security-and-privacy/gnupg
security-and-privacy/gnupg
1
0
Fork 0
mirror of git://git.gnupg.org/gnupg.git synced 2025-07-02 22:46:30 +02:00
Code Activity

Removed files from the HEAD revision, because they are now in another

Browse source 
repository
This commit is contained in:
Werner Koch 2000-12-19 17:20:22 +00:00
parent edb40153c0
commit 07ca4eaa9d
165 changed files with 0 additions and 35058 deletions
Download patch file Download diff file Expand all files Collapse all files

319
mpi/ChangeLog
View file

@ -1,319 +0,0 @@
2000-11-14 Werner Koch <wk@gnupg.org>
* mpi-internal.h, mpi.h: Changed the way they are called and
introduced DID_MPI_LIMP_TYPEDEF hack. Very ugly, should all be
revamped.
* Makefile.am (OMIT_DEPENDENCIES): Hack to work around dependency
problems.
2000-10-11 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.
Fri Jul 28 18:19:11 CEST 2000 Werner Koch <wk@openit.de>
* mpicoder.c (gcry_mpi_scan): Normalize the returned MPI.
Tue Jul 25 17:44:15 CEST 2000 Werner Koch <wk@openit.de>
* config.links: Support for powerpc--netbsd by Gabriel Rosenkoetter.
Mon Jul 17 16:35:47 CEST 2000 Werner Koch <wk@>
* power/: Add all files from GMP for this CPU. Converted comments to
CPP comments because some ASes complain about ' in comments.
* config.links: Support for BSDI 4.x; by Wayne Chapeskie. Add support
for FreeBSD 5 and made the case stmt looking nicer; by Jun Kuriyama.
Add support for NetBSD.
(sparc8): Made the search path the same as sparc9
(sparc64-unknown-linux-gnu): use udiv module; by Adam Mitchell.
* Makefile.am: c/SFLAGS/ASFLAGS/. This has only been used by the
powerpc and actually never passed the -Wa,foo to the cc.
* mpih-div.c (mpihelp_divrem): The MPN_COPY_DECR copied one element
too many. This is a gmp2.0.2p9.txt patch.
* longlong.h (umul_ppmm): Fixes for ARM-4. By Sean MacLennan.
* 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.
2000-03-21 16:17:30 Werner Koch (wk@habibti.openit.de)
* config.links: Add support for FreeBSD 5.
Mon Jan 24 22:24:38 CET 2000 Werner Koch <wk@gnupg.de>
* mpicoder.c (gcry_mpi_aprint): Now really returns the length.
Mon Jan 24 13:04:28 CET 2000 Werner Koch <wk@gnupg.de>
* mpiutil.c: Removed all memory debugging code.
* mpicoder.c (gcry_mpi_aprint): New.
* Replaced all m_ memory functions by g10_ ones.
Fri Dec 31 14:06:56 CET 1999 Werner Koch <wk@gnupg.de>
* mpi-bit.c (gcry_mpi_get_nbits): New.
* mpiutil.c (mpi_set_secure): made static.
(gcry_mpi_get_flag): New.
(gcry_mpi_set_flag): New.
(gcry_mpi_clear_flag): New.
(mpi_set_opaque): renamed to gcry_mpi_set_opaque.
(mpi_get_opaque): renamed to gcry_mpi_get_opaque.
Fri Dec 31 12:48:31 CET 1999 Werner Koch <wk@gnupg.de>
* mpicoder.c (mpi_read_from_buffer): Made static.
(gcry_mpi_print): A buffer of NULL is now allowed to get the required
length back.
(mpi_get_keyid): Removed.
(mpi_print): Made static - should be removed.
Wed Dec 8 21:58:32 CET 1999 Werner Koch <wk@gnupg.de>
* Makefile.am (INCLUDES): Add ../gcrypt.
* g10m.c : Removed.
* mpicoder.c (mpi_write): Removed.
(mpi_read): Removed.
(gcry_mpi_scan): New. Taken from ../gcrypt/mpiapi.c.
(gcry_mpi_print): Ditto.
* mpi-pow.c (mpi_powm): Renamed to ...
(gcry_mpi_powm): ... this.
* mpiutil.c (gcry_mpi_new): New as a wrapper around the old function.
Taken from ../gcrypt/mpiapi.c.
(gcry_mpi_snew): Ditto.
(gcry_mpi_release): Ditto.
(gcry_mpi_copy): Ditto.
(gcry_mpi_set): Ditto.
(gcry_mpi_set_ui): Ditto.
(gcry_mpi_cmp): Ditto.
(gcry_mpi_cmp_ui): Ditto.
(gcry_mpi_randomize): Ditto.
* mpicoder.c (mpi_print): Removed the nbit_info kludge.
* mpi-bits.c (mpi_get_nbits): Replaced the is_protected stuff by
checking whether it is an opaque mpi and then returns it's length
in bits.
* mpiutil.c (mpi_set_opaque): Changed the interface to take a number
of bits for the length. Adjusted all users.
(mpi_get_opaque): Ditto.
Fri Nov 19 17:15:20 CET 1999 Werner Koch <wk@gnupg.de>
* mpicoder.c (g10_log_mpidump): Add a temporary workaround
* mpih-mul.c (mpihelp_mul_n): s/m_is_ecure/g10_is_secure/
* mpiutil.c (mpi_alloc): Remved the debug mode because it has turned
out, that this feature was not very useful in the past. Use the
new alloc functions.
(mpi_alloc_secure): Ditto.
(mpi_alloc_limb_space): Ditto.
(mpi_free_limb_space): Ditto.
(mpi_resize): Ditto.
(mpi_free): Ditto.
(mpi_set_secure): Removed the debug stuff.
(mpi_set_opaque): Ditto.
(mpi_copy): Ditto.
(mpi_alloc_set_ui): Ditto.
(mpi_m_check): Use g10_ wrapper.
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

68
mpi/Makefile.am
View file

@ -1,68 +0,0 @@
## Process this file with automake to produce Makefile.in
INCLUDES = -I$(top_srcdir)/gcrypt
CFLAGS = @CFLAGS@ @MPI_OPT_FLAGS@
ASFLAGS = @MPI_SFLAGS@
EXTRA_DIST = config.links
DISTCLEANFILES = mpih-add1.S mpih-mul1.S mpih-mul2.S mpih-mul3.S \
mpih-lshift.S mpih-rshift.S mpih-sub1.S asm-syntax.h sysdep.h
# Note: we only use .S files so we should delete all left over .s
# CLEANFILES = _*.s
CLEANFILES = *.s
OMIT_DEPENDENCIES = types.h gcrypt.h
noinst_LTLIBRARIES = libmpi.la
libmpi_la_LDFLAGS =
libmpi_la_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 \
mpi.h
# Note this objects are actually links, the sourcefiles are
# distributed by special code in dist-hook
common_asm_objects = mpih-mul1.lo \
mpih-mul2.lo \
mpih-mul3.lo \
mpih-add1.lo \
mpih-sub1.lo \
mpih-lshift.lo \
mpih-rshift.lo
libmpi_la_DEPENDENCIES = $(common_asm_objects) @MPI_EXTRA_ASM_OBJS@
libmpi_la_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:
.S.lo:
.S.s:
$(CPP) $(INCLUDES) $(DEFS) $< | grep -v '^#' >$*.s
# Hmmm, we should use this, so that OSes which do not distinguish
# filename case still work. We have to see how libtool can handle this
# $(CPP) $(INCLUDES) $(DEFS) $< | grep -v '^#' > _$*.s
# $(COMPILE) -c _$*.s
# mv -f _$*.o $*.o

53
mpi/alpha/README
View file

@ -1,53 +0,0 @@
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

11
mpi/alpha/distfiles
View file

@ -1,11 +0,0 @@
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

126
mpi/alpha/mpih-add1.S
View file

@ -1,126 +0,0 @@
/* alpha add_n -- Add two limb vectors of the same length > 0 and store
* sum in a third limb vector.
*
* Copyright (C) 1995 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

122
mpi/alpha/mpih-lshift.S
View file

@ -1,122 +0,0 @@
/* alpha lshift
* Copyright (C) 1994, 1995 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

89
mpi/alpha/mpih-mul1.S
View file

@ -1,89 +0,0 @@
/* 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 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

96
mpi/alpha/mpih-mul2.S
View file

@ -1,96 +0,0 @@
/* 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 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

94
mpi/alpha/mpih-mul3.S
View file

@ -1,94 +0,0 @@
/* 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 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

120
mpi/alpha/mpih-rshift.S
View file

@ -1,120 +0,0 @@
/* alpha rshift
* Copyright (C) 1994, 1995 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

123
mpi/alpha/mpih-sub1.S
View file

@ -1,123 +0,0 @@
/* Alpha sub_n -- Subtract two limb vectors of the same length > 0 and
* store difference in a third 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

161
mpi/alpha/udiv-qrnnd.S
View file

@ -1,161 +0,0 @@
/* Alpha 21064 __udiv_qrnnd
*
* Copyright (C) 1992, 1994, 1995 Free Software Foundation, Inc.
* Copyright (c) 1997 by Werner Koch (dd9jn)
*
* This file is part of G10.
*
* G10 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.
*
* G10 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

291
mpi/config.links
View file

@ -1,291 +0,0 @@
# sourced by ../configure to get the list of files to link
# this should set $mpi_ln_src and mpi_ln_dst.
# Note: this is called from the above directory.
mpi_extra_modules=
mpi_sflags=
test -d ./mpi || mkdir ./mpi
echo '/* created by config.links - do not edit */' >./mpi/asm-syntax.h
if test "$try_asm_modules" = "yes" ; then
case "${target}" in
i[34]86*-*-freebsd*-elf | \
i[34]86*-*-freebsd[3-9]* | \
i[34]86*-*-freebsdelf* | \
i[34]86*-*-netbsd* )
echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h
cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h
path="i386"
;;
i[56]86*-*-freebsd*-elf | \
i[56]86*-*-freebsd[3-9]* | \
i[56]86*-*-freebsdelf* | \
i[56]86*-*-netbsd* | \
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[34]86*-*-linuxaout* | \
i[34]86*-*-linuxoldld* | \
i[34]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"
;;
i[56]86*-*-linuxaout* | \
i[56]86*-*-linuxoldld* | \
i[56]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="i586 i386"
;;
i[34]86*-msdosdjgpp*)
echo '#define BSD_SYNTAX' >>./mpi/asm-syntax.h
cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h
path="i386"
;;
i[56]86*-msdosdjgpp*)
echo '#define BSD_SYNTAX' >>./mpi/asm-syntax.h
cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h
path="i586 i386"
;;
i[34]86*-*-*)
echo '#define ELF_SYNTAX' >>./mpi/asm-syntax.h
cat $srcdir/mpi/i386/syntax.h >>./mpi/asm-syntax.h
path="i386"
;;
i[56]86*-*-* | \
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"
;;
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*)
echo '/* configured NetBSD 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="${mpi_extra_modules} mpih-add1 mpih-mul1 mpih-mul2 mpih-mul3 \
mpih-lshift mpih-rshift mpih-sub1"
mpi_ln_objects=
mpi_ln_src=
mpi_ln_dst=
# try to get file to link from the assembler subdirectory and
# if this fails get it from the generic subdirectory.
path="$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/$fn.S ; then
mpi_ln_src="$mpi_ln_src mpi/$dir/$fn.S"
mpi_ln_dst="$mpi_ln_dst mpi/$fn.S"
break;
elif test -f $srcdir/mpi/$dir/$fn.c ; then
mpi_ln_src="$mpi_ln_src mpi/$dir/$fn.c"
mpi_ln_dst="$mpi_ln_dst mpi/$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_src="$mpi_ln_src mpi/$dir/mpi-asm-defs.h"
mpi_ln_dst="$mpi_ln_dst mpi/mpi-asm-defs.h"
break;
fi
done

10
mpi/generic/distfiles
View file

@ -1,10 +0,0 @@
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

8
mpi/generic/mpi-asm-defs.h
View file

@ -1,8 +0,0 @@
/* 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)

65
mpi/generic/mpih-add1.c
View file

@ -1,65 +0,0 @@
/* mpihelp-add_1.c - MPI helper functions
* Copyright (c) 1997 by Werner Koch (dd9jn)
* Copyright (C) 1994, 1996 Free Software Foundation, Inc.
*
* This file is part of G10.
*
* G10 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.
*
* G10 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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;
}

69
mpi/generic/mpih-lshift.c
View file

@ -1,69 +0,0 @@
/* mpihelp-lshift.c - MPI helper functions
* Copyright (C) 1994, 1996, 1998 Free Software Foundation, Inc.
*
* This file is part of G10.
*
* G10 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.
*
* G10 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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;
}

62
mpi/generic/mpih-mul1.c
View file

@ -1,62 +0,0 @@
/* mpihelp-mul_1.c - MPI helper functions
* Copyright (c) 1997 by Werner Koch (dd9jn)
* Copyright (C) 1994, 1996 Free Software Foundation, Inc.
*
* This file is part of G10.
*
* G10 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.
*
* G10 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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;
}

67
mpi/generic/mpih-mul2.c
View file

@ -1,67 +0,0 @@
/* mpihelp-mul_2.c - MPI helper functions
* Copyright (c) 1997 by Werner Koch (dd9jn)
* Copyright (C) 1994, 1996 Free Software Foundation, Inc.
*
* This file is part of G10.
*
* G10 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.
*
* G10 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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;
}

68
mpi/generic/mpih-mul3.c
View file

@ -1,68 +0,0 @@
/* mpihelp-mul_3.c - MPI helper functions
* Copyright (c) 1997 by Werner Koch (dd9jn)
* Copyright (C) 1994, 1996 Free Software Foundation, Inc.
*
* This file is part of G10.
*
* G10 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.
*
* G10 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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;
}

67
mpi/generic/mpih-rshift.c
View file

@ -1,67 +0,0 @@
/* mpih-rshift.c - MPI helper functions
* Copyright (C) 1994, 1996, 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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;
}

66
mpi/generic/mpih-sub1.c
View file

@ -1,66 +0,0 @@
/* mpihelp-add_2.c - MPI helper functions
* Copyright (c) 1997 by Werner Koch (dd9jn)
* Copyright (C) 1994, 1996 Free Software Foundation, Inc.
*
* This file is part of G10.
*
* G10 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.
*
* G10 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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;
}

135
mpi/generic/udiv-w-sdiv.c
View file

@ -1,135 +0,0 @@
/* mpihelp_udiv_w_sdiv -- implement udiv_qrnnd on machines with only signed
* division.
*
* Contributed by Peter L. Montgomery.
*
* Copyright (C) 1992, 1994, 1996, 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

84
mpi/hppa/README
View file

@ -1,84 +0,0 @@
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

7
mpi/hppa/distfiles
View file

@ -1,7 +0,0 @@
README
udiv-qrnnd.S
mpih-add1.S
mpih-sub1.S
mpih-lshift.S
mpih-rshift.S

70
mpi/hppa/mpih-add1.S
View file

@ -1,70 +0,0 @@
/* hppa add_n -- Add two limb vectors of the same length > 0 and store
* sum in a third limb vector.
*
* Copyright (C) 1992, 1994 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
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$end ; check for (SIZE == 1)
add %r20,%r19,%r28 ; add first limbs ignoring cy
L$loop ldws,ma 4(0,%r25),%r20
ldws,ma 4(0,%r24),%r19
stws,ma %r28,4(0,%r26)
addib,<> -1,%r23,L$loop
addc %r20,%r19,%r28
L$end stws %r28,0(0,%r26)
bv 0(%r2)
addc %r0,%r0,%r28
.exit
.procend

72
mpi/hppa/mpih-lshift.S
View file

@ -1,72 +0,0 @@
/* hppa lshift
*
* Copyright (C) 1992, 1994, 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
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$0002
vshd %r22,%r29,%r20
L$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
L$0002 stws,mb %r20,-4(0,%r26)
vshd %r29,%r0,%r20
bv 0(%r2)
stw %r20,-4(0,%r26)
L$0003 stws,mb %r20,-4(0,%r26)
L$0004 vshd %r22,%r0,%r20
bv 0(%r2)
stw %r20,-4(0,%r26)
.exit
.procend

68
mpi/hppa/mpih-rshift.S
View file

@ -1,68 +0,0 @@
/* hppa rshift
*
* Copyright (C) 1992, 1994, 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
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
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$roop
vshd %r29,%r22,%r20
L$r002 stws,ma %r20,4(0,%r26)
vshd %r0,%r29,%r20
bv 0(%r2)
stw %r20,0(0,%r26)
L$r003 stws,ma %r20,4(0,%r26)
L$r004 vshd %r0,%r22,%r20
bv 0(%r2)
stw %r20,0(0,%r26)
.exit
.procend

77
mpi/hppa/mpih-sub1.S
View file

@ -1,77 +0,0 @@
/* hppa sub_n -- Sub two limb vectors of the same length > 0 and store
* sum in a third limb vector.
* Copyright (C) 1992, 1994 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
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$end ; check for (SIZE == 1)
sub %r20,%r19,%r28 ; subtract first limbs ignoring cy
L$loop ldws,ma 4(0,%r25),%r20
ldws,ma 4(0,%r24),%r19
stws,ma %r28,4(0,%r26)
addib,<> -1,%r23,L$loop
subb %r20,%r19,%r28
L$end stws %r28,0(0,%r26)
addc %r0,%r0,%r28
bv 0(%r2)
subi 1,%r28,%r28
.exit
.procend

297
mpi/hppa/udiv-qrnnd.S
View file

@ -1,297 +0,0 @@
/* HP-PA __udiv_qrnnd division support, used from longlong.h.
* This version runs fast on pre-PA7000 CPUs.
*
* Copyright (C) 1993, 1994 Free Software Foundation, Inc.
* Copyright (c) 1997 by Werner Koch (dd9jn)
*
* This file is part of G10.
*
* G10 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.
*
* G10 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
__udiv_qrnnd
.proc
.callinfo frame=0,no_calls
.entry
comb,< %r23,0,L$largedivisor
sub %r0,%r23,%r1 ; clear cy as side-effect
ds %r0,%r1,%r0
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r24
ds %r25,%r23,%r25
addc %r24,%r24,%r28
ds %r25,%r23,%r25
comclr,>= %r25,%r0,%r0
addl %r25,%r23,%r25
stws %r25,0(0,%r26)
bv 0(%r2)
addc %r28,%r28,%r28
L$largedivisor
extru %r24,31,1,%r19 ; r19 = n0 & 1
bb,< %r23,31,L$odd
extru %r23,30,31,%r22 ; r22 = d >> 1
shd %r25,%r24,1,%r24 ; r24 = new n0
extru %r25,30,31,%r25 ; r25 = new n1
sub %r0,%r22,%r21
ds %r0,%r21,%r0
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
comclr,>= %r25,%r0,%r0
addl %r25,%r22,%r25
sh1addl %r25,%r19,%r25
stws %r25,0(0,%r26)
bv 0(%r2)
addc %r24,%r24,%r28
L$odd addib,sv,n 1,%r22,L$FF.. ; r22 = (d / 2 + 1)
shd %r25,%r24,1,%r24 ; r24 = new n0
extru %r25,30,31,%r25 ; r25 = new n1
sub %r0,%r22,%r21
ds %r0,%r21,%r0
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r24
ds %r25,%r22,%r25
addc %r24,%r24,%r28
comclr,>= %r25,%r0,%r0
addl %r25,%r22,%r25
sh1addl %r25,%r19,%r25
; We have computed (n1,,n0) / (d + 1), q' = r28, r' = r25
add,nuv %r28,%r25,%r25
addl %r25,%r1,%r25
addc %r0,%r28,%r28
sub,<< %r25,%r23,%r0
addl %r25,%r1,%r25
stws %r25,0(0,%r26)
bv 0(%r2)
addc %r0,%r28,%r28
; This is just a special case of the code above.
; We come here when d == 0xFFFFFFFF
L$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

5
mpi/hppa1.1/distfiles
View file

@ -1,5 +0,0 @@
udiv-qrnnd.S
mpih-mul1.S
mpih-mul2.S
mpih-mul3.S

112
mpi/hppa1.1/mpih-mul1.S
View file

@ -1,112 +0,0 @@
/* 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 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
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
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
L$end stws,ma %r19,4(%r26)
addc %r28,%r1,%r19
ldw -16(%r30),%r28
stws,ma %r19,4(%r26)
addc %r0,%r28,%r28
bv 0(%r2)
ldo -64(%r30),%r30
L$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

114
mpi/hppa1.1/mpih-mul2.S
View file

@ -1,114 +0,0 @@
/* 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 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
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
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
L$end ldw 0(%r26),%r29
add %r29,%r19,%r19
stws,ma %r19,4(%r26)
addc %r28,%r1,%r19
ldw -16(%r30),%r28
ldws 0(%r26),%r29
addc %r0,%r28,%r28
add %r29,%r19,%r19
stws,ma %r19,4(%r26)
addc %r0,%r28,%r28
bv 0(%r2)
ldo -64(%r30),%r30
L$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

124
mpi/hppa1.1/mpih-mul3.S
View file

@ -1,124 +0,0 @@
/* 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 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
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
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
L$end ldw 0(%r26),%r29
sub %r29,%r19,%r22
add %r22,%r19,%r0
stws,ma %r22,4(%r26)
addc %r28,%r1,%r19
ldw -16(%r30),%r28
ldws 0(%r26),%r29
addc %r0,%r28,%r28
sub %r29,%r19,%r22
add %r22,%r19,%r0
stws,ma %r22,4(%r26)
addc %r0,%r28,%r28
bv 0(%r2)
ldo -64(%r30),%r30
L$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

88
mpi/hppa1.1/udiv-qrnnd.S
View file

@ -1,88 +0,0 @@
/* HP-PA __udiv_qrnnd division support, used from longlong.h.
* This version runs fast on PA 7000 and later.
*
* Copyright (C) 1993, 1994 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
L$0000 .word 0x43f00000
.word 0x0
.export __udiv_qrnnd
__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
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
L$2 bv 0(%r2)
stws %r22,0(0,%r26)
.exit
.procend

9
mpi/i386/distfiles
View file

@ -1,9 +0,0 @@
mpih-add1.S
mpih-mul1.S
mpih-mul2.S
mpih-mul3.S
mpih-lshift.S
mpih-rshift.S
mpih-sub1.S
syntax.h

118
mpi/i386/mpih-add1.S
View file

@ -1,118 +0,0 @@
/* 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 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

96
mpi/i386/mpih-lshift.S
View file

@ -1,96 +0,0 @@
/* i80386 lshift
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1992, 1994 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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 %cl,%ebx,%eax /* compute carry limb */
decl %edx
jz Lend
pushl %eax /* push carry limb onto stack */
testb $1,%edx
jnz L1 /* enter loop in the middle */
movl %ebx,%eax
ALIGN (3)
Loop: movl (%esi,%edx,4),%ebx /* load next lower limb */
shldl %cl,%ebx,%eax /* compute result limb */
movl %eax,(%edi,%edx,4) /* store it */
decl %edx
L1: movl (%esi,%edx,4),%eax
shldl %cl,%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

86
mpi/i386/mpih-mul1.S
View file

@ -1,86 +0,0 @@
/* i80386 mul_1 -- Multiply a limb vector with a limb and store
* the result in a second limb vector.
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1992, 1994 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

87
mpi/i386/mpih-mul2.S
View file

@ -1,87 +0,0 @@
/* i80386 addmul_1 -- Multiply a limb vector with a limb and add
* the result to a second limb vector.
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1992, 1994 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

87
mpi/i386/mpih-mul3.S
View file

@ -1,87 +0,0 @@
/* i80386 submul_1 -- Multiply a limb vector with a limb and add
* the result to a second limb vector.
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1992, 1994 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

98
mpi/i386/mpih-rshift.S
View file

@ -1,98 +0,0 @@
/* i80386 rshift
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1992, 1994 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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 %cl,%ebx,%eax /* compute carry limb */
incl %edx
jz Lend2
pushl %eax /* push carry limb onto stack */
testb $1,%edx
jnz L2 /* enter loop in the middle */
movl %ebx,%eax
ALIGN (3)
Loop2: movl (%esi,%edx,4),%ebx /* load next higher limb */
shrdl %cl,%ebx,%eax /* compute result limb */
movl %eax,(%edi,%edx,4) /* store it */
incl %edx
L2: movl (%esi,%edx,4),%eax
shrdl %cl,%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

118
mpi/i386/mpih-sub1.S
View file

@ -1,118 +0,0 @@
/* 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 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

62
mpi/i386/syntax.h
View file

@ -1,62 +0,0 @@
/* asm.h -- Definitions for x86 syntax variations.
Copyright (C) 1992, 1994, 1995 Free Software Foundation, Inc.
This file is part of the GNU MP Library.
The GNU MP 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 MP 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 MP Library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
MA 02111-1307, USA. */
#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

26
mpi/i586/README
View file

@ -1,26 +0,0 @@
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...

9
mpi/i586/distfiles
View file

@ -1,9 +0,0 @@
mpih-add1.S
mpih-mul1.S
mpih-mul2.S
mpih-mul3.S
mpih-lshift.S
mpih-rshift.S
mpih-sub1.S
README

134
mpi/i586/mpih-add1.S
View file

@ -1,134 +0,0 @@
/* 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 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

230
mpi/i586/mpih-lshift.S
View file

@ -1,230 +0,0 @@
/* i80586 lshift
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1992, 1994 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

89
mpi/i586/mpih-mul1.S
View file

@ -1,89 +0,0 @@
/* i80586 mul_1 -- Multiply a limb vector with a limb and store
* the result in a second limb vector.
* Copyright (C) 1992, 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

94
mpi/i586/mpih-mul2.S
View file

@ -1,94 +0,0 @@
/* i80586 addmul_1 -- Multiply a limb vector with a limb and add
* the result to a second limb vector.
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1992, 1994 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

94
mpi/i586/mpih-mul3.S
View file

@ -1,94 +0,0 @@
/* i80586 submul_1 -- Multiply a limb vector with a limb and add
* the result to a second limb vector.
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1992, 1994 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

229
mpi/i586/mpih-rshift.S
View file

@ -1,229 +0,0 @@
/* i80586 rshift
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1992, 1994 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

143
mpi/i586/mpih-sub1.S
View file

@ -1,143 +0,0 @@
/* 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 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

1497
mpi/longlong.h
View file

File diff suppressed because it is too large Load diff

8
mpi/m68k/distfiles
View file

@ -1,8 +0,0 @@
syntax.h
mpih-lshift.S
mpih-rshift.S
mpih-add1.S
mpih-sub1.S

3
mpi/m68k/mc68020/distfiles
View file

@ -1,3 +0,0 @@
mpih-mul1.S
mpih-mul2.S
mpih-mul3.S

97
mpi/m68k/mc68020/mpih-mul1.S
View file

@ -1,97 +0,0 @@
/* 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 Free Software Foundation, Inc.
This file is part of the GNU MP Library.
The GNU MP 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 MP 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 MP Library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
MA 02111-1307, USA. */
#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)

87
mpi/m68k/mc68020/mpih-mul2.S
View file

@ -1,87 +0,0 @@
/* 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 Free Software Foundation, Inc.
This file is part of the GNU MP Library.
The GNU MP 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 MP 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 MP Library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
MA 02111-1307, USA. */
#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)

89
mpi/m68k/mc68020/mpih-mul3.S
View file

@ -1,89 +0,0 @@
/* 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 Free Software Foundation, Inc.
This file is part of the GNU MP Library.
The GNU MP 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 MP 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 MP Library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
MA 02111-1307, USA. */
#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)

85
mpi/m68k/mpih-add1.S
View file

@ -1,85 +0,0 @@
/* 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 Free Software Foundation, Inc.
This file is part of the GNU MP Library.
The GNU MP 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 MP 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 MP Library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
MA 02111-1307, 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 + 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)

158
mpi/m68k/mpih-lshift.S
View file

@ -1,158 +0,0 @@
/* mc68020 lshift -- Shift left a low-level natural-number integer.
Copyright (C) 1996, 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 Library 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 Library General Public
License for more details.
You should have received a copy of the GNU Library General Public License
along with the GNU MP Library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
MA 02111-1307, USA.
*/
#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)

156
mpi/m68k/mpih-rshift.S
View file

@ -1,156 +0,0 @@
/* mc68020 rshift -- Shift right a low-level natural-number integer.
Copyright (C) 1996, 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 Library 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 Library General Public
License for more details.
You should have received a copy of the GNU Library General Public License
along with the GNU MP Library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
MA 02111-1307, USA.
*/
#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)

85
mpi/m68k/mpih-sub1.S
View file

@ -1,85 +0,0 @@
/* 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 Free Software Foundation, Inc.
This file is part of the GNU MP Library.
The GNU MP 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 MP 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 MP Library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
MA 02111-1307, USA. */
#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)

177
mpi/m68k/syntax.h
View file

@ -1,177 +0,0 @@
/* asm.h -- Definitions for 68k syntax variations.
Copyright (C) 1992, 1994, 1996 Free Software Foundation, Inc.
This file is part of the GNU MP Library.
The GNU MP 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 MP 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 MP Library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
MA 02111-1307, USA. */
#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

23
mpi/mips3/README
View file

@ -1,23 +0,0 @@
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...

10
mpi/mips3/distfiles
View file

@ -1,10 +0,0 @@
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

10
mpi/mips3/mpi-asm-defs.h
View file

@ -1,10 +0,0 @@
/* 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

122
mpi/mips3/mpih-add1.S
View file

@ -1,122 +0,0 @@
/* mips3 add_n -- Add two limb vectors of the same length > 0 and store
* sum in a third 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

95
mpi/mips3/mpih-lshift.S
View file

@ -1,95 +0,0 @@
/* mips3 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

87
mpi/mips3/mpih-mul1.S
View file

@ -1,87 +0,0 @@
/* 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 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

99
mpi/mips3/mpih-mul2.S
View file

@ -1,99 +0,0 @@
/* 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 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

99
mpi/mips3/mpih-mul3.S
View file

@ -1,99 +0,0 @@
/* 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 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

93
mpi/mips3/mpih-rshift.S
View file

@ -1,93 +0,0 @@
/* mips3 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

123
mpi/mips3/mpih-sub1.S
View file

@ -1,123 +0,0 @@
/* mips3 sub_n -- Subtract two limb vectors of the same length > 0 and
* store difference in a third 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

247
mpi/mpi-add.c
View file

@ -1,247 +0,0 @@
/* mpi-add.c - MPI functions
* Copyright (C) 1998 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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 );
}

264
mpi/mpi-bit.c
View file

@ -1,264 +0,0 @@
/* 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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 int
mpi_get_nbits( MPI a )
{
unsigned n;
if( mpi_is_opaque(a) ) {
return a->sign; /* which holds the number of bits */
}
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;
}
unsigned int
gcry_mpi_get_nbits( MPI a )
{
return mpi_get_nbits( a );
}
/****************
* 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;
}

74
mpi/mpi-cmp.c
View file

@ -1,74 +0,0 @@
/* 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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;
}

322
mpi/mpi-div.c
View file

@ -1,322 +0,0 @@
/* mpi-div.c - MPI functions
* Copyright (C) 1998 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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 );
}

54
mpi/mpi-gcd.c
View file

@ -1,54 +0,0 @@
/* mpi-gcd.c - MPI functions
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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);
}

36
mpi/mpi-inline.c
View file

@ -1,36 +0,0 @@
/* mpi-inline.c
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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"

129
mpi/mpi-inline.h
View file

@ -1,129 +0,0 @@
/* mpi-inline.h - Internal to the Multi Precision Integers
* Copyright (C) 1998, 1999 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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*/

277
mpi/mpi-internal.h
View file

@ -1,277 +0,0 @@
/* mpi-internal.h - Internal to the Multi Precision Integers
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1994, 1996, 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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-asm-defs.h"
#if BYTES_PER_MPI_LIMB == SIZEOF_UNSIGNED_INT
typedef unsigned int mpi_limb_t;
typedef signed int mpi_limb_signed_t;
#elif BYTES_PER_MPI_LIMB == SIZEOF_UNSIGNED_LONG
typedef unsigned long int mpi_limb_t;
typedef signed long int mpi_limb_signed_t;
#elif BYTES_PER_MPI_LIMB == SIZEOF_UNSIGNED_LONG_LONG
typedef unsigned long long int mpi_limb_t;
typedef signed long long int mpi_limb_signed_t;
#elif BYTES_PER_MPI_LIMB == SIZEOF_UNSIGNED_SHORT
typedef unsigned short int mpi_limb_t;
typedef signed short int mpi_limb_signed_t;
#else
#error BYTES_PER_MPI_LIMB does not match any C type
#endif
#define BITS_PER_MPI_LIMB (8*BYTES_PER_MPI_LIMB)
#define DID_MPI_LIMB_TYPEDEF 1
#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*/

270
mpi/mpi-inv.c
View file

@ -1,270 +0,0 @@
/* mpi-inv.c - MPI functions
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
}

222
mpi/mpi-mpow.c
View file

@ -1,222 +0,0 @@
/* mpi-mpow.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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
*/
#include <config.h>
#include <stdio.h>
#include <stdlib.h>
#include "mpi-internal.h"
#include "longlong.h"
#include <assert.h>
/* Barrett is slower than the classical way. It can be tweaked by
* using partial multiplications
*/
/*#define USE_BARRETT*/
#ifdef USE_BARRETT
static void barrett_mulm( MPI w, MPI u, MPI v, MPI m, MPI y, int k, MPI r1, MPI r2 );
static MPI init_barrett( MPI m, int *k, MPI *r1, MPI *r2 );
static int calc_barrett( MPI r, MPI x, MPI m, MPI y, int k, MPI r1, MPI r2 );
#else
#define barrett_mulm( w, u, v, m, y, k, r1, r2 ) mpi_mulm( (w), (u), (v), (m) )
#endif
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;
}
/*log_debug("t=%d i=%d idx=%d\n", t, i, idx );*/
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;
#ifdef USE_BARRETT
MPI barrett_y, barrett_r1, barrett_r2;
int barrett_k;
#endif
for(k=0; basearray[k]; k++ )
;
assert(k);
for(t=0, i=0; (tmp=exparray[i]); i++ ) {
/*log_mpidump("exp: ", tmp );*/
j = mpi_get_nbits(tmp);
if( j > t )
t = j;
}
/*log_mpidump("mod: ", m );*/
assert(i==k);
assert(t);
assert( k < 10 );
G = g10_xcalloc( (1<<k) , sizeof *G );
#ifdef USE_BARRETT
barrett_y = init_barrett( m, &barrett_k, &barrett_r1, &barrett_r2 );
#endif
/* and calculate */
tmp = mpi_alloc( mpi_get_nlimbs(m)+1 );
mpi_set_ui( res, 1 );
for(i = 1; i <= t; i++ ) {
barrett_mulm(tmp, res, res, m, barrett_y, barrett_k,
barrett_r1, barrett_r2 );
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
barrett_mulm( G[idx], G[idx], basearray[j],
m, barrett_y, barrett_k, barrett_r1, barrett_r2 );
}
}
if( !G[idx] )
G[idx] = mpi_alloc(0);
}
}
barrett_mulm(res, tmp, G[idx], m, barrett_y, barrett_k, barrett_r1, barrett_r2 );
}
/* cleanup */
mpi_free(tmp);
#ifdef USE_BARRETT
mpi_free(barrett_y);
mpi_free(barrett_r1);
mpi_free(barrett_r2);
#endif
for(i=0; i < (1<<k); i++ )
mpi_free(G[i]);
g10_free(G);
}
#ifdef USE_BARRETT
static void
barrett_mulm( MPI w, MPI u, MPI v, MPI m, MPI y, int k, MPI r1, MPI r2 )
{
mpi_mul(w, u, v);
if( calc_barrett( w, w, m, y, k, r1, r2 ) )
mpi_fdiv_r( w, w, m );
}
/****************
* Barrett precalculation: y = floor(b^(2k) / m)
*/
static MPI
init_barrett( MPI m, int *k, MPI *r1, MPI *r2 )
{
MPI tmp;
mpi_normalize( m );
*k = mpi_get_nlimbs( m );
tmp = mpi_alloc( *k + 1 );
mpi_set_ui( tmp, 1 );
mpi_lshift_limbs( tmp, 2 * *k );
mpi_fdiv_q( tmp, tmp, m );
*r1 = mpi_alloc( 2* *k + 1 );
*r2 = mpi_alloc( 2* *k + 1 );
return tmp;
}
/****************
* Barrett reduction: We assume that these conditions are met:
* Given x =(x_2k-1 ...x_0)_b
* m =(m_k-1 ....m_0)_b with m_k-1 != 0
* Output r = x mod m
* Before using this function init_barret must be used to calucalte y and k.
* Returns: false = no error
* true = can't perform barret reduction
*/
static int
calc_barrett( MPI r, MPI x, MPI m, MPI y, int k, MPI r1, MPI r2 )
{
int xx = k > 3 ? k-3:0;
mpi_normalize( x );
if( mpi_get_nlimbs(x) > 2*k )
return 1; /* can't do it */
/* 1. q1 = floor( x / b^k-1)
* q2 = q1 * y
* q3 = floor( q2 / b^k+1 )
* Actually, we don't need qx, we can work direct on r2
*/
mpi_set( r2, x );
mpi_rshift_limbs( r2, k-1 );
mpi_mul( r2, r2, y );
mpi_rshift_limbs( r2, k+1 );
/* 2. r1 = x mod b^k+1
* r2 = q3 * m mod b^k+1
* r = r1 - r2
* 3. if r < 0 then r = r + b^k+1
*/
mpi_set( r1, x );
if( r1->nlimbs > k+1 ) /* quick modulo operation */
r1->nlimbs = k+1;
mpi_mul( r2, r2, m );
if( r2->nlimbs > k+1 ) /* quick modulo operation */
r2->nlimbs = k+1;
mpi_sub( r, r1, r2 );
if( mpi_is_neg( r ) ) {
MPI tmp;
tmp = mpi_alloc( k + 2 );
mpi_set_ui( tmp, 1 );
mpi_lshift_limbs( tmp, k+1 );
mpi_add( r, r, tmp );
mpi_free(tmp);
}
/* 4. while r >= m do r = r - m */
while( mpi_cmp( r, m ) >= 0 )
mpi_sub( r, r, m );
return 0;
}
#endif /* USE_BARRETT */

198
mpi/mpi-mul.c
View file

@ -1,198 +0,0 @@
/* mpi-mul.c - MPI functions
* Copyright (C) 1998 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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( 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 )
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 );
}

295
mpi/mpi-pow.c
View file

@ -1,295 +0,0 @@
/* mpi-pow.c - MPI functions
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1994, 1996, 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
gcry_mpi_powm( MPI res, MPI base, MPI exp, 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 = exp->nlimbs;
msize = mod->nlimbs;
size = 2 * msize;
esign = exp->sign;
msign = mod->sign;
esec = mpi_is_secure(exp);
msec = mpi_is_secure(mod);
bsec = mpi_is_secure(base);
rsec = mpi_is_secure(res);
rp = res->d;
ep = exp->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, EXP 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 EXP are identical. Allocate temp. space for EXP. */
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 );
}

132
mpi/mpi-scan.c
View file

@ -1,132 +0,0 @@
/* mpi-scan.c - MPI functions
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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!
*/
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;
}
/****************
* Put a value at position IDX into A. idx counts from lsb to msb
*/
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 */
}
/****************
* Count the number of zerobits at the low end of A
*/
unsigned
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;
}

624
mpi/mpicoder.c
View file

@ -1,624 +0,0 @@
/* mpicoder.c - Coder for the external representation of MPIs
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
*/
#include <config.h>
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <assert.h>
#include "mpi-internal.h"
#include "g10lib.h"
#define MAX_EXTERN_MPI_BITS 16384
static MPI
mpi_read_from_buffer(byte *buffer, unsigned *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_error("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;
j= val->nlimbs = nlimbs;
val->sign = 0;
for( ; j > 0; j-- ) {
a = 0;
for(; i < BYTES_PER_MPI_LIMB; i++ ) {
if( ++nread > *ret_nread )
log_bug("mpi larger than buffer");
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.
* FIXME: Replace this by the more generic gcry_mpi_print()
*/
static 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 = stderr; /* used to be log_stream() */
/* FIXME: Replace this function by a g10_log_xxx one */
fprintf(fp,"%s: ",text);
mpi_print(fp, a, 1 );
fputc('\n', fp);
}
/****************
* Return an m_alloced 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;
if( sign )
*sign = a->sign;
*nbytes = a->nlimbs * BYTES_PER_MPI_LIMB;
p = buffer = force_secure || mpi_is_secure(a) ? g10_xmalloc_secure( *nbytes)
: g10_xmalloc( *nbytes );
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 oepration 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 = *p-- ;
alimb |= *p-- << 8 ;
alimb |= *p-- << 16 ;
alimb |= *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 |= *p-- << 8 ;
if( p >= buffer ) alimb |= *p-- << 16 ;
if( p >= buffer ) alimb |= *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 );
}
int
gcry_mpi_scan( struct gcry_mpi **ret_mpi, enum gcry_mpi_format format,
const char *buffer, size_t *nbytes )
{
struct gcry_mpi *a = NULL;
unsigned int len;
len = nbytes? *nbytes : strlen(buffer);
/* TODO: add a way to allocate the MPI in secure memory
* Hmmm: maybe it is better to retrieve this information from
* the provided buffer. */
#warning secure memory is not used here.
if( format == GCRYMPI_FMT_STD ) {
const byte *s = buffer;
a = mpi_alloc( (len+BYTES_PER_MPI_LIMB-1) / BYTES_PER_MPI_LIMB );
if( len ) { /* not zero */
a->sign = *s & 0x80;
if( a->sign ) {
/* FIXME: we have to convert from 2compl to magnitude format */
mpi_free(a);
return GCRYERR_INTERNAL;
}
else
mpi_set_buffer( a, s, len, 0 );
}
if( ret_mpi ) {
mpi_normalize ( a );
*ret_mpi = a;
}
else
mpi_free(a);
return 0;
}
else if( format == GCRYMPI_FMT_USG ) {
a = mpi_alloc( (len+BYTES_PER_MPI_LIMB-1) / BYTES_PER_MPI_LIMB );
if( len ) /* not zero */
mpi_set_buffer( a, buffer, len, 0 );
if( ret_mpi ) {
mpi_normalize ( a );
*ret_mpi = a;
}
else
mpi_free(a);
return 0;
}
else if( format == GCRYMPI_FMT_PGP ) {
a = mpi_read_from_buffer( (char*)buffer, &len, 0 );
if( nbytes )
*nbytes = len;
if( ret_mpi ) {
mpi_normalize ( a );
*ret_mpi = a;
}
else
mpi_free(a);
return a? 0 : GCRYERR_INV_OBJ;
}
else if( format == GCRYMPI_FMT_SSH ) {
const byte *s = buffer;
size_t n;
if( len < 4 )
return GCRYERR_TOO_SHORT;
n = s[0] << 24 | s[1] << 16 | s[2] << 8 | s[3];
s += 4; len -= 4;
if( n > len )
return GCRYERR_TOO_LARGE; /* or should it be too_short */
a = mpi_alloc( (n+BYTES_PER_MPI_LIMB-1) / BYTES_PER_MPI_LIMB );
if( len ) { /* not zero */
a->sign = *s & 0x80;
if( a->sign ) {
/* FIXME: we have to convert from 2compl to magnitude format */
mpi_free(a);
return GCRYERR_INTERNAL;
}
else
mpi_set_buffer( a, s, n, 0 );
}
if( nbytes )
*nbytes = n+4;
if( ret_mpi ) {
mpi_normalize ( a );
*ret_mpi = a;
}
else
mpi_free(a);
return 0;
}
else if( format == GCRYMPI_FMT_HEX ) {
if( nbytes )
return GCRYERR_INV_ARG; /* can only handle C strings for now */
a = mpi_alloc(0);
if( mpi_fromstr( a, buffer ) )
return GCRYERR_INV_OBJ;
if( ret_mpi ) {
mpi_normalize ( a );
*ret_mpi = a;
}
else
mpi_free(a);
return 0;
}
else
return GCRYERR_INV_ARG;
}
/****************
* Write a using format into buffer which has a length of *NBYTES.
* Returns the number of bytes actually written in nbytes.
* Buffer maybe NULL to query the required length of the buffer
*/
int
gcry_mpi_print( enum gcry_mpi_format format, char *buffer, size_t *nbytes,
struct gcry_mpi *a )
{
unsigned int nbits = mpi_get_nbits(a);
size_t len;
if( !nbytes )
return GCRYERR_INV_ARG;
len = *nbytes;
*nbytes = 0;
if( format == GCRYMPI_FMT_STD ) {
char *tmp;
int extra = 0;
unsigned int n;
if( a->sign )
return GCRYERR_INTERNAL; /* can't handle it yet */
tmp = mpi_get_buffer( a, &n, NULL );
if( n && (*tmp & 0x80) ) {
n++;
extra=1;
}
if( n > len && buffer ) {
g10_free(tmp);
return GCRYERR_TOO_SHORT; /* the provided buffer is too short */
}
if( buffer ) {
byte *s = buffer;
if( extra )
*s++ = 0;
memcpy( s, tmp, n-extra );
}
g10_free(tmp);
*nbytes = n;
return 0;
}
else if( format == GCRYMPI_FMT_USG ) {
unsigned int n = (nbits + 7)/8;
/* we ignore the sign for this format */
/* FIXME: for performance reasons we should put this into
* mpi_aprint becuase we can then use the buffer directly */
if( n > len && buffer )
return GCRYERR_TOO_SHORT; /* the provided buffer is too short */
if( buffer ) {
char *tmp;
tmp = mpi_get_buffer( a, &n, NULL );
memcpy( buffer, tmp, n );
g10_free(tmp);
}
*nbytes = n;
return 0;
}
else if( format == GCRYMPI_FMT_PGP ) {
unsigned int n = (nbits + 7)/8;
if( a->sign )
return GCRYERR_INV_ARG; /* pgp format can only handle unsigned */
if( n+2 > len && buffer )
return GCRYERR_TOO_SHORT; /* the provided buffer is too short */
if( buffer ) {
char *tmp;
byte *s = buffer;
s[0] = nbits >> 8;
s[1] = nbits;
tmp = mpi_get_buffer( a, &n, NULL );
memcpy( s+2, tmp, n );
g10_free(tmp);
}
*nbytes = n+2;
return 0;
}
else if( format == GCRYMPI_FMT_SSH ) {
char *tmp;
int extra = 0;
unsigned int n;
if( a->sign )
return GCRYERR_INTERNAL; /* can't handle it yet */
tmp = mpi_get_buffer( a, &n, NULL );
if( n && (*tmp & 0x80) ) {
n++;
extra=1;
}
if( n+4 > len && buffer ) {
g10_free(tmp);
return GCRYERR_TOO_SHORT; /* the provided buffer is too short */
}
if( buffer ) {
byte *s = buffer;
*s++ = n >> 24;
*s++ = n >> 16;
*s++ = n >> 8;
*s++ = n;
if( extra )
*s++ = 0;
memcpy( s, tmp, n-extra );
}
g10_free(tmp);
*nbytes = 4+n;
return 0;
}
else if( format == GCRYMPI_FMT_HEX ) {
byte *tmp;
int i;
int extra = 0;
unsigned int n=0;
tmp = mpi_get_buffer( a, &n, NULL );
if( !n || (*tmp & 0x80) )
extra=2;
if( 2*n + extra + !!a->sign + 1 > len && buffer ) {
g10_free(tmp);
return GCRYERR_TOO_SHORT; /* the provided buffer is too short */
}
if( buffer ) {
byte *s = buffer;
if( a->sign )
*s++ = '-';
if( extra ) {
*s++ = '0';
*s++ = '0';
}
for(i=0; i < n; i++ ) {
unsigned int c = tmp[i];
*s++ = (c >> 4) < 10? '0'+(c>>4) : 'A'+(c>>4)-10 ;
c &= 15;
*s++ = c < 10? '0'+c : 'A'+c-10 ;
}
*s++ = 0;
*nbytes = (char*)s - buffer;
}
else {
*nbytes = 2*n + extra + !!a->sign + 1;
}
g10_free(tmp);
return 0;
}
else
return GCRYERR_INV_ARG;
}
/****************
* Like gcry_mpi_print but this function allocates the buffer itself.
* The caller has to supply the address of a pointer. nbytes may be
* NULL.
*/
int
gcry_mpi_aprint( enum gcry_mpi_format format, void **buffer, size_t *nbytes,
struct gcry_mpi *a )
{
size_t n;
int rc;
*buffer = NULL;
rc = gcry_mpi_print( format, NULL, &n, a );
if( rc )
return rc;
*buffer = mpi_is_secure(a) ? g10_xmalloc_secure( n ) : g10_xmalloc( n );
rc = gcry_mpi_print( format, *buffer, &n, a );
if( rc ) {
g10_free(*buffer);
*buffer = NULL;
}
else if( nbytes )
*nbytes = n;
return rc;
}

62
mpi/mpih-cmp.c
View file

@ -1,62 +0,0 @@
/* mpihelp-sub.c - MPI helper functions
* Copyright (C) 1998 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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;
}

537
mpi/mpih-div.c
View file

@ -1,537 +0,0 @@
/* mpihelp-div.c - MPI helper functions
* Copyright (C) 1998 Free Software Foundation, Inc.
* Copyright (C) 1994, 1996, 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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;
}
}

527
mpi/mpih-mul.c
View file

@ -1,527 +0,0 @@
/* 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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 "g10lib.h" /* g10_is_secure() */
#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 = g10_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 = g10_is_secure( up ) || g10_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,
g10_is_secure( up ) || g10_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, g10_is_secure( up )
|| g10_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 = g10_xcalloc( 1, 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 );
g10_free( 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;
}

409
mpi/mpiutil.c
View file

@ -1,409 +0,0 @@
/* mpiutil.ac - Utility functions for MPI
* 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA
*/
#include <config.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <assert.h>
#include "g10lib.h"
#include "mpi-internal.h"
#include "memory.h"
/****************
* 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
mpi_alloc( unsigned nlimbs )
{
MPI a;
a = g10_xmalloc( sizeof *a );
a->d = nlimbs? mpi_alloc_limb_space( nlimbs, 0 ) : NULL;
a->alloced = nlimbs;
a->nlimbs = 0;
a->sign = 0;
a->flags = 0;
return a;
}
void
mpi_m_check( MPI a )
{
g10_check_heap(a);
g10_check_heap(a->d);
}
MPI
mpi_alloc_secure( unsigned nlimbs )
{
MPI a;
a = g10_xmalloc( sizeof *a );
a->d = nlimbs? mpi_alloc_limb_space( nlimbs, 1 ) : NULL;
a->alloced = nlimbs;
a->flags = 1;
a->nlimbs = 0;
a->sign = 0;
return a;
}
mpi_ptr_t
mpi_alloc_limb_space( unsigned nlimbs, int secure )
{
size_t len = nlimbs * sizeof(mpi_limb_t);
mpi_ptr_t p;
p = secure? g10_xmalloc_secure( len ) : g10_xmalloc( len );
return p;
}
void
mpi_free_limb_space( mpi_ptr_t a )
{
if( !a )
return;
g10_free(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 g10_realloc()]
*/
void
mpi_resize( MPI a, unsigned nlimbs )
{
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 g10_is_secure
*/
if( a->d )
a->d = g10_xrealloc(a->d, nlimbs * sizeof(mpi_limb_t) );
else /* FIXME: It may not be allocted in secure memory */
a->d = g10_xcalloc( nlimbs , sizeof(mpi_limb_t) );
a->alloced = nlimbs;
}
void
mpi_clear( MPI a )
{
a->nlimbs = 0;
a->flags = 0;
}
void
mpi_free( MPI a )
{
if( !a )
return;
if( a->flags & 4 )
g10_free( a->d );
else {
mpi_free_limb_space(a->d);
}
if( a->flags & ~7 )
log_bug("invalid flag value in mpi\n");
g10_free(a);
}
static 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;
}
bp = mpi_alloc_limb_space( a->nlimbs, 1 );
MPN_COPY( bp, ap, a->nlimbs );
a->d = bp;
mpi_free_limb_space(ap);
}
MPI
gcry_mpi_set_opaque( MPI a, void *p, unsigned int nbits )
{
if( !a ) {
a = mpi_alloc(0);
}
if( a->flags & 4 )
g10_free( a->d );
else {
mpi_free_limb_space(a->d);
}
a->d = p;
a->alloced = 0;
a->nlimbs = 0;
a->sign = nbits;
a->flags = 4;
return a;
}
void *
gcry_mpi_get_opaque( MPI a, unsigned int *nbits )
{
if( !(a->flags & 4) )
log_bug("mpi_get_opaque on normal mpi\n");
if( nbits )
*nbits = a->sign;
return a->d;
}
/****************
* Note: This copy function should not interpret the MPI
* but copy it transparently.
*/
MPI
mpi_copy( MPI a )
{
int i;
MPI b;
if( a && (a->flags & 4) ) {
void *p = g10_is_secure(a->d)? g10_xmalloc_secure( (a->sign+7)/8 )
: g10_xmalloc( (a->sign+7)/8 );
memcpy( p, a->d, (a->sign+7)/8 );
b = gcry_mpi_set_opaque( NULL, p, a->sign );
}
else if( a ) {
b = mpi_is_secure(a)? mpi_alloc_secure( a->nlimbs )
: mpi_alloc( a->nlimbs );
b->nlimbs = a->nlimbs;
b->sign = a->sign;
b->flags = a->flags;
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
mpi_alloc_like( MPI a )
{
MPI b;
if( a && (a->flags & 4) ) {
int n = (a->sign+7)/8;
void *p = g10_is_secure(a->d)? g10_malloc_secure( n )
: g10_malloc( n );
memcpy( p, a->d, n );
b = gcry_mpi_set_opaque( NULL, p, a->sign );
}
else if( a ) {
b = mpi_is_secure(a)? mpi_alloc_secure( a->nlimbs )
: mpi_alloc( a->nlimbs );
b->nlimbs = 0;
b->sign = 0;
b->flags = a->flags;
}
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->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->flags = 0;
}
MPI
mpi_alloc_set_ui( unsigned long u)
{
MPI w = mpi_alloc(1);
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;
}
GCRY_MPI
gcry_mpi_new( unsigned int nbits )
{
return mpi_alloc( (nbits+BITS_PER_MPI_LIMB-1) / BITS_PER_MPI_LIMB );
}
GCRY_MPI
gcry_mpi_snew( unsigned int nbits )
{
return mpi_alloc_secure( (nbits+BITS_PER_MPI_LIMB-1) / BITS_PER_MPI_LIMB );
}
void
gcry_mpi_release( GCRY_MPI a )
{
mpi_free( a );
}
GCRY_MPI
gcry_mpi_copy( const GCRY_MPI a )
{
return mpi_copy( (GCRY_MPI)a );
}
GCRY_MPI
gcry_mpi_set( GCRY_MPI w, const GCRY_MPI u )
{
if( !w )
w = mpi_alloc( mpi_get_nlimbs(u) );
mpi_set( w, (GCRY_MPI)u );
return w;
}
GCRY_MPI
gcry_mpi_set_ui( GCRY_MPI w, unsigned long u )
{
if( !w )
w = mpi_alloc(1);
mpi_set_ui( w, u );
return w;
}
int
gcry_mpi_cmp( const GCRY_MPI u, const GCRY_MPI v )
{
return mpi_cmp( (GCRY_MPI)u, (GCRY_MPI)v );
}
int
gcry_mpi_cmp_ui( const GCRY_MPI u, unsigned long v )
{
return mpi_cmp_ui( (GCRY_MPI)u, v );
}
void
gcry_mpi_randomize( GCRY_MPI w,
unsigned int nbits, enum gcry_random_level level )
{
char *p = mpi_is_secure(w) ? gcry_random_bytes( (nbits+7)/8, level )
: gcry_random_bytes_secure( (nbits+7)/8, level );
mpi_set_buffer( w, p, (nbits+7)/8, 0 );
g10_free(p);
}
void
gcry_mpi_set_flag( GCRY_MPI a, enum gcry_mpi_flag flag )
{
switch( flag ) {
case GCRYMPI_FLAG_SECURE: mpi_set_secure(a); break;
case GCRYMPI_FLAG_OPAQUE:
default: log_bug("invalid flag value\n");
}
}
void
gcry_mpi_clear_flag( GCRY_MPI a, enum gcry_mpi_flag flag )
{
switch( flag ) {
case GCRYMPI_FLAG_SECURE:
case GCRYMPI_FLAG_OPAQUE:
default: log_bug("invalid flag value\n");
}
}
int
gcry_mpi_get_flag( GCRY_MPI a, enum gcry_mpi_flag flag )
{
switch( flag ) {
case GCRYMPI_FLAG_SECURE: return (a->flags & 1);
case GCRYMPI_FLAG_OPAQUE: return (a->flags & 4);
default: log_bug("invalid flag value\n");
}
}

3
mpi/pa7100/distfiles
View file

@ -1,3 +0,0 @@
mpih-lshift.S
mpih-rshift.S

89
mpi/pa7100/mpih-lshift.S
View file

@ -1,89 +0,0 @@
/* hppa lshift
* optimized for the PA7100, where is runs at 3.25 cycles/limb
* Copyright (C) 1992, 1994 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
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
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
L$rest addib,= 4,%r24,L$end1
nop
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$eloop
vshd %r22,%r29,%r20
L$end1 stws,mb %r20,-4(0,%r26)
vshd %r29,%r0,%r20
bv 0(%r2)
stw %r20,-4(0,%r26)
L$end2 stws,mb %r20,-4(0,%r26)
L$0004 vshd %r22,%r0,%r20
bv 0(%r2)
stw %r20,-4(0,%r26)
.exit
.procend

85
mpi/pa7100/mpih-rshift.S
View file

@ -1,85 +0,0 @@
/* hppa rshift
* optimized for the PA7100, where is runs at 3.25 cycles/limb
* Copyright (C) 1992, 1994 Free Software Foundation, Inc.
* Copyright (C) 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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
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
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
L$rrest addib,= 4,%r24,L$rend1
nop
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$eroop
vshd %r29,%r22,%r20
L$rend1 stws,ma %r20,4(0,%r26)
vshd %r0,%r29,%r20
bv 0(%r2)
stw %r20,0(0,%r26)
L$rend2 stws,ma %r20,4(0,%r26)
L$r004 vshd %r0,%r22,%r20
bv 0(%r2)
stw %r20,0(0,%r26)
.exit
.procend

7
mpi/power/distfiles
View file

@ -1,7 +0,0 @@
mpih-add1.S
mpih-lshift.S
mpih-mul1.S
mpih-mul2.S
mpih-mul3.S
mpih-rshift.S
mpih-sub1.S

86
mpi/power/mpih-add1.S
View file

@ -1,86 +0,0 @@
/* IBM POWER add_n -- Add 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

64
mpi/power/mpih-lshift.S
View file

@ -1,64 +0,0 @@
/* 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

115
mpi/power/mpih-mul1.S
View file

@ -1,115 +0,0 @@
/* 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

130
mpi/power/mpih-mul2.S
View file

@ -1,130 +0,0 @@
/* 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., 59 Temple Place - Suite 330, Boston, MA 02111-1307, 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

Some files were not shown because too many files have changed in this diff Show more