)(bl))); \ + else if (__builtin_constant_p (bh) && (bh) ==~(USItype) 0) \ + __asm__ ("{sf%I3|subf%I3c} %1,%4,%3\n\t{aze|addze} %0,%2" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "r" ((USItype)(ah)), \ + "rI" ((USItype)(al)), \ + "r" ((USItype)(bl))); \ + else \ + __asm__ ("{sf%I4|subf%I4c} %1,%5,%4\n\t{sfe|subfe} %0,%3,%2" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "r" ((USItype)(ah)), \ + "r" ((USItype)(bh)), \ + "rI" ((USItype)(al)), \ + "r" ((USItype)(bl))); \ + } while (0) +#define count_leading_zeros(count, x) \ + __asm__ ("{cntlz|cntlzw} %0,%1" \ + : "=r" ((USItype)(count)) \ + : "r" ((USItype)(x))) +#define COUNT_LEADING_ZEROS_0 32 +#if defined (_ARCH_PPC) +#define umul_ppmm(ph, pl, m0, m1) \ + do { \ + USItype __m0 = (m0), __m1 = (m1); \ + __asm__ ("mulhwu %0,%1,%2" \ + : "=r" ((USItype) ph) \ + : "%r" (__m0), \ + "r" (__m1)); \ + (pl) = __m0 * __m1; \ + } while (0) +#define UMUL_TIME 15 +#define smul_ppmm(ph, pl, m0, m1) \ + do { \ + SItype __m0 = (m0), __m1 = (m1); \ + __asm__ ("mulhw %0,%1,%2" \ + : "=r" ((SItype) ph) \ + : "%r" (__m0), \ + "r" (__m1)); \ + (pl) = __m0 * __m1; \ + } while (0) +#define SMUL_TIME 14 +#define UDIV_TIME 120 +#else +#define umul_ppmm(xh, xl, m0, m1) \ + do { \ + USItype __m0 = (m0), __m1 = (m1); \ + __asm__ ("mul %0,%2,%3" \ + : "=r" ((USItype)(xh)), \ + "=q" ((USItype)(xl)) \ + : "r" (__m0), \ + "r" (__m1)); \ + (xh) += ((((SItype) __m0 >> 31) & __m1) \ + + (((SItype) __m1 >> 31) & __m0)); \ + } while (0) +#define UMUL_TIME 8 +#define smul_ppmm(xh, xl, m0, m1) \ + __asm__ ("mul %0,%2,%3" \ + : "=r" ((SItype)(xh)), \ + "=q" ((SItype)(xl)) \ + : "r" (m0), \ + "r" (m1)) +#define SMUL_TIME 4 +#define sdiv_qrnnd(q, r, nh, nl, d) \ + __asm__ ("div %0,%2,%4" \ + : "=r" ((SItype)(q)), "=q" ((SItype)(r)) \ + : "r" ((SItype)(nh)), "1" ((SItype)(nl)), "r" ((SItype)(d))) +#define UDIV_TIME 100 +#endif +#endif /* Power architecture variants. */ + +/* Powerpc 64 bit support taken from gmp-4.1.2. */ +/* We should test _IBMR2 here when we add assembly support for the system + vendor compilers. */ +#if 0 /* Not yet enabled because we don't have hardware for a test. */ +#if (defined (_ARCH_PPC) || defined (__powerpc__)) && W_TYPE_SIZE == 64 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + do { \ + if (__builtin_constant_p (bh) && (bh) == 0) \ + __asm__ ("{a%I4|add%I4c} %1,%3,%4\n\t{aze|addze} %0,%2" \ + : "=r" (sh), "=&r" (sl) : "r" (ah), "%r" (al), "rI" (bl));\ + else if (__builtin_constant_p (bh) && (bh) == ~(UDItype) 0) \ + __asm__ ("{a%I4|add%I4c} %1,%3,%4\n\t{ame|addme} %0,%2" \ + : "=r" (sh), "=&r" (sl) : "r" (ah), "%r" (al), "rI" (bl));\ + else \ + __asm__ ("{a%I5|add%I5c} %1,%4,%5\n\t{ae|adde} %0,%2,%3" \ + : "=r" (sh), "=&r" (sl) \ + : "%r" (ah), "r" (bh), "%r" (al), "rI" (bl)); \ + } while (0) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + do { \ + if (__builtin_constant_p (ah) && (ah) == 0) \ + __asm__ ("{sf%I3|subf%I3c} %1,%4,%3\n\t{sfze|subfze} %0,%2" \ + : "=r" (sh), "=&r" (sl) : "r" (bh), "rI" (al), "r" (bl));\ + else if (__builtin_constant_p (ah) && (ah) == ~(UDItype) 0) \ + __asm__ ("{sf%I3|subf%I3c} %1,%4,%3\n\t{sfme|subfme} %0,%2" \ + : "=r" (sh), "=&r" (sl) : "r" (bh), "rI" (al), "r" (bl));\ + else if (__builtin_constant_p (bh) && (bh) == 0) \ + __asm__ ("{sf%I3|subf%I3c} %1,%4,%3\n\t{ame|addme} %0,%2" \ + : "=r" (sh), "=&r" (sl) : "r" (ah), "rI" (al), "r" (bl));\ + else if (__builtin_constant_p (bh) && (bh) == ~(UDItype) 0) \ + __asm__ ("{sf%I3|subf%I3c} %1,%4,%3\n\t{aze|addze} %0,%2" \ + : "=r" (sh), "=&r" (sl) : "r" (ah), "rI" (al), "r" (bl));\ + else \ + __asm__ ("{sf%I4|subf%I4c} %1,%5,%4\n\t{sfe|subfe} %0,%3,%2" \ + : "=r" (sh), "=&r" (sl) \ + : "r" (ah), "r" (bh), "rI" (al), "r" (bl)); \ + } while (0) +#define count_leading_zeros(count, x) \ + __asm__ ("cntlzd %0,%1" : "=r" (count) : "r" (x)) +#define COUNT_LEADING_ZEROS_0 64 +#define umul_ppmm(ph, pl, m0, m1) \ + do { \ + UDItype __m0 = (m0), __m1 = (m1); \ + __asm__ ("mulhdu %0,%1,%2" : "=r" (ph) : "%r" (m0), "r" (m1)); \ + (pl) = __m0 * __m1; \ + } while (0) +#define UMUL_TIME 15 +#define smul_ppmm(ph, pl, m0, m1) \ + do { \ + DItype __m0 = (m0), __m1 = (m1); \ + __asm__ ("mulhd %0,%1,%2" : "=r" (ph) : "%r" (m0), "r" (m1)); \ + (pl) = __m0 * __m1; \ + } while (0) +#define SMUL_TIME 14 /* ??? */ +#define UDIV_TIME 120 /* ??? */ +#endif /* 64-bit PowerPC. */ +#endif /* if 0 */ + +/*************************************** + ************** PYR ****************** + ***************************************/ +#if defined (__pyr__) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("addw %5,%1 \n" \ + "addwc %3,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%0" ((USItype)(ah)), \ + "g" ((USItype)(bh)), \ + "%1" ((USItype)(al)), \ + "g" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("subw %5,%1 \n" \ + "subwb %3,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "0" ((USItype)(ah)), \ + "g" ((USItype)(bh)), \ + "1" ((USItype)(al)), \ + "g" ((USItype)(bl))) +/* This insn works on Pyramids with AP, XP, or MI CPUs, but not with SP. */ +#define umul_ppmm(w1, w0, u, v) \ + ({union {UDItype __ll; \ + struct {USItype __h, __l;} __i; \ + } __xx; \ + __asm__ ("movw %1,%R0 \n" \ + "uemul %2,%0" \ + : "=&r" (__xx.__ll) \ + : "g" ((USItype) (u)), \ + "g" ((USItype)(v))); \ + (w1) = __xx.__i.__h; (w0) = __xx.__i.__l;}) +#endif /* __pyr__ */ + + +/*************************************** + ************** RT/ROMP ************** + ***************************************/ +#if defined (__ibm032__) /* RT/ROMP */ && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("a %1,%5 \n" \ + "ae %0,%3" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%0" ((USItype)(ah)), \ + "r" ((USItype)(bh)), \ + "%1" ((USItype)(al)), \ + "r" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("s %1,%5\n" \ + "se %0,%3" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "0" ((USItype)(ah)), \ + "r" ((USItype)(bh)), \ + "1" ((USItype)(al)), \ + "r" ((USItype)(bl))) +#define umul_ppmm(ph, pl, m0, m1) \ + do { \ + USItype __m0 = (m0), __m1 = (m1); \ + __asm__ ( \ + "s r2,r2 \n" \ + "mts r10,%2 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "m r2,%3 \n" \ + "cas %0,r2,r0 \n" \ + "mfs r10,%1" \ + : "=r" ((USItype)(ph)), \ + "=r" ((USItype)(pl)) \ + : "%r" (__m0), \ + "r" (__m1) \ + : "r2"); \ + (ph) += ((((SItype) __m0 >> 31) & __m1) \ + + (((SItype) __m1 >> 31) & __m0)); \ + } while (0) +#define UMUL_TIME 20 +#define UDIV_TIME 200 +#define count_leading_zeros(count, x) \ + do { \ + if ((x) >= 0x10000) \ + __asm__ ("clz %0,%1" \ + : "=r" ((USItype)(count)) \ + : "r" ((USItype)(x) >> 16)); \ + else \ + { \ + __asm__ ("clz %0,%1" \ + : "=r" ((USItype)(count)) \ + : "r" ((USItype)(x))); \ + (count) += 16; \ + } \ + } while (0) +#endif /* RT/ROMP */ + + +/*************************************** + ************** SH2 ****************** + ***************************************/ +#if (defined (__sh2__) || defined(__sh3__) || defined(__SH4__) ) \ + && W_TYPE_SIZE == 32 +#define umul_ppmm(w1, w0, u, v) \ + __asm__ ( \ + "dmulu.l %2,%3\n" \ + "sts macl,%1\n" \ + "sts mach,%0" \ + : "=r" ((USItype)(w1)), \ + "=r" ((USItype)(w0)) \ + : "r" ((USItype)(u)), \ + "r" ((USItype)(v)) \ + : "macl", "mach") +#define UMUL_TIME 5 +#endif + +/*************************************** + ************** SPARC **************** + ***************************************/ +#if defined (__sparc__) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("addcc %r4,%5,%1\n" \ + "addx %r2,%3,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "%rJ" ((USItype)(ah)), \ + "rI" ((USItype)(bh)), \ + "%rJ" ((USItype)(al)), \ + "rI" ((USItype)(bl)) \ + __CLOBBER_CC) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("subcc %r4,%5,%1\n" \ + "subx %r2,%3,%0" \ + : "=r" ((USItype)(sh)), \ + "=&r" ((USItype)(sl)) \ + : "rJ" ((USItype)(ah)), \ + "rI" ((USItype)(bh)), \ + "rJ" ((USItype)(al)), \ + "rI" ((USItype)(bl)) \ + __CLOBBER_CC) +#if defined (__sparc_v8__) +/* Don't match immediate range because, 1) it is not often useful, + 2) the 'I' flag thinks of the range as a 13 bit signed interval, + while we want to match a 13 bit interval, sign extended to 32 bits, + but INTERPRETED AS UNSIGNED. */ +#define umul_ppmm(w1, w0, u, v) \ + __asm__ ("umul %2,%3,%1;rd %%y,%0" \ + : "=r" ((USItype)(w1)), \ + "=r" ((USItype)(w0)) \ + : "r" ((USItype)(u)), \ + "r" ((USItype)(v))) +#define UMUL_TIME 5 +#ifndef SUPERSPARC /* SuperSPARC's udiv only handles 53 bit dividends */ +#define udiv_qrnnd(q, r, n1, n0, d) \ + do { \ + USItype __q; \ + __asm__ ("mov %1,%%y;nop;nop;nop;udiv %2,%3,%0" \ + : "=r" ((USItype)(__q)) \ + : "r" ((USItype)(n1)), \ + "r" ((USItype)(n0)), \ + "r" ((USItype)(d))); \ + (r) = (n0) - __q * (d); \ + (q) = __q; \ + } while (0) +#define UDIV_TIME 25 +#endif /* SUPERSPARC */ +#else /* ! __sparc_v8__ */ +#if defined (__sparclite__) +/* This has hardware multiply but not divide. It also has two additional + instructions scan (ffs from high bit) and divscc. */ +#define umul_ppmm(w1, w0, u, v) \ + __asm__ ("umul %2,%3,%1;rd %%y,%0" \ + : "=r" ((USItype)(w1)), \ + "=r" ((USItype)(w0)) \ + : "r" ((USItype)(u)), \ + "r" ((USItype)(v))) +#define UMUL_TIME 5 +#define udiv_qrnnd(q, r, n1, n0, d) \ + __asm__ ("! Inlined udiv_qrnnd \n" \ + " wr %%g0,%2,%%y ! Not a delayed write for sparclite \n" \ + " tst %%g0 \n" \ + " divscc %3,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%%g1 \n" \ + " divscc %%g1,%4,%0 \n" \ + " rd %%y,%1 \n" \ + " bl,a 1f \n" \ + " add %1,%4,%1 \n" \ + "1: ! End of inline udiv_qrnnd" \ + : "=r" ((USItype)(q)), \ + "=r" ((USItype)(r)) \ + : "r" ((USItype)(n1)), \ + "r" ((USItype)(n0)), \ + "rI" ((USItype)(d)) \ + : "%g1" __AND_CLOBBER_CC) +#define UDIV_TIME 37 +#define count_leading_zeros(count, x) \ + __asm__ ("scan %1,0,%0" \ + : "=r" ((USItype)(x)) \ + : "r" ((USItype)(count))) +/* Early sparclites return 63 for an argument of 0, but they warn that future + implementations might change this. Therefore, leave COUNT_LEADING_ZEROS_0 + undefined. */ +#endif /* __sparclite__ */ +#endif /* __sparc_v8__ */ +/* Default to sparc v7 versions of umul_ppmm and udiv_qrnnd. */ +#ifndef umul_ppmm +#define umul_ppmm(w1, w0, u, v) \ + __asm__ ("! Inlined umul_ppmm \n" \ + " wr %%g0,%2,%%y ! SPARC has 0-3 delay insn after a wr \n" \ + " sra %3,31,%%g2 ! Don't move this insn \n" \ + " and %2,%%g2,%%g2 ! Don't move this insn \n" \ + " andcc %%g0,0,%%g1 ! Don't move this insn \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,%3,%%g1 \n" \ + " mulscc %%g1,0,%%g1 \n" \ + " add %%g1,%%g2,%0 \n" \ + " rd %%y,%1" \ + : "=r" ((USItype)(w1)), \ + "=r" ((USItype)(w0)) \ + : "%rI" ((USItype)(u)), \ + "r" ((USItype)(v)) \ + : "%g1", "%g2" __AND_CLOBBER_CC) +#define UMUL_TIME 39 /* 39 instructions */ +#endif +#ifndef udiv_qrnnd +#ifndef LONGLONG_STANDALONE +#define udiv_qrnnd(q, r, n1, n0, d) \ + do { USItype __r; \ + (q) = __udiv_qrnnd (&__r, (n1), (n0), (d)); \ + (r) = __r; \ + } while (0) +extern USItype __udiv_qrnnd (); +#define UDIV_TIME 140 +#endif /* LONGLONG_STANDALONE */ +#endif /* udiv_qrnnd */ +#endif /* __sparc__ */ + + +/*************************************** + ************** VAX ****************** + ***************************************/ +#if defined (__vax__) && W_TYPE_SIZE == 32 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("addl2 %5,%1\n" \ + "adwc %3,%0" \ + : "=g" ((USItype)(sh)), \ + "=&g" ((USItype)(sl)) \ + : "%0" ((USItype)(ah)), \ + "g" ((USItype)(bh)), \ + "%1" ((USItype)(al)), \ + "g" ((USItype)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("subl2 %5,%1\n" \ + "sbwc %3,%0" \ + : "=g" ((USItype)(sh)), \ + "=&g" ((USItype)(sl)) \ + : "0" ((USItype)(ah)), \ + "g" ((USItype)(bh)), \ + "1" ((USItype)(al)), \ + "g" ((USItype)(bl))) +#define umul_ppmm(xh, xl, m0, m1) \ + do { \ + union {UDItype __ll; \ + struct {USItype __l, __h;} __i; \ + } __xx; \ + USItype __m0 = (m0), __m1 = (m1); \ + __asm__ ("emul %1,%2,$0,%0" \ + : "=g" (__xx.__ll) \ + : "g" (__m0), \ + "g" (__m1)); \ + (xh) = __xx.__i.__h; (xl) = __xx.__i.__l; \ + (xh) += ((((SItype) __m0 >> 31) & __m1) \ + + (((SItype) __m1 >> 31) & __m0)); \ + } while (0) +#define sdiv_qrnnd(q, r, n1, n0, d) \ + do { \ + union {DItype __ll; \ + struct {SItype __l, __h;} __i; \ + } __xx; \ + __xx.__i.__h = n1; __xx.__i.__l = n0; \ + __asm__ ("ediv %3,%2,%0,%1" \ + : "=g" (q), "=g" (r) \ + : "g" (__xx.__ll), "g" (d)); \ + } while (0) +#endif /* __vax__ */ + + +/*************************************** + ************** Z8000 **************** + ***************************************/ +#if defined (__z8000__) && W_TYPE_SIZE == 16 +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + __asm__ ("add %H1,%H5\n\tadc %H0,%H3" \ + : "=r" ((unsigned int)(sh)), \ + "=&r" ((unsigned int)(sl)) \ + : "%0" ((unsigned int)(ah)), \ + "r" ((unsigned int)(bh)), \ + "%1" ((unsigned int)(al)), \ + "rQR" ((unsigned int)(bl))) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + __asm__ ("sub %H1,%H5\n\tsbc %H0,%H3" \ + : "=r" ((unsigned int)(sh)), \ + "=&r" ((unsigned int)(sl)) \ + : "0" ((unsigned int)(ah)), \ + "r" ((unsigned int)(bh)), \ + "1" ((unsigned int)(al)), \ + "rQR" ((unsigned int)(bl))) +#define umul_ppmm(xh, xl, m0, m1) \ + do { \ + union {long int __ll; \ + struct {unsigned int __h, __l;} __i; \ + } __xx; \ + unsigned int __m0 = (m0), __m1 = (m1); \ + __asm__ ("mult %S0,%H3" \ + : "=r" (__xx.__i.__h), \ + "=r" (__xx.__i.__l) \ + : "%1" (__m0), \ + "rQR" (__m1)); \ + (xh) = __xx.__i.__h; (xl) = __xx.__i.__l; \ + (xh) += ((((signed int) __m0 >> 15) & __m1) \ + + (((signed int) __m1 >> 15) & __m0)); \ + } while (0) +#endif /* __z8000__ */ + +#endif /* __GNUC__ */ + + +/*************************************** + *********** Generic Versions ******** + ***************************************/ +#if !defined (umul_ppmm) && defined (__umulsidi3) +#define umul_ppmm(ph, pl, m0, m1) \ + { \ + UDWtype __ll = __umulsidi3 (m0, m1); \ + ph = (UWtype) (__ll >> W_TYPE_SIZE); \ + pl = (UWtype) __ll; \ + } +#endif + +#if !defined (__umulsidi3) +#define __umulsidi3(u, v) \ + ({UWtype __hi, __lo; \ + umul_ppmm (__hi, __lo, u, v); \ + ((UDWtype) __hi << W_TYPE_SIZE) | __lo; }) +#endif + +/* If this machine has no inline assembler, use C macros. */ + +#if !defined (add_ssaaaa) +#define add_ssaaaa(sh, sl, ah, al, bh, bl) \ + do { \ + UWtype __x; \ + __x = (al) + (bl); \ + (sh) = (ah) + (bh) + (__x < (al)); \ + (sl) = __x; \ + } while (0) +#endif + +#if !defined (sub_ddmmss) +#define sub_ddmmss(sh, sl, ah, al, bh, bl) \ + do { \ + UWtype __x; \ + __x = (al) - (bl); \ + (sh) = (ah) - (bh) - (__x > (al)); \ + (sl) = __x; \ + } while (0) +#endif + +#if !defined (umul_ppmm) +#define umul_ppmm(w1, w0, u, v) \ + do { \ + UWtype __x0, __x1, __x2, __x3; \ + UHWtype __ul, __vl, __uh, __vh; \ + UWtype __u = (u), __v = (v); \ + \ + __ul = __ll_lowpart (__u); \ + __uh = __ll_highpart (__u); \ + __vl = __ll_lowpart (__v); \ + __vh = __ll_highpart (__v); \ + \ + __x0 = (UWtype) __ul * __vl; \ + __x1 = (UWtype) __ul * __vh; \ + __x2 = (UWtype) __uh * __vl; \ + __x3 = (UWtype) __uh * __vh; \ + \ + __x1 += __ll_highpart (__x0);/* this can't give carry */ \ + __x1 += __x2; /* but this indeed can */ \ + if (__x1 < __x2) /* did we get it? */ \ + __x3 += __ll_B; /* yes, add it in the proper pos. */ \ + \ + (w1) = __x3 + __ll_highpart (__x1); \ + (w0) = (__ll_lowpart (__x1) << W_TYPE_SIZE/2) + __ll_lowpart (__x0);\ + } while (0) +#endif + +#if !defined (umul_ppmm) +#define smul_ppmm(w1, w0, u, v) \ + do { \ + UWtype __w1; \ + UWtype __m0 = (u), __m1 = (v); \ + umul_ppmm (__w1, w0, __m0, __m1); \ + (w1) = __w1 - (-(__m0 >> (W_TYPE_SIZE - 1)) & __m1) \ + - (-(__m1 >> (W_TYPE_SIZE - 1)) & __m0); \ + } while (0) +#endif + +/* Define this unconditionally, so it can be used for debugging. */ +#define __udiv_qrnnd_c(q, r, n1, n0, d) \ + do { \ + UWtype __d1, __d0, __q1, __q0, __r1, __r0, __m; \ + __d1 = __ll_highpart (d); \ + __d0 = __ll_lowpart (d); \ + \ + __r1 = (n1) % __d1; \ + __q1 = (n1) / __d1; \ + __m = (UWtype) __q1 * __d0; \ + __r1 = __r1 * __ll_B | __ll_highpart (n0); \ + if (__r1 < __m) \ + { \ + __q1--, __r1 += (d); \ + if (__r1 >= (d)) /* i.e. we didn't get carry when adding to __r1 */\ + if (__r1 < __m) \ + __q1--, __r1 += (d); \ + } \ + __r1 -= __m; \ + \ + __r0 = __r1 % __d1; \ + __q0 = __r1 / __d1; \ + __m = (UWtype) __q0 * __d0; \ + __r0 = __r0 * __ll_B | __ll_lowpart (n0); \ + if (__r0 < __m) \ + { \ + __q0--, __r0 += (d); \ + if (__r0 >= (d)) \ + if (__r0 < __m) \ + __q0--, __r0 += (d); \ + } \ + __r0 -= __m; \ + \ + (q) = (UWtype) __q1 * __ll_B | __q0; \ + (r) = __r0; \ + } while (0) + +/* If the processor has no udiv_qrnnd but sdiv_qrnnd, go through + __udiv_w_sdiv (defined in libgcc or elsewhere). */ +#if !defined (udiv_qrnnd) && defined (sdiv_qrnnd) +#define udiv_qrnnd(q, r, nh, nl, d) \ + do { \ + UWtype __r; \ + (q) = __MPN(udiv_w_sdiv) (&__r, nh, nl, d); \ + (r) = __r; \ + } while (0) +#endif + +/* If udiv_qrnnd was not defined for this processor, use __udiv_qrnnd_c. */ +#if !defined (udiv_qrnnd) +#define UDIV_NEEDS_NORMALIZATION 1 +#define udiv_qrnnd __udiv_qrnnd_c +#endif + +#if !defined (count_leading_zeros) +extern +#ifdef __STDC__ +const +#endif +unsigned char __clz_tab[]; +#define MPI_INTERNAL_NEED_CLZ_TAB 1 +#define count_leading_zeros(count, x) \ + do { \ + UWtype __xr = (x); \ + UWtype __a; \ + \ + if (W_TYPE_SIZE <= 32) \ + { \ + __a = __xr < ((UWtype) 1 << 2*__BITS4) \ + ? (__xr < ((UWtype) 1 << __BITS4) ? 0 : __BITS4) \ + : (__xr < ((UWtype) 1 << 3*__BITS4) ? 2*__BITS4 : 3*__BITS4);\ + } \ + else \ + { \ + for (__a = W_TYPE_SIZE - 8; __a > 0; __a -= 8) \ + if (((__xr >> __a) & 0xff) != 0) \ + break; \ + } \ + \ + (count) = W_TYPE_SIZE - (__clz_tab[__xr >> __a] + __a); \ + } while (0) +/* This version gives a well-defined value for zero. */ +#define COUNT_LEADING_ZEROS_0 W_TYPE_SIZE +#endif + +#if !defined (count_trailing_zeros) +/* Define count_trailing_zeros using count_leading_zeros. The latter might be + defined in asm, but if it is not, the C version above is good enough. */ +#define count_trailing_zeros(count, x) \ + do { \ + UWtype __ctz_x = (x); \ + UWtype __ctz_c; \ + count_leading_zeros (__ctz_c, __ctz_x & -__ctz_x); \ + (count) = W_TYPE_SIZE - 1 - __ctz_c; \ + } while (0) +#endif + +#ifndef UDIV_NEEDS_NORMALIZATION +#define UDIV_NEEDS_NORMALIZATION 0 +#endif diff --git a/mpi/m68k/distfiles b/mpi/m68k/distfiles new file mode 100644 index 0000000..4c0967b --- /dev/null +++ b/mpi/m68k/distfiles @@ -0,0 +1,8 @@ +syntax.h +mpih-lshift.S +mpih-rshift.S +mpih-add1.S +mpih-sub1.S + + + diff --git a/mpi/m68k/mc68020/distfiles b/mpi/m68k/mc68020/distfiles new file mode 100644 index 0000000..fc7df9f --- /dev/null +++ b/mpi/m68k/mc68020/distfiles @@ -0,0 +1,3 @@ +mpih-mul1.S +mpih-mul2.S +mpih-mul3.S diff --git a/mpi/m68k/mc68020/mpih-mul1.S b/mpi/m68k/mc68020/mpih-mul1.S new file mode 100644 index 0000000..c24305d --- /dev/null +++ b/mpi/m68k/mc68020/mpih-mul1.S @@ -0,0 +1,107 @@ +/* mc68020 __mpn_mul_1 -- Multiply a limb vector with a limb and store + * the result in a second limb vector. + * + * Copyright (C) 1992, 1994, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_mul_1( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_size_t s1_size, (sp + 12) + * mpi_limb_t s2_limb) (sp + 16) + */ + + + TEXT + ALIGN + GLOBL C_SYMBOL_NAME(mpihelp_mul_1) + +C_SYMBOL_NAME(mpihelp_mul_1:) +PROLOG(mpihelp_mul_1) + +#define res_ptr a0 +#define s1_ptr a1 +#define s1_size d2 +#define s2_limb d4 + +/* Save used registers on the stack. */ + moveml R(d2)-R(d4),MEM_PREDEC(sp) +#if 0 + movel R(d2),MEM_PREDEC(sp) + movel R(d3),MEM_PREDEC(sp) + movel R(d4),MEM_PREDEC(sp) +#endif + +/* Copy the arguments to registers. Better use movem? */ + movel MEM_DISP(sp,16),R(res_ptr) + movel MEM_DISP(sp,20),R(s1_ptr) + movel MEM_DISP(sp,24),R(s1_size) + movel MEM_DISP(sp,28),R(s2_limb) + + eorw #1,R(s1_size) + clrl R(d1) + lsrl #1,R(s1_size) + bcc L(L1) + subql #1,R(s1_size) + subl R(d0),R(d0) /* (d0,cy) <= (0,0) */ + +L(Loop:) + movel MEM_POSTINC(s1_ptr),R(d3) + mulul R(s2_limb),R(d1):R(d3) + addxl R(d0),R(d3) + movel R(d3),MEM_POSTINC(res_ptr) +L(L1:) movel MEM_POSTINC(s1_ptr),R(d3) + mulul R(s2_limb),R(d0):R(d3) + addxl R(d1),R(d3) + movel R(d3),MEM_POSTINC(res_ptr) + + dbf R(s1_size),L(Loop) + clrl R(d3) + addxl R(d3),R(d0) + subl #0x10000,R(s1_size) + bcc L(Loop) + +/* Restore used registers from stack frame. */ + moveml MEM_POSTINC(sp),R(d2)-R(d4) +#if 0 + movel MEM_POSTINC(sp),R(d4) + movel MEM_POSTINC(sp),R(d3) + movel MEM_POSTINC(sp),R(d2) +#endif + rts +EPILOG(mpihelp_mul_1) + + diff --git a/mpi/m68k/mc68020/mpih-mul2.S b/mpi/m68k/mc68020/mpih-mul2.S new file mode 100644 index 0000000..2684d7d --- /dev/null +++ b/mpi/m68k/mc68020/mpih-mul2.S @@ -0,0 +1,97 @@ +/* mc68020 __mpn_addmul_1 -- Multiply a limb vector with a limb and add + * the result to a second limb vector. + * + * Copyright (C) 1992, 1994, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + +/******************* + * mpi_limb_t + * mpihelp_addmul_1( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_size_t s1_size, (sp + 12) + * mpi_limb_t s2_limb) (sp + 16) + */ + + + TEXT + ALIGN + GLOBL C_SYMBOL_NAME(mpihelp_addmul_1) + +C_SYMBOL_NAME(mpihelp_addmul_1:) +PROLOG(mpihelp_addmul_1) + +#define res_ptr a0 +#define s1_ptr a1 +#define s1_size d2 +#define s2_limb d4 + +/* Save used registers on the stack. */ + moveml R(d2)-R(d5),MEM_PREDEC(sp) + +/* Copy the arguments to registers. Better use movem? */ + movel MEM_DISP(sp,20),R(res_ptr) + movel MEM_DISP(sp,24),R(s1_ptr) + movel MEM_DISP(sp,28),R(s1_size) + movel MEM_DISP(sp,32),R(s2_limb) + + eorw #1,R(s1_size) + clrl R(d1) + clrl R(d5) + lsrl #1,R(s1_size) + bcc L(L1) + subql #1,R(s1_size) + subl R(d0),R(d0) /* (d0,cy) <= (0,0) */ + +L(Loop:) + movel MEM_POSTINC(s1_ptr),R(d3) + mulul R(s2_limb),R(d1):R(d3) + addxl R(d0),R(d3) + addxl R(d5),R(d1) + addl R(d3),MEM_POSTINC(res_ptr) +L(L1:) movel MEM_POSTINC(s1_ptr),R(d3) + mulul R(s2_limb),R(d0):R(d3) + addxl R(d1),R(d3) + addxl R(d5),R(d0) + addl R(d3),MEM_POSTINC(res_ptr) + + dbf R(s1_size),L(Loop) + addxl R(d5),R(d0) + subl #0x10000,R(s1_size) + bcc L(Loop) + +/* Restore used registers from stack frame. */ + moveml MEM_POSTINC(sp),R(d2)-R(d5) + + rts +EPILOG(mpihelp_addmul_1) + diff --git a/mpi/m68k/mc68020/mpih-mul3.S b/mpi/m68k/mc68020/mpih-mul3.S new file mode 100644 index 0000000..c031152 --- /dev/null +++ b/mpi/m68k/mc68020/mpih-mul3.S @@ -0,0 +1,100 @@ +/* mc68020 __mpn_submul_1 -- Multiply a limb vector with a limb and subtract + * the result from a second limb vector. + * + * Copyright (C) 1992, 1994, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + + +#include "sysdep.h" +#include "asm-syntax.h" + +/******************* + * mpi_limb_t + * mpihelp_submul_1( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_size_t s1_size, (sp + 12) + * mpi_limb_t s2_limb) (sp + 16) + */ + + + TEXT + ALIGN + GLOBL C_SYMBOL_NAME(mpihelp_submul_1) + +C_SYMBOL_NAME(mpihelp_submul_1:) +PROLOG(mpihelp_submul_1) + +#define res_ptr a0 +#define s1_ptr a1 +#define s1_size d2 +#define s2_limb d4 + +/* Save used registers on the stack. */ + moveml R(d2)-R(d5),MEM_PREDEC(sp) + +/* Copy the arguments to registers. Better use movem? */ + movel MEM_DISP(sp,20),R(res_ptr) + movel MEM_DISP(sp,24),R(s1_ptr) + movel MEM_DISP(sp,28),R(s1_size) + movel MEM_DISP(sp,32),R(s2_limb) + + eorw #1,R(s1_size) + clrl R(d1) + clrl R(d5) + lsrl #1,R(s1_size) + bcc L(L1) + subql #1,R(s1_size) + subl R(d0),R(d0) /* (d0,cy) <= (0,0) */ + +L(Loop:) + movel MEM_POSTINC(s1_ptr),R(d3) + mulul R(s2_limb),R(d1):R(d3) + addxl R(d0),R(d3) + addxl R(d5),R(d1) + subl R(d3),MEM_POSTINC(res_ptr) +L(L1:) movel MEM_POSTINC(s1_ptr),R(d3) + mulul R(s2_limb),R(d0):R(d3) + addxl R(d1),R(d3) + addxl R(d5),R(d0) + subl R(d3),MEM_POSTINC(res_ptr) + + dbf R(s1_size),L(Loop) + addxl R(d5),R(d0) + subl #0x10000,R(s1_size) + bcc L(Loop) + +/* Restore used registers from stack frame. */ + moveml MEM_POSTINC(sp),R(d2)-R(d5) + + rts +EPILOG(mpihelp_submul_1) + + diff --git a/mpi/m68k/mpih-add1.S b/mpi/m68k/mpih-add1.S new file mode 100644 index 0000000..c294f9d --- /dev/null +++ b/mpi/m68k/mpih-add1.S @@ -0,0 +1,95 @@ +/* mc68020 __mpn_add_n -- Add two limb vectors of the same length > 0 and store + * sum in a third limb vector. + * + * Copyright (C) 1992, 1994,1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_add_n( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_ptr_t s2_ptr, (sp + 16) + * mpi_size_t size) (sp + 12) + */ + + + TEXT + ALIGN + GLOBL C_SYMBOL_NAME(mpihelp_add_n) + +C_SYMBOL_NAME(mpihelp_add_n:) +PROLOG(mpihelp_add_n) + /* Save used registers on the stack. */ + movel R(d2),MEM_PREDEC(sp) + movel R(a2),MEM_PREDEC(sp) + + /* Copy the arguments to registers. Better use movem? */ + movel MEM_DISP(sp,12),R(a2) + movel MEM_DISP(sp,16),R(a0) + movel MEM_DISP(sp,20),R(a1) + movel MEM_DISP(sp,24),R(d2) + + eorw #1,R(d2) + lsrl #1,R(d2) + bcc L(L1) + subql #1,R(d2) /* clears cy as side effect */ + +L(Loop:) + movel MEM_POSTINC(a0),R(d0) + movel MEM_POSTINC(a1),R(d1) + addxl R(d1),R(d0) + movel R(d0),MEM_POSTINC(a2) +L(L1:) movel MEM_POSTINC(a0),R(d0) + movel MEM_POSTINC(a1),R(d1) + addxl R(d1),R(d0) + movel R(d0),MEM_POSTINC(a2) + + dbf R(d2),L(Loop) /* loop until 16 lsb of %4 == -1 */ + subxl R(d0),R(d0) /* d0 <= -cy; save cy as 0 or -1 in d0 */ + subl #0x10000,R(d2) + bcs L(L2) + addl R(d0),R(d0) /* restore cy */ + bra L(Loop) + +L(L2:) + negl R(d0) + + /* Restore used registers from stack frame. */ + movel MEM_POSTINC(sp),R(a2) + movel MEM_POSTINC(sp),R(d2) + + rts +EPILOG(mpihelp_add_n) + + diff --git a/mpi/m68k/mpih-lshift.S b/mpi/m68k/mpih-lshift.S new file mode 100644 index 0000000..2256828 --- /dev/null +++ b/mpi/m68k/mpih-lshift.S @@ -0,0 +1,167 @@ +/* mc68020 lshift -- Shift left a low-level natural-number integer. + * + * Copyright (C) 1996, 1998, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_lshift( mpi_ptr_t wp, (sp + 4) + * mpi_ptr_t up, (sp + 8) + * mpi_size_t usize, (sp + 12) + * unsigned cnt) (sp + 16) + */ + +#define res_ptr a1 +#define s_ptr a0 +#define s_size d6 +#define cnt d4 + + TEXT + ALIGN + GLOBL C_SYMBOL_NAME(mpihelp_lshift) + +C_SYMBOL_NAME(mpihelp_lshift:) +PROLOG(mpihelp_lshift) + + /* Save used registers on the stack. */ + moveml R(d2)-R(d6)/R(a2),MEM_PREDEC(sp) + + /* Copy the arguments to registers. */ + movel MEM_DISP(sp,28),R(res_ptr) + movel MEM_DISP(sp,32),R(s_ptr) + movel MEM_DISP(sp,36),R(s_size) + movel MEM_DISP(sp,40),R(cnt) + + moveql #1,R(d5) + cmpl R(d5),R(cnt) + bne L(Lnormal) + cmpl R(s_ptr),R(res_ptr) + bls L(Lspecial) /* jump if s_ptr >= res_ptr */ +#if (defined (__mc68020__) || defined (__NeXT__) || defined(mc68020)) + lea MEM_INDX1(s_ptr,s_size,l,4),R(a2) +#else /* not mc68020 */ + movel R(s_size),R(d0) + asll #2,R(d0) + lea MEM_INDX(s_ptr,d0,l),R(a2) +#endif + cmpl R(res_ptr),R(a2) + bls L(Lspecial) /* jump if res_ptr >= s_ptr + s_size */ + +L(Lnormal:) + moveql #32,R(d5) + subl R(cnt),R(d5) + +#if (defined (__mc68020__) || defined (__NeXT__) || defined(mc68020)) + lea MEM_INDX1(s_ptr,s_size,l,4),R(s_ptr) + lea MEM_INDX1(res_ptr,s_size,l,4),R(res_ptr) +#else /* not mc68000 */ + movel R(s_size),R(d0) + asll #2,R(d0) + addl R(s_size),R(s_ptr) + addl R(s_size),R(res_ptr) +#endif + movel MEM_PREDEC(s_ptr),R(d2) + movel R(d2),R(d0) + lsrl R(d5),R(d0) /* compute carry limb */ + + lsll R(cnt),R(d2) + movel R(d2),R(d1) + subql #1,R(s_size) + beq L(Lend) + lsrl #1,R(s_size) + bcs L(L1) + subql #1,R(s_size) + +L(Loop:) + movel MEM_PREDEC(s_ptr),R(d2) + movel R(d2),R(d3) + lsrl R(d5),R(d3) + orl R(d3),R(d1) + movel R(d1),MEM_PREDEC(res_ptr) + lsll R(cnt),R(d2) +L(L1:) + movel MEM_PREDEC(s_ptr),R(d1) + movel R(d1),R(d3) + lsrl R(d5),R(d3) + orl R(d3),R(d2) + movel R(d2),MEM_PREDEC(res_ptr) + lsll R(cnt),R(d1) + + dbf R(s_size),L(Loop) + subl #0x10000,R(s_size) + bcc L(Loop) + +L(Lend:) + movel R(d1),MEM_PREDEC(res_ptr) /* store least significant limb */ + +/* Restore used registers from stack frame. */ + moveml MEM_POSTINC(sp),R(d2)-R(d6)/R(a2) + rts + +/* We loop from least significant end of the arrays, which is only + permissable if the source and destination don't overlap, since the + function is documented to work for overlapping source and destination. */ + +L(Lspecial:) + clrl R(d0) /* initialize carry */ + eorw #1,R(s_size) + lsrl #1,R(s_size) + bcc L(LL1) + subql #1,R(s_size) + +L(LLoop:) + movel MEM_POSTINC(s_ptr),R(d2) + addxl R(d2),R(d2) + movel R(d2),MEM_POSTINC(res_ptr) +L(LL1:) + movel MEM_POSTINC(s_ptr),R(d2) + addxl R(d2),R(d2) + movel R(d2),MEM_POSTINC(res_ptr) + + dbf R(s_size),L(LLoop) + addxl R(d0),R(d0) /* save cy in lsb */ + subl #0x10000,R(s_size) + bcs L(LLend) + lsrl #1,R(d0) /* restore cy */ + bra L(LLoop) + +L(LLend:) +/* Restore used registers from stack frame. */ + moveml MEM_POSTINC(sp),R(d2)-R(d6)/R(a2) + rts +EPILOG(mpihelp_lshift) + + + + + diff --git a/mpi/m68k/mpih-rshift.S b/mpi/m68k/mpih-rshift.S new file mode 100644 index 0000000..e9c20d2 --- /dev/null +++ b/mpi/m68k/mpih-rshift.S @@ -0,0 +1,165 @@ +/* mc68020 rshift -- Shift right a low-level natural-number integer. + * + * Copyright (C) 1996, 1998, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_rshift( mpi_ptr_t wp, (sp + 4) + * mpi_ptr_t up, (sp + 8) + * mpi_size_t usize, (sp + 12) + * unsigned cnt) (sp + 16) + */ + +#define res_ptr a1 +#define s_ptr a0 +#define s_size d6 +#define cnt d4 + + TEXT + ALIGN + GLOBL C_SYMBOL_NAME(mpihelp_rshift) + +C_SYMBOL_NAME(mpihelp_rshift:) +PROLOG(mpihelp_rshift) + /* Save used registers on the stack. */ + moveml R(d2)-R(d6)/R(a2),MEM_PREDEC(sp) + + /* Copy the arguments to registers. */ + movel MEM_DISP(sp,28),R(res_ptr) + movel MEM_DISP(sp,32),R(s_ptr) + movel MEM_DISP(sp,36),R(s_size) + movel MEM_DISP(sp,40),R(cnt) + + moveql #1,R(d5) + cmpl R(d5),R(cnt) + bne L(Rnormal) + cmpl R(res_ptr),R(s_ptr) + bls L(Rspecial) /* jump if res_ptr >= s_ptr */ +#if (defined (__mc68020__) || defined (__NeXT__) || defined(mc68020)) + lea MEM_INDX1(res_ptr,s_size,l,4),R(a2) +#else /* not mc68020 */ + movel R(s_size),R(d0) + asll #2,R(d0) + lea MEM_INDX(res_ptr,d0,l),R(a2) +#endif + cmpl R(s_ptr),R(a2) + bls L(Rspecial) /* jump if s_ptr >= res_ptr + s_size */ + +L(Rnormal:) + moveql #32,R(d5) + subl R(cnt),R(d5) + movel MEM_POSTINC(s_ptr),R(d2) + movel R(d2),R(d0) + lsll R(d5),R(d0) /* compute carry limb */ + + lsrl R(cnt),R(d2) + movel R(d2),R(d1) + subql #1,R(s_size) + beq L(Rend) + lsrl #1,R(s_size) + bcs L(R1) + subql #1,R(s_size) + +L(Roop:) + movel MEM_POSTINC(s_ptr),R(d2) + movel R(d2),R(d3) + lsll R(d5),R(d3) + orl R(d3),R(d1) + movel R(d1),MEM_POSTINC(res_ptr) + lsrl R(cnt),R(d2) +L(R1:) + movel MEM_POSTINC(s_ptr),R(d1) + movel R(d1),R(d3) + lsll R(d5),R(d3) + orl R(d3),R(d2) + movel R(d2),MEM_POSTINC(res_ptr) + lsrl R(cnt),R(d1) + + dbf R(s_size),L(Roop) + subl #0x10000,R(s_size) + bcc L(Roop) + +L(Rend:) + movel R(d1),MEM(res_ptr) /* store most significant limb */ + +/* Restore used registers from stack frame. */ + moveml MEM_POSTINC(sp),R(d2)-R(d6)/R(a2) + rts + +/* We loop from most significant end of the arrays, which is only + permissable if the source and destination don't overlap, since the + function is documented to work for overlapping source and destination. */ + +L(Rspecial:) +#if (defined (__mc68020__) || defined (__NeXT__) || defined(mc68020)) + lea MEM_INDX1(s_ptr,s_size,l,4),R(s_ptr) + lea MEM_INDX1(res_ptr,s_size,l,4),R(res_ptr) +#else /* not mc68000 */ + movel R(s_size),R(d0) + asll #2,R(d0) + addl R(s_size),R(s_ptr) + addl R(s_size),R(res_ptr) +#endif + + clrl R(d0) /* initialize carry */ + eorw #1,R(s_size) + lsrl #1,R(s_size) + bcc L(LR1) + subql #1,R(s_size) + +L(LRoop:) + movel MEM_PREDEC(s_ptr),R(d2) + roxrl #1,R(d2) + movel R(d2),MEM_PREDEC(res_ptr) +L(LR1:) + movel MEM_PREDEC(s_ptr),R(d2) + roxrl #1,R(d2) + movel R(d2),MEM_PREDEC(res_ptr) + + dbf R(s_size),L(LRoop) + roxrl #1,R(d0) /* save cy in msb */ + subl #0x10000,R(s_size) + bcs L(LRend) + addl R(d0),R(d0) /* restore cy */ + bra L(LRoop) + +L(LRend:) +/* Restore used registers from stack frame. */ + moveml MEM_POSTINC(sp),R(d2)-R(d6)/R(a2) + rts +EPILOG(mpihelp_rshift) + + + + diff --git a/mpi/m68k/mpih-sub1.S b/mpi/m68k/mpih-sub1.S new file mode 100644 index 0000000..b5de0b9 --- /dev/null +++ b/mpi/m68k/mpih-sub1.S @@ -0,0 +1,94 @@ +/* mc68020 __mpn_sub_n -- Subtract two limb vectors of the same length > 0 and + * store difference in a third limb vector. + * + * Copyright (C) 1992, 1994, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + + +/******************* + * mpi_limb_t + * mpihelp_sub_n( mpi_ptr_t res_ptr, (sp + 4) + * mpi_ptr_t s1_ptr, (sp + 8) + * mpi_ptr_t s2_ptr, (sp + 16) + * mpi_size_t size) (sp + 12) + */ + + + TEXT + ALIGN + GLOBL C_SYMBOL_NAME(mpihelp_sub_n) + +C_SYMBOL_NAME(mpihelp_sub_n:) +PROLOG(mpihelp_sub_n) +/* Save used registers on the stack. */ + movel R(d2),MEM_PREDEC(sp) + movel R(a2),MEM_PREDEC(sp) + +/* Copy the arguments to registers. Better use movem? */ + movel MEM_DISP(sp,12),R(a2) + movel MEM_DISP(sp,16),R(a0) + movel MEM_DISP(sp,20),R(a1) + movel MEM_DISP(sp,24),R(d2) + + eorw #1,R(d2) + lsrl #1,R(d2) + bcc L(L1) + subql #1,R(d2) /* clears cy as side effect */ + +L(Loop:) + movel MEM_POSTINC(a0),R(d0) + movel MEM_POSTINC(a1),R(d1) + subxl R(d1),R(d0) + movel R(d0),MEM_POSTINC(a2) +L(L1:) movel MEM_POSTINC(a0),R(d0) + movel MEM_POSTINC(a1),R(d1) + subxl R(d1),R(d0) + movel R(d0),MEM_POSTINC(a2) + + dbf R(d2),L(Loop) /* loop until 16 lsb of %4 == -1 */ + subxl R(d0),R(d0) /* d0 <= -cy; save cy as 0 or -1 in d0 */ + subl #0x10000,R(d2) + bcs L(L2) + addl R(d0),R(d0) /* restore cy */ + bra L(Loop) + +L(L2:) + negl R(d0) + +/* Restore used registers from stack frame. */ + movel MEM_POSTINC(sp),R(a2) + movel MEM_POSTINC(sp),R(d2) + + rts +EPILOG(mpihelp_sub_n) + + diff --git a/mpi/m68k/syntax.h b/mpi/m68k/syntax.h new file mode 100644 index 0000000..ad27a32 --- /dev/null +++ b/mpi/m68k/syntax.h @@ -0,0 +1,188 @@ +/* asm.h -- Definitions for 68k syntax variations. + * + * Copyright (C) 1992, 1994, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +#undef ALIGN + +#ifdef MIT_SYNTAX +#define PROLOG(name) +#define EPILOG(name) +#define R(r)r +#define MEM(base)base@ +#define MEM_DISP(base,displacement)base@(displacement) +#define MEM_INDX(base,idx,size_suffix)base@(idx:size_suffix) +#define MEM_INDX1(base,idx,size_suffix,scale)base@(idx:size_suffix:scale) +#define MEM_PREDEC(memory_base)memory_base@- +#define MEM_POSTINC(memory_base)memory_base@+ +#define L(label) label +#define TEXT .text +#define ALIGN .even +#define GLOBL .globl +#define moveql moveq +/* Use variable sized opcodes. */ +#define bcc jcc +#define bcs jcs +#define bls jls +#define beq jeq +#define bne jne +#define bra jra +#endif + +#ifdef SONY_SYNTAX +#define PROLOG(name) +#define EPILOG(name) +#define R(r)r +#define MEM(base)(base) +#define MEM_DISP(base,displacement)(displacement,base) +#define MEM_INDX(base,idx,size_suffix)(base,idx.size_suffix) +#define MEM_INDX1(base,idx,size_suffix,scale)(base,idx.size_suffix*scale) +#define MEM_PREDEC(memory_base)-(memory_base) +#define MEM_POSTINC(memory_base)(memory_base)+ +#define L(label) label +#define TEXT .text +#define ALIGN .even +#define GLOBL .globl +#endif + +#ifdef MOTOROLA_SYNTAX +#define PROLOG(name) +#define EPILOG(name) +#define R(r)r +#define MEM(base)(base) +#define MEM_DISP(base,displacement)(displacement,base) +#define MEM_INDX(base,idx,size_suffix)(base,idx.size_suffix) +#define MEM_INDX1(base,idx,size_suffix,scale)(base,idx.size_suffix*scale) +#define MEM_PREDEC(memory_base)-(memory_base) +#define MEM_POSTINC(memory_base)(memory_base)+ +#define L(label) label +#define TEXT +#define ALIGN +#define GLOBL XDEF +#define lea LEA +#define movel MOVE.L +#define moveml MOVEM.L +#define moveql MOVEQ.L +#define cmpl CMP.L +#define orl OR.L +#define clrl CLR.L +#define eorw EOR.W +#define lsrl LSR.L +#define lsll LSL.L +#define roxrl ROXR.L +#define roxll ROXL.L +#define addl ADD.L +#define addxl ADDX.L +#define addql ADDQ.L +#define subl SUB.L +#define subxl SUBX.L +#define subql SUBQ.L +#define negl NEG.L +#define mulul MULU.L +#define bcc BCC +#define bcs BCS +#define bls BLS +#define beq BEQ +#define bne BNE +#define bra BRA +#define dbf DBF +#define rts RTS +#define d0 D0 +#define d1 D1 +#define d2 D2 +#define d3 D3 +#define d4 D4 +#define d5 D5 +#define d6 D6 +#define d7 D7 +#define a0 A0 +#define a1 A1 +#define a2 A2 +#define a3 A3 +#define a4 A4 +#define a5 A5 +#define a6 A6 +#define a7 A7 +#define sp SP +#endif + +#ifdef ELF_SYNTAX +#define PROLOG(name) .type name,@function +#define EPILOG(name) .size name,.-name +#define MEM(base)(R(base)) +#define MEM_DISP(base,displacement)(displacement,R(base)) +#define MEM_PREDEC(memory_base)-(R(memory_base)) +#define MEM_POSTINC(memory_base)(R(memory_base))+ +#ifdef __STDC__ +#define R_(r)%##r +#define R(r)R_(r) +#define MEM_INDX_(base,idx,size_suffix)(R(base),R(idx##.##size_suffix)) +#define MEM_INDX(base,idx,size_suffix)MEM_INDX_(base,idx,size_suffix) +#define MEM_INDX1_(base,idx,size_suffix,scale)(R(base),R(idx##.##size_suffix*scale)) +#define MEM_INDX1(base,idx,size_suffix,scale)MEM_INDX1_(base,idx,size_suffix,scale) +#define L(label) .##label +#else +#define R(r)%/**/r +#define MEM_INDX(base,idx,size_suffix)(R(base),R(idx).size_suffix) +#define MEM_INDX1(base,idx,size_suffix,scale)(R(base),R(idx).size_suffix*scale) +#define L(label) ./**/label +#endif +#define TEXT .text +#define ALIGN .align 2 +#define GLOBL .globl +#define bcc jbcc +#define bcs jbcs +#define bls jbls +#define beq jbeq +#define bne jbne +#define bra jbra +#endif + +#if defined (SONY_SYNTAX) || defined (ELF_SYNTAX) +#define movel move.l +#define moveml movem.l +#define moveql moveq.l +#define cmpl cmp.l +#define orl or.l +#define clrl clr.l +#define eorw eor.w +#define lsrl lsr.l +#define lsll lsl.l +#define roxrl roxr.l +#define roxll roxl.l +#define addl add.l +#define addxl addx.l +#define addql addq.l +#define subl sub.l +#define subxl subx.l +#define subql subq.l +#define negl neg.l +#define mulul mulu.l +#endif diff --git a/mpi/mips3/README b/mpi/mips3/README new file mode 100644 index 0000000..e94b2c7 --- /dev/null +++ b/mpi/mips3/README @@ -0,0 +1,23 @@ +This directory contains mpn functions optimized for MIPS3. Example of +processors that implement MIPS3 are R4000, R4400, R4600, R4700, and R8000. + +RELEVANT OPTIMIZATION ISSUES + +1. On the R4000 and R4400, branches, both the plain and the "likely" ones, + take 3 cycles to execute. (The fastest possible loop will take 4 cycles, + because of the delay insn.) + + On the R4600, branches takes a single cycle + + On the R8000, branches often take no noticable cycles, as they are + executed in a separate function unit.. + +2. The R4000 and R4400 have a load latency of 4 cycles. + +3. On the R4000 and R4400, multiplies take a data-dependent number of + cycles, contrary to the SGI documentation. There seem to be 3 or 4 + possible latencies. + +STATUS + +Good... diff --git a/mpi/mips3/distfiles b/mpi/mips3/distfiles new file mode 100644 index 0000000..85260fc --- /dev/null +++ b/mpi/mips3/distfiles @@ -0,0 +1,10 @@ +README +mpih-add1.S +mpih-sub1.S +mpih-mul1.S +mpih-mul2.S +mpih-mul3.S +mpih-lshift.S +mpih-rshift.S +mpi-asm-defs.h + diff --git a/mpi/mips3/mpi-asm-defs.h b/mpi/mips3/mpi-asm-defs.h new file mode 100644 index 0000000..2d9a9c1 --- /dev/null +++ b/mpi/mips3/mpi-asm-defs.h @@ -0,0 +1,10 @@ +/* This file defines some basic constants for the MPI machinery. We + * need to define the types on a per-CPU basis, so it is done with + * this file here. */ +#define BYTES_PER_MPI_LIMB 8 + + + + + + diff --git a/mpi/mips3/mpih-add1.S b/mpi/mips3/mpih-add1.S new file mode 100644 index 0000000..d3ea91d --- /dev/null +++ b/mpi/mips3/mpih-add1.S @@ -0,0 +1,125 @@ +/* mips3 add_n -- Add two limb vectors of the same length > 0 and store + * sum in a third limb vector. + * + * Copyright (C) 1995, 1998, 2000 + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +/******************* + * mpi_limb_t + * mpihelp_add_n( mpi_ptr_t res_ptr, ($4) + * mpi_ptr_t s1_ptr, ($5) + * mpi_ptr_t s2_ptr, ($6) + * mpi_size_t size) ($7) + */ + + .text + .align 2 + .globl mpihelp_add_n + .ent mpihelp_add_n +mpihelp_add_n: + .set noreorder + .set nomacro + + ld $10,0($5) + ld $11,0($6) + + daddiu $7,$7,-1 + and $9,$7,4-1 # number of limbs in first loop + beq $9,$0,.L0 # if multiple of 4 limbs, skip first loop + move $2,$0 + + dsubu $7,$7,$9 + +.Loop0: daddiu $9,$9,-1 + ld $12,8($5) + daddu $11,$11,$2 + ld $13,8($6) + sltu $8,$11,$2 + daddu $11,$10,$11 + sltu $2,$11,$10 + sd $11,0($4) + or $2,$2,$8 + + daddiu $5,$5,8 + daddiu $6,$6,8 + move $10,$12 + move $11,$13 + bne $9,$0,.Loop0 + daddiu $4,$4,8 + +.L0: beq $7,$0,.Lend + nop + +.Loop: daddiu $7,$7,-4 + + ld $12,8($5) + daddu $11,$11,$2 + ld $13,8($6) + sltu $8,$11,$2 + daddu $11,$10,$11 + sltu $2,$11,$10 + sd $11,0($4) + or $2,$2,$8 + + ld $10,16($5) + daddu $13,$13,$2 + ld $11,16($6) + sltu $8,$13,$2 + daddu $13,$12,$13 + sltu $2,$13,$12 + sd $13,8($4) + or $2,$2,$8 + + ld $12,24($5) + daddu $11,$11,$2 + ld $13,24($6) + sltu $8,$11,$2 + daddu $11,$10,$11 + sltu $2,$11,$10 + sd $11,16($4) + or $2,$2,$8 + + ld $10,32($5) + daddu $13,$13,$2 + ld $11,32($6) + sltu $8,$13,$2 + daddu $13,$12,$13 + sltu $2,$13,$12 + sd $13,24($4) + or $2,$2,$8 + + daddiu $5,$5,32 + daddiu $6,$6,32 + + bne $7,$0,.Loop + daddiu $4,$4,32 + +.Lend: daddu $11,$11,$2 + sltu $8,$11,$2 + daddu $11,$10,$11 + sltu $2,$11,$10 + sd $11,0($4) + j $31 + or $2,$2,$8 + + .end mpihelp_add_n + diff --git a/mpi/mips3/mpih-lshift.S b/mpi/mips3/mpih-lshift.S new file mode 100644 index 0000000..7eca648 --- /dev/null +++ b/mpi/mips3/mpih-lshift.S @@ -0,0 +1,98 @@ +/* mips3 lshift + * + * Copyright (C) 1995, 1998, 2000, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +/******************* + * mpi_limb_t + * mpihelp_lshift( mpi_ptr_t wp, ($4) + * mpi_ptr_t up, ($5) + * mpi_size_t usize, ($6) + * unsigned cnt) ($7) + */ + + .text + .align 2 + .globl mpihelp_lshift + .ent mpihelp_lshift +mpihelp_lshift: + .set noreorder + .set nomacro + + dsll $2,$6,3 + daddu $5,$5,$2 # make r5 point at end of src + ld $10,-8($5) # load first limb + dsubu $13,$0,$7 + daddu $4,$4,$2 # make r4 point at end of res + daddiu $6,$6,-1 + and $9,$6,4-1 # number of limbs in first loop + beq $9,$0,.L0 # if multiple of 4 limbs, skip first loop + dsrl $2,$10,$13 # compute function result + + dsubu $6,$6,$9 + +.Loop0: ld $3,-16($5) + daddiu $4,$4,-8 + daddiu $5,$5,-8 + daddiu $9,$9,-1 + dsll $11,$10,$7 + dsrl $12,$3,$13 + move $10,$3 + or $8,$11,$12 + bne $9,$0,.Loop0 + sd $8,0($4) + +.L0: beq $6,$0,.Lend + nop + +.Loop: ld $3,-16($5) + daddiu $4,$4,-32 + daddiu $6,$6,-4 + dsll $11,$10,$7 + dsrl $12,$3,$13 + + ld $10,-24($5) + dsll $14,$3,$7 + or $8,$11,$12 + sd $8,24($4) + dsrl $9,$10,$13 + + ld $3,-32($5) + dsll $11,$10,$7 + or $8,$14,$9 + sd $8,16($4) + dsrl $12,$3,$13 + + ld $10,-40($5) + dsll $14,$3,$7 + or $8,$11,$12 + sd $8,8($4) + dsrl $9,$10,$13 + + daddiu $5,$5,-32 + or $8,$14,$9 + bgtz $6,.Loop + sd $8,0($4) + +.Lend: dsll $8,$10,$7 + j $31 + sd $8,-8($4) + .end mpihelp_lshift diff --git a/mpi/mips3/mpih-mul1.S b/mpi/mips3/mpih-mul1.S new file mode 100644 index 0000000..0562f4f --- /dev/null +++ b/mpi/mips3/mpih-mul1.S @@ -0,0 +1,90 @@ +/* mips3 mpih-mul1.S -- Multiply a limb vector with a limb and store + * the result in a second limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1998, 2000 + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +/******************* + * mpi_limb_t + * mpihelp_mul_1( mpi_ptr_t res_ptr, (r4) + * mpi_ptr_t s1_ptr, (r5) + * mpi_size_t s1_size, (r6) + * mpi_limb_t s2_limb) (r7) + */ + + .text + .align 4 + .globl mpihelp_mul_1 + .ent mpihelp_mul_1 +mpihelp_mul_1: + .set noreorder + .set nomacro + +/* # warm up phase 0 */ + ld $8,0($5) + +/* # warm up phase 1 */ + daddiu $5,$5,8 + dmultu $8,$7 + + daddiu $6,$6,-1 + beq $6,$0,$LC0 + move $2,$0 # zero cy2 + + daddiu $6,$6,-1 + beq $6,$0,$LC1 + ld $8,0($5) # load new s1 limb as early as possible + +Loop: mflo $10 + mfhi $9 + daddiu $5,$5,8 + daddu $10,$10,$2 # add old carry limb to low product limb + dmultu $8,$7 + ld $8,0($5) # load new s1 limb as early as possible + daddiu $6,$6,-1 # decrement loop counter + sltu $2,$10,$2 # carry from previous addition -> $2 + sd $10,0($4) + daddiu $4,$4,8 + bne $6,$0,Loop + daddu $2,$9,$2 # add high product limb and carry from addition + +/* # cool down phase 1 */ +$LC1: mflo $10 + mfhi $9 + daddu $10,$10,$2 + sltu $2,$10,$2 + dmultu $8,$7 + sd $10,0($4) + daddiu $4,$4,8 + daddu $2,$9,$2 # add high product limb and carry from addition + +/* # cool down phase 0 */ +$LC0: mflo $10 + mfhi $9 + daddu $10,$10,$2 + sltu $2,$10,$2 + sd $10,0($4) + j $31 + daddu $2,$9,$2 # add high product limb and carry from addition + + .end mpihelp_mul_1 + diff --git a/mpi/mips3/mpih-mul2.S b/mpi/mips3/mpih-mul2.S new file mode 100644 index 0000000..de56355 --- /dev/null +++ b/mpi/mips3/mpih-mul2.S @@ -0,0 +1,102 @@ +/* MIPS3 addmul_1 -- Multiply a limb vector with a single limb and + * add the product to a second limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1998, 2000 + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +/******************* + * mpi_limb_t + * mpihelp_addmul_1( mpi_ptr_t res_ptr, (r4) + * mpi_ptr_t s1_ptr, (r5) + * mpi_size_t s1_size, (r6) + * mpi_limb_t s2_limb) (r7) + */ + + .text + .align 4 + .globl mpihelp_addmul_1 + .ent mpihelp_addmul_1 +mpihelp_addmul_1: + .set noreorder + .set nomacro + +/* # warm up phase 0 */ + ld $8,0($5) + +/* # warm up phase 1 */ + daddiu $5,$5,8 + dmultu $8,$7 + + daddiu $6,$6,-1 + beq $6,$0,$LC0 + move $2,$0 # zero cy2 + + daddiu $6,$6,-1 + beq $6,$0,$LC1 + ld $8,0($5) # load new s1 limb as early as possible + +Loop: ld $10,0($4) + mflo $3 + mfhi $9 + daddiu $5,$5,8 + daddu $3,$3,$2 # add old carry limb to low product limb + dmultu $8,$7 + ld $8,0($5) # load new s1 limb as early as possible + daddiu $6,$6,-1 # decrement loop counter + sltu $2,$3,$2 # carry from previous addition -> $2 + daddu $3,$10,$3 + sltu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + daddiu $4,$4,8 + bne $6,$0,Loop + daddu $2,$9,$2 # add high product limb and carry from addition + +/* # cool down phase 1 */ +$LC1: ld $10,0($4) + mflo $3 + mfhi $9 + daddu $3,$3,$2 + sltu $2,$3,$2 + dmultu $8,$7 + daddu $3,$10,$3 + sltu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + daddiu $4,$4,8 + daddu $2,$9,$2 # add high product limb and carry from addition + +/* # cool down phase 0 */ +$LC0: ld $10,0($4) + mflo $3 + mfhi $9 + daddu $3,$3,$2 + sltu $2,$3,$2 + daddu $3,$10,$3 + sltu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + j $31 + daddu $2,$9,$2 # add high product limb and carry from addition + + .end mpihelp_addmul_1 + diff --git a/mpi/mips3/mpih-mul3.S b/mpi/mips3/mpih-mul3.S new file mode 100644 index 0000000..960c5c2 --- /dev/null +++ b/mpi/mips3/mpih-mul3.S @@ -0,0 +1,102 @@ +/* MIPS3 submul_1 -- Multiply a limb vector with a single limb and + * subtract the product from a second limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1998, 2000 + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +/******************* + * mpi_limb_t + * mpihelp_submul_1( mpi_ptr_t res_ptr, (r4) + * mpi_ptr_t s1_ptr, (r5) + * mpi_size_t s1_size, (r6) + * mpi_limb_t s2_limb) (r7) + */ + + .text + .align 4 + .globl mpihelp_submul_1 + .ent mpihelp_submul_1 +mpihelp_submul_1: + .set noreorder + .set nomacro + +/* # warm up phase 0 */ + ld $8,0($5) + +/* # warm up phase 1 */ + daddiu $5,$5,8 + dmultu $8,$7 + + daddiu $6,$6,-1 + beq $6,$0,$LC0 + move $2,$0 # zero cy2 + + daddiu $6,$6,-1 + beq $6,$0,$LC1 + ld $8,0($5) # load new s1 limb as early as possible + +Loop: ld $10,0($4) + mflo $3 + mfhi $9 + daddiu $5,$5,8 + daddu $3,$3,$2 # add old carry limb to low product limb + dmultu $8,$7 + ld $8,0($5) # load new s1 limb as early as possible + daddiu $6,$6,-1 # decrement loop counter + sltu $2,$3,$2 # carry from previous addition -> $2 + dsubu $3,$10,$3 + sgtu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + daddiu $4,$4,8 + bne $6,$0,Loop + daddu $2,$9,$2 # add high product limb and carry from addition + +/* # cool down phase 1 */ +$LC1: ld $10,0($4) + mflo $3 + mfhi $9 + daddu $3,$3,$2 + sltu $2,$3,$2 + dmultu $8,$7 + dsubu $3,$10,$3 + sgtu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + daddiu $4,$4,8 + daddu $2,$9,$2 # add high product limb and carry from addition + +/* # cool down phase 0 */ +$LC0: ld $10,0($4) + mflo $3 + mfhi $9 + daddu $3,$3,$2 + sltu $2,$3,$2 + dsubu $3,$10,$3 + sgtu $10,$3,$10 + daddu $2,$2,$10 + sd $3,0($4) + j $31 + daddu $2,$9,$2 # add high product limb and carry from addition + + .end mpihelp_submul_1 + diff --git a/mpi/mips3/mpih-rshift.S b/mpi/mips3/mpih-rshift.S new file mode 100644 index 0000000..2229c50 --- /dev/null +++ b/mpi/mips3/mpih-rshift.S @@ -0,0 +1,96 @@ +/* mips3 rshift + * + * Copyright (C) 1995, 1998, 2000 + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +/******************* + * mpi_limb_t + * mpihelp_rshift( mpi_ptr_t wp, ($4) + * mpi_ptr_t up, ($5) + * mpi_size_t usize, ($6) + * unsigned cnt) ($7) + */ + + .text + .align 2 + .globl mpihelp_rshift + .ent mpihelp_rshift +mpihelp_rshift: + .set noreorder + .set nomacro + + ld $10,0($5) # load first limb + dsubu $13,$0,$7 + daddiu $6,$6,-1 + and $9,$6,4-1 # number of limbs in first loop + beq $9,$0,.L0 # if multiple of 4 limbs, skip first loop + dsll $2,$10,$13 # compute function result + + dsubu $6,$6,$9 + +.Loop0: ld $3,8($5) + daddiu $4,$4,8 + daddiu $5,$5,8 + daddiu $9,$9,-1 + dsrl $11,$10,$7 + dsll $12,$3,$13 + move $10,$3 + or $8,$11,$12 + bne $9,$0,.Loop0 + sd $8,-8($4) + +.L0: beq $6,$0,.Lend + nop + +.Loop: ld $3,8($5) + daddiu $4,$4,32 + daddiu $6,$6,-4 + dsrl $11,$10,$7 + dsll $12,$3,$13 + + ld $10,16($5) + dsrl $14,$3,$7 + or $8,$11,$12 + sd $8,-32($4) + dsll $9,$10,$13 + + ld $3,24($5) + dsrl $11,$10,$7 + or $8,$14,$9 + sd $8,-24($4) + dsll $12,$3,$13 + + ld $10,32($5) + dsrl $14,$3,$7 + or $8,$11,$12 + sd $8,-16($4) + dsll $9,$10,$13 + + daddiu $5,$5,32 + or $8,$14,$9 + bgtz $6,.Loop + sd $8,-8($4) + +.Lend: dsrl $8,$10,$7 + j $31 + sd $8,0($4) + .end mpihelp_rshift + diff --git a/mpi/mips3/mpih-sub1.S b/mpi/mips3/mpih-sub1.S new file mode 100644 index 0000000..d23ff9b --- /dev/null +++ b/mpi/mips3/mpih-sub1.S @@ -0,0 +1,126 @@ +/* mips3 sub_n -- Subtract two limb vectors of the same length > 0 and + * store difference in a third limb vector. + * + * Copyright (C) 1995, 1998, 1999, 2000, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +/******************* + * mpi_limb_t + * mpihelp_sub_n( mpi_ptr_t res_ptr, (r4) + * mpi_ptr_t s1_ptr, (r5) + * mpi_ptr_t s2_ptr, (r6) + * mpi_size_t size) (r7) + */ + + + .text + .align 2 + .globl mpihelp_sub_n + .ent mpihelp_sub_n +mpihelp_sub_n: + .set noreorder + .set nomacro + + ld $10,0($5) + ld $11,0($6) + + daddiu $7,$7,-1 + and $9,$7,4-1 # number of limbs in first loop + beq $9,$0,.L0 # if multiple of 4 limbs, skip first loop + move $2,$0 + + dsubu $7,$7,$9 + +.Loop0: daddiu $9,$9,-1 + ld $12,8($5) + daddu $11,$11,$2 + ld $13,8($6) + sltu $8,$11,$2 + dsubu $11,$10,$11 + sltu $2,$10,$11 + sd $11,0($4) + or $2,$2,$8 + + daddiu $5,$5,8 + daddiu $6,$6,8 + move $10,$12 + move $11,$13 + bne $9,$0,.Loop0 + daddiu $4,$4,8 + +.L0: beq $7,$0,.Lend + nop + +.Loop: daddiu $7,$7,-4 + + ld $12,8($5) + daddu $11,$11,$2 + ld $13,8($6) + sltu $8,$11,$2 + dsubu $11,$10,$11 + sltu $2,$10,$11 + sd $11,0($4) + or $2,$2,$8 + + ld $10,16($5) + daddu $13,$13,$2 + ld $11,16($6) + sltu $8,$13,$2 + dsubu $13,$12,$13 + sltu $2,$12,$13 + sd $13,8($4) + or $2,$2,$8 + + ld $12,24($5) + daddu $11,$11,$2 + ld $13,24($6) + sltu $8,$11,$2 + dsubu $11,$10,$11 + sltu $2,$10,$11 + sd $11,16($4) + or $2,$2,$8 + + ld $10,32($5) + daddu $13,$13,$2 + ld $11,32($6) + sltu $8,$13,$2 + dsubu $13,$12,$13 + sltu $2,$12,$13 + sd $13,24($4) + or $2,$2,$8 + + daddiu $5,$5,32 + daddiu $6,$6,32 + + bne $7,$0,.Loop + daddiu $4,$4,32 + +.Lend: daddu $11,$11,$2 + sltu $8,$11,$2 + dsubu $11,$10,$11 + sltu $2,$10,$11 + sd $11,0($4) + j $31 + or $2,$2,$8 + + .end mpihelp_sub_n + diff --git a/mpi/mpi-add.c b/mpi/mpi-add.c new file mode 100644 index 0000000..d8bd46b --- /dev/null +++ b/mpi/mpi-add.c @@ -0,0 +1,248 @@ +/* mpi-add.c - MPI functions + * Copyright (C) 1998, 1999, 2000, 2001 Free Software Foundation, Inc. + * Copyright (C) 1994, 1996 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> + +#include "mpi-internal.h" + + +/**************** + * Add the unsigned integer V to the mpi-integer U and store the + * result in W. U and V may be the same. + */ +void +mpi_add_ui(MPI w, MPI u, unsigned long v ) +{ + mpi_ptr_t wp, up; + mpi_size_t usize, wsize; + int usign, wsign; + + usize = u->nlimbs; + usign = u->sign; + wsign = 0; + + /* If not space for W (and possible carry), increase space. */ + wsize = usize + 1; + if( w->alloced < wsize ) + mpi_resize(w, wsize); + + /* These must be after realloc (U may be the same as W). */ + up = u->d; + wp = w->d; + + if( !usize ) { /* simple */ + wp[0] = v; + wsize = v? 1:0; + } + else if( !usign ) { /* mpi is not negative */ + mpi_limb_t cy; + cy = mpihelp_add_1(wp, up, usize, v); + wp[usize] = cy; + wsize = usize + cy; + } + else { /* The signs are different. Need exact comparison to determine + * which operand to subtract from which. */ + if( usize == 1 && up[0] < v ) { + wp[0] = v - up[0]; + wsize = 1; + } + else { + mpihelp_sub_1(wp, up, usize, v); + /* Size can decrease with at most one limb. */ + wsize = usize - (wp[usize-1]==0); + wsign = 1; + } + } + + w->nlimbs = wsize; + w->sign = wsign; +} + + +void +mpi_add(MPI w, MPI u, MPI v) +{ + mpi_ptr_t wp, up, vp; + mpi_size_t usize, vsize, wsize; + int usign, vsign, wsign; + + if( u->nlimbs < v->nlimbs ) { /* Swap U and V. */ + usize = v->nlimbs; + usign = v->sign; + vsize = u->nlimbs; + vsign = u->sign; + wsize = usize + 1; + RESIZE_IF_NEEDED(w, wsize); + /* These must be after realloc (u or v may be the same as w). */ + up = v->d; + vp = u->d; + } + else { + usize = u->nlimbs; + usign = u->sign; + vsize = v->nlimbs; + vsign = v->sign; + wsize = usize + 1; + RESIZE_IF_NEEDED(w, wsize); + /* These must be after realloc (u or v may be the same as w). */ + up = u->d; + vp = v->d; + } + wp = w->d; + wsign = 0; + + if( !vsize ) { /* simple */ + MPN_COPY(wp, up, usize ); + wsize = usize; + wsign = usign; + } + else if( usign != vsign ) { /* different sign */ + /* This test is right since USIZE >= VSIZE */ + if( usize != vsize ) { + mpihelp_sub(wp, up, usize, vp, vsize); + wsize = usize; + MPN_NORMALIZE(wp, wsize); + wsign = usign; + } + else if( mpihelp_cmp(up, vp, usize) < 0 ) { + mpihelp_sub_n(wp, vp, up, usize); + wsize = usize; + MPN_NORMALIZE(wp, wsize); + if( !usign ) + wsign = 1; + } + else { + mpihelp_sub_n(wp, up, vp, usize); + wsize = usize; + MPN_NORMALIZE(wp, wsize); + if( usign ) + wsign = 1; + } + } + else { /* U and V have same sign. Add them. */ + mpi_limb_t cy = mpihelp_add(wp, up, usize, vp, vsize); + wp[usize] = cy; + wsize = usize + cy; + if( usign ) + wsign = 1; + } + + w->nlimbs = wsize; + w->sign = wsign; +} + + +/**************** + * Subtract the unsigned integer V from the mpi-integer U and store the + * result in W. + */ +void +mpi_sub_ui(MPI w, MPI u, unsigned long v ) +{ + mpi_ptr_t wp, up; + mpi_size_t usize, wsize; + int usign, wsign; + + usize = u->nlimbs; + usign = u->sign; + wsign = 0; + + /* If not space for W (and possible carry), increase space. */ + wsize = usize + 1; + if( w->alloced < wsize ) + mpi_resize(w, wsize); + + /* These must be after realloc (U may be the same as W). */ + up = u->d; + wp = w->d; + + if( !usize ) { /* simple */ + wp[0] = v; + wsize = v? 1:0; + wsign = 1; + } + else if( usign ) { /* mpi and v are negative */ + mpi_limb_t cy; + cy = mpihelp_add_1(wp, up, usize, v); + wp[usize] = cy; + wsize = usize + cy; + } + else { /* The signs are different. Need exact comparison to determine + * which operand to subtract from which. */ + if( usize == 1 && up[0] < v ) { + wp[0] = v - up[0]; + wsize = 1; + wsign = 1; + } + else { + mpihelp_sub_1(wp, up, usize, v); + /* Size can decrease with at most one limb. */ + wsize = usize - (wp[usize-1]==0); + } + } + + w->nlimbs = wsize; + w->sign = wsign; +} + +void +mpi_sub(MPI w, MPI u, MPI v) +{ + if( w == v ) { + MPI vv = mpi_copy(v); + vv->sign = !vv->sign; + mpi_add( w, u, vv ); + mpi_free(vv); + } + else { + /* fixme: this is not thread-save (we temp. modify v) */ + v->sign = !v->sign; + mpi_add( w, u, v ); + v->sign = !v->sign; + } +} + + +void +mpi_addm( MPI w, MPI u, MPI v, MPI m) +{ + mpi_add(w, u, v); + mpi_fdiv_r( w, w, m ); +} + +void +mpi_subm( MPI w, MPI u, MPI v, MPI m) +{ + mpi_sub(w, u, v); + mpi_fdiv_r( w, w, m ); +} + diff --git a/mpi/mpi-bit.c b/mpi/mpi-bit.c new file mode 100644 index 0000000..da9d009 --- /dev/null +++ b/mpi/mpi-bit.c @@ -0,0 +1,256 @@ +/* mpi-bit.c - MPI bit level fucntions + * Copyright (C) 1998, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include <assert.h> +#include "mpi-internal.h" +#include "longlong.h" + + +#ifdef MPI_INTERNAL_NEED_CLZ_TAB +#ifdef __STDC__ +const +#endif +unsigned char +__clz_tab[] = +{ + 0,1,2,2,3,3,3,3,4,4,4,4,4,4,4,4,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5, + 6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6, + 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, + 7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, + 8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8, +}; +#endif + + +#define A_LIMB_1 ((mpi_limb_t)1) + + +/**************** + * Sometimes we have MSL (most significant limbs) which are 0; + * this is for some reasons not good, so this function removes them. + */ +void +mpi_normalize( MPI a ) +{ + if( mpi_is_opaque (a) ) + return; + + for( ; a->nlimbs && !a->d[a->nlimbs-1]; a->nlimbs-- ) + ; +} + + + +/**************** + * Return the number of bits in A. + */ +unsigned +mpi_get_nbits( MPI a ) +{ + unsigned n; + + mpi_normalize( a ); + if( a->nlimbs ) { + mpi_limb_t alimb = a->d[a->nlimbs-1]; + if( alimb ) + count_leading_zeros( n, alimb ); + else + n = BITS_PER_MPI_LIMB; + n = BITS_PER_MPI_LIMB - n + (a->nlimbs-1) * BITS_PER_MPI_LIMB; + } + else + n = 0; + return n; +} + + +/**************** + * Test whether bit N is set. + */ +int +mpi_test_bit( MPI a, unsigned n ) +{ + unsigned limbno, bitno; + mpi_limb_t limb; + + limbno = n / BITS_PER_MPI_LIMB; + bitno = n % BITS_PER_MPI_LIMB; + + if( limbno >= a->nlimbs ) + return 0; /* too far left: this is a 0 */ + limb = a->d[limbno]; + return (limb & (A_LIMB_1 << bitno))? 1: 0; +} + + +/**************** + * Set bit N of A. + */ +void +mpi_set_bit( MPI a, unsigned n ) +{ + unsigned limbno, bitno; + + limbno = n / BITS_PER_MPI_LIMB; + bitno = n % BITS_PER_MPI_LIMB; + + if( limbno >= a->nlimbs ) { /* resize */ + if( a->alloced >= limbno ) + mpi_resize(a, limbno+1 ); + a->nlimbs = limbno+1; + } + a->d[limbno] |= (A_LIMB_1<<bitno); +} + +/**************** + * Set bit N of A. and clear all bits above + */ +void +mpi_set_highbit( MPI a, unsigned n ) +{ + unsigned limbno, bitno; + + limbno = n / BITS_PER_MPI_LIMB; + bitno = n % BITS_PER_MPI_LIMB; + + if( limbno >= a->nlimbs ) { /* resize */ + if( a->alloced >= limbno ) + mpi_resize(a, limbno+1 ); + a->nlimbs = limbno+1; + } + a->d[limbno] |= (A_LIMB_1<<bitno); + for( bitno++; bitno < BITS_PER_MPI_LIMB; bitno++ ) + a->d[limbno] &= ~(A_LIMB_1 << bitno); + a->nlimbs = limbno+1; +} + +/**************** + * clear bit N of A and all bits above + */ +void +mpi_clear_highbit( MPI a, unsigned n ) +{ + unsigned limbno, bitno; + + limbno = n / BITS_PER_MPI_LIMB; + bitno = n % BITS_PER_MPI_LIMB; + + if( limbno >= a->nlimbs ) + return; /* not allocated, so need to clear bits :-) */ + + for( ; bitno < BITS_PER_MPI_LIMB; bitno++ ) + a->d[limbno] &= ~(A_LIMB_1 << bitno); + a->nlimbs = limbno+1; +} + +/**************** + * Clear bit N of A. + */ +void +mpi_clear_bit( MPI a, unsigned n ) +{ + unsigned limbno, bitno; + + limbno = n / BITS_PER_MPI_LIMB; + bitno = n % BITS_PER_MPI_LIMB; + + if( limbno >= a->nlimbs ) + return; /* don't need to clear this bit, it's to far to left */ + a->d[limbno] &= ~(A_LIMB_1 << bitno); +} + + +/**************** + * Shift A by N bits to the right + * FIXME: should use alloc_limb if X and A are same. + */ +void +mpi_rshift( MPI x, MPI a, unsigned n ) +{ + mpi_ptr_t xp; + mpi_size_t xsize; + + xsize = a->nlimbs; + x->sign = a->sign; + RESIZE_IF_NEEDED(x, xsize); + xp = x->d; + + if( xsize ) { + mpihelp_rshift( xp, a->d, xsize, n); + MPN_NORMALIZE( xp, xsize); + } + x->nlimbs = xsize; +} + + +/**************** + * Shift A by COUNT limbs to the left + * This is used only within the MPI library + */ +void +mpi_lshift_limbs( MPI a, unsigned int count ) +{ + mpi_ptr_t ap = a->d; + int n = a->nlimbs; + int i; + + if( !count || !n ) + return; + + RESIZE_IF_NEEDED( a, n+count ); + + for( i = n-1; i >= 0; i-- ) + ap[i+count] = ap[i]; + for(i=0; i < count; i++ ) + ap[i] = 0; + a->nlimbs += count; +} + + +/**************** + * Shift A by COUNT limbs to the right + * This is used only within the MPI library + */ +void +mpi_rshift_limbs( MPI a, unsigned int count ) +{ + mpi_ptr_t ap = a->d; + mpi_size_t n = a->nlimbs; + unsigned int i; + + if( count >= n ) { + a->nlimbs = 0; + return; + } + + for( i = 0; i < n - count; i++ ) + ap[i] = ap[i+count]; + ap[i] = 0; + a->nlimbs -= count; +} + + diff --git a/mpi/mpi-cmp.c b/mpi/mpi-cmp.c new file mode 100644 index 0000000..083edbf --- /dev/null +++ b/mpi/mpi-cmp.c @@ -0,0 +1,75 @@ +/* mpi-cmp.c - MPI functions + * Copyright (C) 1998, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" + +int +mpi_cmp_ui( MPI u, unsigned long v ) +{ + mpi_limb_t limb = v; + + mpi_normalize( u ); + if( !u->nlimbs && !limb ) + return 0; + if( u->sign ) + return -1; + if( u->nlimbs > 1 ) + return 1; + + if( u->d[0] == limb ) + return 0; + else if( u->d[0] > limb ) + return 1; + else + return -1; +} + +int +mpi_cmp( MPI u, MPI v ) +{ + mpi_size_t usize, vsize; + int cmp; + + mpi_normalize( u ); + mpi_normalize( v ); + usize = u->nlimbs; + vsize = v->nlimbs; + if( !u->sign && v->sign ) + return 1; + if( u->sign && !v->sign ) + return -1; + if( usize != vsize && !u->sign && !v->sign ) + return usize - vsize; + if( usize != vsize && u->sign && v->sign ) + return vsize + usize; + if( !usize ) + return 0; + if( !(cmp=mpihelp_cmp( u->d, v->d, usize )) ) + return 0; + if( (cmp < 0?1:0) == (u->sign?1:0)) + return 1; + return -1; +} + + diff --git a/mpi/mpi-div.c b/mpi/mpi-div.c new file mode 100644 index 0000000..f29493e --- /dev/null +++ b/mpi/mpi-div.c @@ -0,0 +1,323 @@ +/* mpi-div.c - MPI functions + * Copyright (C) 1994, 1996 Free Software Foundation, Inc. + * Copyright (C) 1998, 1999, 2000, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" +#include "longlong.h" + + + +void +mpi_fdiv_r( MPI rem, MPI dividend, MPI divisor ) +{ + int divisor_sign = divisor->sign; + MPI temp_divisor = NULL; + + /* We need the original value of the divisor after the remainder has been + * preliminary calculated. We have to copy it to temporary space if it's + * the same variable as REM. */ + if( rem == divisor ) { + temp_divisor = mpi_copy( divisor ); + divisor = temp_divisor; + } + + mpi_tdiv_r( rem, dividend, divisor ); + + if( ((divisor_sign?1:0) ^ (dividend->sign?1:0)) && rem->nlimbs ) + mpi_add( rem, rem, divisor); + + if( temp_divisor ) + mpi_free(temp_divisor); +} + + + +/**************** + * Division rounding the quotient towards -infinity. + * The remainder gets the same sign as the denominator. + * rem is optional + */ + +ulong +mpi_fdiv_r_ui( MPI rem, MPI dividend, ulong divisor ) +{ + mpi_limb_t rlimb; + + rlimb = mpihelp_mod_1( dividend->d, dividend->nlimbs, divisor ); + if( rlimb && dividend->sign ) + rlimb = divisor - rlimb; + + if( rem ) { + rem->d[0] = rlimb; + rem->nlimbs = rlimb? 1:0; + } + return rlimb; +} + + +void +mpi_fdiv_q( MPI quot, MPI dividend, MPI divisor ) +{ + MPI tmp = mpi_alloc( mpi_get_nlimbs(quot) ); + mpi_fdiv_qr( quot, tmp, dividend, divisor); + mpi_free(tmp); +} + +void +mpi_fdiv_qr( MPI quot, MPI rem, MPI dividend, MPI divisor ) +{ + int divisor_sign = divisor->sign; + MPI temp_divisor = NULL; + + if( quot == divisor || rem == divisor ) { + temp_divisor = mpi_copy( divisor ); + divisor = temp_divisor; + } + + mpi_tdiv_qr( quot, rem, dividend, divisor ); + + if( (divisor_sign ^ dividend->sign) && rem->nlimbs ) { + mpi_sub_ui( quot, quot, 1 ); + mpi_add( rem, rem, divisor); + } + + if( temp_divisor ) + mpi_free(temp_divisor); +} + + +/* If den == quot, den needs temporary storage. + * If den == rem, den needs temporary storage. + * If num == quot, num needs temporary storage. + * If den has temporary storage, it can be normalized while being copied, + * i.e no extra storage should be allocated. + */ + +void +mpi_tdiv_r( MPI rem, MPI num, MPI den) +{ + mpi_tdiv_qr(NULL, rem, num, den ); +} + +void +mpi_tdiv_qr( MPI quot, MPI rem, MPI num, MPI den) +{ + mpi_ptr_t np, dp; + mpi_ptr_t qp, rp; + mpi_size_t nsize = num->nlimbs; + mpi_size_t dsize = den->nlimbs; + mpi_size_t qsize, rsize; + mpi_size_t sign_remainder = num->sign; + mpi_size_t sign_quotient = num->sign ^ den->sign; + unsigned normalization_steps; + mpi_limb_t q_limb; + mpi_ptr_t marker[5]; + int markidx=0; + + /* Ensure space is enough for quotient and remainder. + * We need space for an extra limb in the remainder, because it's + * up-shifted (normalized) below. */ + rsize = nsize + 1; + mpi_resize( rem, rsize); + + qsize = rsize - dsize; /* qsize cannot be bigger than this. */ + if( qsize <= 0 ) { + if( num != rem ) { + rem->nlimbs = num->nlimbs; + rem->sign = num->sign; + MPN_COPY(rem->d, num->d, nsize); + } + if( quot ) { + /* This needs to follow the assignment to rem, in case the + * numerator and quotient are the same. */ + quot->nlimbs = 0; + quot->sign = 0; + } + return; + } + + if( quot ) + mpi_resize( quot, qsize); + + /* Read pointers here, when reallocation is finished. */ + np = num->d; + dp = den->d; + rp = rem->d; + + /* Optimize division by a single-limb divisor. */ + if( dsize == 1 ) { + mpi_limb_t rlimb; + if( quot ) { + qp = quot->d; + rlimb = mpihelp_divmod_1( qp, np, nsize, dp[0] ); + qsize -= qp[qsize - 1] == 0; + quot->nlimbs = qsize; + quot->sign = sign_quotient; + } + else + rlimb = mpihelp_mod_1( np, nsize, dp[0] ); + rp[0] = rlimb; + rsize = rlimb != 0?1:0; + rem->nlimbs = rsize; + rem->sign = sign_remainder; + return; + } + + + if( quot ) { + qp = quot->d; + /* Make sure QP and NP point to different objects. Otherwise the + * numerator would be gradually overwritten by the quotient limbs. */ + if(qp == np) { /* Copy NP object to temporary space. */ + np = marker[markidx++] = mpi_alloc_limb_space(nsize, + mpi_is_secure(quot)); + MPN_COPY(np, qp, nsize); + } + } + else /* Put quotient at top of remainder. */ + qp = rp + dsize; + + count_leading_zeros( normalization_steps, dp[dsize - 1] ); + + /* Normalize the denominator, i.e. make its most significant bit set by + * shifting it NORMALIZATION_STEPS bits to the left. Also shift the + * numerator the same number of steps (to keep the quotient the same!). + */ + if( normalization_steps ) { + mpi_ptr_t tp; + mpi_limb_t nlimb; + + /* Shift up the denominator setting the most significant bit of + * the most significant word. Use temporary storage not to clobber + * the original contents of the denominator. */ + tp = marker[markidx++] = mpi_alloc_limb_space(dsize,mpi_is_secure(den)); + mpihelp_lshift( tp, dp, dsize, normalization_steps ); + dp = tp; + + /* Shift up the numerator, possibly introducing a new most + * significant word. Move the shifted numerator in the remainder + * meanwhile. */ + nlimb = mpihelp_lshift(rp, np, nsize, normalization_steps); + if( nlimb ) { + rp[nsize] = nlimb; + rsize = nsize + 1; + } + else + rsize = nsize; + } + else { + /* The denominator is already normalized, as required. Copy it to + * temporary space if it overlaps with the quotient or remainder. */ + if( dp == rp || (quot && (dp == qp))) { + mpi_ptr_t tp; + + tp = marker[markidx++] = mpi_alloc_limb_space(dsize, mpi_is_secure(den)); + MPN_COPY( tp, dp, dsize ); + dp = tp; + } + + /* Move the numerator to the remainder. */ + if( rp != np ) + MPN_COPY(rp, np, nsize); + + rsize = nsize; + } + + q_limb = mpihelp_divrem( qp, 0, rp, rsize, dp, dsize ); + + if( quot ) { + qsize = rsize - dsize; + if(q_limb) { + qp[qsize] = q_limb; + qsize += 1; + } + + quot->nlimbs = qsize; + quot->sign = sign_quotient; + } + + rsize = dsize; + MPN_NORMALIZE (rp, rsize); + + if( normalization_steps && rsize ) { + mpihelp_rshift(rp, rp, rsize, normalization_steps); + rsize -= rp[rsize - 1] == 0?1:0; + } + + rem->nlimbs = rsize; + rem->sign = sign_remainder; + while( markidx ) + mpi_free_limb_space(marker[--markidx]); +} + +void +mpi_tdiv_q_2exp( MPI w, MPI u, unsigned count ) +{ + mpi_size_t usize, wsize; + mpi_size_t limb_cnt; + + usize = u->nlimbs; + limb_cnt = count / BITS_PER_MPI_LIMB; + wsize = usize - limb_cnt; + if( limb_cnt >= usize ) + w->nlimbs = 0; + else { + mpi_ptr_t wp; + mpi_ptr_t up; + + RESIZE_IF_NEEDED( w, wsize ); + wp = w->d; + up = u->d; + + count %= BITS_PER_MPI_LIMB; + if( count ) { + mpihelp_rshift( wp, up + limb_cnt, wsize, count ); + wsize -= !wp[wsize - 1]; + } + else { + MPN_COPY_INCR( wp, up + limb_cnt, wsize); + } + + w->nlimbs = wsize; + } +} + +/**************** + * Check whether dividend is divisible by divisor + * (note: divisor must fit into a limb) + */ +int +mpi_divisible_ui(MPI dividend, ulong divisor ) +{ + return !mpihelp_mod_1( dividend->d, dividend->nlimbs, divisor ); +} + diff --git a/mpi/mpi-gcd.c b/mpi/mpi-gcd.c new file mode 100644 index 0000000..6f384ce --- /dev/null +++ b/mpi/mpi-gcd.c @@ -0,0 +1,55 @@ +/* mpi-gcd.c - MPI functions + * Copyright (C) 1998, 1999, 2000, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" + +/**************** + * Find the greatest common divisor G of A and B. + * Return: true if this 1, false in all other cases + */ +int +mpi_gcd( MPI g, MPI xa, MPI xb ) +{ + MPI a, b; + + a = mpi_copy(xa); + b = mpi_copy(xb); + + /* TAOCP Vol II, 4.5.2, Algorithm A */ + a->sign = 0; + b->sign = 0; + while( mpi_cmp_ui( b, 0 ) ) { + mpi_fdiv_r( g, a, b ); /* g used as temorary variable */ + mpi_set(a,b); + mpi_set(b,g); + } + mpi_set(g, a); + + mpi_free(a); + mpi_free(b); + return !mpi_cmp_ui( g, 1); +} + + + diff --git a/mpi/mpi-inline.c b/mpi/mpi-inline.c new file mode 100644 index 0000000..8e29b06 --- /dev/null +++ b/mpi/mpi-inline.c @@ -0,0 +1,37 @@ +/* mpi-inline.c + * Copyright (C) 1999, 2000, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> + +/* put the inline functions as real functions into the lib */ +#define G10_MPI_INLINE_DECL + +#include "mpi-internal.h" + +/* always include the header becuase it is only + * included by mpi-internal if __GCC__ is defined but we + * need it here in all cases and the above definition of + * of the macro allows us to do so + */ +#include "mpi-inline.h" + diff --git a/mpi/mpi-inline.h b/mpi/mpi-inline.h new file mode 100644 index 0000000..db2c6a9 --- /dev/null +++ b/mpi/mpi-inline.h @@ -0,0 +1,128 @@ +/* mpi-inline.h - Internal to the Multi Precision Integers + * Copyright (C) 1994, 1996, 1998, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#ifndef G10_MPI_INLINE_H +#define G10_MPI_INLINE_H + +#ifndef G10_MPI_INLINE_DECL +#define G10_MPI_INLINE_DECL extern __inline__ +#endif + +G10_MPI_INLINE_DECL mpi_limb_t +mpihelp_add_1( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_size_t s1_size, mpi_limb_t s2_limb) +{ + mpi_limb_t x; + + x = *s1_ptr++; + s2_limb += x; + *res_ptr++ = s2_limb; + if( s2_limb < x ) { /* sum is less than the left operand: handle carry */ + while( --s1_size ) { + x = *s1_ptr++ + 1; /* add carry */ + *res_ptr++ = x; /* and store */ + if( x ) /* not 0 (no overflow): we can stop */ + goto leave; + } + return 1; /* return carry (size of s1 to small) */ + } + + leave: + if( res_ptr != s1_ptr ) { /* not the same variable */ + mpi_size_t i; /* copy the rest */ + for( i=0; i < s1_size-1; i++ ) + res_ptr[i] = s1_ptr[i]; + } + return 0; /* no carry */ +} + + + +G10_MPI_INLINE_DECL mpi_limb_t +mpihelp_add(mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, mpi_size_t s1_size, + mpi_ptr_t s2_ptr, mpi_size_t s2_size) +{ + mpi_limb_t cy = 0; + + if( s2_size ) + cy = mpihelp_add_n( res_ptr, s1_ptr, s2_ptr, s2_size ); + + if( s1_size - s2_size ) + cy = mpihelp_add_1( res_ptr + s2_size, s1_ptr + s2_size, + s1_size - s2_size, cy); + return cy; +} + + +G10_MPI_INLINE_DECL mpi_limb_t +mpihelp_sub_1(mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_size_t s1_size, mpi_limb_t s2_limb ) +{ + mpi_limb_t x; + + x = *s1_ptr++; + s2_limb = x - s2_limb; + *res_ptr++ = s2_limb; + if( s2_limb > x ) { + while( --s1_size ) { + x = *s1_ptr++; + *res_ptr++ = x - 1; + if( x ) + goto leave; + } + return 1; + } + + leave: + if( res_ptr != s1_ptr ) { + mpi_size_t i; + for( i=0; i < s1_size-1; i++ ) + res_ptr[i] = s1_ptr[i]; + } + return 0; +} + + + +G10_MPI_INLINE_DECL mpi_limb_t +mpihelp_sub( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, mpi_size_t s1_size, + mpi_ptr_t s2_ptr, mpi_size_t s2_size) +{ + mpi_limb_t cy = 0; + + if( s2_size ) + cy = mpihelp_sub_n(res_ptr, s1_ptr, s2_ptr, s2_size); + + if( s1_size - s2_size ) + cy = mpihelp_sub_1(res_ptr + s2_size, s1_ptr + s2_size, + s1_size - s2_size, cy); + return cy; +} + +#endif /*G10_MPI_INLINE_H*/ diff --git a/mpi/mpi-internal.h b/mpi/mpi-internal.h new file mode 100644 index 0000000..4676fcd --- /dev/null +++ b/mpi/mpi-internal.h @@ -0,0 +1,260 @@ +/* mpi-internal.h - Internal to the Multi Precision Integers + * Copyright (C) 1994, 1996 Free Software Foundation, Inc. + * Copyright (C) 1998, 2000 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#ifndef G10_MPI_INTERNAL_H +#define G10_MPI_INTERNAL_H + +#include "mpi.h" + +/* If KARATSUBA_THRESHOLD is not already defined, define it to a + * value which is good on most machines. */ + +/* tested 4, 16, 32 and 64, where 16 gave the best performance when + * checking a 768 and a 1024 bit ElGamal signature. + * (wk 22.12.97) */ +#ifndef KARATSUBA_THRESHOLD +#define KARATSUBA_THRESHOLD 16 +#endif + +/* The code can't handle KARATSUBA_THRESHOLD smaller than 2. */ +#if KARATSUBA_THRESHOLD < 2 +#undef KARATSUBA_THRESHOLD +#define KARATSUBA_THRESHOLD 2 +#endif + + +typedef mpi_limb_t *mpi_ptr_t; /* pointer to a limb */ +typedef int mpi_size_t; /* (must be a signed type) */ + +#define ABS(x) (x >= 0 ? x : -x) +#define MIN(l,o) ((l) < (o) ? (l) : (o)) +#define MAX(h,i) ((h) > (i) ? (h) : (i)) +#define RESIZE_IF_NEEDED(a,b) \ + do { \ + if( (a)->alloced < (b) ) \ + mpi_resize((a), (b)); \ + } while(0) + +/* Copy N limbs from S to D. */ +#define MPN_COPY( d, s, n) \ + do { \ + mpi_size_t _i; \ + for( _i = 0; _i < (n); _i++ ) \ + (d)[_i] = (s)[_i]; \ + } while(0) + +#define MPN_COPY_INCR( d, s, n) \ + do { \ + mpi_size_t _i; \ + for( _i = 0; _i < (n); _i++ ) \ + (d)[_i] = (d)[_i]; \ + } while (0) + +#define MPN_COPY_DECR( d, s, n ) \ + do { \ + mpi_size_t _i; \ + for( _i = (n)-1; _i >= 0; _i--) \ + (d)[_i] = (s)[_i]; \ + } while(0) + +/* Zero N limbs at D */ +#define MPN_ZERO(d, n) \ + do { \ + int _i; \ + for( _i = 0; _i < (n); _i++ ) \ + (d)[_i] = 0; \ + } while (0) + +#define MPN_NORMALIZE(d, n) \ + do { \ + while( (n) > 0 ) { \ + if( (d)[(n)-1] ) \ + break; \ + (n)--; \ + } \ + } while(0) + +#define MPN_NORMALIZE_NOT_ZERO(d, n) \ + do { \ + for(;;) { \ + if( (d)[(n)-1] ) \ + break; \ + (n)--; \ + } \ + } while(0) + +#define MPN_MUL_N_RECURSE(prodp, up, vp, size, tspace) \ + do { \ + if( (size) < KARATSUBA_THRESHOLD ) \ + mul_n_basecase (prodp, up, vp, size); \ + else \ + mul_n (prodp, up, vp, size, tspace); \ + } while (0); + + +/* Divide the two-limb number in (NH,,NL) by D, with DI being the largest + * limb not larger than (2**(2*BITS_PER_MP_LIMB))/D - (2**BITS_PER_MP_LIMB). + * If this would yield overflow, DI should be the largest possible number + * (i.e., only ones). For correct operation, the most significant bit of D + * has to be set. Put the quotient in Q and the remainder in R. + */ +#define UDIV_QRNND_PREINV(q, r, nh, nl, d, di) \ + do { \ + mpi_limb_t _q, _ql, _r; \ + mpi_limb_t _xh, _xl; \ + umul_ppmm (_q, _ql, (nh), (di)); \ + _q += (nh); /* DI is 2**BITS_PER_MPI_LIMB too small */ \ + umul_ppmm (_xh, _xl, _q, (d)); \ + sub_ddmmss (_xh, _r, (nh), (nl), _xh, _xl); \ + if( _xh ) { \ + sub_ddmmss (_xh, _r, _xh, _r, 0, (d)); \ + _q++; \ + if( _xh) { \ + sub_ddmmss (_xh, _r, _xh, _r, 0, (d)); \ + _q++; \ + } \ + } \ + if( _r >= (d) ) { \ + _r -= (d); \ + _q++; \ + } \ + (r) = _r; \ + (q) = _q; \ + } while (0) + + +/*-- mpiutil.c --*/ +#ifdef M_DEBUG +#define mpi_alloc_limb_space(n,f) mpi_debug_alloc_limb_space((n),(f), M_DBGINFO( __LINE__ ) ) +#define mpi_free_limb_space(n) mpi_debug_free_limb_space((n), M_DBGINFO( __LINE__ ) ) + mpi_ptr_t mpi_debug_alloc_limb_space( unsigned nlimbs, int sec, const char *info ); + void mpi_debug_free_limb_space( mpi_ptr_t a, const char *info ); +#else + mpi_ptr_t mpi_alloc_limb_space( unsigned nlimbs, int sec ); + void mpi_free_limb_space( mpi_ptr_t a ); +#endif +void mpi_assign_limb_space( MPI a, mpi_ptr_t ap, unsigned nlimbs ); + +/*-- mpi-bit.c --*/ +void mpi_rshift_limbs( MPI a, unsigned int count ); +void mpi_lshift_limbs( MPI a, unsigned int count ); + + +/*-- mpihelp-add.c --*/ +mpi_limb_t mpihelp_add_1(mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_size_t s1_size, mpi_limb_t s2_limb ); +mpi_limb_t mpihelp_add_n( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_ptr_t s2_ptr, mpi_size_t size); +mpi_limb_t mpihelp_add(mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, mpi_size_t s1_size, + mpi_ptr_t s2_ptr, mpi_size_t s2_size); + +/*-- mpihelp-sub.c --*/ +mpi_limb_t mpihelp_sub_1( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_size_t s1_size, mpi_limb_t s2_limb ); +mpi_limb_t mpihelp_sub_n( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_ptr_t s2_ptr, mpi_size_t size); +mpi_limb_t mpihelp_sub(mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, mpi_size_t s1_size, + mpi_ptr_t s2_ptr, mpi_size_t s2_size); + +/*-- mpihelp-cmp.c --*/ +int mpihelp_cmp( mpi_ptr_t op1_ptr, mpi_ptr_t op2_ptr, mpi_size_t size ); + +/*-- mpihelp-mul.c --*/ + +struct karatsuba_ctx { + struct karatsuba_ctx *next; + mpi_ptr_t tspace; + mpi_size_t tspace_size; + mpi_ptr_t tp; + mpi_size_t tp_size; +}; + +void mpihelp_release_karatsuba_ctx( struct karatsuba_ctx *ctx ); + +mpi_limb_t mpihelp_addmul_1( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_size_t s1_size, mpi_limb_t s2_limb); +mpi_limb_t mpihelp_submul_1( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_size_t s1_size, mpi_limb_t s2_limb); +void mpihelp_mul_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_ptr_t vp, + mpi_size_t size); +mpi_limb_t mpihelp_mul( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t usize, + mpi_ptr_t vp, mpi_size_t vsize); +void mpih_sqr_n_basecase( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size ); +void mpih_sqr_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size, + mpi_ptr_t tspace); + +void mpihelp_mul_karatsuba_case( mpi_ptr_t prodp, + mpi_ptr_t up, mpi_size_t usize, + mpi_ptr_t vp, mpi_size_t vsize, + struct karatsuba_ctx *ctx ); + + +/*-- mpihelp-mul_1.c (or xxx/cpu/ *.S) --*/ +mpi_limb_t mpihelp_mul_1( mpi_ptr_t res_ptr, mpi_ptr_t s1_ptr, + mpi_size_t s1_size, mpi_limb_t s2_limb); + +/*-- mpihelp-div.c --*/ +mpi_limb_t mpihelp_mod_1(mpi_ptr_t dividend_ptr, mpi_size_t dividend_size, + mpi_limb_t divisor_limb); +mpi_limb_t mpihelp_divrem( mpi_ptr_t qp, mpi_size_t qextra_limbs, + mpi_ptr_t np, mpi_size_t nsize, + mpi_ptr_t dp, mpi_size_t dsize); +mpi_limb_t mpihelp_divmod_1( mpi_ptr_t quot_ptr, + mpi_ptr_t dividend_ptr, mpi_size_t dividend_size, + mpi_limb_t divisor_limb); + +/*-- mpihelp-shift.c --*/ +mpi_limb_t mpihelp_lshift( mpi_ptr_t wp, mpi_ptr_t up, mpi_size_t usize, + unsigned cnt); +mpi_limb_t mpihelp_rshift( mpi_ptr_t wp, mpi_ptr_t up, mpi_size_t usize, + unsigned cnt); + + +/* Define stuff for longlong.h. */ +#define W_TYPE_SIZE BITS_PER_MPI_LIMB + typedef mpi_limb_t UWtype; + typedef unsigned int UHWtype; +#if defined (__GNUC__) + typedef unsigned int UQItype __attribute__ ((mode (QI))); + typedef int SItype __attribute__ ((mode (SI))); + typedef unsigned int USItype __attribute__ ((mode (SI))); + typedef int DItype __attribute__ ((mode (DI))); + typedef unsigned int UDItype __attribute__ ((mode (DI))); +#else + typedef unsigned char UQItype; + typedef long SItype; + typedef unsigned long USItype; +#endif + +#ifdef __GNUC__ +#include "mpi-inline.h" +#endif + +#endif /*G10_MPI_INTERNAL_H*/ diff --git a/mpi/mpi-inv.c b/mpi/mpi-inv.c new file mode 100644 index 0000000..bc4d589 --- /dev/null +++ b/mpi/mpi-inv.c @@ -0,0 +1,268 @@ +/* mpi-inv.c - MPI functions + * Copyright (C) 1998, 1999, 2000, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" + + +/**************** + * Calculate the multiplicative inverse X of A mod N + * That is: Find the solution x for + * 1 = (a*x) mod n + */ +void +mpi_invm( MPI x, MPI a, MPI n ) +{ +#if 0 + MPI u, v, u1, u2, u3, v1, v2, v3, q, t1, t2, t3; + MPI ta, tb, tc; + + u = mpi_copy(a); + v = mpi_copy(n); + u1 = mpi_alloc_set_ui(1); + u2 = mpi_alloc_set_ui(0); + u3 = mpi_copy(u); + v1 = mpi_alloc_set_ui(0); + v2 = mpi_alloc_set_ui(1); + v3 = mpi_copy(v); + q = mpi_alloc( mpi_get_nlimbs(u)+1 ); + t1 = mpi_alloc( mpi_get_nlimbs(u)+1 ); + t2 = mpi_alloc( mpi_get_nlimbs(u)+1 ); + t3 = mpi_alloc( mpi_get_nlimbs(u)+1 ); + while( mpi_cmp_ui( v3, 0 ) ) { + mpi_fdiv_q( q, u3, v3 ); + mpi_mul(t1, v1, q); mpi_mul(t2, v2, q); mpi_mul(t3, v3, q); + mpi_sub(t1, u1, t1); mpi_sub(t2, u2, t2); mpi_sub(t3, u3, t3); + mpi_set(u1, v1); mpi_set(u2, v2); mpi_set(u3, v3); + mpi_set(v1, t1); mpi_set(v2, t2); mpi_set(v3, t3); + } + /* log_debug("result:\n"); + log_mpidump("q =", q ); + log_mpidump("u1=", u1); + log_mpidump("u2=", u2); + log_mpidump("u3=", u3); + log_mpidump("v1=", v1); + log_mpidump("v2=", v2); */ + mpi_set(x, u1); + + mpi_free(u1); + mpi_free(u2); + mpi_free(u3); + mpi_free(v1); + mpi_free(v2); + mpi_free(v3); + mpi_free(q); + mpi_free(t1); + mpi_free(t2); + mpi_free(t3); + mpi_free(u); + mpi_free(v); +#elif 0 + /* Extended Euclid's algorithm (See TAOPC Vol II, 4.5.2, Alg X) + * modified according to Michael Penk's solution for Exercice 35 */ + + /* FIXME: we can simplify this in most cases (see Knuth) */ + MPI u, v, u1, u2, u3, v1, v2, v3, t1, t2, t3; + unsigned k; + int sign; + + u = mpi_copy(a); + v = mpi_copy(n); + for(k=0; !mpi_test_bit(u,0) && !mpi_test_bit(v,0); k++ ) { + mpi_rshift(u, u, 1); + mpi_rshift(v, v, 1); + } + + + u1 = mpi_alloc_set_ui(1); + u2 = mpi_alloc_set_ui(0); + u3 = mpi_copy(u); + v1 = mpi_copy(v); /* !-- used as const 1 */ + v2 = mpi_alloc( mpi_get_nlimbs(u) ); mpi_sub( v2, u1, u ); + v3 = mpi_copy(v); + if( mpi_test_bit(u, 0) ) { /* u is odd */ + t1 = mpi_alloc_set_ui(0); + t2 = mpi_alloc_set_ui(1); t2->sign = 1; + t3 = mpi_copy(v); t3->sign = !t3->sign; + goto Y4; + } + else { + t1 = mpi_alloc_set_ui(1); + t2 = mpi_alloc_set_ui(0); + t3 = mpi_copy(u); + } + do { + do { + if( mpi_test_bit(t1, 0) || mpi_test_bit(t2, 0) ) { /* one is odd */ + mpi_add(t1, t1, v); + mpi_sub(t2, t2, u); + } + mpi_rshift(t1, t1, 1); + mpi_rshift(t2, t2, 1); + mpi_rshift(t3, t3, 1); + Y4: + ; + } while( !mpi_test_bit( t3, 0 ) ); /* while t3 is even */ + + if( !t3->sign ) { + mpi_set(u1, t1); + mpi_set(u2, t2); + mpi_set(u3, t3); + } + else { + mpi_sub(v1, v, t1); + sign = u->sign; u->sign = !u->sign; + mpi_sub(v2, u, t2); + u->sign = sign; + sign = t3->sign; t3->sign = !t3->sign; + mpi_set(v3, t3); + t3->sign = sign; + } + mpi_sub(t1, u1, v1); + mpi_sub(t2, u2, v2); + mpi_sub(t3, u3, v3); + if( t1->sign ) { + mpi_add(t1, t1, v); + mpi_sub(t2, t2, u); + } + } while( mpi_cmp_ui( t3, 0 ) ); /* while t3 != 0 */ + /* mpi_lshift( u3, k ); */ + mpi_set(x, u1); + + mpi_free(u1); + mpi_free(u2); + mpi_free(u3); + mpi_free(v1); + mpi_free(v2); + mpi_free(v3); + mpi_free(t1); + mpi_free(t2); + mpi_free(t3); +#else + /* Extended Euclid's algorithm (See TAOPC Vol II, 4.5.2, Alg X) + * modified according to Michael Penk's solution for Exercice 35 + * with further enhancement */ + MPI u, v, u1, u2=NULL, u3, v1, v2=NULL, v3, t1, t2=NULL, t3; + unsigned k; + int sign; + int odd ; + + u = mpi_copy(a); + v = mpi_copy(n); + + for(k=0; !mpi_test_bit(u,0) && !mpi_test_bit(v,0); k++ ) { + mpi_rshift(u, u, 1); + mpi_rshift(v, v, 1); + } + odd = mpi_test_bit(v,0); + + u1 = mpi_alloc_set_ui(1); + if( !odd ) + u2 = mpi_alloc_set_ui(0); + u3 = mpi_copy(u); + v1 = mpi_copy(v); + if( !odd ) { + v2 = mpi_alloc( mpi_get_nlimbs(u) ); + mpi_sub( v2, u1, u ); /* U is used as const 1 */ + } + v3 = mpi_copy(v); + if( mpi_test_bit(u, 0) ) { /* u is odd */ + t1 = mpi_alloc_set_ui(0); + if( !odd ) { + t2 = mpi_alloc_set_ui(1); t2->sign = 1; + } + t3 = mpi_copy(v); t3->sign = !t3->sign; + goto Y4; + } + else { + t1 = mpi_alloc_set_ui(1); + if( !odd ) + t2 = mpi_alloc_set_ui(0); + t3 = mpi_copy(u); + } + do { + do { + if( !odd ) { + if( mpi_test_bit(t1, 0) || mpi_test_bit(t2, 0) ) { /* one is odd */ + mpi_add(t1, t1, v); + mpi_sub(t2, t2, u); + } + mpi_rshift(t1, t1, 1); + mpi_rshift(t2, t2, 1); + mpi_rshift(t3, t3, 1); + } + else { + if( mpi_test_bit(t1, 0) ) + mpi_add(t1, t1, v); + mpi_rshift(t1, t1, 1); + mpi_rshift(t3, t3, 1); + } + Y4: + ; + } while( !mpi_test_bit( t3, 0 ) ); /* while t3 is even */ + + if( !t3->sign ) { + mpi_set(u1, t1); + if( !odd ) + mpi_set(u2, t2); + mpi_set(u3, t3); + } + else { + mpi_sub(v1, v, t1); + sign = u->sign; u->sign = !u->sign; + if( !odd ) + mpi_sub(v2, u, t2); + u->sign = sign; + sign = t3->sign; t3->sign = !t3->sign; + mpi_set(v3, t3); + t3->sign = sign; + } + mpi_sub(t1, u1, v1); + if( !odd ) + mpi_sub(t2, u2, v2); + mpi_sub(t3, u3, v3); + if( t1->sign ) { + mpi_add(t1, t1, v); + if( !odd ) + mpi_sub(t2, t2, u); + } + } while( mpi_cmp_ui( t3, 0 ) ); /* while t3 != 0 */ + /* mpi_lshift( u3, k ); */ + mpi_set(x, u1); + + mpi_free(u1); + mpi_free(v1); + mpi_free(t1); + if( !odd ) { + mpi_free(u2); + mpi_free(v2); + mpi_free(t2); + } + mpi_free(u3); + mpi_free(v3); + mpi_free(t3); + + mpi_free(u); + mpi_free(v); +#endif +} diff --git a/mpi/mpi-mpow.c b/mpi/mpi-mpow.c new file mode 100644 index 0000000..9a18610 --- /dev/null +++ b/mpi/mpi-mpow.c @@ -0,0 +1,100 @@ +/* mpi-mpow.c - MPI functions + * Copyright (C) 1998, 1999, 2000 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" +#include "longlong.h" +#include <assert.h> + +static int +build_index( MPI *exparray, int k, int i, int t ) +{ + int j, bitno; + int idx = 0; + + bitno = t-i; + for(j=k-1; j >= 0; j-- ) { + idx <<= 1; + if( mpi_test_bit( exparray[j], bitno ) ) + idx |= 1; + } + return idx; +} + +/**************** + * RES = (BASE[0] ^ EXP[0]) * (BASE[1] ^ EXP[1]) * ... * mod M + */ +void +mpi_mulpowm( MPI res, MPI *basearray, MPI *exparray, MPI m) +{ + int k; /* number of elements */ + int t; /* bit size of largest exponent */ + int i, j, idx; + MPI *G; /* table with precomputed values of size 2^k */ + MPI tmp; + + for(k=0; basearray[k]; k++ ) + ; + assert(k); + for(t=0, i=0; (tmp=exparray[i]); i++ ) { + j = mpi_get_nbits(tmp); + if( j > t ) + t = j; + } + assert(i==k); + assert(t); + assert( k < 10 ); + + G = xmalloc_clear( (1<<k) * sizeof *G ); + /* and calculate */ + tmp = mpi_alloc( mpi_get_nlimbs(m)+1 ); + mpi_set_ui( res, 1 ); + for(i = 1; i <= t; i++ ) { + mpi_mulm(tmp, res, res, m ); + idx = build_index( exparray, k, i, t ); + assert( idx >= 0 && idx < (1<<k) ); + if( !G[idx] ) { + if( !idx ) + G[0] = mpi_alloc_set_ui( 1 ); + else { + for(j=0; j < k; j++ ) { + if( (idx & (1<<j) ) ) { + if( !G[idx] ) + G[idx] = mpi_copy( basearray[j] ); + else + mpi_mulm( G[idx], G[idx], basearray[j], m ); + } + } + if( !G[idx] ) + G[idx] = mpi_alloc(0); + } + } + mpi_mulm(res, tmp, G[idx], m ); + } + + /* cleanup */ + mpi_free(tmp); + for(i=0; i < (1<<k); i++ ) + mpi_free(G[i]); + xfree(G); +} diff --git a/mpi/mpi-mul.c b/mpi/mpi-mul.c new file mode 100644 index 0000000..97e08ec --- /dev/null +++ b/mpi/mpi-mul.c @@ -0,0 +1,216 @@ +/* mpi-mul.c - MPI functions + * Copyright (C) 1994, 1996 Free Software Foundation, Inc. + * Copyright (C) 1998, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" + + +void +mpi_mul_ui( MPI prod, MPI mult, unsigned long small_mult ) +{ + mpi_size_t size, prod_size; + mpi_ptr_t prod_ptr; + mpi_limb_t cy; + int sign; + + size = mult->nlimbs; + sign = mult->sign; + + if( !size || !small_mult ) { + prod->nlimbs = 0; + prod->sign = 0; + return; + } + + prod_size = size + 1; + if( prod->alloced < prod_size ) + mpi_resize( prod, prod_size ); + prod_ptr = prod->d; + + cy = mpihelp_mul_1( prod_ptr, mult->d, size, (mpi_limb_t)small_mult ); + if( cy ) + prod_ptr[size++] = cy; + prod->nlimbs = size; + prod->sign = sign; +} + + +void +mpi_mul_2exp( MPI w, MPI u, unsigned long cnt) +{ + mpi_size_t usize, wsize, limb_cnt; + mpi_ptr_t wp; + mpi_limb_t wlimb; + int usign, wsign; + + usize = u->nlimbs; + usign = u->sign; + + if( !usize ) { + w->nlimbs = 0; + w->sign = 0; + return; + } + + limb_cnt = cnt / BITS_PER_MPI_LIMB; + wsize = usize + limb_cnt + 1; + if( w->alloced < wsize ) + mpi_resize(w, wsize ); + wp = w->d; + wsize = usize + limb_cnt; + wsign = usign; + + cnt %= BITS_PER_MPI_LIMB; + if( cnt ) { + wlimb = mpihelp_lshift( wp + limb_cnt, u->d, usize, cnt ); + if( wlimb ) { + wp[wsize] = wlimb; + wsize++; + } + } + else { + MPN_COPY_DECR( wp + limb_cnt, u->d, usize ); + } + + /* Zero all whole limbs at low end. Do it here and not before calling + * mpn_lshift, not to lose for U == W. */ + MPN_ZERO( wp, limb_cnt ); + + w->nlimbs = wsize; + w->sign = wsign; +} + + + +void +mpi_mul( MPI w, MPI u, MPI v) +{ + mpi_size_t usize, vsize, wsize; + mpi_ptr_t up, vp, wp; + mpi_limb_t cy; + int usign, vsign, usecure, vsecure, sign_product; + int assign_wp=0; + mpi_ptr_t tmp_limb=NULL; + + + if( u->nlimbs < v->nlimbs ) { /* Swap U and V. */ + usize = v->nlimbs; + usign = v->sign; + usecure = mpi_is_secure(v); + up = v->d; + vsize = u->nlimbs; + vsign = u->sign; + vsecure = mpi_is_secure(u); + vp = u->d; + } + else { + usize = u->nlimbs; + usign = u->sign; + usecure = mpi_is_secure(u); + up = u->d; + vsize = v->nlimbs; + vsign = v->sign; + vsecure = mpi_is_secure(v); + vp = v->d; + } + sign_product = usign ^ vsign; + wp = w->d; + + /* Ensure W has space enough to store the result. */ + wsize = usize + vsize; + if ( !mpi_is_secure (w) && (mpi_is_secure (u) || mpi_is_secure (v)) ) { + /* w is not allocated in secure space but u or v is. To make sure + * that no temporray results are stored in w, we temporary use + * a newly allocated limb space for w */ + wp = mpi_alloc_limb_space( wsize, 1 ); + assign_wp = 2; /* mark it as 2 so that we can later copy it back to + * mormal memory */ + } + else if( w->alloced < wsize ) { + if( wp == up || wp == vp ) { + wp = mpi_alloc_limb_space( wsize, mpi_is_secure(w) ); + assign_wp = 1; + } + else { + mpi_resize(w, wsize ); + wp = w->d; + } + } + else { /* Make U and V not overlap with W. */ + if( wp == up ) { + /* W and U are identical. Allocate temporary space for U. */ + up = tmp_limb = mpi_alloc_limb_space( usize, usecure ); + /* Is V identical too? Keep it identical with U. */ + if( wp == vp ) + vp = up; + /* Copy to the temporary space. */ + MPN_COPY( up, wp, usize ); + } + else if( wp == vp ) { + /* W and V are identical. Allocate temporary space for V. */ + vp = tmp_limb = mpi_alloc_limb_space( vsize, vsecure ); + /* Copy to the temporary space. */ + MPN_COPY( vp, wp, vsize ); + } + } + + if( !vsize ) + wsize = 0; + else { + cy = mpihelp_mul( wp, up, usize, vp, vsize ); + wsize -= cy? 0:1; + } + + if( assign_wp ) { + if (assign_wp == 2) { + /* copy the temp wp from secure memory back to normal memory */ + mpi_ptr_t tmp_wp = mpi_alloc_limb_space (wsize, 0); + MPN_COPY (tmp_wp, wp, wsize); + mpi_free_limb_space (wp); + wp = tmp_wp; + } + mpi_assign_limb_space( w, wp, wsize ); + } + w->nlimbs = wsize; + w->sign = sign_product; + if( tmp_limb ) + mpi_free_limb_space( tmp_limb ); +} + + +void +mpi_mulm( MPI w, MPI u, MPI v, MPI m) +{ + mpi_mul(w, u, v); + mpi_fdiv_r( w, w, m ); +} + diff --git a/mpi/mpi-pow.c b/mpi/mpi-pow.c new file mode 100644 index 0000000..8fcce5c --- /dev/null +++ b/mpi/mpi-pow.c @@ -0,0 +1,296 @@ +/* mpi-pow.c - MPI functions + * Copyright (C) 1994, 1996, 1998, 2000 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include "mpi-internal.h" +#include "longlong.h" +#include <assert.h> + + +/**************** + * RES = BASE ^ EXP mod MOD + */ +void +mpi_powm( MPI res, MPI base, MPI exponent, MPI mod) +{ + mpi_ptr_t rp, ep, mp, bp; + mpi_size_t esize, msize, bsize, rsize; + int esign, msign, bsign, rsign; + int esec, msec, bsec, rsec; + mpi_size_t size; + int mod_shift_cnt; + int negative_result; + mpi_ptr_t mp_marker=NULL, bp_marker=NULL, ep_marker=NULL; + mpi_ptr_t xp_marker=NULL; + int assign_rp=0; + mpi_ptr_t tspace = NULL; + mpi_size_t tsize=0; /* to avoid compiler warning */ + /* fixme: we should check that the warning is void*/ + + esize = exponent->nlimbs; + msize = mod->nlimbs; + size = 2 * msize; + esign = exponent->sign; + msign = mod->sign; + + esec = mpi_is_secure(exponent); + msec = mpi_is_secure(mod); + bsec = mpi_is_secure(base); + rsec = mpi_is_secure(res); + + rp = res->d; + ep = exponent->d; + + if( !msize ) + msize = 1 / msize; /* provoke a signal */ + + if( !esize ) { + /* Exponent is zero, result is 1 mod MOD, i.e., 1 or 0 + * depending on if MOD equals 1. */ + rp[0] = 1; + res->nlimbs = (msize == 1 && mod->d[0] == 1) ? 0 : 1; + res->sign = 0; + goto leave; + } + + /* Normalize MOD (i.e. make its most significant bit set) as required by + * mpn_divrem. This will make the intermediate values in the calculation + * slightly larger, but the correct result is obtained after a final + * reduction using the original MOD value. */ + mp = mp_marker = mpi_alloc_limb_space(msize, msec); + count_leading_zeros( mod_shift_cnt, mod->d[msize-1] ); + if( mod_shift_cnt ) + mpihelp_lshift( mp, mod->d, msize, mod_shift_cnt ); + else + MPN_COPY( mp, mod->d, msize ); + + bsize = base->nlimbs; + bsign = base->sign; + if( bsize > msize ) { /* The base is larger than the module. Reduce it. */ + /* Allocate (BSIZE + 1) with space for remainder and quotient. + * (The quotient is (bsize - msize + 1) limbs.) */ + bp = bp_marker = mpi_alloc_limb_space( bsize + 1, bsec ); + MPN_COPY( bp, base->d, bsize ); + /* We don't care about the quotient, store it above the remainder, + * at BP + MSIZE. */ + mpihelp_divrem( bp + msize, 0, bp, bsize, mp, msize ); + bsize = msize; + /* Canonicalize the base, since we are going to multiply with it + * quite a few times. */ + MPN_NORMALIZE( bp, bsize ); + } + else + bp = base->d; + + if( !bsize ) { + res->nlimbs = 0; + res->sign = 0; + goto leave; + } + + if( res->alloced < size ) { + /* We have to allocate more space for RES. If any of the input + * parameters are identical to RES, defer deallocation of the old + * space. */ + if( rp == ep || rp == mp || rp == bp ) { + rp = mpi_alloc_limb_space( size, rsec ); + assign_rp = 1; + } + else { + mpi_resize( res, size ); + rp = res->d; + } + } + else { /* Make BASE, EXPONENT and MOD not overlap with RES. */ + if( rp == bp ) { + /* RES and BASE are identical. Allocate temp. space for BASE. */ + assert( !bp_marker ); + bp = bp_marker = mpi_alloc_limb_space( bsize, bsec ); + MPN_COPY(bp, rp, bsize); + } + if( rp == ep ) { + /* RES and EXPONENT are identical. + Allocate temp. space for EXPONENT. */ + ep = ep_marker = mpi_alloc_limb_space( esize, esec ); + MPN_COPY(ep, rp, esize); + } + if( rp == mp ) { + /* RES and MOD are identical. Allocate temporary space for MOD.*/ + assert( !mp_marker ); + mp = mp_marker = mpi_alloc_limb_space( msize, msec ); + MPN_COPY(mp, rp, msize); + } + } + + MPN_COPY( rp, bp, bsize ); + rsize = bsize; + rsign = bsign; + + { + mpi_size_t i; + mpi_ptr_t xp = xp_marker = mpi_alloc_limb_space( 2 * (msize + 1), msec ); + int c; + mpi_limb_t e; + mpi_limb_t carry_limb; + struct karatsuba_ctx karactx; + + memset( &karactx, 0, sizeof karactx ); + negative_result = (ep[0] & 1) && base->sign; + + i = esize - 1; + e = ep[i]; + count_leading_zeros (c, e); + e = (e << c) << 1; /* shift the exp bits to the left, lose msb */ + c = BITS_PER_MPI_LIMB - 1 - c; + + /* Main loop. + * + * Make the result be pointed to alternately by XP and RP. This + * helps us avoid block copying, which would otherwise be necessary + * with the overlap restrictions of mpihelp_divmod. With 50% probability + * the result after this loop will be in the area originally pointed + * by RP (==RES->d), and with 50% probability in the area originally + * pointed to by XP. + */ + + for(;;) { + while( c ) { + mpi_ptr_t tp; + mpi_size_t xsize; + + /*mpihelp_mul_n(xp, rp, rp, rsize);*/ + if( rsize < KARATSUBA_THRESHOLD ) + mpih_sqr_n_basecase( xp, rp, rsize ); + else { + if( !tspace ) { + tsize = 2 * rsize; + tspace = mpi_alloc_limb_space( tsize, 0 ); + } + else if( tsize < (2*rsize) ) { + mpi_free_limb_space( tspace ); + tsize = 2 * rsize; + tspace = mpi_alloc_limb_space( tsize, 0 ); + } + mpih_sqr_n( xp, rp, rsize, tspace ); + } + + xsize = 2 * rsize; + if( xsize > msize ) { + mpihelp_divrem(xp + msize, 0, xp, xsize, mp, msize); + xsize = msize; + } + + tp = rp; rp = xp; xp = tp; + rsize = xsize; + + if( (mpi_limb_signed_t)e < 0 ) { + /*mpihelp_mul( xp, rp, rsize, bp, bsize );*/ + if( bsize < KARATSUBA_THRESHOLD ) { + mpihelp_mul( xp, rp, rsize, bp, bsize ); + } + else { + mpihelp_mul_karatsuba_case( + xp, rp, rsize, bp, bsize, &karactx ); + } + + xsize = rsize + bsize; + if( xsize > msize ) { + mpihelp_divrem(xp + msize, 0, xp, xsize, mp, msize); + xsize = msize; + } + + tp = rp; rp = xp; xp = tp; + rsize = xsize; + } + e <<= 1; + c--; + } + + i--; + if( i < 0 ) + break; + e = ep[i]; + c = BITS_PER_MPI_LIMB; + } + + /* We shifted MOD, the modulo reduction argument, left MOD_SHIFT_CNT + * steps. Adjust the result by reducing it with the original MOD. + * + * Also make sure the result is put in RES->d (where it already + * might be, see above). + */ + if( mod_shift_cnt ) { + carry_limb = mpihelp_lshift( res->d, rp, rsize, mod_shift_cnt); + rp = res->d; + if( carry_limb ) { + rp[rsize] = carry_limb; + rsize++; + } + } + else { + MPN_COPY( res->d, rp, rsize); + rp = res->d; + } + + if( rsize >= msize ) { + mpihelp_divrem(rp + msize, 0, rp, rsize, mp, msize); + rsize = msize; + } + + /* Remove any leading zero words from the result. */ + if( mod_shift_cnt ) + mpihelp_rshift( rp, rp, rsize, mod_shift_cnt); + MPN_NORMALIZE (rp, rsize); + + mpihelp_release_karatsuba_ctx( &karactx ); + } + + if( negative_result && rsize ) { + if( mod_shift_cnt ) + mpihelp_rshift( mp, mp, msize, mod_shift_cnt); + mpihelp_sub( rp, mp, msize, rp, rsize); + rsize = msize; + rsign = msign; + MPN_NORMALIZE(rp, rsize); + } + res->nlimbs = rsize; + res->sign = rsign; + + leave: + if( assign_rp ) mpi_assign_limb_space( res, rp, size ); + if( mp_marker ) mpi_free_limb_space( mp_marker ); + if( bp_marker ) mpi_free_limb_space( bp_marker ); + if( ep_marker ) mpi_free_limb_space( ep_marker ); + if( xp_marker ) mpi_free_limb_space( xp_marker ); + if( tspace ) mpi_free_limb_space( tspace ); +} + diff --git a/mpi/mpi-scan.c b/mpi/mpi-scan.c new file mode 100644 index 0000000..c76f4bb --- /dev/null +++ b/mpi/mpi-scan.c @@ -0,0 +1,140 @@ +/* mpi-scan.c - MPI functions + * Copyright (C) 1998, 1999, 2000, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" +#include "longlong.h" + +/**************** + * Scan through an mpi and return byte for byte. a -1 is returned to indicate + * the end of the mpi. Scanning is done from the lsb to the msb, returned + * values are in the range of 0 .. 255. + * + * FIXME: This code is VERY ugly! + */ +#if 0 /* Code is not used */ +int +mpi_getbyte( MPI a, unsigned idx ) +{ + int i, j; + unsigned n; + mpi_ptr_t ap; + mpi_limb_t limb; + + ap = a->d; + for(n=0,i=0; i < a->nlimbs; i++ ) { + limb = ap[i]; + for( j=0; j < BYTES_PER_MPI_LIMB; j++, n++ ) + if( n == idx ) + return (limb >> j*8) & 0xff; + } + return -1; +} +#endif /* Code is not used */ + + +/**************** + * Put a value at position IDX into A. idx counts from lsb to msb + */ +/* FIXME: There is a problem with the long constants which should have +a LL prefix or better the macros we use at other places. */ +#if 0 /* Code is not used */ +void +mpi_putbyte( MPI a, unsigned idx, int xc ) +{ + + int i, j; + unsigned n; + mpi_ptr_t ap; + mpi_limb_t limb, c; + + c = xc & 0xff; + ap = a->d; + for(n=0,i=0; i < a->alloced; i++ ) { + limb = ap[i]; + for( j=0; j < BYTES_PER_MPI_LIMB; j++, n++ ) + if( n == idx ) { +#if BYTES_PER_MPI_LIMB == 4 + if( j == 0 ) + limb = (limb & 0xffffff00) | c; + else if( j == 1 ) + limb = (limb & 0xffff00ff) | (c<<8); + else if( j == 2 ) + limb = (limb & 0xff00ffff) | (c<<16); + else + limb = (limb & 0x00ffffff) | (c<<24); +#elif BYTES_PER_MPI_LIMB == 8 + if( j == 0 ) + limb = (limb & 0xffffffffffffff00) | c; + else if( j == 1 ) + limb = (limb & 0xffffffffffff00ff) | (c<<8); + else if( j == 2 ) + limb = (limb & 0xffffffffff00ffff) | (c<<16); + else if( j == 3 ) + limb = (limb & 0xffffffff00ffffff) | (c<<24); + else if( j == 4 ) + limb = (limb & 0xffffff00ffffffff) | (c<<32); + else if( j == 5 ) + limb = (limb & 0xffff00ffffffffff) | (c<<40); + else if( j == 6 ) + limb = (limb & 0xff00ffffffffffff) | (c<<48); + else + limb = (limb & 0x00ffffffffffffff) | (c<<56); +#else +#error please enhance this function, its ugly - i know. +#endif + if( a->nlimbs <= i ) + a->nlimbs = i+1; + ap[i] = limb; + return; + } + } + abort(); /* index out of range */ +} +#endif /* Code is not used */ + + +/**************** + * Count the number of zerobits at the low end of A + */ +unsigned int +mpi_trailing_zeros( MPI a ) +{ + unsigned n, count = 0; + + for(n=0; n < a->nlimbs; n++ ) { + if( a->d[n] ) { + unsigned nn; + mpi_limb_t alimb = a->d[n]; + + count_trailing_zeros( nn, alimb ); + count += nn; + break; + } + count += BITS_PER_MPI_LIMB; + } + return count; + +} + + diff --git a/mpi/mpicoder.c b/mpi/mpicoder.c new file mode 100644 index 0000000..d5ce999 --- /dev/null +++ b/mpi/mpicoder.c @@ -0,0 +1,474 @@ +/* mpicoder.c - Coder for the external representation of MPIs + * Copyright (C) 1998, 1999, 2005 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include <config.h> +#include <stdio.h> +#include <string.h> +#include <stdlib.h> +#include <assert.h> + +#include "mpi.h" +#include "mpi-internal.h" +#include "iobuf.h" +#include "memory.h" +#include "util.h" + +#ifdef M_DEBUG +#undef mpi_read +#endif + +#define MAX_EXTERN_MPI_BITS 16384 + +/**************** + * write an mpi to out. + */ +int +mpi_write( IOBUF out, MPI a ) +{ + int rc; + unsigned nbits = mpi_get_nbits(a); + byte *p, *buf; + unsigned n; + + if( nbits > MAX_EXTERN_MPI_BITS ) + log_bug("mpi_encode: mpi too large (%u bits)\n", nbits); + + iobuf_put(out, (nbits >>8) ); + iobuf_put(out, (nbits) ); + + p = buf = mpi_get_buffer( a, &n, NULL ); + rc = iobuf_write( out, p, n ); + xfree(buf); + return rc; +} + + +/**************** + * Read an external representation of an mpi and return the MPI + * The external format is a 16 bit unsigned value stored in network byte order, + * giving the number of bits for the following integer. The integer is stored + * with MSB first (left padded with zeroes to align on a byte boundary). + */ +MPI +#ifdef M_DEBUG +mpi_debug_read(IOBUF inp, unsigned *ret_nread, int secure, const char *info) +#else +mpi_read(IOBUF inp, unsigned *ret_nread, int secure) +#endif +{ + int c, i, j; + unsigned int nmax = *ret_nread; + unsigned nbits, nbytes, nlimbs, nread=0; + mpi_limb_t a; + MPI val = MPI_NULL; + + if (nread == nmax) + goto overflow; + if( (c = iobuf_get(inp)) == -1 ) + goto leave; + nread++; + nbits = c << 8; + + if (nread == nmax) + goto overflow; + if( (c = iobuf_get(inp)) == -1 ) + goto leave; + nread++; + nbits |= c; + + if( nbits > MAX_EXTERN_MPI_BITS ) { + log_error("mpi too large for this implementation (%u bits)\n", nbits); + goto leave; + } + + nbytes = (nbits+7) / 8; + nlimbs = (nbytes+BYTES_PER_MPI_LIMB-1) / BYTES_PER_MPI_LIMB; +#ifdef M_DEBUG + val = secure? mpi_debug_alloc_secure( nlimbs, info ) + : mpi_debug_alloc( nlimbs, info ); +#else + val = secure? mpi_alloc_secure( nlimbs ) + : mpi_alloc( nlimbs ); +#endif + i = BYTES_PER_MPI_LIMB - nbytes % BYTES_PER_MPI_LIMB; + i %= BYTES_PER_MPI_LIMB; + val->nbits = nbits; + j= val->nlimbs = nlimbs; + val->sign = 0; + for( ; j > 0; j-- ) { + a = 0; + for(; i < BYTES_PER_MPI_LIMB; i++ ) { + if (nread == nmax) { +#ifdef M_DEBUG + mpi_debug_free (val); +#else + mpi_free (val); +#endif + val = NULL; + goto overflow; + } + a <<= 8; + a |= iobuf_get(inp) & 0xff; nread++; + } + i = 0; + val->d[j-1] = a; + } + + leave: + *ret_nread = nread; + return val; + overflow: + log_error ("mpi larger than indicated length (%u bytes)\n", nmax); + *ret_nread = nread; + return val; +} + + +MPI +mpi_read_from_buffer(byte *buffer, unsigned int *ret_nread, int secure) +{ + int i, j; + unsigned nbits, nbytes, nlimbs, nread=0; + mpi_limb_t a; + MPI val = MPI_NULL; + + if( *ret_nread < 2 ) + goto leave; + nbits = buffer[0] << 8 | buffer[1]; + if( nbits > MAX_EXTERN_MPI_BITS ) { + log_info ("mpi too large (%u bits)\n", nbits); + goto leave; + } + buffer += 2; + nread = 2; + + nbytes = (nbits+7) / 8; + nlimbs = (nbytes+BYTES_PER_MPI_LIMB-1) / BYTES_PER_MPI_LIMB; + val = secure? mpi_alloc_secure( nlimbs ) + : mpi_alloc( nlimbs ); + i = BYTES_PER_MPI_LIMB - nbytes % BYTES_PER_MPI_LIMB; + i %= BYTES_PER_MPI_LIMB; + val->nbits = nbits; + j= val->nlimbs = nlimbs; + val->sign = 0; + for( ; j > 0; j-- ) { + a = 0; + for(; i < BYTES_PER_MPI_LIMB; i++ ) { + if( ++nread > *ret_nread ) { + /* This (as well as the above error condition) may + happen if we use this function to parse a decrypted + MPI which didn't turn out to be a real MPI - possible + because the supplied key was wrong but the OpenPGP + checksum didn't caught it. */ + log_info ("mpi larger than buffer\n"); + mpi_free (val); + val = MPI_NULL; + goto leave; + } + a <<= 8; + a |= *buffer++; + } + i = 0; + val->d[j-1] = a; + } + + leave: + *ret_nread = nread; + return val; +} + + +/**************** + * Make an mpi from a character string. + */ +int +mpi_fromstr(MPI val, const char *str) +{ + int hexmode=0, sign=0, prepend_zero=0, i, j, c, c1, c2; + unsigned nbits, nbytes, nlimbs; + mpi_limb_t a; + + if( *str == '-' ) { + sign = 1; + str++; + } + if( *str == '0' && str[1] == 'x' ) + hexmode = 1; + else + return 1; /* other bases are not yet supported */ + str += 2; + + nbits = strlen(str)*4; + if( nbits % 8 ) + prepend_zero = 1; + nbytes = (nbits+7) / 8; + nlimbs = (nbytes+BYTES_PER_MPI_LIMB-1) / BYTES_PER_MPI_LIMB; + if( val->alloced < nlimbs ) + mpi_resize(val, nlimbs ); + i = BYTES_PER_MPI_LIMB - nbytes % BYTES_PER_MPI_LIMB; + i %= BYTES_PER_MPI_LIMB; + j= val->nlimbs = nlimbs; + val->sign = sign; + for( ; j > 0; j-- ) { + a = 0; + for(; i < BYTES_PER_MPI_LIMB; i++ ) { + if( prepend_zero ) { + c1 = '0'; + prepend_zero = 0; + } + else + c1 = *str++; + assert(c1); + c2 = *str++; + assert(c2); + if( c1 >= '0' && c1 <= '9' ) + c = c1 - '0'; + else if( c1 >= 'a' && c1 <= 'f' ) + c = c1 - 'a' + 10; + else if( c1 >= 'A' && c1 <= 'F' ) + c = c1 - 'A' + 10; + else { + mpi_clear(val); + return 1; + } + c <<= 4; + if( c2 >= '0' && c2 <= '9' ) + c |= c2 - '0'; + else if( c2 >= 'a' && c2 <= 'f' ) + c |= c2 - 'a' + 10; + else if( c2 >= 'A' && c2 <= 'F' ) + c |= c2 - 'A' + 10; + else { + mpi_clear(val); + return 1; + } + a <<= 8; + a |= c; + } + i = 0; + val->d[j-1] = a; + } + + return 0; +} + + +/**************** + * print an MPI to the given stream and return the number of characters + * printed. + */ +int +mpi_print( FILE *fp, MPI a, int mode ) +{ + int i, n=0; + + if( a == MPI_NULL ) + return fprintf(fp, "[MPI_NULL]"); + if( !mode ) { + unsigned int n1; + + n1 = mpi_get_nbits(a); + n += fprintf(fp, "[%u bits]", n1); + } + else { + if( a->sign ) + putc('-', fp); +#if BYTES_PER_MPI_LIMB == 2 +#define X "4" +#elif BYTES_PER_MPI_LIMB == 4 +#define X "8" +#elif BYTES_PER_MPI_LIMB == 8 +#define X "16" +#else +#error please define the format here +#endif + for(i=a->nlimbs; i > 0 ; i-- ) { + n += fprintf(fp, i!=a->nlimbs? "%0" X "lX":"%lX", (ulong)a->d[i-1]); +#undef X + } + if( !a->nlimbs ) + putc('0', fp ); + } + return n; +} + + +void +g10_log_mpidump( const char *text, MPI a ) +{ + FILE *fp = log_stream(); + + g10_log_print_prefix(text); + mpi_print(fp, a, 1 ); + fputc('\n', fp); +} + +/**************** + * Special function to get the low 8 bytes from an mpi. + * This can be used as a keyid; KEYID is an 2 element array. + * Return the low 4 bytes. + */ +u32 +mpi_get_keyid( MPI a, u32 *keyid ) +{ +#if BYTES_PER_MPI_LIMB == 4 + if( keyid ) { + keyid[0] = a->nlimbs >= 2? a->d[1] : 0; + keyid[1] = a->nlimbs >= 1? a->d[0] : 0; + } + return a->nlimbs >= 1? a->d[0] : 0; +#elif BYTES_PER_MPI_LIMB == 8 + if( keyid ) { + keyid[0] = a->nlimbs? (u32)(a->d[0] >> 32) : 0; + keyid[1] = a->nlimbs? (u32)(a->d[0] & 0xffffffff) : 0; + } + return a->nlimbs? (u32)(a->d[0] & 0xffffffff) : 0; +#else +#error Make this function work with other LIMB sizes +#endif +} + + +/**************** + * Return an xmalloced buffer with the MPI (msb first). + * NBYTES receives the length of this buffer. Caller must free the + * return string (This function does return a 0 byte buffer with NBYTES + * set to zero if the value of A is zero. If sign is not NULL, it will + * be set to the sign of the A. + */ +static byte * +do_get_buffer( MPI a, unsigned *nbytes, int *sign, int force_secure ) +{ + byte *p, *buffer; + mpi_limb_t alimb; + int i; + unsigned int n; + + if( sign ) + *sign = a->sign; + *nbytes = n = a->nlimbs * BYTES_PER_MPI_LIMB; + if (!n) + n++; /* avoid zero length allocation */ + p = buffer = force_secure || mpi_is_secure(a) ? xmalloc_secure(n) + : xmalloc(n); + + for(i=a->nlimbs-1; i >= 0; i-- ) { + alimb = a->d[i]; +#if BYTES_PER_MPI_LIMB == 4 + *p++ = alimb >> 24; + *p++ = alimb >> 16; + *p++ = alimb >> 8; + *p++ = alimb ; +#elif BYTES_PER_MPI_LIMB == 8 + *p++ = alimb >> 56; + *p++ = alimb >> 48; + *p++ = alimb >> 40; + *p++ = alimb >> 32; + *p++ = alimb >> 24; + *p++ = alimb >> 16; + *p++ = alimb >> 8; + *p++ = alimb ; +#else +#error please implement for this limb size. +#endif + } + + /* this is sub-optimal but we need to do the shift operation + * because the caller has to free the returned buffer */ + for(p=buffer; !*p && *nbytes; p++, --*nbytes ) + ; + if( p != buffer ) + memmove(buffer,p, *nbytes); + + return buffer; +} + + +byte * +mpi_get_buffer( MPI a, unsigned *nbytes, int *sign ) +{ + return do_get_buffer( a, nbytes, sign, 0 ); +} + +byte * +mpi_get_secure_buffer( MPI a, unsigned *nbytes, int *sign ) +{ + return do_get_buffer( a, nbytes, sign, 1 ); +} + +/**************** + * Use BUFFER to update MPI. + */ +void +mpi_set_buffer( MPI a, const byte *buffer, unsigned nbytes, int sign ) +{ + const byte *p; + mpi_limb_t alimb; + int nlimbs; + int i; + + nlimbs = (nbytes + BYTES_PER_MPI_LIMB - 1) / BYTES_PER_MPI_LIMB; + RESIZE_IF_NEEDED(a, nlimbs); + a->sign = sign; + + for(i=0, p = buffer+nbytes-1; p >= buffer+BYTES_PER_MPI_LIMB; ) { +#if BYTES_PER_MPI_LIMB == 4 + alimb = (mpi_limb_t)*p-- ; + alimb |= (mpi_limb_t)*p-- << 8 ; + alimb |= (mpi_limb_t)*p-- << 16 ; + alimb |= (mpi_limb_t)*p-- << 24 ; +#elif BYTES_PER_MPI_LIMB == 8 + alimb = (mpi_limb_t)*p-- ; + alimb |= (mpi_limb_t)*p-- << 8 ; + alimb |= (mpi_limb_t)*p-- << 16 ; + alimb |= (mpi_limb_t)*p-- << 24 ; + alimb |= (mpi_limb_t)*p-- << 32 ; + alimb |= (mpi_limb_t)*p-- << 40 ; + alimb |= (mpi_limb_t)*p-- << 48 ; + alimb |= (mpi_limb_t)*p-- << 56 ; +#else +#error please implement for this limb size. +#endif + a->d[i++] = alimb; + } + if( p >= buffer ) { +#if BYTES_PER_MPI_LIMB == 4 + alimb = *p-- ; + if( p >= buffer ) alimb |= (mpi_limb_t)*p-- << 8 ; + if( p >= buffer ) alimb |= (mpi_limb_t)*p-- << 16 ; + if( p >= buffer ) alimb |= (mpi_limb_t)*p-- << 24 ; +#elif BYTES_PER_MPI_LIMB == 8 + alimb = (mpi_limb_t)*p-- ; + if( p >= buffer ) alimb |= (mpi_limb_t)*p-- << 8 ; + if( p >= buffer ) alimb |= (mpi_limb_t)*p-- << 16 ; + if( p >= buffer ) alimb |= (mpi_limb_t)*p-- << 24 ; + if( p >= buffer ) alimb |= (mpi_limb_t)*p-- << 32 ; + if( p >= buffer ) alimb |= (mpi_limb_t)*p-- << 40 ; + if( p >= buffer ) alimb |= (mpi_limb_t)*p-- << 48 ; + if( p >= buffer ) alimb |= (mpi_limb_t)*p-- << 56 ; +#else +#error please implement for this limb size. +#endif + a->d[i++] = alimb; + } + a->nlimbs = i; + assert( i == nlimbs ); +} diff --git a/mpi/mpih-cmp.c b/mpi/mpih-cmp.c new file mode 100644 index 0000000..4c12c71 --- /dev/null +++ b/mpi/mpih-cmp.c @@ -0,0 +1,63 @@ +/* mpihelp-sub.c - MPI helper functions + * Copyright (C) 1994, 1996 Free Software Foundation, Inc. + * Copyright (C) 1998, 1999, 2000, 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> + +#include "mpi-internal.h" + +/**************** + * Compare OP1_PTR/OP1_SIZE with OP2_PTR/OP2_SIZE. + * There are no restrictions on the relative sizes of + * the two arguments. + * Return 1 if OP1 > OP2, 0 if they are equal, and -1 if OP1 < OP2. + */ +int +mpihelp_cmp( mpi_ptr_t op1_ptr, mpi_ptr_t op2_ptr, mpi_size_t size ) +{ + mpi_size_t i; + mpi_limb_t op1_word, op2_word; + + for( i = size - 1; i >= 0 ; i--) { + op1_word = op1_ptr[i]; + op2_word = op2_ptr[i]; + if( op1_word != op2_word ) + goto diff; + } + return 0; + + diff: + /* This can *not* be simplified to + * op2_word - op2_word + * since that expression might give signed overflow. */ + return (op1_word > op2_word) ? 1 : -1; +} + diff --git a/mpi/mpih-div.c b/mpi/mpih-div.c new file mode 100644 index 0000000..1e962a1 --- /dev/null +++ b/mpi/mpih-div.c @@ -0,0 +1,536 @@ +/* mpihelp-div.c - MPI helper functions + * Copyright (C) 1994, 1996 Free Software Foundation, Inc. + * Copyright (C) 1998, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include "mpi-internal.h" +#include "longlong.h" + +#ifndef UMUL_TIME +#define UMUL_TIME 1 +#endif +#ifndef UDIV_TIME +#define UDIV_TIME UMUL_TIME +#endif + +/* FIXME: We should be using invert_limb (or invert_normalized_limb) + * here (not udiv_qrnnd). + */ + +mpi_limb_t +mpihelp_mod_1(mpi_ptr_t dividend_ptr, mpi_size_t dividend_size, + mpi_limb_t divisor_limb) +{ + mpi_size_t i; + mpi_limb_t n1, n0, r; + int dummy; + + /* Botch: Should this be handled at all? Rely on callers? */ + if( !dividend_size ) + return 0; + + /* If multiplication is much faster than division, and the + * dividend is large, pre-invert the divisor, and use + * only multiplications in the inner loop. + * + * This test should be read: + * Does it ever help to use udiv_qrnnd_preinv? + * && Does what we save compensate for the inversion overhead? + */ + if( UDIV_TIME > (2 * UMUL_TIME + 6) + && (UDIV_TIME - (2 * UMUL_TIME + 6)) * dividend_size > UDIV_TIME ) { + int normalization_steps; + + count_leading_zeros( normalization_steps, divisor_limb ); + if( normalization_steps ) { + mpi_limb_t divisor_limb_inverted; + + divisor_limb <<= normalization_steps; + + /* Compute (2**2N - 2**N * DIVISOR_LIMB) / DIVISOR_LIMB. The + * result is a (N+1)-bit approximation to 1/DIVISOR_LIMB, with the + * most significant bit (with weight 2**N) implicit. + * + * Special case for DIVISOR_LIMB == 100...000. + */ + if( !(divisor_limb << 1) ) + divisor_limb_inverted = ~(mpi_limb_t)0; + else + udiv_qrnnd(divisor_limb_inverted, dummy, + -divisor_limb, 0, divisor_limb); + + n1 = dividend_ptr[dividend_size - 1]; + r = n1 >> (BITS_PER_MPI_LIMB - normalization_steps); + + /* Possible optimization: + * if (r == 0 + * && divisor_limb > ((n1 << normalization_steps) + * | (dividend_ptr[dividend_size - 2] >> ...))) + * ...one division less... + */ + for( i = dividend_size - 2; i >= 0; i--) { + n0 = dividend_ptr[i]; + UDIV_QRNND_PREINV(dummy, r, r, + ((n1 << normalization_steps) + | (n0 >> (BITS_PER_MPI_LIMB - normalization_steps))), + divisor_limb, divisor_limb_inverted); + n1 = n0; + } + UDIV_QRNND_PREINV(dummy, r, r, + n1 << normalization_steps, + divisor_limb, divisor_limb_inverted); + return r >> normalization_steps; + } + else { + mpi_limb_t divisor_limb_inverted; + + /* Compute (2**2N - 2**N * DIVISOR_LIMB) / DIVISOR_LIMB. The + * result is a (N+1)-bit approximation to 1/DIVISOR_LIMB, with the + * most significant bit (with weight 2**N) implicit. + * + * Special case for DIVISOR_LIMB == 100...000. + */ + if( !(divisor_limb << 1) ) + divisor_limb_inverted = ~(mpi_limb_t)0; + else + udiv_qrnnd(divisor_limb_inverted, dummy, + -divisor_limb, 0, divisor_limb); + + i = dividend_size - 1; + r = dividend_ptr[i]; + + if( r >= divisor_limb ) + r = 0; + else + i--; + + for( ; i >= 0; i--) { + n0 = dividend_ptr[i]; + UDIV_QRNND_PREINV(dummy, r, r, + n0, divisor_limb, divisor_limb_inverted); + } + return r; + } + } + else { + if( UDIV_NEEDS_NORMALIZATION ) { + int normalization_steps; + + count_leading_zeros(normalization_steps, divisor_limb); + if( normalization_steps ) { + divisor_limb <<= normalization_steps; + + n1 = dividend_ptr[dividend_size - 1]; + r = n1 >> (BITS_PER_MPI_LIMB - normalization_steps); + + /* Possible optimization: + * if (r == 0 + * && divisor_limb > ((n1 << normalization_steps) + * | (dividend_ptr[dividend_size - 2] >> ...))) + * ...one division less... + */ + for(i = dividend_size - 2; i >= 0; i--) { + n0 = dividend_ptr[i]; + udiv_qrnnd (dummy, r, r, + ((n1 << normalization_steps) + | (n0 >> (BITS_PER_MPI_LIMB - normalization_steps))), + divisor_limb); + n1 = n0; + } + udiv_qrnnd (dummy, r, r, + n1 << normalization_steps, + divisor_limb); + return r >> normalization_steps; + } + } + /* No normalization needed, either because udiv_qrnnd doesn't require + * it, or because DIVISOR_LIMB is already normalized. */ + i = dividend_size - 1; + r = dividend_ptr[i]; + + if(r >= divisor_limb) + r = 0; + else + i--; + + for(; i >= 0; i--) { + n0 = dividend_ptr[i]; + udiv_qrnnd (dummy, r, r, n0, divisor_limb); + } + return r; + } +} + +/* Divide num (NP/NSIZE) by den (DP/DSIZE) and write + * the NSIZE-DSIZE least significant quotient limbs at QP + * and the DSIZE long remainder at NP. If QEXTRA_LIMBS is + * non-zero, generate that many fraction bits and append them after the + * other quotient limbs. + * Return the most significant limb of the quotient, this is always 0 or 1. + * + * Preconditions: + * 0. NSIZE >= DSIZE. + * 1. The most significant bit of the divisor must be set. + * 2. QP must either not overlap with the input operands at all, or + * QP + DSIZE >= NP must hold true. (This means that it's + * possible to put the quotient in the high part of NUM, right after the + * remainder in NUM. + * 3. NSIZE >= DSIZE, even if QEXTRA_LIMBS is non-zero. + */ + +mpi_limb_t +mpihelp_divrem( mpi_ptr_t qp, mpi_size_t qextra_limbs, + mpi_ptr_t np, mpi_size_t nsize, + mpi_ptr_t dp, mpi_size_t dsize) +{ + mpi_limb_t most_significant_q_limb = 0; + + switch(dsize) { + case 0: + /* We are asked to divide by zero, so go ahead and do it! (To make + the compiler not remove this statement, return the value.) */ + return 1 / dsize; + + case 1: + { + mpi_size_t i; + mpi_limb_t n1; + mpi_limb_t d; + + d = dp[0]; + n1 = np[nsize - 1]; + + if( n1 >= d ) { + n1 -= d; + most_significant_q_limb = 1; + } + + qp += qextra_limbs; + for( i = nsize - 2; i >= 0; i--) + udiv_qrnnd( qp[i], n1, n1, np[i], d ); + qp -= qextra_limbs; + + for( i = qextra_limbs - 1; i >= 0; i-- ) + udiv_qrnnd (qp[i], n1, n1, 0, d); + + np[0] = n1; + } + break; + + case 2: + { + mpi_size_t i; + mpi_limb_t n1, n0, n2; + mpi_limb_t d1, d0; + + np += nsize - 2; + d1 = dp[1]; + d0 = dp[0]; + n1 = np[1]; + n0 = np[0]; + + if( n1 >= d1 && (n1 > d1 || n0 >= d0) ) { + sub_ddmmss (n1, n0, n1, n0, d1, d0); + most_significant_q_limb = 1; + } + + for( i = qextra_limbs + nsize - 2 - 1; i >= 0; i-- ) { + mpi_limb_t q; + mpi_limb_t r; + + if( i >= qextra_limbs ) + np--; + else + np[0] = 0; + + if( n1 == d1 ) { + /* Q should be either 111..111 or 111..110. Need special + * treatment of this rare case as normal division would + * give overflow. */ + q = ~(mpi_limb_t)0; + + r = n0 + d1; + if( r < d1 ) { /* Carry in the addition? */ + add_ssaaaa( n1, n0, r - d0, np[0], 0, d0 ); + qp[i] = q; + continue; + } + n1 = d0 - (d0 != 0?1:0); + n0 = -d0; + } + else { + udiv_qrnnd (q, r, n1, n0, d1); + umul_ppmm (n1, n0, d0, q); + } + + n2 = np[0]; + q_test: + if( n1 > r || (n1 == r && n0 > n2) ) { + /* The estimated Q was too large. */ + q--; + sub_ddmmss (n1, n0, n1, n0, 0, d0); + r += d1; + if( r >= d1 ) /* If not carry, test Q again. */ + goto q_test; + } + + qp[i] = q; + sub_ddmmss (n1, n0, r, n2, n1, n0); + } + np[1] = n1; + np[0] = n0; + } + break; + + default: + { + mpi_size_t i; + mpi_limb_t dX, d1, n0; + + np += nsize - dsize; + dX = dp[dsize - 1]; + d1 = dp[dsize - 2]; + n0 = np[dsize - 1]; + + if( n0 >= dX ) { + if(n0 > dX || mpihelp_cmp(np, dp, dsize - 1) >= 0 ) { + mpihelp_sub_n(np, np, dp, dsize); + n0 = np[dsize - 1]; + most_significant_q_limb = 1; + } + } + + for( i = qextra_limbs + nsize - dsize - 1; i >= 0; i--) { + mpi_limb_t q; + mpi_limb_t n1, n2; + mpi_limb_t cy_limb; + + if( i >= qextra_limbs ) { + np--; + n2 = np[dsize]; + } + else { + n2 = np[dsize - 1]; + MPN_COPY_DECR (np + 1, np, dsize - 1); + np[0] = 0; + } + + if( n0 == dX ) { + /* This might over-estimate q, but it's probably not worth + * the extra code here to find out. */ + q = ~(mpi_limb_t)0; + } + else { + mpi_limb_t r; + + udiv_qrnnd(q, r, n0, np[dsize - 1], dX); + umul_ppmm(n1, n0, d1, q); + + while( n1 > r || (n1 == r && n0 > np[dsize - 2])) { + q--; + r += dX; + if( r < dX ) /* I.e. "carry in previous addition?" */ + break; + n1 -= n0 < d1; + n0 -= d1; + } + } + + /* Possible optimization: We already have (q * n0) and (1 * n1) + * after the calculation of q. Taking advantage of that, we + * could make this loop make two iterations less. */ + cy_limb = mpihelp_submul_1(np, dp, dsize, q); + + if( n2 != cy_limb ) { + mpihelp_add_n(np, np, dp, dsize); + q--; + } + + qp[i] = q; + n0 = np[dsize - 1]; + } + } + } + + return most_significant_q_limb; +} + + +/**************** + * Divide (DIVIDEND_PTR,,DIVIDEND_SIZE) by DIVISOR_LIMB. + * Write DIVIDEND_SIZE limbs of quotient at QUOT_PTR. + * Return the single-limb remainder. + * There are no constraints on the value of the divisor. + * + * QUOT_PTR and DIVIDEND_PTR might point to the same limb. + */ + +mpi_limb_t +mpihelp_divmod_1( mpi_ptr_t quot_ptr, + mpi_ptr_t dividend_ptr, mpi_size_t dividend_size, + mpi_limb_t divisor_limb) +{ + mpi_size_t i; + mpi_limb_t n1, n0, r; + int dummy; + + if( !dividend_size ) + return 0; + + /* If multiplication is much faster than division, and the + * dividend is large, pre-invert the divisor, and use + * only multiplications in the inner loop. + * + * This test should be read: + * Does it ever help to use udiv_qrnnd_preinv? + * && Does what we save compensate for the inversion overhead? + */ + if( UDIV_TIME > (2 * UMUL_TIME + 6) + && (UDIV_TIME - (2 * UMUL_TIME + 6)) * dividend_size > UDIV_TIME ) { + int normalization_steps; + + count_leading_zeros( normalization_steps, divisor_limb ); + if( normalization_steps ) { + mpi_limb_t divisor_limb_inverted; + + divisor_limb <<= normalization_steps; + + /* Compute (2**2N - 2**N * DIVISOR_LIMB) / DIVISOR_LIMB. The + * result is a (N+1)-bit approximation to 1/DIVISOR_LIMB, with the + * most significant bit (with weight 2**N) implicit. + */ + /* Special case for DIVISOR_LIMB == 100...000. */ + if( !(divisor_limb << 1) ) + divisor_limb_inverted = ~(mpi_limb_t)0; + else + udiv_qrnnd(divisor_limb_inverted, dummy, + -divisor_limb, 0, divisor_limb); + + n1 = dividend_ptr[dividend_size - 1]; + r = n1 >> (BITS_PER_MPI_LIMB - normalization_steps); + + /* Possible optimization: + * if (r == 0 + * && divisor_limb > ((n1 << normalization_steps) + * | (dividend_ptr[dividend_size - 2] >> ...))) + * ...one division less... + */ + for( i = dividend_size - 2; i >= 0; i--) { + n0 = dividend_ptr[i]; + UDIV_QRNND_PREINV( quot_ptr[i + 1], r, r, + ((n1 << normalization_steps) + | (n0 >> (BITS_PER_MPI_LIMB - normalization_steps))), + divisor_limb, divisor_limb_inverted); + n1 = n0; + } + UDIV_QRNND_PREINV( quot_ptr[0], r, r, + n1 << normalization_steps, + divisor_limb, divisor_limb_inverted); + return r >> normalization_steps; + } + else { + mpi_limb_t divisor_limb_inverted; + + /* Compute (2**2N - 2**N * DIVISOR_LIMB) / DIVISOR_LIMB. The + * result is a (N+1)-bit approximation to 1/DIVISOR_LIMB, with the + * most significant bit (with weight 2**N) implicit. + */ + /* Special case for DIVISOR_LIMB == 100...000. */ + if( !(divisor_limb << 1) ) + divisor_limb_inverted = ~(mpi_limb_t) 0; + else + udiv_qrnnd(divisor_limb_inverted, dummy, + -divisor_limb, 0, divisor_limb); + + i = dividend_size - 1; + r = dividend_ptr[i]; + + if( r >= divisor_limb ) + r = 0; + else + quot_ptr[i--] = 0; + + for( ; i >= 0; i-- ) { + n0 = dividend_ptr[i]; + UDIV_QRNND_PREINV( quot_ptr[i], r, r, + n0, divisor_limb, divisor_limb_inverted); + } + return r; + } + } + else { + if(UDIV_NEEDS_NORMALIZATION) { + int normalization_steps; + + count_leading_zeros (normalization_steps, divisor_limb); + if( normalization_steps ) { + divisor_limb <<= normalization_steps; + + n1 = dividend_ptr[dividend_size - 1]; + r = n1 >> (BITS_PER_MPI_LIMB - normalization_steps); + + /* Possible optimization: + * if (r == 0 + * && divisor_limb > ((n1 << normalization_steps) + * | (dividend_ptr[dividend_size - 2] >> ...))) + * ...one division less... + */ + for( i = dividend_size - 2; i >= 0; i--) { + n0 = dividend_ptr[i]; + udiv_qrnnd (quot_ptr[i + 1], r, r, + ((n1 << normalization_steps) + | (n0 >> (BITS_PER_MPI_LIMB - normalization_steps))), + divisor_limb); + n1 = n0; + } + udiv_qrnnd (quot_ptr[0], r, r, + n1 << normalization_steps, + divisor_limb); + return r >> normalization_steps; + } + } + /* No normalization needed, either because udiv_qrnnd doesn't require + * it, or because DIVISOR_LIMB is already normalized. */ + i = dividend_size - 1; + r = dividend_ptr[i]; + + if(r >= divisor_limb) + r = 0; + else + quot_ptr[i--] = 0; + + for(; i >= 0; i--) { + n0 = dividend_ptr[i]; + udiv_qrnnd( quot_ptr[i], r, r, n0, divisor_limb ); + } + return r; + } +} diff --git a/mpi/mpih-mul.c b/mpi/mpih-mul.c new file mode 100644 index 0000000..397ec60 --- /dev/null +++ b/mpi/mpih-mul.c @@ -0,0 +1,529 @@ +/* mpihelp-mul.c - MPI helper functions + * Copyright (C) 1994, 1996, 1998, 1999, + * 2000 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include "mpi-internal.h" +#include "longlong.h" + + + +#define MPN_MUL_N_RECURSE(prodp, up, vp, size, tspace) \ + do { \ + if( (size) < KARATSUBA_THRESHOLD ) \ + mul_n_basecase (prodp, up, vp, size); \ + else \ + mul_n (prodp, up, vp, size, tspace); \ + } while (0); + +#define MPN_SQR_N_RECURSE(prodp, up, size, tspace) \ + do { \ + if ((size) < KARATSUBA_THRESHOLD) \ + mpih_sqr_n_basecase (prodp, up, size); \ + else \ + mpih_sqr_n (prodp, up, size, tspace); \ + } while (0); + + + + +/* Multiply the natural numbers u (pointed to by UP) and v (pointed to by VP), + * both with SIZE limbs, and store the result at PRODP. 2 * SIZE limbs are + * always stored. Return the most significant limb. + * + * Argument constraints: + * 1. PRODP != UP and PRODP != VP, i.e. the destination + * must be distinct from the multiplier and the multiplicand. + * + * + * Handle simple cases with traditional multiplication. + * + * This is the most critical code of multiplication. All multiplies rely + * on this, both small and huge. Small ones arrive here immediately. Huge + * ones arrive here as this is the base case for Karatsuba's recursive + * algorithm below. + */ + +static mpi_limb_t +mul_n_basecase( mpi_ptr_t prodp, mpi_ptr_t up, + mpi_ptr_t vp, mpi_size_t size) +{ + mpi_size_t i; + mpi_limb_t cy; + mpi_limb_t v_limb; + + /* Multiply by the first limb in V separately, as the result can be + * stored (not added) to PROD. We also avoid a loop for zeroing. */ + v_limb = vp[0]; + if( v_limb <= 1 ) { + if( v_limb == 1 ) + MPN_COPY( prodp, up, size ); + else + MPN_ZERO( prodp, size ); + cy = 0; + } + else + cy = mpihelp_mul_1( prodp, up, size, v_limb ); + + prodp[size] = cy; + prodp++; + + /* For each iteration in the outer loop, multiply one limb from + * U with one limb from V, and add it to PROD. */ + for( i = 1; i < size; i++ ) { + v_limb = vp[i]; + if( v_limb <= 1 ) { + cy = 0; + if( v_limb == 1 ) + cy = mpihelp_add_n(prodp, prodp, up, size); + } + else + cy = mpihelp_addmul_1(prodp, up, size, v_limb); + + prodp[size] = cy; + prodp++; + } + + return cy; +} + + +static void +mul_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_ptr_t vp, + mpi_size_t size, mpi_ptr_t tspace ) +{ + if( size & 1 ) { + /* The size is odd, and the code below doesn't handle that. + * Multiply the least significant (size - 1) limbs with a recursive + * call, and handle the most significant limb of S1 and S2 + * separately. + * A slightly faster way to do this would be to make the Karatsuba + * code below behave as if the size were even, and let it check for + * odd size in the end. I.e., in essence move this code to the end. + * Doing so would save us a recursive call, and potentially make the + * stack grow a lot less. + */ + mpi_size_t esize = size - 1; /* even size */ + mpi_limb_t cy_limb; + + MPN_MUL_N_RECURSE( prodp, up, vp, esize, tspace ); + cy_limb = mpihelp_addmul_1( prodp + esize, up, esize, vp[esize] ); + prodp[esize + esize] = cy_limb; + cy_limb = mpihelp_addmul_1( prodp + esize, vp, size, up[esize] ); + prodp[esize + size] = cy_limb; + } + else { + /* Anatolij Alekseevich Karatsuba's divide-and-conquer algorithm. + * + * Split U in two pieces, U1 and U0, such that + * U = U0 + U1*(B**n), + * and V in V1 and V0, such that + * V = V0 + V1*(B**n). + * + * UV is then computed recursively using the identity + * + * 2n n n n + * UV = (B + B )U V + B (U -U )(V -V ) + (B + 1)U V + * 1 1 1 0 0 1 0 0 + * + * Where B = 2**BITS_PER_MP_LIMB. + */ + mpi_size_t hsize = size >> 1; + mpi_limb_t cy; + int negflg; + + /* Product H. ________________ ________________ + * |_____U1 x V1____||____U0 x V0_____| + * Put result in upper part of PROD and pass low part of TSPACE + * as new TSPACE. + */ + MPN_MUL_N_RECURSE(prodp + size, up + hsize, vp + hsize, hsize, tspace); + + /* Product M. ________________ + * |_(U1-U0)(V0-V1)_| + */ + if( mpihelp_cmp(up + hsize, up, hsize) >= 0 ) { + mpihelp_sub_n(prodp, up + hsize, up, hsize); + negflg = 0; + } + else { + mpihelp_sub_n(prodp, up, up + hsize, hsize); + negflg = 1; + } + if( mpihelp_cmp(vp + hsize, vp, hsize) >= 0 ) { + mpihelp_sub_n(prodp + hsize, vp + hsize, vp, hsize); + negflg ^= 1; + } + else { + mpihelp_sub_n(prodp + hsize, vp, vp + hsize, hsize); + /* No change of NEGFLG. */ + } + /* Read temporary operands from low part of PROD. + * Put result in low part of TSPACE using upper part of TSPACE + * as new TSPACE. + */ + MPN_MUL_N_RECURSE(tspace, prodp, prodp + hsize, hsize, tspace + size); + + /* Add/copy product H. */ + MPN_COPY (prodp + hsize, prodp + size, hsize); + cy = mpihelp_add_n( prodp + size, prodp + size, + prodp + size + hsize, hsize); + + /* Add product M (if NEGFLG M is a negative number) */ + if(negflg) + cy -= mpihelp_sub_n(prodp + hsize, prodp + hsize, tspace, size); + else + cy += mpihelp_add_n(prodp + hsize, prodp + hsize, tspace, size); + + /* Product L. ________________ ________________ + * |________________||____U0 x V0_____| + * Read temporary operands from low part of PROD. + * Put result in low part of TSPACE using upper part of TSPACE + * as new TSPACE. + */ + MPN_MUL_N_RECURSE(tspace, up, vp, hsize, tspace + size); + + /* Add/copy Product L (twice) */ + + cy += mpihelp_add_n(prodp + hsize, prodp + hsize, tspace, size); + if( cy ) + mpihelp_add_1(prodp + hsize + size, prodp + hsize + size, hsize, cy); + + MPN_COPY(prodp, tspace, hsize); + cy = mpihelp_add_n(prodp + hsize, prodp + hsize, tspace + hsize, hsize); + if( cy ) + mpihelp_add_1(prodp + size, prodp + size, size, 1); + } +} + + +void +mpih_sqr_n_basecase( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size ) +{ + mpi_size_t i; + mpi_limb_t cy_limb; + mpi_limb_t v_limb; + + /* Multiply by the first limb in V separately, as the result can be + * stored (not added) to PROD. We also avoid a loop for zeroing. */ + v_limb = up[0]; + if( v_limb <= 1 ) { + if( v_limb == 1 ) + MPN_COPY( prodp, up, size ); + else + MPN_ZERO(prodp, size); + cy_limb = 0; + } + else + cy_limb = mpihelp_mul_1( prodp, up, size, v_limb ); + + prodp[size] = cy_limb; + prodp++; + + /* For each iteration in the outer loop, multiply one limb from + * U with one limb from V, and add it to PROD. */ + for( i=1; i < size; i++) { + v_limb = up[i]; + if( v_limb <= 1 ) { + cy_limb = 0; + if( v_limb == 1 ) + cy_limb = mpihelp_add_n(prodp, prodp, up, size); + } + else + cy_limb = mpihelp_addmul_1(prodp, up, size, v_limb); + + prodp[size] = cy_limb; + prodp++; + } +} + + +void +mpih_sqr_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t size, mpi_ptr_t tspace) +{ + if( size & 1 ) { + /* The size is odd, and the code below doesn't handle that. + * Multiply the least significant (size - 1) limbs with a recursive + * call, and handle the most significant limb of S1 and S2 + * separately. + * A slightly faster way to do this would be to make the Karatsuba + * code below behave as if the size were even, and let it check for + * odd size in the end. I.e., in essence move this code to the end. + * Doing so would save us a recursive call, and potentially make the + * stack grow a lot less. + */ + mpi_size_t esize = size - 1; /* even size */ + mpi_limb_t cy_limb; + + MPN_SQR_N_RECURSE( prodp, up, esize, tspace ); + cy_limb = mpihelp_addmul_1( prodp + esize, up, esize, up[esize] ); + prodp[esize + esize] = cy_limb; + cy_limb = mpihelp_addmul_1( prodp + esize, up, size, up[esize] ); + + prodp[esize + size] = cy_limb; + } + else { + mpi_size_t hsize = size >> 1; + mpi_limb_t cy; + + /* Product H. ________________ ________________ + * |_____U1 x U1____||____U0 x U0_____| + * Put result in upper part of PROD and pass low part of TSPACE + * as new TSPACE. + */ + MPN_SQR_N_RECURSE(prodp + size, up + hsize, hsize, tspace); + + /* Product M. ________________ + * |_(U1-U0)(U0-U1)_| + */ + if( mpihelp_cmp( up + hsize, up, hsize) >= 0 ) + mpihelp_sub_n( prodp, up + hsize, up, hsize); + else + mpihelp_sub_n (prodp, up, up + hsize, hsize); + + /* Read temporary operands from low part of PROD. + * Put result in low part of TSPACE using upper part of TSPACE + * as new TSPACE. */ + MPN_SQR_N_RECURSE(tspace, prodp, hsize, tspace + size); + + /* Add/copy product H */ + MPN_COPY(prodp + hsize, prodp + size, hsize); + cy = mpihelp_add_n(prodp + size, prodp + size, + prodp + size + hsize, hsize); + + /* Add product M (if NEGFLG M is a negative number). */ + cy -= mpihelp_sub_n (prodp + hsize, prodp + hsize, tspace, size); + + /* Product L. ________________ ________________ + * |________________||____U0 x U0_____| + * Read temporary operands from low part of PROD. + * Put result in low part of TSPACE using upper part of TSPACE + * as new TSPACE. */ + MPN_SQR_N_RECURSE (tspace, up, hsize, tspace + size); + + /* Add/copy Product L (twice). */ + cy += mpihelp_add_n (prodp + hsize, prodp + hsize, tspace, size); + if( cy ) + mpihelp_add_1(prodp + hsize + size, prodp + hsize + size, + hsize, cy); + + MPN_COPY(prodp, tspace, hsize); + cy = mpihelp_add_n (prodp + hsize, prodp + hsize, tspace + hsize, hsize); + if( cy ) + mpihelp_add_1 (prodp + size, prodp + size, size, 1); + } +} + + +/* This should be made into an inline function in gmp.h. */ +void +mpihelp_mul_n( mpi_ptr_t prodp, mpi_ptr_t up, mpi_ptr_t vp, mpi_size_t size) +{ + int secure; + + if( up == vp ) { + if( size < KARATSUBA_THRESHOLD ) + mpih_sqr_n_basecase( prodp, up, size ); + else { + mpi_ptr_t tspace; + secure = m_is_secure( up ); + tspace = mpi_alloc_limb_space( 2 * size, secure ); + mpih_sqr_n( prodp, up, size, tspace ); + mpi_free_limb_space( tspace ); + } + } + else { + if( size < KARATSUBA_THRESHOLD ) + mul_n_basecase( prodp, up, vp, size ); + else { + mpi_ptr_t tspace; + secure = m_is_secure( up ) || m_is_secure( vp ); + tspace = mpi_alloc_limb_space( 2 * size, secure ); + mul_n (prodp, up, vp, size, tspace); + mpi_free_limb_space( tspace ); + } + } +} + + + +void +mpihelp_mul_karatsuba_case( mpi_ptr_t prodp, + mpi_ptr_t up, mpi_size_t usize, + mpi_ptr_t vp, mpi_size_t vsize, + struct karatsuba_ctx *ctx ) +{ + mpi_limb_t cy; + + if( !ctx->tspace || ctx->tspace_size < vsize ) { + if( ctx->tspace ) + mpi_free_limb_space( ctx->tspace ); + ctx->tspace = mpi_alloc_limb_space( 2 * vsize, + m_is_secure( up ) || m_is_secure( vp ) ); + ctx->tspace_size = vsize; + } + + MPN_MUL_N_RECURSE( prodp, up, vp, vsize, ctx->tspace ); + + prodp += vsize; + up += vsize; + usize -= vsize; + if( usize >= vsize ) { + if( !ctx->tp || ctx->tp_size < vsize ) { + if( ctx->tp ) + mpi_free_limb_space( ctx->tp ); + ctx->tp = mpi_alloc_limb_space( 2 * vsize, m_is_secure( up ) + || m_is_secure( vp ) ); + ctx->tp_size = vsize; + } + + do { + MPN_MUL_N_RECURSE( ctx->tp, up, vp, vsize, ctx->tspace ); + cy = mpihelp_add_n( prodp, prodp, ctx->tp, vsize ); + mpihelp_add_1( prodp + vsize, ctx->tp + vsize, vsize, cy ); + prodp += vsize; + up += vsize; + usize -= vsize; + } while( usize >= vsize ); + } + + if( usize ) { + if( usize < KARATSUBA_THRESHOLD ) { + mpihelp_mul( ctx->tspace, vp, vsize, up, usize ); + } + else { + if( !ctx->next ) { + ctx->next = xmalloc_clear( sizeof *ctx ); + } + mpihelp_mul_karatsuba_case( ctx->tspace, + vp, vsize, + up, usize, + ctx->next ); + } + + cy = mpihelp_add_n( prodp, prodp, ctx->tspace, vsize); + mpihelp_add_1( prodp + vsize, ctx->tspace + vsize, usize, cy ); + } +} + + +void +mpihelp_release_karatsuba_ctx( struct karatsuba_ctx *ctx ) +{ + struct karatsuba_ctx *ctx2; + + if( ctx->tp ) + mpi_free_limb_space( ctx->tp ); + if( ctx->tspace ) + mpi_free_limb_space( ctx->tspace ); + for( ctx=ctx->next; ctx; ctx = ctx2 ) { + ctx2 = ctx->next; + if( ctx->tp ) + mpi_free_limb_space( ctx->tp ); + if( ctx->tspace ) + mpi_free_limb_space( ctx->tspace ); + xfree( ctx ); + } +} + +/* Multiply the natural numbers u (pointed to by UP, with USIZE limbs) + * and v (pointed to by VP, with VSIZE limbs), and store the result at + * PRODP. USIZE + VSIZE limbs are always stored, but if the input + * operands are normalized. Return the most significant limb of the + * result. + * + * NOTE: The space pointed to by PRODP is overwritten before finished + * with U and V, so overlap is an error. + * + * Argument constraints: + * 1. USIZE >= VSIZE. + * 2. PRODP != UP and PRODP != VP, i.e. the destination + * must be distinct from the multiplier and the multiplicand. + */ + +mpi_limb_t +mpihelp_mul( mpi_ptr_t prodp, mpi_ptr_t up, mpi_size_t usize, + mpi_ptr_t vp, mpi_size_t vsize) +{ + mpi_ptr_t prod_endp = prodp + usize + vsize - 1; + mpi_limb_t cy; + struct karatsuba_ctx ctx; + + if( vsize < KARATSUBA_THRESHOLD ) { + mpi_size_t i; + mpi_limb_t v_limb; + + if( !vsize ) + return 0; + + /* Multiply by the first limb in V separately, as the result can be + * stored (not added) to PROD. We also avoid a loop for zeroing. */ + v_limb = vp[0]; + if( v_limb <= 1 ) { + if( v_limb == 1 ) + MPN_COPY( prodp, up, usize ); + else + MPN_ZERO( prodp, usize ); + cy = 0; + } + else + cy = mpihelp_mul_1( prodp, up, usize, v_limb ); + + prodp[usize] = cy; + prodp++; + + /* For each iteration in the outer loop, multiply one limb from + * U with one limb from V, and add it to PROD. */ + for( i = 1; i < vsize; i++ ) { + v_limb = vp[i]; + if( v_limb <= 1 ) { + cy = 0; + if( v_limb == 1 ) + cy = mpihelp_add_n(prodp, prodp, up, usize); + } + else + cy = mpihelp_addmul_1(prodp, up, usize, v_limb); + + prodp[usize] = cy; + prodp++; + } + + return cy; + } + + memset( &ctx, 0, sizeof ctx ); + mpihelp_mul_karatsuba_case( prodp, up, usize, vp, vsize, &ctx ); + mpihelp_release_karatsuba_ctx( &ctx ); + return *prod_endp; +} + + diff --git a/mpi/mpiutil.c b/mpi/mpiutil.c new file mode 100644 index 0000000..fdbd354 --- /dev/null +++ b/mpi/mpiutil.c @@ -0,0 +1,470 @@ +/* mpiutil.ac - Utility functions for MPI + * Copyright (C) 1998, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + * 02110-1301, USA. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <assert.h> + +#include "mpi.h" +#include "mpi-internal.h" +#include "memory.h" +#include "util.h" + + +#ifdef M_DEBUG +#undef mpi_alloc +#undef mpi_alloc_secure +#undef mpi_free +#endif + +/**************** + * Note: It was a bad idea to use the number of limbs to allocate + * because on a alpha the limbs are large but we normally need + * integers of n bits - So we should chnage this to bits (or bytes). + * + * But mpi_alloc is used in a lot of places :-) + */ +MPI +#ifdef M_DEBUG +mpi_debug_alloc( unsigned nlimbs, const char *info ) +#else +mpi_alloc( unsigned nlimbs ) +#endif +{ + MPI a; + + if( DBG_MEMORY ) + log_debug("mpi_alloc(%u)\n", nlimbs*BITS_PER_MPI_LIMB ); +#ifdef M_DEBUG + a = m_debug_alloc( sizeof *a, info ); + a->d = nlimbs? mpi_debug_alloc_limb_space( nlimbs, 0, info ) : NULL; +#else + a = xmalloc( sizeof *a ); + a->d = nlimbs? mpi_alloc_limb_space( nlimbs, 0 ) : NULL; +#endif + a->alloced = nlimbs; + a->nlimbs = 0; + a->sign = 0; + a->flags = 0; + a->nbits = 0; + return a; +} + +void +mpi_m_check( MPI a ) +{ + m_check(a); + m_check(a->d); +} + +MPI +#ifdef M_DEBUG +mpi_debug_alloc_secure( unsigned nlimbs, const char *info ) +#else +mpi_alloc_secure( unsigned nlimbs ) +#endif +{ + MPI a; + + if( DBG_MEMORY ) + log_debug("mpi_alloc_secure(%u)\n", nlimbs*BITS_PER_MPI_LIMB ); +#ifdef M_DEBUG + a = m_debug_alloc( sizeof *a, info ); + a->d = nlimbs? mpi_debug_alloc_limb_space( nlimbs, 1, info ) : NULL; +#else + a = xmalloc( sizeof *a ); + a->d = nlimbs? mpi_alloc_limb_space( nlimbs, 1 ) : NULL; +#endif + a->alloced = nlimbs; + a->flags = 1; + a->nlimbs = 0; + a->sign = 0; + a->nbits = 0; + return a; +} + + +#if 0 +static void *unused_limbs_5; +static void *unused_limbs_32; +static void *unused_limbs_64; +#endif + +mpi_ptr_t +#ifdef M_DEBUG +mpi_debug_alloc_limb_space( unsigned nlimbs, int secure, const char *info ) +#else +mpi_alloc_limb_space( unsigned nlimbs, int secure ) +#endif +{ + size_t len = nlimbs * sizeof(mpi_limb_t); + mpi_ptr_t p; + + if( DBG_MEMORY ) + log_debug("mpi_alloc_limb_space(%u)\n", (unsigned)len*8 ); +#if 0 + if( !secure ) { + if( nlimbs == 5 && unused_limbs_5 ) { /* DSA 160 bits */ + p = unused_limbs_5; + unused_limbs_5 = *p; + return p; + } + else if( nlimbs == 32 && unused_limbs_32 ) { /* DSA 1024 bits */ + p = unused_limbs_32; + unused_limbs_32 = *p; + return p; + } + else if( nlimbs == 64 && unused_limbs_64 ) { /* DSA 2*1024 bits */ + p = unused_limbs_64; + unused_limbs_64 = *p; + return p; + } + } +#endif + +#ifdef M_DEBUG + p = secure? m_debug_alloc_secure(len, info):m_debug_alloc( len, info ); +#else + p = secure? xmalloc_secure( len ):xmalloc( len ); +#endif + + return p; +} + +void +#ifdef M_DEBUG +mpi_debug_free_limb_space( mpi_ptr_t a, const char *info ) +#else +mpi_free_limb_space( mpi_ptr_t a ) +#endif +{ + if( !a ) + return; + if( DBG_MEMORY ) + log_debug("mpi_free_limb_space of size %lu\n", (ulong)m_size(a)*8 ); + +#if 0 + if( !m_is_secure(a) ) { + size_t nlimbs = m_size(a) / 4 ; + void *p = a; + + if( nlimbs == 5 ) { /* DSA 160 bits */ + *a = unused_limbs_5; + unused_limbs_5 = a; + return; + } + else if( nlimbs == 32 ) { /* DSA 1024 bits */ + *a = unused_limbs_32; + unused_limbs_32 = a; + return; + } + else if( nlimbs == 64 ) { /* DSA 2*1024 bits */ + *a = unused_limbs_64; + unused_limbs_64 = a; + return; + } + } +#endif + + xfree(a); +} + + +void +mpi_assign_limb_space( MPI a, mpi_ptr_t ap, unsigned nlimbs ) +{ + mpi_free_limb_space(a->d); + a->d = ap; + a->alloced = nlimbs; +} + + + +/**************** + * Resize the array of A to NLIMBS. the additional space is cleared + * (set to 0) [done by xrealloc()] + */ +void +#ifdef M_DEBUG +mpi_debug_resize( MPI a, unsigned nlimbs, const char *info ) +#else +mpi_resize( MPI a, unsigned nlimbs ) +#endif +{ + if( nlimbs <= a->alloced ) + return; /* no need to do it */ + /* Note: a->secure is not used - instead the realloc functions + * take care of it. Maybe we should drop a->secure completely + * and rely on a mpi_is_secure function, which would be + * a wrapper around m_is_secure + */ +#ifdef M_DEBUG + if( a->d ) + a->d = m_debug_realloc(a->d, nlimbs * sizeof(mpi_limb_t), info ); + else + a->d = m_debug_alloc_clear( nlimbs * sizeof(mpi_limb_t), info ); +#else + if( a->d ) + a->d = xrealloc(a->d, nlimbs * sizeof(mpi_limb_t) ); + else + a->d = xmalloc_clear( nlimbs * sizeof(mpi_limb_t) ); +#endif + a->alloced = nlimbs; +} + +void +mpi_clear( MPI a ) +{ + a->nlimbs = 0; + a->nbits = 0; + a->flags = 0; +} + + +void +#ifdef M_DEBUG +mpi_debug_free( MPI a, const char *info ) +#else +mpi_free( MPI a ) +#endif +{ + if( !a ) + return; + if( DBG_MEMORY ) + log_debug("mpi_free\n" ); + if( a->flags & 4 ) + xfree( a->d ); + else { +#ifdef M_DEBUG + mpi_debug_free_limb_space(a->d, info); +#else + mpi_free_limb_space(a->d); +#endif + } + if( a->flags & ~7 ) + log_bug("invalid flag value in mpi\n"); + xfree(a); +} + + +void +mpi_set_secure( MPI a ) +{ + mpi_ptr_t ap, bp; + + if( (a->flags & 1) ) + return; + a->flags |= 1; + ap = a->d; + if( !a->nlimbs ) { + assert(!ap); + return; + } +#ifdef M_DEBUG + bp = mpi_debug_alloc_limb_space( a->nlimbs, 1, "set_secure" ); +#else + bp = mpi_alloc_limb_space( a->nlimbs, 1 ); +#endif + MPN_COPY( bp, ap, a->nlimbs ); + a->d = bp; +#ifdef M_DEBUG + mpi_debug_free_limb_space(ap, "set_secure"); +#else + mpi_free_limb_space(ap); +#endif +} + + +MPI +mpi_set_opaque( MPI a, void *p, unsigned int len ) +{ + if( !a ) { +#ifdef M_DEBUG + a = mpi_debug_alloc(0,"alloc_opaque"); +#else + a = mpi_alloc(0); +#endif + } + + if( a->flags & 4 ) + xfree( a->d ); + else { +#ifdef M_DEBUG + mpi_debug_free_limb_space(a->d, "alloc_opaque"); +#else + mpi_free_limb_space(a->d); +#endif + } + + a->d = p; + a->alloced = 0; + a->nlimbs = 0; + a->nbits = len; + a->flags = 4; + return a; +} + + +void * +mpi_get_opaque( MPI a, unsigned int *len ) +{ + if( !(a->flags & 4) ) + log_bug("mpi_get_opaque on normal mpi\n"); + if( len ) + *len = a->nbits; + return a->d; +} + + +/**************** + * Note: This copy function should not interpret the MPI + * but copy it transparently. + */ +MPI +#ifdef M_DEBUG +mpi_debug_copy( MPI a, const char *info ) +#else +mpi_copy( MPI a ) +#endif +{ + int i; + MPI b; + + if( a && (a->flags & 4) ) { + void *p = m_is_secure(a->d)? xmalloc_secure( a->nbits ) + : xmalloc( a->nbits ); + memcpy( p, a->d, a->nbits ); + b = mpi_set_opaque( NULL, p, a->nbits ); + } + else if( a ) { +#ifdef M_DEBUG + b = mpi_is_secure(a)? mpi_debug_alloc_secure( a->nlimbs, info ) + : mpi_debug_alloc( a->nlimbs, info ); +#else + b = mpi_is_secure(a)? mpi_alloc_secure( a->nlimbs ) + : mpi_alloc( a->nlimbs ); +#endif + b->nlimbs = a->nlimbs; + b->sign = a->sign; + b->flags = a->flags; + b->nbits = a->nbits; + for(i=0; i < b->nlimbs; i++ ) + b->d[i] = a->d[i]; + } + else + b = NULL; + return b; +} + + +/**************** + * This function allocates an MPI which is optimized to hold + * a value as large as the one given in the arhgument and allocates it + * with the same flags as A. + */ +MPI +#ifdef M_DEBUG +mpi_debug_alloc_like( MPI a, const char *info ) +#else +mpi_alloc_like( MPI a ) +#endif +{ + MPI b; + + if( a && (a->flags & 4) ) { + void *p = m_is_secure(a->d)? xmalloc_secure( a->nbits ) + : xmalloc( a->nbits ); + memcpy( p, a->d, a->nbits ); + b = mpi_set_opaque( NULL, p, a->nbits ); + } + else if( a ) { +#ifdef M_DEBUG + b = mpi_is_secure(a)? mpi_debug_alloc_secure( a->nlimbs, info ) + : mpi_debug_alloc( a->nlimbs, info ); +#else + b = mpi_is_secure(a)? mpi_alloc_secure( a->nlimbs ) + : mpi_alloc( a->nlimbs ); +#endif + b->nlimbs = 0; + b->sign = 0; + b->flags = a->flags; + b->nbits = 0; + } + else + b = NULL; + return b; +} + + +void +mpi_set( MPI w, MPI u) +{ + mpi_ptr_t wp, up; + mpi_size_t usize = u->nlimbs; + int usign = u->sign; + + RESIZE_IF_NEEDED(w, usize); + wp = w->d; + up = u->d; + MPN_COPY( wp, up, usize ); + w->nlimbs = usize; + w->nbits = u->nbits; + w->flags = u->flags; + w->sign = usign; +} + + +void +mpi_set_ui( MPI w, unsigned long u) +{ + RESIZE_IF_NEEDED(w, 1); + w->d[0] = u; + w->nlimbs = u? 1:0; + w->sign = 0; + w->nbits = 0; + w->flags = 0; +} + + +MPI +mpi_alloc_set_ui( unsigned long u) +{ +#ifdef M_DEBUG + MPI w = mpi_debug_alloc(1,"alloc_set_ui"); +#else + MPI w = mpi_alloc(1); +#endif + w->d[0] = u; + w->nlimbs = u? 1:0; + w->sign = 0; + return w; +} + + +void +mpi_swap( MPI a, MPI b) +{ + struct gcry_mpi tmp; + + tmp = *a; *a = *b; *b = tmp; +} diff --git a/mpi/pa7100/distfiles b/mpi/pa7100/distfiles new file mode 100644 index 0000000..fece943 --- /dev/null +++ b/mpi/pa7100/distfiles @@ -0,0 +1,3 @@ +mpih-lshift.S +mpih-rshift.S + diff --git a/mpi/pa7100/mpih-lshift.S b/mpi/pa7100/mpih-lshift.S new file mode 100644 index 0000000..55b48fa --- /dev/null +++ b/mpi/pa7100/mpih-lshift.S @@ -0,0 +1,97 @@ +/* hppa lshift + * optimized for the PA7100, where it runs at 3.25 cycles/limb + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + + +/******************* + * mpi_limb_t + * mpihelp_lshift( mpi_ptr_t wp, (gr26) + * mpi_ptr_t up, (gr25) + * mpi_size_t usize, (gr24) + * unsigned cnt) (gr23) + */ + + .code + .export mpihelp_lshift + .label mpihelp_lshift + .proc + .callinfo frame=64,no_calls + .entry + + sh2add %r24,%r25,%r25 + sh2add %r24,%r26,%r26 + ldws,mb -4(0,%r25),%r22 + subi 32,%r23,%r1 + mtsar %r1 + addib,= -1,%r24,L$0004 + vshd %r0,%r22,%r28 ; compute carry out limb + ldws,mb -4(0,%r25),%r29 + addib,<= -5,%r24,L$rest + vshd %r22,%r29,%r20 + + .label L$loop + ldws,mb -4(0,%r25),%r22 + stws,mb %r20,-4(0,%r26) + vshd %r29,%r22,%r20 + ldws,mb -4(0,%r25),%r29 + stws,mb %r20,-4(0,%r26) + vshd %r22,%r29,%r20 + ldws,mb -4(0,%r25),%r22 + stws,mb %r20,-4(0,%r26) + vshd %r29,%r22,%r20 + ldws,mb -4(0,%r25),%r29 + stws,mb %r20,-4(0,%r26) + addib,> -4,%r24,L$loop + vshd %r22,%r29,%r20 + + .label L$rest + addib,= 4,%r24,L$end1 + nop + .label L$eloop + ldws,mb -4(0,%r25),%r22 + stws,mb %r20,-4(0,%r26) + addib,<= -1,%r24,L$end2 + vshd %r29,%r22,%r20 + ldws,mb -4(0,%r25),%r29 + stws,mb %r20,-4(0,%r26) + addib,> -1,%r24,L$eloop + vshd %r22,%r29,%r20 + + .label L$end1 + stws,mb %r20,-4(0,%r26) + vshd %r29,%r0,%r20 + bv 0(%r2) + stw %r20,-4(0,%r26) + .label L$end2 + stws,mb %r20,-4(0,%r26) + .label L$0004 + vshd %r22,%r0,%r20 + bv 0(%r2) + stw %r20,-4(0,%r26) + + .exit + .procend + + + diff --git a/mpi/pa7100/mpih-rshift.S b/mpi/pa7100/mpih-rshift.S new file mode 100644 index 0000000..81906f9 --- /dev/null +++ b/mpi/pa7100/mpih-rshift.S @@ -0,0 +1,93 @@ +/* hppa rshift + * optimized for the PA7100, where it runs at 3.25 cycles/limb + * + * Copyright (C) 1992, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + + +/******************* + * mpi_limb_t + * mpihelp_rshift( mpi_ptr_t wp, (gr26) + * mpi_ptr_t up, (gr25) + * mpi_size_t usize, (gr24) + * unsigned cnt) (gr23) + */ + + .code + .export mpihelp_rshift + .label mpihelp_rshift + .proc + .callinfo frame=64,no_calls + .entry + + ldws,ma 4(0,%r25),%r22 + mtsar %r23 + addib,= -1,%r24,L$r004 + vshd %r22,%r0,%r28 ; compute carry out limb + ldws,ma 4(0,%r25),%r29 + addib,<= -5,%r24,L$rrest + vshd %r29,%r22,%r20 + + .label L$roop + ldws,ma 4(0,%r25),%r22 + stws,ma %r20,4(0,%r26) + vshd %r22,%r29,%r20 + ldws,ma 4(0,%r25),%r29 + stws,ma %r20,4(0,%r26) + vshd %r29,%r22,%r20 + ldws,ma 4(0,%r25),%r22 + stws,ma %r20,4(0,%r26) + vshd %r22,%r29,%r20 + ldws,ma 4(0,%r25),%r29 + stws,ma %r20,4(0,%r26) + addib,> -4,%r24,L$roop + vshd %r29,%r22,%r20 + + .label L$rrest + addib,= 4,%r24,L$rend1 + nop + .label L$eroop + ldws,ma 4(0,%r25),%r22 + stws,ma %r20,4(0,%r26) + addib,<= -1,%r24,L$rend2 + vshd %r22,%r29,%r20 + ldws,ma 4(0,%r25),%r29 + stws,ma %r20,4(0,%r26) + addib,> -1,%r24,L$eroop + vshd %r29,%r22,%r20 + + .label L$rend1 + stws,ma %r20,4(0,%r26) + vshd %r0,%r29,%r20 + bv 0(%r2) + stw %r20,0(0,%r26) + .label L$rend2 + stws,ma %r20,4(0,%r26) + .label L$r004 + vshd %r0,%r22,%r20 + bv 0(%r2) + stw %r20,0(0,%r26) + + .exit + .procend + + diff --git a/mpi/power/distfiles b/mpi/power/distfiles new file mode 100644 index 0000000..e664c8d --- /dev/null +++ b/mpi/power/distfiles @@ -0,0 +1,7 @@ +mpih-add1.S +mpih-lshift.S +mpih-mul1.S +mpih-mul2.S +mpih-mul3.S +mpih-rshift.S +mpih-sub1.S diff --git a/mpi/power/mpih-add1.S b/mpi/power/mpih-add1.S new file mode 100644 index 0000000..ae8b1a6 --- /dev/null +++ b/mpi/power/mpih-add1.S @@ -0,0 +1,87 @@ +/* IBM POWER add_n -- Add two limb vectors of equal, non-zero length. + * + * Copyright (C) 1992, 1994, 1996, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + +/* +# INPUT PARAMETERS +# res_ptr r3 +# s1_ptr r4 +# s2_ptr r5 +# size r6 + */ + + .toc + .extern mpihelp_add_n[DS] + .extern .mpihelp_add_n +.csect [PR] + .align 2 + .globl mpihelp_add_n + .globl .mpihelp_add_n + .csect mpihelp_add_n[DS] +mpihelp_add_n: + .long .mpihelp_add_n, TOC[tc0], 0 + .csect [PR] +.mpihelp_add_n: + andil. 10,6,1 # odd or even number of limbs? + l 8,0(4) # load least significant s1 limb + l 0,0(5) # load least significant s2 limb + cal 3,-4(3) # offset res_ptr, it's updated before it's used + sri 10,6,1 # count for unrolled loop + a 7,0,8 # add least significant limbs, set cy + mtctr 10 # copy count into CTR + beq 0,Leven # branch if even # of limbs (# of limbs >= 2) + +# We have an odd # of limbs. Add the first limbs separately. + cmpi 1,10,0 # is count for unrolled loop zero? + bne 1,L1 # branch if not + st 7,4(3) + aze 3,10 # use the fact that r10 is zero... + br # return + +# We added least significant limbs. Now reload the next limbs to enter loop. +L1: lu 8,4(4) # load s1 limb and update s1_ptr + lu 0,4(5) # load s2 limb and update s2_ptr + stu 7,4(3) + ae 7,0,8 # add limbs, set cy +Leven: lu 9,4(4) # load s1 limb and update s1_ptr + lu 10,4(5) # load s2 limb and update s2_ptr + bdz Lend # If done, skip loop + +Loop: lu 8,4(4) # load s1 limb and update s1_ptr + lu 0,4(5) # load s2 limb and update s2_ptr + ae 11,9,10 # add previous limbs with cy, set cy + stu 7,4(3) # + lu 9,4(4) # load s1 limb and update s1_ptr + lu 10,4(5) # load s2 limb and update s2_ptr + ae 7,0,8 # add previous limbs with cy, set cy + stu 11,4(3) # + bdn Loop # decrement CTR and loop back + +Lend: ae 11,9,10 # add limbs with cy, set cy + st 7,4(3) # + st 11,8(3) # + lil 3,0 # load cy into ... + aze 3,3 # ... return value register + br + diff --git a/mpi/power/mpih-lshift.S b/mpi/power/mpih-lshift.S new file mode 100644 index 0000000..07eb929 --- /dev/null +++ b/mpi/power/mpih-lshift.S @@ -0,0 +1,65 @@ +/* IBM POWER lshift + * + * Copyright (C) 1992, 1994, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + +/* +# INPUT PARAMETERS +# res_ptr r3 +# s_ptr r4 +# size r5 +# cnt r6 + */ + + .toc + .extern mpihelp_lshift[DS] + .extern .mpihelp_lshift +.csect [PR] + .align 2 + .globl mpihelp_lshift + .globl .mpihelp_lshift + .csect mpihelp_lshift[DS] +mpihelp_lshift: + .long .mpihelp_lshift, TOC[tc0], 0 + .csect [PR] +.mpihelp_lshift: + sli 0,5,2 + cax 9,3,0 + cax 4,4,0 + sfi 8,6,32 + mtctr 5 # put limb count in CTR loop register + lu 0,-4(4) # read most significant limb + sre 3,0,8 # compute carry out limb, and init MQ register + bdz Lend2 # if just one limb, skip loop + lu 0,-4(4) # read 2:nd most significant limb + sreq 7,0,8 # compute most significant limb of result + bdz Lend # if just two limb, skip loop +Loop: lu 0,-4(4) # load next lower limb + stu 7,-4(9) # store previous result during read latency + sreq 7,0,8 # compute result limb + bdn Loop # loop back until CTR is zero +Lend: stu 7,-4(9) # store 2:nd least significant limb +Lend2: sle 7,0,6 # compute least significant limb + st 7,-4(9) # store it + br + diff --git a/mpi/power/mpih-mul1.S b/mpi/power/mpih-mul1.S new file mode 100644 index 0000000..214dbce --- /dev/null +++ b/mpi/power/mpih-mul1.S @@ -0,0 +1,116 @@ +/* IBM POWER mul_1 -- Multiply a limb vector with a limb and store + * the result in a second limb vector. + * + * Copyright (C) 1992, 1994, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + +/* +# INPUT PARAMETERS +# res_ptr r3 +# s1_ptr r4 +# size r5 +# s2_limb r6 + +# The RS/6000 has no unsigned 32x32->64 bit multiplication instruction. To +# obtain that operation, we have to use the 32x32->64 signed multiplication +# instruction, and add the appropriate compensation to the high limb of the +# result. We add the multiplicand if the multiplier has its most significant +# bit set, and we add the multiplier if the multiplicand has its most +# significant bit set. We need to preserve the carry flag between each +# iteration, so we have to compute the compensation carefully (the natural, +# srai+and doesn't work). Since the POWER architecture has a branch unit +# we can branch in zero cycles, so that's how we perform the additions. + */ + + .toc + .csect .mpihelp_mul_1[PR] + .align 2 + .globl mpihelp_mul_1 + .globl .mpihelp_mul_1 + .csect mpihelp_mul_1[DS] +mpihelp_mul_1: + .long .mpihelp_mul_1[PR], TOC[tc0], 0 + .csect .mpihelp_mul_1[PR] +.mpihelp_mul_1: + + cal 3,-4(3) + l 0,0(4) + cmpi 0,6,0 + mtctr 5 + mul 9,0,6 + srai 7,0,31 + and 7,7,6 + mfmq 8 + ai 0,0,0 # reset carry + cax 9,9,7 + blt Lneg +Lpos: bdz Lend +Lploop: lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 10,0,6 + mfmq 0 + ae 8,0,9 + bge Lp0 + cax 10,10,6 # adjust high limb for negative limb from s1 +Lp0: bdz Lend0 + lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 9,0,6 + mfmq 0 + ae 8,0,10 + bge Lp1 + cax 9,9,6 # adjust high limb for negative limb from s1 +Lp1: bdn Lploop + b Lend + +Lneg: cax 9,9,0 + bdz Lend +Lnloop: lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 10,0,6 + cax 10,10,0 # adjust high limb for negative s2_limb + mfmq 0 + ae 8,0,9 + bge Ln0 + cax 10,10,6 # adjust high limb for negative limb from s1 +Ln0: bdz Lend0 + lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 9,0,6 + cax 9,9,0 # adjust high limb for negative s2_limb + mfmq 0 + ae 8,0,10 + bge Ln1 + cax 9,9,6 # adjust high limb for negative limb from s1 +Ln1: bdn Lnloop + b Lend + +Lend0: cal 9,0(10) +Lend: st 8,4(3) + aze 3,9 + br + diff --git a/mpi/power/mpih-mul2.S b/mpi/power/mpih-mul2.S new file mode 100644 index 0000000..571d54e --- /dev/null +++ b/mpi/power/mpih-mul2.S @@ -0,0 +1,131 @@ +/* IBM POWER addmul_1 -- Multiply a limb vector with a limb and add + * the result to a second limb vector. + * + * Copyright (C) 1992, 1994, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + + + +/* +# INPUT PARAMETERS +# res_ptr r3 +# s1_ptr r4 +# size r5 +# s2_limb r6 + +# The RS/6000 has no unsigned 32x32->64 bit multiplication instruction. To +# obtain that operation, we have to use the 32x32->64 signed multiplication +# instruction, and add the appropriate compensation to the high limb of the +# result. We add the multiplicand if the multiplier has its most significant +# bit set, and we add the multiplier if the multiplicand has its most +# significant bit set. We need to preserve the carry flag between each +# iteration, so we have to compute the compensation carefully (the natural, +# srai+and doesn't work). Since the POWER architecture has a branch unit +# we can branch in zero cycles, so that's how we perform the additions. + */ + + .toc + .csect .mpihelp_addmul_1[PR] + .align 2 + .globl mpihelp_addmul_1 + .globl .mpihelp_addmul_1 + .csect mpihelp_addmul_1[DS] +mpihelp_addmul_1: + .long .mpihelp_addmul_1[PR], TOC[tc0], 0 + .csect .mpihelp_addmul_1[PR] +.mpihelp_addmul_1: + + cal 3,-4(3) + l 0,0(4) + cmpi 0,6,0 + mtctr 5 + mul 9,0,6 + srai 7,0,31 + and 7,7,6 + mfmq 8 + cax 9,9,7 + l 7,4(3) + a 8,8,7 # add res_limb + blt Lneg +Lpos: bdz Lend + +Lploop: lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 10,0,6 + mfmq 0 + ae 8,0,9 # low limb + old_cy_limb + old cy + l 7,4(3) + aze 10,10 # propagate cy to new cy_limb + a 8,8,7 # add res_limb + bge Lp0 + cax 10,10,6 # adjust high limb for negative limb from s1 +Lp0: bdz Lend0 + lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 9,0,6 + mfmq 0 + ae 8,0,10 + l 7,4(3) + aze 9,9 + a 8,8,7 + bge Lp1 + cax 9,9,6 # adjust high limb for negative limb from s1 +Lp1: bdn Lploop + + b Lend + +Lneg: cax 9,9,0 + bdz Lend +Lnloop: lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 10,0,6 + mfmq 7 + ae 8,7,9 + l 7,4(3) + ae 10,10,0 # propagate cy to new cy_limb + a 8,8,7 # add res_limb + bge Ln0 + cax 10,10,6 # adjust high limb for negative limb from s1 +Ln0: bdz Lend0 + lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 9,0,6 + mfmq 7 + ae 8,7,10 + l 7,4(3) + ae 9,9,0 # propagate cy to new cy_limb + a 8,8,7 # add res_limb + bge Ln1 + cax 9,9,6 # adjust high limb for negative limb from s1 +Ln1: bdn Lnloop + b Lend + +Lend0: cal 9,0(10) +Lend: st 8,4(3) + aze 3,9 + br + diff --git a/mpi/power/mpih-mul3.S b/mpi/power/mpih-mul3.S new file mode 100644 index 0000000..192c6f4 --- /dev/null +++ b/mpi/power/mpih-mul3.S @@ -0,0 +1,136 @@ +/* IBM POWER submul_1 -- Multiply a limb vector with a limb and subtract + * the result from a second limb vector. + * + * Copyright (C) 1992, 1994, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + + +/* + +# INPUT PARAMETERS +# res_ptr r3 +# s1_ptr r4 +# size r5 +# s2_limb r6 + +# The RS/6000 has no unsigned 32x32->64 bit multiplication instruction. To +# obtain that operation, we have to use the 32x32->64 signed multiplication +# instruction, and add the appropriate compensation to the high limb of the +# result. We add the multiplicand if the multiplier has its most significant +# bit set, and we add the multiplier if the multiplicand has its most +# significant bit set. We need to preserve the carry flag between each +# iteration, so we have to compute the compensation carefully (the natural, +# srai+and doesn't work). Since the POWER architecture has a branch unit +# we can branch in zero cycles, so that's how we perform the additions. + */ + + .toc + .csect .mpihelp_submul_1[PR] + .align 2 + .globl mpihelp_submul_1 + .globl .mpihelp_submul_1 + .csect mpihelp_submul_1[DS] +mpihelp_submul_1: + .long .mpihelp_submul_1[PR], TOC[tc0], 0 + .csect .mpihelp_submul_1[PR] +.mpihelp_submul_1: + + cal 3,-4(3) + l 0,0(4) + cmpi 0,6,0 + mtctr 5 + mul 9,0,6 + srai 7,0,31 + and 7,7,6 + mfmq 11 + cax 9,9,7 + l 7,4(3) + sf 8,11,7 # add res_limb + a 11,8,11 # invert cy (r11 is junk) + blt Lneg +Lpos: bdz Lend + +Lploop: lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 10,0,6 + mfmq 0 + ae 11,0,9 # low limb + old_cy_limb + old cy + l 7,4(3) + aze 10,10 # propagate cy to new cy_limb + sf 8,11,7 # add res_limb + a 11,8,11 # invert cy (r11 is junk) + bge Lp0 + cax 10,10,6 # adjust high limb for negative limb from s1 +Lp0: bdz Lend0 + lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 9,0,6 + mfmq 0 + ae 11,0,10 + l 7,4(3) + aze 9,9 + sf 8,11,7 + a 11,8,11 # invert cy (r11 is junk) + bge Lp1 + cax 9,9,6 # adjust high limb for negative limb from s1 +Lp1: bdn Lploop + + b Lend + +Lneg: cax 9,9,0 + bdz Lend +Lnloop: lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 10,0,6 + mfmq 7 + ae 11,7,9 + l 7,4(3) + ae 10,10,0 # propagate cy to new cy_limb + sf 8,11,7 # add res_limb + a 11,8,11 # invert cy (r11 is junk) + bge Ln0 + cax 10,10,6 # adjust high limb for negative limb from s1 +Ln0: bdz Lend0 + lu 0,4(4) + stu 8,4(3) + cmpi 0,0,0 + mul 9,0,6 + mfmq 7 + ae 11,7,10 + l 7,4(3) + ae 9,9,0 # propagate cy to new cy_limb + sf 8,11,7 # add res_limb + a 11,8,11 # invert cy (r11 is junk) + bge Ln1 + cax 9,9,6 # adjust high limb for negative limb from s1 +Ln1: bdn Lnloop + b Lend + +Lend0: cal 9,0(10) +Lend: st 8,4(3) + aze 3,9 + br + diff --git a/mpi/power/mpih-rshift.S b/mpi/power/mpih-rshift.S new file mode 100644 index 0000000..3214f86 --- /dev/null +++ b/mpi/power/mpih-rshift.S @@ -0,0 +1,65 @@ +/* IBM POWER rshift + * + * Copyright (C) 1992, 1994, 1999 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + + +/* +# INPUT PARAMETERS +# res_ptr r3 +# s_ptr r4 +# size r5 +# cnt r6 +*/ + + .toc + .extern mpihelp_rshift[DS] + .extern .mpihelp_rshift +.csect [PR] + .align 2 + .globl mpihelp_rshift + .globl .mpihelp_rshift + .csect mpihelp_rshift[DS] +mpihelp_rshift: + .long .mpihelp_rshift, TOC[tc0], 0 + .csect [PR] +.mpihelp_rshift: + sfi 8,6,32 + mtctr 5 # put limb count in CTR loop register + l 0,0(4) # read least significant limb + ai 9,3,-4 # adjust res_ptr since it's offset in the stu:s + sle 3,0,8 # compute carry limb, and init MQ register + bdz Lend2 # if just one limb, skip loop + lu 0,4(4) # read 2:nd least significant limb + sleq 7,0,8 # compute least significant limb of result + bdz Lend # if just two limb, skip loop +Loop: lu 0,4(4) # load next higher limb + stu 7,4(9) # store previous result during read latency + sleq 7,0,8 # compute result limb + bdn Loop # loop back until CTR is zero +Lend: stu 7,4(9) # store 2:nd most significant limb +Lend2: sre 7,0,6 # compute most significant limb + st 7,4(9) # store it + br + + diff --git a/mpi/power/mpih-sub1.S b/mpi/power/mpih-sub1.S new file mode 100644 index 0000000..d7703ee --- /dev/null +++ b/mpi/power/mpih-sub1.S @@ -0,0 +1,89 @@ +/* IBM POWER sub_n -- Subtract two limb vectors of equal, non-zero length. + * + * Copyright (C) 1992, 1994, 1995, 1996, 1999 + * Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + +/* +# INPUT PARAMETERS +# res_ptr r3 +# s1_ptr r4 +# s2_ptr r5 +# size r6 + */ + + .toc + .extern mpihelp_sub_n[DS] + .extern .mpihelp_sub_n +.csect [PR] + .align 2 + .globl mpihelp_sub_n + .globl .mpihelp_sub_n + .csect mpihelp_sub_n[DS] +mpihelp_sub_n: + .long .mpihelp_sub_n, TOC[tc0], 0 + .csect [PR] +.mpihelp_sub_n: + andil. 10,6,1 # odd or even number of limbs? + l 8,0(4) # load least significant s1 limb + l 0,0(5) # load least significant s2 limb + cal 3,-4(3) # offset res_ptr, it's updated before it's used + sri 10,6,1 # count for unrolled loop + sf 7,0,8 # subtract least significant limbs, set cy + mtctr 10 # copy count into CTR + beq 0,Leven # branch if even # of limbs (# of limbs >= 2) + +# We have an odd # of limbs. Add the first limbs separately. + cmpi 1,10,0 # is count for unrolled loop zero? + bne 1,L1 # branch if not + st 7,4(3) + sfe 3,0,0 # load !cy into ... + sfi 3,3,0 # ... return value register + br # return + +# We added least significant limbs. Now reload the next limbs to enter loop. +L1: lu 8,4(4) # load s1 limb and update s1_ptr + lu 0,4(5) # load s2 limb and update s2_ptr + stu 7,4(3) + sfe 7,0,8 # subtract limbs, set cy +Leven: lu 9,4(4) # load s1 limb and update s1_ptr + lu 10,4(5) # load s2 limb and update s2_ptr + bdz Lend # If done, skip loop + +Loop: lu 8,4(4) # load s1 limb and update s1_ptr + lu 0,4(5) # load s2 limb and update s2_ptr + sfe 11,10,9 # subtract previous limbs with cy, set cy + stu 7,4(3) # + lu 9,4(4) # load s1 limb and update s1_ptr + lu 10,4(5) # load s2 limb and update s2_ptr + sfe 7,0,8 # subtract previous limbs with cy, set cy + stu 11,4(3) # + bdn Loop # decrement CTR and loop back + +Lend: sfe 11,10,9 # subtract limbs with cy, set cy + st 7,4(3) # + st 11,8(3) # + sfe 3,0,0 # load !cy into ... + sfi 3,3,0 # ... return value register + br + diff --git a/mpi/powerpc32/distfiles b/mpi/powerpc32/distfiles new file mode 100644 index 0000000..af10d79 --- /dev/null +++ b/mpi/powerpc32/distfiles @@ -0,0 +1,9 @@ +mpih-add1.S +mpih-sub1.S +mpih-mul1.S +mpih-mul2.S +mpih-mul3.S +mpih-lshift.S +mpih-rshift.S +syntax.h + diff --git a/mpi/powerpc32/mpih-add1.S b/mpi/powerpc32/mpih-add1.S new file mode 100644 index 0000000..d8f78da --- /dev/null +++ b/mpi/powerpc32/mpih-add1.S @@ -0,0 +1,137 @@ +/* PowerPC-32 add_n -- Add two limb vectors of equal, non-zero length. + * + * Copyright (C) 1992, 1994, 1995, 1998 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + + +#ifndef USE_PPC_PATCHES + +/******************* + * mpi_limb_t + * mpihelp_add_n( mpi_ptr_t res_ptr, (r3) + * mpi_ptr_t s1_ptr, (r4) + * mpi_ptr_t s2_ptr, (r5) + * mpi_size_t size) (r6) + */ + + .toc + .extern mpihelp_add_n[DS] + .extern .mpihelp_add_n +.csect [PR] + .align 2 + .globl mpihelp_add_n + .globl .mpihelp_add_n + .csect mpihelp_add_n[DS] +mpihelp_add_n: + .long .mpihelp_add_n, TOC[tc0], 0 + .csect [PR] +.mpihelp_add_n: + mtctr 6 # copy size into CTR + lwz 8,0(4) # load least significant s1 limb + lwz 0,0(5) # load least significant s2 limb + addi 3,3,-4 # offset res_ptr, it is updated before used + addc 7,0,8 # add least significant limbs, set cy + bdz Lend # If done, skip loop +Loop: lwzu 8,4(4) # load s1 limb and update s1_ptr + lwzu 0,4(5) # load s2 limb and update s2_ptr + stwu 7,4(3) # store previous limb in load latency slot + adde 7,0,8 # add new limbs with cy, set cy + bdnz Loop # decrement CTR and loop back +Lend: stw 7,4(3) # store ultimate result limb + li 3,0 # load cy into ... + addze 3,3 # ... return value register + blr + +#else +/* Add two limb vectors of equal, non-zero length for PowerPC. + Copyright (C) 1997 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Library General Public License as + published by the Free Software Foundation; either version 2 of the + License, or (at your option) any later version. + + 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 + Library General Public License for more details. + + You should have received a copy of the GNU Library General Public + License along with the GNU C Library; see the file COPYING.LIB. If not, + write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, + Boston, MA 02110-1301, USA. */ + +#include "sysdep.h" +#include "asm-syntax.h" + + +/* mp_limb_t mpn_add_n (mp_ptr res_ptr, mp_srcptr s1_ptr, mp_srcptr s2_ptr, + mp_size_t size) + Calculate s1+s2 and put result in res_ptr; return carry, 0 or 1. */ + +/* Note on optimisation: This code is optimal for the 601. Almost every other + possible 2-unrolled inner loop will not be. Also, watch out for the + alignment... */ + +EALIGN(mpihelp_add_n,3,0) +/* Set up for loop below. */ + mtcrf 0x01,%r6 + srwi. %r7,%r6,1 + li %r10,0 + mtctr %r7 + bt 31,2f + +/* Clear the carry. */ + addic %r0,%r0,0 +/* Adjust pointers for loop. */ + addi %r3,%r3,-4 + addi %r4,%r4,-4 + addi %r5,%r5,-4 + b 0f + +2: lwz %r7,0(%r5) + lwz %r6,0(%r4) + addc %r6,%r6,%r7 + stw %r6,0(%r3) + beq 1f + +/* The loop. */ + +/* Align start of loop to an odd word boundary to guarantee that the + last two words can be fetched in one access (for 601). */ +0: lwz %r9,4(%r4) + lwz %r8,4(%r5) + lwzu %r6,8(%r4) + lwzu %r7,8(%r5) + adde %r8,%r9,%r8 + stw %r8,4(%r3) + adde %r6,%r6,%r7 + stwu %r6,8(%r3) + bdnz 0b +/* Return the carry. */ +1: addze %r3,%r10 + blr +END(mpihelp_add_n) +#endif + diff --git a/mpi/powerpc32/mpih-lshift.S b/mpi/powerpc32/mpih-lshift.S new file mode 100644 index 0000000..4e5da5b --- /dev/null +++ b/mpi/powerpc32/mpih-lshift.S @@ -0,0 +1,199 @@ +/* PowerPC-32 lshift + * + * Copyright (C) 1995, 1998 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + + +#ifndef USE_PPC_PATCHES + +/******************* + * mpi_limb_t + * mpihelp_lshift( mpi_ptr_t wp, (r3) + * mpi_ptr_t up, (r4) + * mpi_size_t usize, (r5) + * unsigned cnt) (r6) + */ + + .toc +.csect .text[PR] + .align 2 + .globl mpihelp_lshift + .globl .mpihelp_lshift + .csect mpihelp_lshift[DS] +mpihelp_lshift: + .long .mpihelp_lshift, TOC[tc0], 0 + .csect .text[PR] +.mpihelp_lshift: + mtctr 5 # copy size into CTR + slwi 0,5,2 + add 7,3,0 # make r7 point at end of res + add 4,4,0 # make r4 point at end of s1 + subfic 8,6,32 + lwzu 11,-4(4) # load first s1 limb + srw 3,11,8 # compute function return value + bdz Lend1 + +Loop: lwzu 10,-4(4) + slw 9,11,6 + srw 12,10,8 + or 9,9,12 + stwu 9,-4(7) + bdz Lend2 + lwzu 11,-4(4) + slw 9,10,6 + srw 12,11,8 + or 9,9,12 + stwu 9,-4(7) + bdnz Loop + +Lend1: slw 0,11,6 + stw 0,-4(7) + blr + +Lend2: slw 0,10,6 + stw 0,-4(7) + blr + +#else +/* Shift a limb left, low level routine. + Copyright (C) 1996, 1997 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Library General Public License as + published by the Free Software Foundation; either version 2 of the + License, or (at your option) any later version. + + 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 + Library General Public License for more details. + + You should have received a copy of the GNU Library General Public + License along with the GNU C Library; see the file COPYING.LIB. If not, + write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, + Boston, MA 02110-1301, USA. */ + +/* mp_limb_t mpn_lshift (mp_ptr wp, mp_srcptr up, mp_size_t usize, + unsigned int cnt) */ + +EALIGN(mpihelp_lshift,3,0) + mtctr %r5 # copy size into CTR + cmplwi %cr0,%r5,16 # is size < 16 + slwi %r0,%r5,2 + add %r7,%r3,%r0 # make r7 point at end of res + add %r4,%r4,%r0 # make r4 point at end of s1 + lwzu %r11,-4(%r4) # load first s1 limb + subfic %r8,%r6,32 + srw %r3,%r11,%r8 # compute function return value + bge %cr0,L(big) # branch if size >= 16 + + bdz L(end1) + +0: lwzu %r10,-4(%r4) + slw %r9,%r11,%r6 + srw %r12,%r10,%r8 + or %r9,%r9,%r12 + stwu %r9,-4(%r7) + bdz L(end2) + lwzu %r11,-4(%r4) + slw %r9,%r10,%r6 + srw %r12,%r11,%r8 + or %r9,%r9,%r12 + stwu %r9,-4(%r7) + bdnz 0b + +L(end1):slw %r0,%r11,%r6 + stw %r0,-4(%r7) + blr + + +/* Guaranteed not to succeed. */ +L(boom): tweq %r0,%r0 + +/* We imitate a case statement, by using (yuk!) fixed-length code chunks, + of size 4*12 bytes. We have to do this (or something) to make this PIC. */ +L(big): mflr %r9 + bltl- %cr0,L(boom) # Never taken, only used to set LR. + slwi %r10,%r6,4 + mflr %r12 + add %r10,%r12,%r10 + slwi %r8,%r6,5 + add %r10,%r8,%r10 + mtctr %r10 + addi %r5,%r5,-1 + mtlr %r9 + bctr + +L(end2):slw %r0,%r10,%r6 + stw %r0,-4(%r7) + blr + +#define DO_LSHIFT(n) \ + mtctr %r5; \ +0: lwzu %r10,-4(%r4); \ + slwi %r9,%r11,n; \ + inslwi %r9,%r10,n,32-n; \ + stwu %r9,-4(%r7); \ + bdz- L(end2); \ + lwzu %r11,-4(%r4); \ + slwi %r9,%r10,n; \ + inslwi %r9,%r11,n,32-n; \ + stwu %r9,-4(%r7); \ + bdnz 0b; \ + b L(end1) + + DO_LSHIFT(1) + DO_LSHIFT(2) + DO_LSHIFT(3) + DO_LSHIFT(4) + DO_LSHIFT(5) + DO_LSHIFT(6) + DO_LSHIFT(7) + DO_LSHIFT(8) + DO_LSHIFT(9) + DO_LSHIFT(10) + DO_LSHIFT(11) + DO_LSHIFT(12) + DO_LSHIFT(13) + DO_LSHIFT(14) + DO_LSHIFT(15) + DO_LSHIFT(16) + DO_LSHIFT(17) + DO_LSHIFT(18) + DO_LSHIFT(19) + DO_LSHIFT(20) + DO_LSHIFT(21) + DO_LSHIFT(22) + DO_LSHIFT(23) + DO_LSHIFT(24) + DO_LSHIFT(25) + DO_LSHIFT(26) + DO_LSHIFT(27) + DO_LSHIFT(28) + DO_LSHIFT(29) + DO_LSHIFT(30) + DO_LSHIFT(31) + +END(mpihelp_lshift) +#endif diff --git a/mpi/powerpc32/mpih-mul1.S b/mpi/powerpc32/mpih-mul1.S new file mode 100644 index 0000000..9f251f1 --- /dev/null +++ b/mpi/powerpc32/mpih-mul1.S @@ -0,0 +1,121 @@ +/* PowerPC-32 mul_1 -- Multiply a limb vector with a limb and store + * the result in a second limb vector. + * + * Copyright (C) 1992, 1993, 1994, 1995, + * 1998 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + + +#ifndef USE_PPC_PATCHES + +/******************* + * mpi_limb_t + * mpihelp_mul_1( mpi_ptr_t res_ptr, (r3) + * mpi_ptr_t s1_ptr, (r4) + * mpi_size_t s1_size, (r5) + * mpi_limb_t s2_limb) (r6) + * + * This is a fairly straightforward implementation. The timing of the PC601 + * is hard to understand, so I will wait to optimize this until I have some + * hardware to play with. + * + * The code trivially generalizes to 64 bit limbs for the PC620. + */ + + .toc + .csect .mpihelp_mul_1[PR] + .align 2 + .globl mpihelp_mul_1 + .globl .mpihelp_mul_1 + .csect mpihelp_mul_1[DS] +mpihelp_mul_1: + .long .mpihelp_mul_1[PR], TOC[tc0], 0 + .csect .mpihelp_mul_1[PR] +.mpihelp_mul_1: + mtctr 5 + + lwz 0,0(4) + mullw 7,0,6 + mulhwu 10,0,6 + addi 3,3,-4 # adjust res_ptr + addic 5,5,0 # clear cy with dummy insn + bdz Lend + +Loop: lwzu 0,4(4) + stwu 7,4(3) + mullw 8,0,6 + adde 7,8,10 + mulhwu 10,0,6 + bdnz Loop + +Lend: stw 7,4(3) + addze 3,10 + blr + +#else +/* Multiply a limb vector by a limb, for PowerPC. + Copyright (C) 1993, 1994, 1995, 1997 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Library General Public License as + published by the Free Software Foundation; either version 2 of the + License, or (at your option) any later version. + + 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 + Library General Public License for more details. + + You should have received a copy of the GNU Library General Public + License along with the GNU C Library; see the file COPYING.LIB. If not, + write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, + Boston, MA 02110-1301, USA. */ + + +/* mp_limb_t mpn_mul_1 (mp_ptr res_ptr, mp_srcptr s1_ptr, + mp_size_t s1_size, mp_limb_t s2_limb) + Calculate s1*s2 and put result in res_ptr; return carry. */ + +ENTRY(mpihelp_mul_1) + mtctr %r5 + + lwz %r0,0(%r4) + mullw %r7,%r0,%r6 + mulhwu %r10,%r0,%r6 + addi %r3,%r3,-4 # adjust res_ptr + addic %r5,%r5,0 # clear cy with dummy insn + bdz 1f + +0: lwzu %r0,4(%r4) + stwu %r7,4(%r3) + mullw %r8,%r0,%r6 + adde %r7,%r8,%r10 + mulhwu %r10,%r0,%r6 + bdnz 0b + +1: stw %r7,4(%r3) + addze %r3,%r10 + blr +END(mpihelp_mul_1) +#endif diff --git a/mpi/powerpc32/mpih-mul2.S b/mpi/powerpc32/mpih-mul2.S new file mode 100644 index 0000000..b562ea6 --- /dev/null +++ b/mpi/powerpc32/mpih-mul2.S @@ -0,0 +1,128 @@ +/* PowerPC-32 addmul_1 -- Multiply a limb vector with a limb and add + * the result to a second limb vector. + * + * Copyright (C) 1995, 1998 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + + +#ifndef USE_PPC_PATCHES + +/******************* + * mpi_limb_t + * mpihelp_addmul_1( mpi_ptr_t res_ptr, (r3) + * mpi_ptr_t s1_ptr, (r4) + * mpi_size_t s1_size, (r5) + * mpi_limb_t s2_limb) (r6) + * + * This is a fairly straightforward implementation. The timing of the PC601 + * is hard to understand, so I will wait to optimize this until I have some + * hardware to play with. + * + * The code trivially generalizes to 64 bit limbs for the PC620. + */ + + + .toc + .csect .mpihelp_addmul_1[PR] + .align 2 + .globl mpihelp_addmul_1 + .globl .mpihelp_addmul_1 + .csect mpihelp_addmul_1[DS] +mpihelp_addmul_1: + .long .mpihelp_addmul_1[PR], TOC[tc0], 0 + .csect .mpihelp_addmul_1[PR] +.mpihelp_addmul_1: + mtctr 5 + + lwz 0,0(4) + mullw 7,0,6 + mulhwu 10,0,6 + lwz 9,0(3) + addc 8,7,9 + addi 3,3,-4 + bdz Lend + +Loop: lwzu 0,4(4) + stwu 8,4(3) + mullw 8,0,6 + adde 7,8,10 + mulhwu 10,0,6 + lwz 9,4(3) + addze 10,10 + addc 8,7,9 + bdnz Loop + +Lend: stw 8,4(3) + addze 3,10 + blr + +#else +/* Multiply a limb vector by a single limb, for PowerPC. + Copyright (C) 1993, 1994, 1995, 1997 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Library General Public License as + published by the Free Software Foundation; either version 2 of the + License, or (at your option) any later version. + + 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 + Library General Public License for more details. + + You should have received a copy of the GNU Library General Public + License along with the GNU C Library; see the file COPYING.LIB. If not, + write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, + Boston, MA 02110-1301, USA. */ + + +/* mp_limb_t mpn_addmul_1 (mp_ptr res_ptr, mp_srcptr s1_ptr, + mp_size_t s1_size, mp_limb_t s2_limb) + Calculate res+s1*s2 and put result back in res; return carry. */ +ENTRY(mpihelp_addmul_1) + mtctr %r5 + + lwz %r0,0(%r4) + mullw %r7,%r0,%r6 + mulhwu %r10,%r0,%r6 + lwz %r9,0(%r3) + addc %r8,%r7,%r9 + addi %r3,%r3,-4 /* adjust res_ptr */ + bdz 1f + +0: lwzu %r0,4(%r4) + stwu %r8,4(%r3) + mullw %r8,%r0,%r6 + adde %r7,%r8,%r10 + mulhwu %r10,%r0,%r6 + lwz %r9,4(%r3) + addze %r10,%r10 + addc %r8,%r7,%r9 + bdnz 0b + +1: stw %r8,4(%r3) + addze %r3,%r10 + blr +END(mpihelp_addmul_1) +#endif diff --git a/mpi/powerpc32/mpih-mul3.S b/mpi/powerpc32/mpih-mul3.S new file mode 100644 index 0000000..3b813f9 --- /dev/null +++ b/mpi/powerpc32/mpih-mul3.S @@ -0,0 +1,131 @@ +/* PowerPC-32 submul_1 -- Multiply a limb vector with a limb and subtract + * the result from a second limb vector. + * + * Copyright (C) 1995, 1998 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + + +#ifndef USE_PPC_PATCHES + +/******************* + * mpi_limb_t + * mpihelp_submul_1( mpi_ptr_t res_ptr, (r3) + * mpi_ptr_t s1_ptr, (r4) + * mpi_size_t s1_size, (r5) + * mpi_limb_t s2_limb) (r6) + * + * This is a fairly straightforward implementation. The timing of the PC601 + * is hard to understand, so I will wait to optimize this until I have some + * hardware to play with. + * + * The code trivially generalizes to 64 bit limbs for the PC620. + */ + + .toc + .csect .mpihelp_submul_1[PR] + .align 2 + .globl mpihelp_submul_1 + .globl .mpihelp_submul_1 + .csect mpihelp_submul_1[DS] +mpihelp_submul_1: + .long .mpihelp_submul_1[PR], TOC[tc0], 0 + .csect .mpihelp_submul_1[PR] +.mpihelp_submul_1: + mtctr 5 + + lwz 0,0(4) + mullw 7,0,6 + mulhwu 10,0,6 + lwz 9,0(3) + subfc 8,7,9 + addc 7,7,8 # invert cy (r7 is junk) + addi 3,3,-4 + bdz Lend + +Loop: lwzu 0,4(4) + stwu 8,4(3) + mullw 8,0,6 + adde 7,8,10 + mulhwu 10,0,6 + lwz 9,4(3) + addze 10,10 + subfc 8,7,9 + addc 7,7,8 # invert cy (r7 is junk) + bdnz Loop + +Lend: stw 8,4(3) + addze 3,10 + blr + +#else +/* Multiply a limb vector by a single limb, for PowerPC. + Copyright (C) 1993, 1994, 1995, 1997 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Library General Public License as + published by the Free Software Foundation; either version 2 of the + License, or (at your option) any later version. + + 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 + Library General Public License for more details. + + You should have received a copy of the GNU Library General Public + License along with the GNU C Library; see the file COPYING.LIB. If not, + write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, + Boston, MA 02110-1301, USA. */ + +/* mp_limb_t mpn_submul_1 (mp_ptr res_ptr, mp_srcptr s1_ptr, + mp_size_t s1_size, mp_limb_t s2_limb) + Calculate res-s1*s2 and put result back in res; return carry. */ + +ENTRY(mpihelp_submul_1) + mtctr %r5 + + lwz %r0,0(%r4) + mullw %r7,%r0,%r6 + mulhwu %r10,%r0,%r6 + lwz %r9,0(%r3) + subf %r8,%r7,%r9 + addc %r7,%r7,%r8 # invert cy (r7 is junk) + addi %r3,%r3,-4 # adjust res_ptr + bdz 1f + +0: lwzu %r0,4(%r4) + stwu %r8,4(%r3) + mullw %r8,%r0,%r6 + adde %r7,%r8,%r10 + mulhwu %r10,%r0,%r6 + lwz %r9,4(%r3) + addze %r10,%r10 + subf %r8,%r7,%r9 + addc %r7,%r7,%r8 # invert cy (r7 is junk) + bdnz 0b + +1: stw %r8,4(%r3) + addze %r3,%r10 + blr +END(mpihelp_submul_1) +#endif diff --git a/mpi/powerpc32/mpih-rshift.S b/mpi/powerpc32/mpih-rshift.S new file mode 100644 index 0000000..4a826b4 --- /dev/null +++ b/mpi/powerpc32/mpih-rshift.S @@ -0,0 +1,132 @@ +/* PowerPC-32 rshift + * + * Copyright (C) 1995, 1998 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +#include "sysdep.h" +#include "asm-syntax.h" + + +#ifndef USE_PPC_PATCHES + +/******************* + * mpi_limb_t + * mpihelp_rshift( mpi_ptr_t wp, (r3) + * mpi_ptr_t up, (r4) + * mpi_size_t usize, (r5) + * unsigned cnt) (r6) + */ + + .toc +.csect .text[PR] + .align 2 + .globl mpihelp_rshift + .globl .mpihelp_rshift + .csect mpihelp_rshift[DS] +mpihelp_rshift: + .long .mpihelp_rshift, TOC[tc0], 0 + .csect .text[PR] +.mpihelp_rshift: + mtctr 5 # copy size into CTR + addi 7,3,-4 # move adjusted res_ptr to free return reg + subfic 8,6,32 + lwz 11,0(4) # load first s1 limb + slw 3,11,8 # compute function return value + bdz Lend1 + +Loop: lwzu 10,4(4) + srw 9,11,6 + slw 12,10,8 + or 9,9,12 + stwu 9,4(7) + bdz Lend2 + lwzu 11,4(4) + srw 9,10,6 + slw 12,11,8 + or 9,9,12 + stwu 9,4(7) + bdnz Loop + +Lend1: srw 0,11,6 + stw 0,4(7) + blr + +Lend2: srw 0,10,6 + stw 0,4(7) + blr + +#else +/* Shift a limb right, low level routine. + Copyright (C) 1995, 1997 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Library General Public License as + published by the Free Software Foundation; either version 2 of the + License, or (at your option) any later version. + + 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 + Library General Public License for more details. + + You should have received a copy of the GNU Library General Public + License along with the GNU C Library; see the file COPYING.LIB. If not, + write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, + Boston, MA 02110-1301, USA. */ + + +/* INPUT PARAMETERS + res_ptr r3 + s1_ptr r4 + size r5 + cnt r6 */ + +ENTRY(mpihelp_rshift) + mtctr 5 # copy size into CTR + addi 7,3,-4 # move adjusted res_ptr to free return reg + subfic 8,6,32 + lwz 11,0(4) # load first s1 limb + slw 3,11,8 # compute function return value + bdz 1f + +0: lwzu 10,4(4) + srw 9,11,6 + slw 12,10,8 + or 9,9,12 + stwu 9,4(7) + bdz 2f + lwzu 11,4(4) + srw 9,10,6 + slw 12,11,8 + or 9,9,12 + stwu 9,4(7) + bdnz 0b + +1: srw 0,11,6 + stw 0,4(7) + blr + +2: srw 0,10,6 + stw 0,4(7) + blr +END(mpihelp_rshift) +#endif diff --git a/mpi/powerpc32/mpih-sub1.S b/mpi/powerpc32/mpih-sub1.S new file mode 100644 index 0000000..8902864 --- /dev/null +++ b/mpi/powerpc32/mpih-sub1.S @@ -0,0 +1,133 @@ +/* PowerPC-32 sub_n -- Subtract two limb vectors of the same length > 0 + * and store difference in a third limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1998 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +#include "sysdep.h" +#include "asm-syntax.h" + + +#ifndef USE_PPC_PATCHES + +/******************* + * mpi_limb_t + * mpihelp_sub_n( mpi_ptr_t res_ptr, (r3) + * mpi_ptr_t s1_ptr, (r4) + * mpi_ptr_t s2_ptr, (r5) + * mpi_size_t size) (r6) + */ + + .toc + .extern mpihelp_sub_n[DS] + .extern .mpihelp_sub_n +.csect [PR] + .align 2 + .globl mpihelp_sub_n + .globl .mpihelp_sub_n + .csect mpihelp_sub_n[DS] +mpihelp_sub_n: + .long .mpihelp_sub_n, TOC[tc0], 0 + .csect [PR] +.mpihelp_sub_n: + mtctr 6 # copy size into CTR + lwz 8,0(4) # load least significant s1 limb + lwz 0,0(5) # load least significant s2 limb + addi 3,3,-4 # offset res_ptr, it is updated before used + subfc 7,0,8 # add least significant limbs, set cy + bdz Lend # If done, skip loop +Loop: lwzu 8,4(4) # load s1 limb and update s1_ptr + lwzu 0,4(5) # load s2 limb and update s2_ptr + stwu 7,4(3) # store previous limb in load latency slot + subfe 7,0,8 # add new limbs with cy, set cy + bdnz Loop # decrement CTR and loop back +Lend: stw 7,4(3) # store ultimate result limb + subfe 3,0,0 # load !cy into ... + subfic 3,3,0 # ... return value register + blr + +#else +/* Subtract two limb vectors of equal, non-zero length for PowerPC. + Copyright (C) 1997 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Library General Public License as + published by the Free Software Foundation; either version 2 of the + License, or (at your option) any later version. + + 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 + Library General Public License for more details. + + You should have received a copy of the GNU Library General Public + License along with the GNU C Library; see the file COPYING.LIB. If not, + write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, + Boston, MA 02110-1301, USA. */ + +/* mp_limb_t mpn_sub_n (mp_ptr res_ptr, mp_srcptr s1_ptr, mp_srcptr s2_ptr, + mp_size_t size) + Calculate s1-s2 and put result in res_ptr; return borrow, 0 or 1. */ + +/* Note on optimisation: This code is optimal for the 601. Almost every other + possible 2-unrolled inner loop will not be. Also, watch out for the + alignment... */ + +EALIGN(mpihelp_sub_n,3,1) +/* Set up for loop below. */ + mtcrf 0x01,%r6 + srwi. %r7,%r6,1 + mtctr %r7 + bt 31,2f + +/* Set the carry (clear the borrow). */ + subfc %r0,%r0,%r0 +/* Adjust pointers for loop. */ + addi %r3,%r3,-4 + addi %r4,%r4,-4 + addi %r5,%r5,-4 + b 0f + +2: lwz %r7,0(%r5) + lwz %r6,0(%r4) + subfc %r6,%r7,%r6 + stw %r6,0(%r3) + beq 1f + +/* Align start of loop to an odd word boundary to guarantee that the + last two words can be fetched in one access (for 601). This turns + out to be important. */ +0: + lwz %r9,4(%r4) + lwz %r8,4(%r5) + lwzu %r6,8(%r4) + lwzu %r7,8(%r5) + subfe %r8,%r8,%r9 + stw %r8,4(%r3) + subfe %r6,%r7,%r6 + stwu %r6,8(%r3) + bdnz 0b +/* Return the borrow. */ +1: subfe %r3,%r3,%r3 + neg %r3,%r3 + blr +END(mpihelp_sub_n) +#endif diff --git a/mpi/powerpc32/syntax.h b/mpi/powerpc32/syntax.h new file mode 100644 index 0000000..64d9c86 --- /dev/null +++ b/mpi/powerpc32/syntax.h @@ -0,0 +1,75 @@ +/* gmp2-2.0.2-ppc/mpn/powerpc-linux/syntax.h Tue Oct 6 19:27:01 1998 */ +/* From glibc's sysdeps/unix/sysv/linux/powerpc/sysdep.h */ + +/* Copyright (C) 1992, 1997, 1998 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Library General Public License as + published by the Free Software Foundation; either version 2 of the + License, or (at your option) any later version. + + 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 + Library General Public License for more details. + + You should have received a copy of the GNU Library General Public + License along with the GNU C Library; see the file COPYING.LIB. If not, + write to the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, + Boston, MA 02110-1301, USA. */ + + +#define USE_PPC_PATCHES 1 + +/* This seems to always be the case on PPC. */ +#define ALIGNARG(log2) log2 +/* For ELF we need the `.type' directive to make shared libs work right. */ +#define ASM_TYPE_DIRECTIVE(name,typearg) .type name,typearg; +#define ASM_SIZE_DIRECTIVE(name) .size name,.-name +#define ASM_GLOBAL_DIRECTIVE .globl + +#ifdef __STDC__ +# define C_LABEL(name) C_SYMBOL_NAME(name)##: +#else +# define C_LABEL(name) C_SYMBOL_NAME(name)/**/: +#endif + +#ifdef __STDC__ +# define L(body) .L##body +#else +# define L(body) .L/**/body +#endif + +/* No profiling of gmp's assembly for now... */ +#define CALL_MCOUNT /* no profiling */ + +#define ENTRY(name) \ + ASM_GLOBAL_DIRECTIVE C_SYMBOL_NAME(name); \ + ASM_TYPE_DIRECTIVE (C_SYMBOL_NAME(name),@function) \ + .align ALIGNARG(2); \ + C_LABEL(name) \ + CALL_MCOUNT + +#define EALIGN_W_0 /* No words to insert. */ +#define EALIGN_W_1 nop +#define EALIGN_W_2 nop;nop +#define EALIGN_W_3 nop;nop;nop +#define EALIGN_W_4 EALIGN_W_3;nop +#define EALIGN_W_5 EALIGN_W_4;nop +#define EALIGN_W_6 EALIGN_W_5;nop +#define EALIGN_W_7 EALIGN_W_6;nop + +/* EALIGN is like ENTRY, but does alignment to 'words'*4 bytes + past a 2^align boundary. */ +#define EALIGN(name, alignt, words) \ + ASM_GLOBAL_DIRECTIVE C_SYMBOL_NAME(name); \ + ASM_TYPE_DIRECTIVE (C_SYMBOL_NAME(name),@function) \ + .align ALIGNARG(alignt); \ + EALIGN_W_##words; \ + C_LABEL(name) + +#undef END +#define END(name) \ + ASM_SIZE_DIRECTIVE(name) + diff --git a/mpi/powerpc64/distfiles b/mpi/powerpc64/distfiles new file mode 100644 index 0000000..e69de29 --- /dev/null +++ b/mpi/powerpc64/distfiles diff --git a/mpi/sparc32/distfiles b/mpi/sparc32/distfiles new file mode 100644 index 0000000..95ff428 --- /dev/null +++ b/mpi/sparc32/distfiles @@ -0,0 +1,6 @@ + +mpih-lshift.S +mpih-rshift.S +mpih-add1.S +udiv.S + diff --git a/mpi/sparc32/mpih-add1.S b/mpi/sparc32/mpih-add1.S new file mode 100644 index 0000000..8cd6db6 --- /dev/null +++ b/mpi/sparc32/mpih-add1.S @@ -0,0 +1,240 @@ +/* SPARC _add_n -- Add two limb vectors of the same length > 0 and store + * sum in a third limb vector. + * + * Copyright (C) 1995, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + + +/******************* + * mpi_limb_t + * mpihelp_add_n( mpi_ptr_t res_ptr, + * mpi_ptr_t s1_ptr, + * mpi_ptr_t s2_ptr, + * mpi_size_t size) + */ + +! INPUT PARAMETERS +#define res_ptr %o0 +#define s1_ptr %o1 +#define s2_ptr %o2 +#define size %o3 + +#include "sysdep.h" + + .text + .align 4 + .global C_SYMBOL_NAME(mpihelp_add_n) +C_SYMBOL_NAME(mpihelp_add_n): + xor s2_ptr,res_ptr,%g1 + andcc %g1,4,%g0 + bne L1 ! branch if alignment differs + nop +! ** V1a ** +L0: andcc res_ptr,4,%g0 ! res_ptr unaligned? Side effect: cy=0 + be L_v1 ! if no, branch + nop +/* Add least significant limb separately to align res_ptr and s2_ptr */ + ld [s1_ptr],%g4 + add s1_ptr,4,s1_ptr + ld [s2_ptr],%g2 + add s2_ptr,4,s2_ptr + add size,-1,size + addcc %g4,%g2,%o4 + st %o4,[res_ptr] + add res_ptr,4,res_ptr +L_v1: addx %g0,%g0,%o4 ! save cy in register + cmp size,2 ! if size < 2 ... + bl Lend2 ! ... branch to tail code + subcc %g0,%o4,%g0 ! restore cy + + ld [s1_ptr+0],%g4 + addcc size,-10,size + ld [s1_ptr+4],%g1 + ldd [s2_ptr+0],%g2 + blt Lfin1 + subcc %g0,%o4,%g0 ! restore cy +/* Add blocks of 8 limbs until less than 8 limbs remain */ +Loop1: addxcc %g4,%g2,%o4 + ld [s1_ptr+8],%g4 + addxcc %g1,%g3,%o5 + ld [s1_ptr+12],%g1 + ldd [s2_ptr+8],%g2 + std %o4,[res_ptr+0] + addxcc %g4,%g2,%o4 + ld [s1_ptr+16],%g4 + addxcc %g1,%g3,%o5 + ld [s1_ptr+20],%g1 + ldd [s2_ptr+16],%g2 + std %o4,[res_ptr+8] + addxcc %g4,%g2,%o4 + ld [s1_ptr+24],%g4 + addxcc %g1,%g3,%o5 + ld [s1_ptr+28],%g1 + ldd [s2_ptr+24],%g2 + std %o4,[res_ptr+16] + addxcc %g4,%g2,%o4 + ld [s1_ptr+32],%g4 + addxcc %g1,%g3,%o5 + ld [s1_ptr+36],%g1 + ldd [s2_ptr+32],%g2 + std %o4,[res_ptr+24] + addx %g0,%g0,%o4 ! save cy in register + addcc size,-8,size + add s1_ptr,32,s1_ptr + add s2_ptr,32,s2_ptr + add res_ptr,32,res_ptr + bge Loop1 + subcc %g0,%o4,%g0 ! restore cy + +Lfin1: addcc size,8-2,size + blt Lend1 + subcc %g0,%o4,%g0 ! restore cy +/* Add blocks of 2 limbs until less than 2 limbs remain */ +Loope1: addxcc %g4,%g2,%o4 + ld [s1_ptr+8],%g4 + addxcc %g1,%g3,%o5 + ld [s1_ptr+12],%g1 + ldd [s2_ptr+8],%g2 + std %o4,[res_ptr+0] + addx %g0,%g0,%o4 ! save cy in register + addcc size,-2,size + add s1_ptr,8,s1_ptr + add s2_ptr,8,s2_ptr + add res_ptr,8,res_ptr + bge Loope1 + subcc %g0,%o4,%g0 ! restore cy +Lend1: addxcc %g4,%g2,%o4 + addxcc %g1,%g3,%o5 + std %o4,[res_ptr+0] + addx %g0,%g0,%o4 ! save cy in register + + andcc size,1,%g0 + be Lret1 + subcc %g0,%o4,%g0 ! restore cy +/* Add last limb */ + ld [s1_ptr+8],%g4 + ld [s2_ptr+8],%g2 + addxcc %g4,%g2,%o4 + st %o4,[res_ptr+8] + +Lret1: retl + addx %g0,%g0,%o0 ! return carry-out from most sign. limb + +L1: xor s1_ptr,res_ptr,%g1 + andcc %g1,4,%g0 + bne L2 + nop +! ** V1b ** + mov s2_ptr,%g1 + mov s1_ptr,s2_ptr + b L0 + mov %g1,s1_ptr + +! ** V2 ** +/* If we come here, the alignment of s1_ptr and res_ptr as well as the + alignment of s2_ptr and res_ptr differ. Since there are only two ways + things can be aligned (that we care about) we now know that the alignment + of s1_ptr and s2_ptr are the same. */ + +L2: cmp size,1 + be Ljone + nop + andcc s1_ptr,4,%g0 ! s1_ptr unaligned? Side effect: cy=0 + be L_v2 ! if no, branch + nop +/* Add least significant limb separately to align s1_ptr and s2_ptr */ + ld [s1_ptr],%g4 + add s1_ptr,4,s1_ptr + ld [s2_ptr],%g2 + add s2_ptr,4,s2_ptr + add size,-1,size + addcc %g4,%g2,%o4 + st %o4,[res_ptr] + add res_ptr,4,res_ptr + +L_v2: addx %g0,%g0,%o4 ! save cy in register + addcc size,-8,size + blt Lfin2 + subcc %g0,%o4,%g0 ! restore cy +/* Add blocks of 8 limbs until less than 8 limbs remain */ +Loop2: ldd [s1_ptr+0],%g2 + ldd [s2_ptr+0],%o4 + addxcc %g2,%o4,%g2 + st %g2,[res_ptr+0] + addxcc %g3,%o5,%g3 + st %g3,[res_ptr+4] + ldd [s1_ptr+8],%g2 + ldd [s2_ptr+8],%o4 + addxcc %g2,%o4,%g2 + st %g2,[res_ptr+8] + addxcc %g3,%o5,%g3 + st %g3,[res_ptr+12] + ldd [s1_ptr+16],%g2 + ldd [s2_ptr+16],%o4 + addxcc %g2,%o4,%g2 + st %g2,[res_ptr+16] + addxcc %g3,%o5,%g3 + st %g3,[res_ptr+20] + ldd [s1_ptr+24],%g2 + ldd [s2_ptr+24],%o4 + addxcc %g2,%o4,%g2 + st %g2,[res_ptr+24] + addxcc %g3,%o5,%g3 + st %g3,[res_ptr+28] + addx %g0,%g0,%o4 ! save cy in register + addcc size,-8,size + add s1_ptr,32,s1_ptr + add s2_ptr,32,s2_ptr + add res_ptr,32,res_ptr + bge Loop2 + subcc %g0,%o4,%g0 ! restore cy + +Lfin2: addcc size,8-2,size + blt Lend2 + subcc %g0,%o4,%g0 ! restore cy +Loope2: ldd [s1_ptr+0],%g2 + ldd [s2_ptr+0],%o4 + addxcc %g2,%o4,%g2 + st %g2,[res_ptr+0] + addxcc %g3,%o5,%g3 + st %g3,[res_ptr+4] + addx %g0,%g0,%o4 ! save cy in register + addcc size,-2,size + add s1_ptr,8,s1_ptr + add s2_ptr,8,s2_ptr + add res_ptr,8,res_ptr + bge Loope2 + subcc %g0,%o4,%g0 ! restore cy +Lend2: andcc size,1,%g0 + be Lret2 + subcc %g0,%o4,%g0 ! restore cy +/* Add last limb */ +Ljone: ld [s1_ptr],%g4 + ld [s2_ptr],%g2 + addxcc %g4,%g2,%o4 + st %o4,[res_ptr] + +Lret2: retl + addx %g0,%g0,%o0 ! return carry-out from most sign. limb + + + diff --git a/mpi/sparc32/mpih-lshift.S b/mpi/sparc32/mpih-lshift.S new file mode 100644 index 0000000..2099574 --- /dev/null +++ b/mpi/sparc32/mpih-lshift.S @@ -0,0 +1,98 @@ +/* sparc lshift + * + * Copyright (C) 1995, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + + +! INPUT PARAMETERS +! res_ptr %o0 +! src_ptr %o1 +! size %o2 +! cnt %o3 + +#include "sysdep.h" + + .text + .align 4 + .global C_SYMBOL_NAME(mpihelp_lshift) +C_SYMBOL_NAME(mpihelp_lshift): + sll %o2,2,%g1 + add %o1,%g1,%o1 ! make %o1 point at end of src + ld [%o1-4],%g2 ! load first limb + sub %g0,%o3,%o5 ! negate shift count + add %o0,%g1,%o0 ! make %o0 point at end of res + add %o2,-1,%o2 + andcc %o2,4-1,%g4 ! number of limbs in first loop + srl %g2,%o5,%g1 ! compute function result + be L0 ! if multiple of 4 limbs, skip first loop + st %g1,[%sp+80] + + sub %o2,%g4,%o2 ! adjust count for main loop + +Loop0: ld [%o1-8],%g3 + add %o0,-4,%o0 + add %o1,-4,%o1 + addcc %g4,-1,%g4 + sll %g2,%o3,%o4 + srl %g3,%o5,%g1 + mov %g3,%g2 + or %o4,%g1,%o4 + bne Loop0 + st %o4,[%o0+0] + +L0: tst %o2 + be Lend + nop + +Loop: ld [%o1-8],%g3 + add %o0,-16,%o0 + addcc %o2,-4,%o2 + sll %g2,%o3,%o4 + srl %g3,%o5,%g1 + + ld [%o1-12],%g2 + sll %g3,%o3,%g4 + or %o4,%g1,%o4 + st %o4,[%o0+12] + srl %g2,%o5,%g1 + + ld [%o1-16],%g3 + sll %g2,%o3,%o4 + or %g4,%g1,%g4 + st %g4,[%o0+8] + srl %g3,%o5,%g1 + + ld [%o1-20],%g2 + sll %g3,%o3,%g4 + or %o4,%g1,%o4 + st %o4,[%o0+4] + srl %g2,%o5,%g1 + + add %o1,-16,%o1 + or %g4,%g1,%g4 + bne Loop + st %g4,[%o0+0] + +Lend: sll %g2,%o3,%g2 + st %g2,[%o0-4] + retl + ld [%sp+80],%o0 + diff --git a/mpi/sparc32/mpih-rshift.S b/mpi/sparc32/mpih-rshift.S new file mode 100644 index 0000000..b20dba9 --- /dev/null +++ b/mpi/sparc32/mpih-rshift.S @@ -0,0 +1,94 @@ +/* sparc rshift + * + * Copyright (C) 1995, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + */ + +! INPUT PARAMETERS +! res_ptr %o0 +! src_ptr %o1 +! size %o2 +! cnt %o3 + +#include "sysdep.h" + + .text + .align 4 + .global C_SYMBOL_NAME(mpihelp_rshift) +C_SYMBOL_NAME(mpihelp_rshift): + ld [%o1],%g2 ! load first limb + sub %g0,%o3,%o5 ! negate shift count + add %o2,-1,%o2 + andcc %o2,4-1,%g4 ! number of limbs in first loop + sll %g2,%o5,%g1 ! compute function result + be L0 ! if multiple of 4 limbs, skip first loop + st %g1,[%sp+80] + + sub %o2,%g4,%o2 ! adjust count for main loop + +Loop0: ld [%o1+4],%g3 + add %o0,4,%o0 + add %o1,4,%o1 + addcc %g4,-1,%g4 + srl %g2,%o3,%o4 + sll %g3,%o5,%g1 + mov %g3,%g2 + or %o4,%g1,%o4 + bne Loop0 + st %o4,[%o0-4] + +L0: tst %o2 + be Lend + nop + +Loop: ld [%o1+4],%g3 + add %o0,16,%o0 + addcc %o2,-4,%o2 + srl %g2,%o3,%o4 + sll %g3,%o5,%g1 + + ld [%o1+8],%g2 + srl %g3,%o3,%g4 + or %o4,%g1,%o4 + st %o4,[%o0-16] + sll %g2,%o5,%g1 + + ld [%o1+12],%g3 + srl %g2,%o3,%o4 + or %g4,%g1,%g4 + st %g4,[%o0-12] + sll %g3,%o5,%g1 + + ld [%o1+16],%g2 + srl %g3,%o3,%g4 + or %o4,%g1,%o4 + st %o4,[%o0-8] + sll %g2,%o5,%g1 + + add %o1,16,%o1 + or %g4,%g1,%g4 + bne Loop + st %g4,[%o0-4] + +Lend: srl %g2,%o3,%g2 + st %g2,[%o0-0] + retl + ld [%sp+80],%o0 + diff --git a/mpi/sparc32/udiv.S b/mpi/sparc32/udiv.S new file mode 100644 index 0000000..e4028a6 --- /dev/null +++ b/mpi/sparc32/udiv.S @@ -0,0 +1,198 @@ +/* SPARC v7 __udiv_qrnnd division support, used from longlong.h. + * This is for v7 CPUs without a floating-point unit. + * + * Copyright (C) 1993, 1994, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +! INPUT PARAMETERS +! rem_ptr o0 +! n1 o1 +! n0 o2 +! d o3 + +#include "sysdep.h" + + .text + .align 4 + .global C_SYMBOL_NAME(__udiv_qrnnd) +C_SYMBOL_NAME(__udiv_qrnnd): + tst %o3 + bneg Largedivisor + mov 8,%g1 + + b Lp1 + addxcc %o2,%o2,%o2 + +Lplop: bcc Ln1 + addxcc %o2,%o2,%o2 +Lp1: addx %o1,%o1,%o1 + subcc %o1,%o3,%o4 + bcc Ln2 + addxcc %o2,%o2,%o2 +Lp2: addx %o1,%o1,%o1 + subcc %o1,%o3,%o4 + bcc Ln3 + addxcc %o2,%o2,%o2 +Lp3: addx %o1,%o1,%o1 + subcc %o1,%o3,%o4 + bcc Ln4 + addxcc %o2,%o2,%o2 +Lp4: addx %o1,%o1,%o1 + addcc %g1,-1,%g1 + bne Lplop + subcc %o1,%o3,%o4 + bcc Ln5 + addxcc %o2,%o2,%o2 +Lp5: st %o1,[%o0] + retl + xnor %g0,%o2,%o0 + +Lnlop: bcc Lp1 + addxcc %o2,%o2,%o2 +Ln1: addx %o4,%o4,%o4 + subcc %o4,%o3,%o1 + bcc Lp2 + addxcc %o2,%o2,%o2 +Ln2: addx %o4,%o4,%o4 + subcc %o4,%o3,%o1 + bcc Lp3 + addxcc %o2,%o2,%o2 +Ln3: addx %o4,%o4,%o4 + subcc %o4,%o3,%o1 + bcc Lp4 + addxcc %o2,%o2,%o2 +Ln4: addx %o4,%o4,%o4 + addcc %g1,-1,%g1 + bne Lnlop + subcc %o4,%o3,%o1 + bcc Lp5 + addxcc %o2,%o2,%o2 +Ln5: st %o4,[%o0] + retl + xnor %g0,%o2,%o0 + +Largedivisor: + and %o2,1,%o5 ! %o5 = n0 & 1 + + srl %o2,1,%o2 + sll %o1,31,%g2 + or %g2,%o2,%o2 ! %o2 = lo(n1n0 >> 1) + srl %o1,1,%o1 ! %o1 = hi(n1n0 >> 1) + + and %o3,1,%g2 + srl %o3,1,%g3 ! %g3 = floor(d / 2) + add %g3,%g2,%g3 ! %g3 = ceil(d / 2) + + b LLp1 + addxcc %o2,%o2,%o2 + +LLplop: bcc LLn1 + addxcc %o2,%o2,%o2 +LLp1: addx %o1,%o1,%o1 + subcc %o1,%g3,%o4 + bcc LLn2 + addxcc %o2,%o2,%o2 +LLp2: addx %o1,%o1,%o1 + subcc %o1,%g3,%o4 + bcc LLn3 + addxcc %o2,%o2,%o2 +LLp3: addx %o1,%o1,%o1 + subcc %o1,%g3,%o4 + bcc LLn4 + addxcc %o2,%o2,%o2 +LLp4: addx %o1,%o1,%o1 + addcc %g1,-1,%g1 + bne LLplop + subcc %o1,%g3,%o4 + bcc LLn5 + addxcc %o2,%o2,%o2 +LLp5: add %o1,%o1,%o1 ! << 1 + tst %g2 + bne Oddp + add %o5,%o1,%o1 + st %o1,[%o0] + retl + xnor %g0,%o2,%o0 + +LLnlop: bcc LLp1 + addxcc %o2,%o2,%o2 +LLn1: addx %o4,%o4,%o4 + subcc %o4,%g3,%o1 + bcc LLp2 + addxcc %o2,%o2,%o2 +LLn2: addx %o4,%o4,%o4 + subcc %o4,%g3,%o1 + bcc LLp3 + addxcc %o2,%o2,%o2 +LLn3: addx %o4,%o4,%o4 + subcc %o4,%g3,%o1 + bcc LLp4 + addxcc %o2,%o2,%o2 +LLn4: addx %o4,%o4,%o4 + addcc %g1,-1,%g1 + bne LLnlop + subcc %o4,%g3,%o1 + bcc LLp5 + addxcc %o2,%o2,%o2 +LLn5: add %o4,%o4,%o4 ! << 1 + tst %g2 + bne Oddn + add %o5,%o4,%o4 + st %o4,[%o0] + retl + xnor %g0,%o2,%o0 + +Oddp: xnor %g0,%o2,%o2 + ! q' in %o2. r' in %o1 + addcc %o1,%o2,%o1 + bcc LLp6 + addx %o2,0,%o2 + sub %o1,%o3,%o1 +LLp6: subcc %o1,%o3,%g0 + bcs LLp7 + subx %o2,-1,%o2 + sub %o1,%o3,%o1 +LLp7: st %o1,[%o0] + retl + mov %o2,%o0 + +Oddn: xnor %g0,%o2,%o2 + ! q' in %o2. r' in %o4 + addcc %o4,%o2,%o4 + bcc LLn6 + addx %o2,0,%o2 + sub %o4,%o3,%o4 +LLn6: subcc %o4,%o3,%g0 + bcs LLn7 + subx %o2,-1,%o2 + sub %o4,%o3,%o4 +LLn7: st %o4,[%o0] + retl + mov %o2,%o0 diff --git a/mpi/sparc32v8/distfiles b/mpi/sparc32v8/distfiles new file mode 100644 index 0000000..5789f35 --- /dev/null +++ b/mpi/sparc32v8/distfiles @@ -0,0 +1,5 @@ + +mpih-mul1.S +mpih-mul2.S +mpih-mul3.S + diff --git a/mpi/sparc32v8/mpih-mul1.S b/mpi/sparc32v8/mpih-mul1.S new file mode 100644 index 0000000..52c911a --- /dev/null +++ b/mpi/sparc32v8/mpih-mul1.S @@ -0,0 +1,112 @@ +/* SPARC v8 __mpn_mul_1 -- Multiply a limb vector with a single limb and + * store the product in a second limb vector. + * + * Copyright (C) 1992, 1994, 1995, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + +! INPUT PARAMETERS +! res_ptr o0 +! s1_ptr o1 +! size o2 +! s2_limb o3 + +#include "sysdep.h" + +.text + .align 8 + .global C_SYMBOL_NAME(mpihelp_mul_1) +C_SYMBOL_NAME(mpihelp_mul_1): + sll %o2,4,%g1 + and %g1,(4-1)<<4,%g1 +#if PIC + mov %o7,%g4 ! Save return address register + call 1f + add %o7,LL-1f,%g3 +1: mov %g4,%o7 ! Restore return address register +#else + sethi %hi(LL),%g3 + or %g3,%lo(LL),%g3 +#endif + jmp %g3+%g1 + ld [%o1+0],%o4 ! 1 +LL: +LL00: add %o0,-4,%o0 + add %o1,-4,%o1 + b Loop00 /* 4, 8, 12, ... */ + orcc %g0,%g0,%g2 +LL01: b Loop01 /* 1, 5, 9, ... */ + orcc %g0,%g0,%g2 + nop + nop +LL10: add %o0,-12,%o0 /* 2, 6, 10, ... */ + add %o1,4,%o1 + b Loop10 + orcc %g0,%g0,%g2 + nop +LL11: add %o0,-8,%o0 /* 3, 7, 11, ... */ + add %o1,-8,%o1 + b Loop11 + orcc %g0,%g0,%g2 + +Loop: addcc %g3,%g2,%g3 ! 1 + ld [%o1+4],%o4 ! 2 + st %g3,[%o0+0] ! 1 + rd %y,%g2 ! 1 +Loop00: umul %o4,%o3,%g3 ! 2 + addxcc %g3,%g2,%g3 ! 2 + ld [%o1+8],%o4 ! 3 + st %g3,[%o0+4] ! 2 + rd %y,%g2 ! 2 +Loop11: umul %o4,%o3,%g3 ! 3 + addxcc %g3,%g2,%g3 ! 3 + ld [%o1+12],%o4 ! 4 + add %o1,16,%o1 + st %g3,[%o0+8] ! 3 + rd %y,%g2 ! 3 +Loop10: umul %o4,%o3,%g3 ! 4 + addxcc %g3,%g2,%g3 ! 4 + ld [%o1+0],%o4 ! 1 + st %g3,[%o0+12] ! 4 + add %o0,16,%o0 + rd %y,%g2 ! 4 + addx %g0,%g2,%g2 +Loop01: addcc %o2,-4,%o2 + bg Loop + umul %o4,%o3,%g3 ! 1 + + addcc %g3,%g2,%g3 ! 4 + st %g3,[%o0+0] ! 4 + rd %y,%g2 ! 4 + + retl + addx %g0,%g2,%o0 + + diff --git a/mpi/sparc32v8/mpih-mul2.S b/mpi/sparc32v8/mpih-mul2.S new file mode 100644 index 0000000..641b76a --- /dev/null +++ b/mpi/sparc32v8/mpih-mul2.S @@ -0,0 +1,135 @@ +/* SPARC v8 __mpn_addmul_1 -- Multiply a limb vector with a limb and + * add the result to a second limb vector. + * + * Copyright (C) 1992, 1993, 1994, 1995, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + +! INPUT PARAMETERS +! res_ptr o0 +! s1_ptr o1 +! size o2 +! s2_limb o3 + +#include "sysdep.h" + +.text + .align 4 + .global C_SYMBOL_NAME(mpihelp_addmul_1) +C_SYMBOL_NAME(mpihelp_addmul_1): + orcc %g0,%g0,%g2 + ld [%o1+0],%o4 ! 1 + + sll %o2,4,%g1 + and %g1,(4-1)<<4,%g1 +#if PIC + mov %o7,%g4 ! Save return address register + call 1f + add %o7,LL-1f,%g3 +1: mov %g4,%o7 ! Restore return address register +#else + sethi %hi(LL),%g3 + or %g3,%lo(LL),%g3 +#endif + jmp %g3+%g1 + nop +LL: +LL00: add %o0,-4,%o0 + b Loop00 /* 4, 8, 12, ... */ + add %o1,-4,%o1 + nop +LL01: b Loop01 /* 1, 5, 9, ... */ + nop + nop + nop +LL10: add %o0,-12,%o0 /* 2, 6, 10, ... */ + b Loop10 + add %o1,4,%o1 + nop +LL11: add %o0,-8,%o0 /* 3, 7, 11, ... */ + b Loop11 + add %o1,-8,%o1 + nop + +1: addcc %g3,%g2,%g3 ! 1 + ld [%o1+4],%o4 ! 2 + rd %y,%g2 ! 1 + addx %g0,%g2,%g2 + ld [%o0+0],%g1 ! 2 + addcc %g1,%g3,%g3 + st %g3,[%o0+0] ! 1 +Loop00: umul %o4,%o3,%g3 ! 2 + ld [%o0+4],%g1 ! 2 + addxcc %g3,%g2,%g3 ! 2 + ld [%o1+8],%o4 ! 3 + rd %y,%g2 ! 2 + addx %g0,%g2,%g2 + nop + addcc %g1,%g3,%g3 + st %g3,[%o0+4] ! 2 +Loop11: umul %o4,%o3,%g3 ! 3 + addxcc %g3,%g2,%g3 ! 3 + ld [%o1+12],%o4 ! 4 + rd %y,%g2 ! 3 + add %o1,16,%o1 + addx %g0,%g2,%g2 + ld [%o0+8],%g1 ! 2 + addcc %g1,%g3,%g3 + st %g3,[%o0+8] ! 3 +Loop10: umul %o4,%o3,%g3 ! 4 + addxcc %g3,%g2,%g3 ! 4 + ld [%o1+0],%o4 ! 1 + rd %y,%g2 ! 4 + addx %g0,%g2,%g2 + ld [%o0+12],%g1 ! 2 + addcc %g1,%g3,%g3 + st %g3,[%o0+12] ! 4 + add %o0,16,%o0 + addx %g0,%g2,%g2 +Loop01: addcc %o2,-4,%o2 + bg 1b + umul %o4,%o3,%g3 ! 1 + + addcc %g3,%g2,%g3 ! 4 + rd %y,%g2 ! 4 + addx %g0,%g2,%g2 + ld [%o0+0],%g1 ! 2 + addcc %g1,%g3,%g3 + st %g3,[%o0+0] ! 4 + addx %g0,%g2,%o0 + + retl + nop + + +! umul, ld, addxcc, rd, st + +! umul, ld, addxcc, rd, ld, addcc, st, addx + diff --git a/mpi/sparc32v8/mpih-mul3.S b/mpi/sparc32v8/mpih-mul3.S new file mode 100644 index 0000000..b11e05a --- /dev/null +++ b/mpi/sparc32v8/mpih-mul3.S @@ -0,0 +1,70 @@ +/* SPARC v8 __mpn_submul_1 -- Multiply a limb vector with a limb and + * subtract the result from a second limb vector. + * + * Copyright (C) 1992, 1993, 1994, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + +! INPUT PARAMETERS +! res_ptr o0 +! s1_ptr o1 +! size o2 +! s2_limb o3 + +#include "sysdep.h" + +.text + .align 4 + .global C_SYMBOL_NAME(mpihelp_submul_1) +C_SYMBOL_NAME(mpihelp_submul_1): + sub %g0,%o2,%o2 ! negate ... + sll %o2,2,%o2 ! ... and scale size + sub %o1,%o2,%o1 ! o1 is offset s1_ptr + sub %o0,%o2,%g1 ! g1 is offset res_ptr + + mov 0,%o0 ! clear cy_limb + +Loop: ld [%o1+%o2],%o4 + ld [%g1+%o2],%g2 + umul %o4,%o3,%o5 + rd %y,%g3 + addcc %o5,%o0,%o5 + addx %g3,0,%o0 + subcc %g2,%o5,%g2 + addx %o0,0,%o0 + st %g2,[%g1+%o2] + + addcc %o2,4,%o2 + bne Loop + nop + + retl + nop + + diff --git a/mpi/supersparc/distfiles b/mpi/supersparc/distfiles new file mode 100644 index 0000000..5506c41 --- /dev/null +++ b/mpi/supersparc/distfiles @@ -0,0 +1,3 @@ + +udiv.S + diff --git a/mpi/supersparc/udiv.S b/mpi/supersparc/udiv.S new file mode 100644 index 0000000..9a12348 --- /dev/null +++ b/mpi/supersparc/udiv.S @@ -0,0 +1,121 @@ +/* SuperSPARC __udiv_qrnnd division support, used from longlong.h. + * This is for SuperSPARC only, to compensate for its + * semi-functional udiv instruction. + * + * Copyright (C) 1993, 1994, 1996, 1998, + * 2001 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 2 of the License, or + * (at your option) any later version. + * + * GnuPG 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 General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, + * USA. + * + * Note: This code is heavily based on the GNU MP Library. + * Actually it's the same code with only minor changes in the + * way the data is stored; this is to support the abstraction + * of an optional secure memory allocation which may be used + * to avoid revealing of sensitive data due to paging etc. + * The GNU MP Library itself is published under the LGPL; + * however I decided to publish this code under the plain GPL. + */ + + + +! INPUT PARAMETERS +! rem_ptr i0 +! n1 i1 +! n0 i2 +! d i3 + +#include "sysdep.h" +#undef ret /* Kludge for glibc */ + + .text + .align 8 +LC0: .double 0r4294967296 +LC1: .double 0r2147483648 + + .align 4 + .global C_SYMBOL_NAME(__udiv_qrnnd) +C_SYMBOL_NAME(__udiv_qrnnd): + !#PROLOGUE# 0 + save %sp,-104,%sp + !#PROLOGUE# 1 + st %i1,[%fp-8] + ld [%fp-8],%f10 + sethi %hi(LC0),%o7 + fitod %f10,%f4 + ldd [%o7+%lo(LC0)],%f8 + cmp %i1,0 + bge L248 + mov %i0,%i5 + faddd %f4,%f8,%f4 +L248: + st %i2,[%fp-8] + ld [%fp-8],%f10 + fmuld %f4,%f8,%f6 + cmp %i2,0 + bge L249 + fitod %f10,%f2 + faddd %f2,%f8,%f2 +L249: + st %i3,[%fp-8] + faddd %f6,%f2,%f2 + ld [%fp-8],%f10 + cmp %i3,0 + bge L250 + fitod %f10,%f4 + faddd %f4,%f8,%f4 +L250: + fdivd %f2,%f4,%f2 + sethi %hi(LC1),%o7 + ldd [%o7+%lo(LC1)],%f4 + fcmped %f2,%f4 + nop + fbge,a L251 + fsubd %f2,%f4,%f2 + fdtoi %f2,%f2 + st %f2,[%fp-8] + b L252 + ld [%fp-8],%i4 +L251: + fdtoi %f2,%f2 + st %f2,[%fp-8] + ld [%fp-8],%i4 + sethi %hi(-2147483648),%g2 + xor %i4,%g2,%i4 +L252: + umul %i3,%i4,%g3 + rd %y,%i0 + subcc %i2,%g3,%o7 + subxcc %i1,%i0,%g0 + be L253 + cmp %o7,%i3 + + add %i4,-1,%i0 + add %o7,%i3,%o7 + st %o7,[%i5] + ret + restore +L253: + blu L246 + mov %i4,%i0 + add %i4,1,%i0 + sub %o7,%i3,%o7 +L246: + st %o7,[%i5] + ret + restore + |