d27b33644d
handle leading zeroes in q bug (66558). suggested by nicolson; r=mcgreer,a=wtc
2001-01-29 14:24:57 +00:00
752be879db
Bugzilla bug #66286 : link libnss3.so with -R '$ORIGIN' on Solaris so that
...
it searches for libfreebl_*.so in the same directory where it resides.
As a result, libnss3.so, not libnspr4.so, must be the one calling dlopen,
so libnss3.so can't call PR_LoadLibraryWithFlags.
Modified files: lib/freebl/loader.c, lib/nss/config.mk
2001-01-24 05:26:19 +00:00
73586118bb
fixes for prime generatation:
...
1. Increase maximum tries to 600 (based on 2*(approx. frequency of 1024-bit primes))
2. Lower the number of Miller-Rabin iterations for p and q based on HAC table 4.4
r=nelsonb,a=wtc
2001-01-18 16:39:10 +00:00
fe9fa27179
When searching for a prime, the number of Miller-Rabin tests to be
...
performed will be done in accordance with a table published in the
Handbook of Applied Cryptography. See Bug 65151. Also, changes to
test program for this function.
2001-01-18 01:39:17 +00:00
03f5d37be2
remove unspecified arg check (bug 65122)
2001-01-12 14:30:44 +00:00
3f0e83ae0c
bug 64706
...
when prime generation step fails, retry up to a specified number of times
2001-01-12 14:29:47 +00:00
a6efa3c2c5
Save and restore %ebx register in all functions. Fixes bug 64940.
2001-01-12 01:37:02 +00:00
729c505bf3
Changed HP assembler routine multacc512 in hpma512.s so that it
...
propagates carries exactly the same way that maxpy_little does.
Re-enabled use of multacc512 in mpi_hp.c.
2001-01-08 05:58:34 +00:00
c974f7d877
Fix infinite loop by making "i" be signed again.
...
Also match up braces for vi editor.
2001-01-08 03:26:50 +00:00
1f1f72480e
Generation of prime numbers has been failing on HP since MPI began to
...
use the newest HP assembler function, multacc512. So, that function
has been disabled. Further investigation is needed.
2001-01-08 01:01:35 +00:00
a12fc02b67
reflect function declaration changes to loader module (const stuff)
2001-01-06 17:07:07 +00:00
ffeb56bbd2
on hypbrid cpu models (except sparc), use long long instead of long
2001-01-06 17:06:26 +00:00
50c492d08c
fix broken build on bigendian platforms
2001-01-06 16:34:23 +00:00
ec87616c36
widespread changes to reduce compiler warnings. liberal application of const.
2001-01-05 22:37:50 +00:00
97cc48fda6
Reduce warnings by adding missing #includes to eliminate calls to
...
undeclared functions.
2001-01-04 08:21:15 +00:00
351f30a205
Bugzilla: 64132. NSS lock instrumentation
2001-01-03 19:51:22 +00:00
2266954d60
fix problem in decrypt
2001-01-02 16:57:47 +00:00
dfe078513f
Don't build libfreebl_*_3.a . Just build the .so's.
2000-12-29 23:21:51 +00:00
16a28a548d
This file has been superseded by mpv_sparcv8.s and mpv_sparcv9.s
2000-12-29 01:07:00 +00:00
165ce1171a
Implement new mpi function mpi_read_variable_radix(). It's like
...
mpi_read_radix(), except that if the input stream begins with a leading
zero or "0x", it will interpret the input as octal or hex, respectively.
2000-12-29 01:01:37 +00:00
99f5c067fe
Optimized version of these functions, contributed by Sun.
2000-12-29 00:36:31 +00:00
2701ecb476
Add missing release_md target. This will cause the new DSOs to be
...
included in the released jar files.
2000-12-28 21:35:55 +00:00
30ce773929
Define NSS_USE_HYBRID in hybrid builds.
2000-12-28 04:13:02 +00:00
19aaa707ef
Separate the floating point and integer implementations of modular
...
exponentiation into two functions that can be chosen at run time.
2000-12-28 03:41:50 +00:00
d84a97a8ac
Radical change to freebl "32-bit" builds on Solaris and HP-UX "32-bit".
...
Build two DSOs, both conforming to the platform's 32-bit ABI, one of
which uses only 32-bit instructions, the other uses 64-bit instructions.
Then load the best one for the local CPU at run time. This allows
greatly enhanced performance on 64-bit CPUs while still supporting older
32-bit CPUs.
2000-12-28 03:39:01 +00:00
3cf4961be7
mapfile defines the public ABI for the new freebl DSOs.
2000-12-28 03:35:43 +00:00
4a9ba412bb
Verify vector version and length before relying on the DSO's vector.
2000-12-28 03:26:29 +00:00
46ef2e65b7
Bugzilla bug #63807 : fixed function prototype of SHA1_TraceState.
2000-12-27 18:47:57 +00:00
b41b82ce9d
This code permits multiple implementations of freebl to coexist for a
...
single platform. The best implementation for the local CPU is dynamicly
loaded at run time. Requires DSOs to be in the LD_LIBRARY_PATH or
equivalent (e.g. SHLIB_PATH).
2000-12-27 03:20:04 +00:00
855cec4af9
Stubs for RC5 functions. This is NOT a working implementation of RC5.
2000-12-27 03:16:59 +00:00
f7c24847c7
Attempt to clarify code to reduce confusion about side effect of
...
MP_CHECKOK macro.
2000-12-27 03:14:28 +00:00
29c09b0208
Ensure function SHA1_TraceState is always defined, even when TRACING_SSL
...
is not defined.
2000-12-27 02:41:59 +00:00
3dc39fe215
Change HP 64-bit and hybrid builds to use new HP assembly code.
...
Fix uses of $(LIB).
2000-12-27 02:08:54 +00:00
9e517f5d33
The Initial Developer of the Original Code is Netscape
...
Communications Corporation. Portions created by Netscape are
Copyright (C) 2000 Netscape Communications Corporation. All Rights
Reserved.
2000-12-27 01:48:25 +00:00
ed7c69d604
The Initial Developer of the Original Code is the Hewlett-Packard Company.
...
Portions created by Hewlett-Packard Company are
Copyright (C) 1999 Hewlett-Packard Company. All Rights Reserved.
2000-12-27 01:45:23 +00:00
14f9d75af8
The Initial Developer of the Original Code is the Hewlett-Packard Company.
...
Portions created by Hewlett-Packard Company are
Copyright (C) 1997 Hewlett-Packard Company. All Rights Reserved.
2000-12-27 01:44:44 +00:00
2b04b7346c
Revise mp_div and s_mp_div substantially. Reduce the number of temporary
...
variables. Eliminate inner loop of shifts from s_mp_div.
2000-12-20 05:54:18 +00:00
49ed8cb93a
initial checkin of Rijndael (AES) implementation. Only blocksize == 128 bits
...
has been tested on this checkin.
Added make target "rijndael_tables", which will generate rijndael32.tab.
2000-12-19 23:43:12 +00:00
f6d5a84b79
Correct make variable name for AIX.
2000-12-15 06:00:55 +00:00
a758f782ec
Change AIX 32-bit options to use new fine-grained feature test macros.
2000-12-15 05:54:08 +00:00
fd321489ff
Implement new feature test macros for controlling the use of mp_word over
...
individual types of operations, e.g., multiply, divide, add, subtract.
Use a specific combination of these for 32-bit on AIX.
2000-12-15 05:38:09 +00:00
cbcc0658f2
Add missing Makefile lines to build mpi optimally for AIX 32 and 64-bit.
2000-12-15 03:01:23 +00:00
a99008bc8d
Fix mpi-test for machines that have long or long long digits.
2000-12-15 03:00:00 +00:00
6502891944
Fix hybrid model builds for HP-UX.
2000-12-13 04:36:28 +00:00
3be31901cc
Start building freebl with MPI's new floating point acceleration code.
2000-12-13 01:26:48 +00:00
d50f40cfc7
Add support for unsigned long long mp_digits.
...
Revise and simplify the scheme of feature test macros used in mpi.h.
Remove all NSS_ and SOLARIS ifdefs from the source.
2000-12-13 01:22:22 +00:00
18d92f7b48
This file is the output of Sun's WorkShop 5.0 compiler for mpv_sparc.c,
...
when compiled for the v8+ or v9 architecture. This file may be built
with WorkShop 4.2 and will produce the same .o as WorkShop 5.0.
2000-12-13 01:19:55 +00:00
5073b4c76e
On DEC Alpha/OSF1, following the examples in <c_asm.h>, use inline
...
assembly code to get 64-bit products and squares from 32-bit multipliers
in just two instructions. Speeds up multiplication and squaring a lot.
2000-12-09 03:36:41 +00:00
04688e0249
When building without the inline assembly macros, use a simple cast
...
rather than the math lib's floor() function. Also, don't and with a
mask of all 1 bits when casting to a smaller unsigned type will suffice.
2000-12-09 03:32:41 +00:00
7a0b5a0ef2
This modification, contributed by Sun, avoids computing values >= 2^52.
...
This code runs much faster than the previous version on UltraSparc-3 CPUs,
and slightly slower than the previous version on UltraSparc-II CPUs.
2000-12-08 00:20:36 +00:00
885d29d0b4
Use the new floating point Montgomery multiply code from Sun on Solaris.
2000-12-02 02:37:22 +00:00
f49b012cd4
Back out a previous change to one line of this code for performance.
2000-12-01 03:58:17 +00:00
6abe6c9e6c
Add new optional third argument which specifies modulus size in bytes.
...
Don't print values for powers 0, 1, and 2 unless the repetition count
(the optional second argument) is 1.
2000-12-01 03:56:12 +00:00
32b43a3ae6
Replace compiler-generated inner loop code with optimized code.
...
This optimized code was contributed by Sun Microsystems Inc.
2000-12-01 03:53:13 +00:00
196b9b582c
* The Original Code is SPARC hand-optimized Montgomery multiply functions.
...
*
* The Initial Developer of the Original Code is Sun Microsystems Inc.
* Portions created by Sun Microsystems Inc. are
* Copyright (C) 1999-2000 Sun Microsystems Inc. All Rights Reserved.
2000-12-01 03:18:27 +00:00
291aa8039a
* inline macros for SPARC Montgomery multiply functions.
...
*
* The Initial Developer of the Original Code is Sun Microsystems Inc.
* Portions created by Sun Microsystems Inc. are
* Copyright (C) 1999-2000 Sun Microsystems Inc. All Rights Reserved.
2000-12-01 03:17:53 +00:00
8f9113a6de
Add missing error detection to identest. Change the order in which
...
mp_ints are destroyed to make debugging easier. These changes
facilitated detecting and fixing bug 61394.
2000-11-29 04:07:10 +00:00
4a90b4de0b
Fix bug 61394. Don't increase length of input vector by 1. Avoid
...
copying the vector twice.
2000-11-29 01:30:28 +00:00
2889df8598
Get the montmulf code to build, in preparation for integration into MPI.
2000-11-17 20:15:29 +00:00
1342aa1546
checkin of blinding implementation
2000-11-17 17:58:35 +00:00
a5f98d63db
return SEC_ERR_NEED_RANDOM when the RNG does not have sufficient seeding data to proceed.
2000-11-17 01:01:36 +00:00
0c02bcc920
these functions should be static
2000-11-17 00:59:51 +00:00
7a6c3be3e4
Add values for test modulus and exponent for benchmarking purposes.
2000-11-08 02:44:17 +00:00
c029359f7f
Fix mpp_pprime (Miller-Rabin Pseudo Primality test) to report MP_NO,
...
instead of MP_OKAY, for numbers == 3 mod 4 that are composite.
Also added two test cases to mpi-test.c. Fixes bug 59438.
2000-11-08 01:52:52 +00:00
ae93637c3d
Make this file buildable on Mac
2000-11-08 01:11:23 +00:00
0bc2e73f90
Changes to RSA keygen:
...
* use mp_invmod instead of mp_xgcd. nelsonb optimized mp_invmod for even moduli, and this will take advantage of it.
* Assure that p > q.
* Assure that n is actually keysizeInBits long by setting the two high-order bits of p and q during generation. Previously, it was possible for n to be 1 bit short.
2000-10-31 16:52:31 +00:00
575f23cc64
Performance benchmark for modular exponentiation.
...
Test modulus and exponent to be added later.
2000-10-26 23:30:50 +00:00
667d4a7e36
New implementation of mp_invmod for even moduli. 3x-500x faster than
...
xgcd for even moduli.
2000-10-24 21:32:53 +00:00
e7bf810f5e
fix for 64-bit platforms.
2000-10-18 15:41:23 +00:00
6534f684da
* The Original Code is SPARC hand-optimized Montgomery multiply functions.
...
*
* The Initial Developer of the Original Code is Sun Microsystems Inc.
* Portions created by Sun Microsystems Inc. are
* Copyright (C) 1999-2000 Sun Microsystems Inc. All Rights Reserved.
2000-10-13 00:30:36 +00:00
00b0412ddb
* inline macros for SPARC Montgomery multiply functions.
...
*
* The Initial Developer of the Original Code is Sun Microsystems Inc.
* Portions created by Sun Microsystems Inc. are
* Copyright (C) 1999-2000 Sun Microsystems Inc. All Rights Reserved.
2000-10-13 00:28:48 +00:00
67dc4d8c49
* interface file for SPARC Montgomery multiply functions.
...
*
* The Initial Developer of the Original Code is Sun Microsystems Inc.
* Portions created by Sun Microsystems Inc. are
* Copyright (C) 1999-2000 Sun Microsystems Inc. All Rights Reserved.
2000-10-13 00:28:08 +00:00
5e81658537
* The Original Code is SPARC optimized Montgomery multiply functions.
...
*
* The Initial Developer of the Original Code is Sun Microsystems Inc.
* Portions created by Sun Microsystems Inc. are
* Copyright (C) 1999-2000 Sun Microsystems Inc. All Rights Reserved.
2000-10-13 00:27:04 +00:00
8882eb9358
HPUX 64-bit also failed.
2000-10-05 02:47:43 +00:00
4979496e5a
workaround for bug #55234 . optimized arcfour is not working on solaris 64-bit.
2000-10-05 01:33:54 +00:00
1f2ad178df
Bugzilla bug #54869 : support gcc for 32-bit Solaris/SPARC build.
2000-10-02 22:52:43 +00:00
544ec0f11a
fix 3.1 bugs: change && to || for arg check, two new functions are static
2000-10-02 17:39:37 +00:00
2444aa8c5e
Added missing closing comment delimiter "*/".
2000-09-30 04:30:08 +00:00
e118ef4f63
Incorporate multiplication performance enhancements for UltraSparc CPUs
...
with the VIS instructions.
2000-09-30 01:46:30 +00:00
c206d3e3c8
The Initial Developer of the Original Code is Netscape Communications
...
Corporation. Portions created by Netscape are
Copyright (C) 2000 Netscape Communications Corporation.
All Rights Reserved.
2000-09-30 01:43:27 +00:00
191a128491
The Initial Developer of the Original Code is Sun Microsystems Inc.
...
Portions created by Sun Microsystems Inc. are
Copyright (C) 1999-2000 Sun Microsystems Inc. All Rights Reserved.
2000-09-29 23:39:30 +00:00
431bb85395
The Initial Developer of the Original Code is Sun Microsystems Inc.
...
Portions created by Sun Microsystems Inc. are
Copyright (C) 1995-2000 Sun Microsystems Inc. All Rights Reserved.
2000-09-29 23:38:05 +00:00
81cb1fd3ce
change for BSAFE. on windows, nsinstall will do copy instead of link.
...
need to forcibly remove library before attempting to reinstall.
2000-09-29 22:30:09 +00:00
9b1d01f3fd
Don't use type "ulong". Use "unsigned long" instead.
2000-09-29 19:51:32 +00:00
50098f2c22
Fix some bugs from the last checkin.
2000-09-29 04:03:34 +00:00
5a93a08faa
changes to allow build with RSA BSAFE Crypto-C library
2000-09-29 02:15:42 +00:00
181442bd0a
fixes based on review by nelsonb:
...
use mp_sub instead of mp_mod when possible.
increment generator candidate for DH instead of creating a new random one each time.
fix potential bugs when DH_Derive and KEA_Derive, where the secret is longer than the buffer the user requests.
check input key size to RSA_NewKey.
implement private key op with no CRT for PKCS12.
reduce base mod p and q before using mp_exptmod.
use mpl_lsh in pqg code (in place of exponent and multiply).
change PQG_Verify to return SECSuccess unless there is a library failure (not if verification fails).
2000-09-29 02:10:24 +00:00
58b6dcc398
Add mp_set_long and mp_set_ulong to the public API.
2000-09-28 22:53:45 +00:00
96a66d4965
revert changes to DH param gen. MPI does compute q=2p+1 for strong primes, but returns q not p. #ifdef in rsa for testing mp_exptmod_d vs. mp_exptmod, mp_exptmod is considerably faster.
2000-09-22 16:24:16 +00:00
2195f46fe1
Fix DH param gen, MPI returns Sophie-Germain prime q, then compute 2q + 1. Initial checkin of two KEA functions.
2000-09-21 23:45:24 +00:00
9bd5c43fa0
OS/2 build changes.
2000-09-20 17:07:22 +00:00
7d42df7e4b
don't set modulus octets until good primes are generated.
2000-09-19 06:18:04 +00:00
d5a2efabb1
don't set prime data until good primes are generated.
2000-09-19 06:03:52 +00:00
82b40af680
Bugzilla bug #53143 : removed the definition of NSPR internal macros.
...
Modified files: FreeBSD.mk, HP-UXB.10.10.mk, HP-UXB.10.20.mk, Linux.mk,
OSF1.mk, SunOS5.mk, cmd/SSLsample/Makefile.NSS, lib/freebl/Makefile.
2000-09-19 02:24:31 +00:00
c620564ee8
just some general cleanup
2000-09-16 15:56:06 +00:00
082f53f177
Eliminate extraneous "oop!" output at end of script on some platforms.
2000-09-14 00:32:24 +00:00
c4fe227bb9
Changes to make mpi use 64-bit digits on machines where unsigned longs
...
are 64-bits. Renamed s_mp_invmod_32b to s_mp_invmod_radix since radix
can now exceed 32 bits. Improved performance on all platforms except
MIPS and X86.
2000-09-14 00:31:01 +00:00
bdcc2c1a39
fix incorrect buffer zero
2000-09-13 18:44:11 +00:00
3d68e7a779
failure to save cx->i and cx->j caused second arcfour encrypt/decrypt to return garbage: bug 52192 r=mcgreer
2000-09-12 23:02:24 +00:00
4d941adcb4
leading zero problems fixed; thanks to nelsonb for pointing out that I needed mp_to_fixlen_octets
2000-09-12 20:32:43 +00:00
cac7876f1e
Fix two bugs in mp_invmod: a) returned wrong result if output var is also
...
one of the input vars, b) sometimes returned negative answer. Added a test
case to mpi-test.c with a value that previously produced a negative answer.
2000-09-12 00:41:09 +00:00
527ef9f038
initial checkin of Diffie-Hellman implementation
2000-09-11 17:11:50 +00:00
9aeb25aa0d
Changed the PR_GetError() call to PORT_GetError() because we use
...
PORT_SetError() to set error codes.
Modified files: dsa.c, rsa.c
2000-09-11 04:17:02 +00:00
d236df63ea
fix bugs where some nss code expects leading 0's (once used for signs). allows pkcs11 pairwise consistency check to operate correctly.
2000-09-08 22:41:51 +00:00
d57dc8ee4e
gcc does not recognize the -xchip=ultra2 option on Solaris/Sparc.
2000-09-08 01:05:23 +00:00
3e85ae179d
make compare_key do something (even if it will go away soon)
2000-09-07 23:18:40 +00:00
4135f08e8f
first draft of rsa CRT decryption
2000-09-07 16:29:23 +00:00
2230de4f0a
separate keygen from primegen, to facilitate testing (by using the self-test key from blapitest). using this verified the keygen process (against that self-test, anyway). leaving a testing function in temporarily.
2000-09-07 07:33:34 +00:00
8ddaa1f3d0
fix bad change made in keygen
2000-09-07 06:44:57 +00:00
112bbb432e
call extended euclidean algorithm directly instead of using mp_invmod
2000-09-07 03:14:16 +00:00
6e7fa5d7ee
Remove PQG_DestroyParams(), PQG_DestroyVerify from pqg.c because they
...
duplicate the functions in lib/util/pqgutil.c. Also, the functions in
pqgutil correctly handle the case where the Params or Verify was
constructed without an arena (a fact which is now more obvious when
reading the code there).
2000-09-07 01:24:52 +00:00
6db674e674
first draft of rsa public key op (tested against blapi self-test)
2000-09-06 23:58:41 +00:00
61e4b81333
first draft of rsa keygen
...
secmpi.h is defines used by nss code calling mpi
2000-09-06 23:27:34 +00:00
5de79419b1
private header file to talk between mpi and nss
2000-09-06 21:46:43 +00:00
865fd868f2
*added pqg code based on fips 186-1
...
*added stubs to rsa functions
2000-09-06 17:43:42 +00:00
fa26c513b0
When building on NT for X86, use new mpi_x86.asm implementation.
2000-09-02 05:41:21 +00:00
54ce559f68
Makefile.win - gmake Makefile for building MPI with MSVC on NT.
2000-09-02 05:38:51 +00:00
dab570427d
mpi_x86.asm - assembly language implementation of s_mpv_ functions.
2000-09-02 05:33:15 +00:00
0bbb06a9ca
Use new assembler files.
...
Define MP_NO_MP_WORD on platforms that don't have 64-bit registers.
2000-09-02 01:40:41 +00:00
5fe44c9349
Checkin a big speedup for sparc v8 CPUs when compiled with Workshop
...
compilers. Be sure to use -DMP_NO_MP_WORD with v8 CPUs.
2000-08-31 04:59:05 +00:00
eb2a2982e3
Fix mpi.h so that it compiles when MP_NO_MP_WORD is defined on the command
...
line (or prior to including mpi.h).
2000-08-31 03:59:48 +00:00
1b6110f143
Add missing copyright notice. Doh!
2000-08-31 03:45:39 +00:00
9b2c335e20
Some cleanup. Check error codes for all functions that return them.
2000-08-31 02:52:23 +00:00
8b0c1dcf1e
Big changes for performance. mp_digits are now always unsigned ints.
...
mp_words are used only on machines that support long long arithmetic.
s_mp_mod_d() was deleted. It was not being used and was not part of the
public API. The code that computes squares in s_mp_sqr was broken out
into a separate new function s_mpv_sqr_add_prop(), which is a target for
assembly language optimization. New function s_mpv_div_2dx1d(), also a
target for assembly optimization. These changes made X86 benchmark time
go from 22.5 seconds to 8.3 seconds on my reference test system.
2000-08-31 02:51:23 +00:00
52652e08a5
Add support for Linux on X86 CPUs. Use assembler code on MIPS CPUs.
2000-08-31 02:43:56 +00:00
5e06b71329
Gnu assembler implementations of all s_mpv_ functions for Intel X86 CPUs.
2000-08-31 02:41:20 +00:00
118024af9e
Add implementation of s_mpv_sqr_add_prop.
2000-08-31 02:40:32 +00:00
41b3d16d17
Change mpi-test to check the return values from most functions-under-test
...
explicitly for errors.
2000-08-29 04:26:23 +00:00
bb088780f3
Initial Checkin.
2000-08-26 23:50:58 +00:00
609c9bf7d4
don't zero input to random update (RNG_SystemInfoForRNG uses environment as input). Fix for bug 49898.
2000-08-23 18:11:47 +00:00
3431bd1245
Fix two uninitialized variables, found while porting to Linux.
2000-08-23 01:20:15 +00:00
7c623513e6
des.c wants _X86_ defined for intel CPUs.
...
coreconf does this for windows, but not for Linux, FreeBSD, etc.
2000-08-23 00:06:08 +00:00
f28d74b9ab
Big changes for performance.
...
- In mpi-priv.h, declare new 3 argument versions of s_mp_add and s_mp_sub.
Also declare new set of s_mpv_ functions that operate on vectors (arrays)
of mp_digits instead of on mp_ints. These functions are candidates for
implementation in assembler.
- In mpi.c reimplement mp_add and mp_sub using the new 3arg functions.
Implement 3 argument versions of s_mp_add and s_mp_sub.
This eliminates all need for temporary variables in mp_add and mp_sub.
Implement c language reference implementations of new s_mpv vector multiply
and multiply and add functions. Change mp_mul and mp_sqr so they no longer
pre-zero the output variable. It's no longer nececssary with the new s_mpv
functions. s_mp_pad no longer zeros out the new padded space.
-In mpmontg.c, implement variable width exponetiation windows. Implement
a new function to compute the multiply and Montgomery reduction in a
single pass. This is "Improvement 2" from Dusse' and Kaliski's paper
"A Cryptographic Library for the Motorola DSP56000". Performance impact
is negligible in this c implementation. However, this function is another
target for assembly language optimization.
2000-08-22 01:57:34 +00:00
a61f1b73ab
Overhaul Makefile. Change it to use gmake's "static pattern rules",
...
instead of having explicit individual rules for every program. Also,
build .o files for programs, and link them in a separate step. This
speeds building after changing a .c file in the library.
2000-08-22 01:45:25 +00:00
4cf2feb015
Get these tests to build with the current mpi API.
2000-08-22 01:19:26 +00:00
67da52ab59
Go back to using gmake.
2000-08-22 01:15:41 +00:00
ddb4786c4a
MIPS assembler code to optimize inner multiply loops for mips3 CPUs.
2000-08-22 00:55:10 +00:00
5585f66cd2
initial checkin of DSA implementation.
2000-08-18 18:32:33 +00:00
99c2fa3f33
Build des.c and desblapi.c.
2000-08-14 22:28:10 +00:00
792d99fbaa
The Initial Developer of the Original Code is Nelson B. Bolyard,
...
nelsonb@iname.com . Portions created by Nelson B. Bolyard are
Copyright (C) 1990, 2000 Nelson B. Bolyard, All Rights Reserved.
2000-08-14 22:27:20 +00:00
9cc24caab8
"clean" target will also remove primes.c from mpi.
2000-08-14 21:04:46 +00:00
c8c44aff3b
Add commented-out lines to build on other platforms.
...
Clean target now removes libmpi.a also.
2000-08-11 05:01:59 +00:00
8f6b77ce4d
Change mp_add and mp_sub to not allocate a temporary variable unless it's
...
really necessary.
2000-08-11 01:58:20 +00:00
d2a30db497
Replace s_mp_ispow2d with faster version. Change s_mp_mul_d to detect
...
multiplication by 0, 1, and powers of 2, and handle them without actual
multiplication.
2000-08-10 21:43:16 +00:00
67cf3588f2
Eliminate one more warning. Add missing f suffix to float constant.
2000-08-09 22:45:51 +00:00
5826c388d2
- Change mp_mul to minimize number of outer loops when multipliers are
...
of unequal length.
- Fix s_mp_fixup_reciprocal to not truncate the result.
- Use new macros in mpi-priv.h
2000-08-09 20:55:39 +00:00
203d540004
Use new MP_HOWMANY macro instead of equivalent code.
2000-08-09 20:53:25 +00:00
3f16031cb2
Comment out call to unimplemented function, so this will compile.
2000-08-09 20:46:59 +00:00
60e7af5b8b
Some unix platform kshells don't understand export VAR=value
...
So instead use VAR=value; export VAR
2000-08-09 20:44:05 +00:00
6e65d05e51
Add new macros: MP_MIN, MP_MAX, MP_HOWMANY, MP_ROUNDUP.
2000-08-09 20:42:18 +00:00
a4c0bd7e17
zero the correct amount of input bytes
2000-08-08 20:30:56 +00:00
1a81db9b8a
Fix crash caused by zeroing a larger buffer than the caller passed in.
2000-08-08 05:56:56 +00:00
53ac6c5c98
- Change mpi private function s_mp_mul_d_add_offset into a macro.
...
- Declare and implement new function s_mp_mul_add, which is a candidate
for replacement with assembler code.
- Convert mp_mul, mp_sqr, etc. to use s_mp_mul_add.
- New implementation of mp_invmod for odd moduli. Algorithm from paper
"Fast Modular Reciprocals" by Richard Schroeppel (a.k.a. Captain Nemo).
- New function s_mp_invmod_32b in mpi.c, computes inverse mod 2**32, also
from same paper. Used in mp_invmod and mp_exptmod.
2000-08-08 03:20:35 +00:00
0bade04ba3
Platform independent performance enhancements to functions that multiply,
...
square, subtract, right shift, compare, mul_d_add_offset. This lib's
Modular Exponentiation performance now compares favorably with most (not
all) other open source bignum libs on IRIX/R5000. No assembler code is
presently being used. Comparison on other platforms will now commence.
2000-08-05 03:37:46 +00:00
e30a05c7bf
Minor performance tweaks.
2000-08-04 19:58:20 +00:00
f5478e8ae0
Add new signed mp_word type, mp_sword. Used to get arithmetic right
...
shifts when doing subtraction.
2000-08-04 19:57:24 +00:00
a8b9681335
mpl_rsh now calls s_mp_div_2d instead of duplicating its logic.
...
mpl_lsh now calls s_mp_mul_2d instead of duplicating its logic.
2000-08-02 20:52:17 +00:00
26a0e342f4
- Implement s_mp_mul_2d, which complements s_mp_div_2d.
...
- Rewrite and speed up mp_sqr, most-frequently used in mp_exptmod().
- Speed up mp_sqrt by starting with a much better initial estimate.
2000-08-02 20:50:57 +00:00
da54a36647
Accept input args in hex, not decimal. Line up output into columns.
2000-08-02 20:48:28 +00:00
9620ef0ca4
Declare s_mp_mul_2d(), which is complementary to s_mp_div_2d().
2000-08-02 20:47:05 +00:00
886543402f
Reduced the number of temporary variables allocated and freed during a
...
modular exponentiation by over 99%. Modified mp_mul and mp_sqr to only
allocate temporary variables when absolutely needed. Changed mp_copy
and mp_init_copy to allocate space according to the amount allocated
in the source, reducing the need to grow the variable later.
2000-08-02 01:03:14 +00:00
147b7451ec
Investigating allocation of temporary variables.
2000-08-02 01:01:01 +00:00
e24f34f2c8
Declare mp_int function arguments "const" as appropriate.
2000-08-01 01:38:30 +00:00
9caea97bd1
Change default value for MP_ARGCHK to depend on whether or not DEBUG is
...
defined. Raise the default "precision" to reduce reallocations.
2000-08-01 01:36:30 +00:00
cb72925c4a
Fix initalizers in logtab.h to be "float" values, not "double" values.
2000-07-31 20:36:41 +00:00
1ebd55b401
Fix build problem caused by the fact that including stdlib.h implicitly
...
includes stdio.h on some platforms, but not on others.
2000-07-31 20:08:46 +00:00
4e92230a97
Use Windowed computation of exponent in mp_exptmod().
2000-07-30 06:37:14 +00:00
71be9cf03c
Add function mpl_get_bits(). Returns value of contiguous subset of bits
...
of bignum. Useful for windowed modular exponentiation.
2000-07-30 06:35:38 +00:00
78256a8fa7
MPI Arbitrary Precision Integer Arithmetic library.
...
The Initial Developer of the Original Code is
Michael J. Fromberger <sting@linguist.dartmouth.edu>
2000-07-30 02:05:19 +00:00
97ef19c3b8
MPI Arbitrary Precision Integer Arithmetic library.
...
The Initial Developer of the Original Code is
Michael J. Fromberger <sting@linguist.dartmouth.edu>
2000-07-30 01:56:35 +00:00
792ad45230
Build mpi library files in freebl.
2000-07-29 19:07:22 +00:00
df9c21a9de
Separate DH implementation for use with BSAFE and mpi-based implementation
...
into two files.
2000-07-29 19:06:54 +00:00
534ec52daa
Build libmpi.a. Include mpmontg.o in libmpi.a. Build programs with
...
libmpi.a. Fix dependencies so lib is rebuilt when headers change.
2000-07-28 23:11:12 +00:00
6b24794435
Modular exponentiation with Montgomery reduction for MPI.
...
The Initial Developer of the Original Code is
Netscape Communications Corporation. Portions created by Netscape are
Copyright (C) 2000 Netscape Communications Corporation.
All Rights Reserved.
2000-07-28 23:09:02 +00:00
71fb2fa5e4
Disambiguoate error messages in mod-expt test.
2000-07-28 23:04:23 +00:00
385fd4fc86
- Renamed mp_exptmod to s_mp_exptmod. This function is now used only in
...
when the modulus is even. mp_exptmod is now in mpmontg.c.
- Declare and define new function s_mp_mul_d_add_offset(), which computes
the product of an mp_digit and an mp_int, shifts the product to the left
by a number of digits, and adds the shifted product to another mp_int.
Used in multiplication and Montgomery reduction. Numerous platform
vendors have assembly language versions of this function.
- Removed some ANDs with DIGIT_MAX. Casting to mp_digit is necessary
and sufficient. The AND doesn't help.
2000-07-28 23:03:12 +00:00
f554a212b4
Use unsigned int for mp_digit instead of unsigned long when both are
...
32-bits. This reduced warnings. Use mp_size instead of unsigned int
in function declarations. Fix MP_DIGIT_FMT for 32-bit digits.
2000-07-28 22:55:56 +00:00
a38b5ec52f
Use mp_size instead of unsigned int in function prototypes.
2000-07-28 22:49:42 +00:00
99f3a8bf86
Add new function s_mp_add_offset().
...
Very handy in Montgomery reduction using Dusse' and Kalski's method.
2000-07-27 03:02:42 +00:00
b05581f7ec
Moved the declarations of many private defines and functions out of mpi.c
...
and into mpi-priv.h. Previously, mplogic.c and mpprime.c had duplicate
copies of some of those declarations. Now, they include mpi-priv.h
2000-07-27 02:38:54 +00:00
bf4e60e7b3
Fix sprintf format, can't use DIGIT_FMT.
2000-07-27 00:23:17 +00:00
ba8b232867
Move enhanced prime finder function mpp_make_prime from primegen utility
...
program into mpprime.c. declared in mpprime.h.
2000-07-26 05:41:59 +00:00
8bf87ce6bb
Fix bug in mp_div_2d that set r = q % d when a and q are the same mp_int.
...
Minor performance enhancement to s_mp_div_2d when DIGIT_BIT divides d.
2000-07-26 05:39:53 +00:00
7bb8634a3c
Fix a bug in recently added function mp_set_bit().
2000-07-26 05:32:30 +00:00
b9d1a404d2
Update nTires via a pointer argument to mpp_make_prime().
...
Change some printfs.
2000-07-25 00:16:57 +00:00
aef9190b64
Add new function mpp_fermat_list, which performs the fermat test on a
...
list of "witness" numbers. Simplify and accelerate mpp_pprime().
2000-07-25 00:15:06 +00:00
cbc85c762b
Fix assertion. Comparison sense was inverted.
2000-07-25 00:12:57 +00:00
f16cc60f1a
Add -fullwarn to some build flags. Change dependencies for primegen.
2000-07-22 07:22:37 +00:00
175cab0f15
Implement sieving to accelerate elimination of composites.
...
This sieve eliminates all but about 3500 numbers from each range of 64k.
2000-07-22 05:54:21 +00:00
1f040d3f9f
Use large prime table for faster prime gen. Make prime table "const".
...
Change functions that use prime table to use const arguments.
2000-07-22 05:36:45 +00:00
41121574a7
Change s_mp_norm to always use a power of two as the normalizing factor.
...
This enhancement performance by turning the normalizing and denormalizing
operations into shifts.
2000-07-21 21:06:08 +00:00
4b4e6bccd9
Implement mpl_get_bit, mpl_set_bit, mpl_significant_bits.
2000-07-21 21:04:14 +00:00
3b292e3f60
Fix symbol prefixes in some comments.
2000-07-21 20:50:31 +00:00
36c22a2a15
some compilers don't like the void* arithmetic
2000-07-20 16:58:32 +00:00
350eea5535
Use unsigned ints for length parameters to "octets" functions.
2000-07-20 04:47:24 +00:00
7f70f07983
Eliminate lots of warnings found by -fullwarn.
...
Add sanity test to s_mp_div().
2000-07-20 04:21:37 +00:00
04007f8bb5
Eliminate warnings found with -fullwarn.
2000-07-20 04:20:21 +00:00
dbede1a951
Change the startup/shutdown procedure. Now a minimum amount of seed bytes, MIN_SEED_COUNT, are required before output bytes can be extracted. Also, it is possible to do a full restart by calling RNG_RNGInit() and RNG_RNGShutdown() in succession. This is noted in the blapi.h comments for those functions.
2000-07-19 23:54:43 +00:00
5a1cb37329
Build all mpi .c files with -DMP_API_COMPATIBLE
2000-07-19 23:23:24 +00:00
94e36d6f32
Implement 5 new functions for dealing with variable length integers
...
represented as unsigned and signed octet strings.
mp_read_unsigned_octets(), mp_unsigned_octet_size(),
mp_to_unsigned_octets(), mp_to_signed_octets(), mp_to_fixlen_octets()
See mpi.h for more info.
2000-07-19 23:22:43 +00:00
mcgreer%netscape.com
wtc%netscape.com
nelsonb%netscape.com
larryh%netscape.com
javi%netscape.com
relyea%netscape.com
jgmyers%netscape.com