From d8e7bba357efba5726f6d6797b47293a74549b7f Mon Sep 17 00:00:00 2001 From: rofl0r Date: Sun, 31 Dec 2017 20:27:39 +0000 Subject: [PATCH] use tomsfastmath instead of tommath hack from wpa_supplicant using latest git version from https://github.com/libtom/tomsfastmath --- Makefile | 12 +- src/crypto/bignum.c | 207 - src/crypto/crypto_internal-modexp.c | 49 +- src/crypto/libtommath.c | 3397 ----------------- src/crypto/tfm/fp_2expt.c | 39 + src/crypto/tfm/fp_add.c | 43 + src/crypto/tfm/fp_cmp.c | 31 + src/crypto/tfm/fp_cmp_d.c | 38 + src/crypto/tfm/fp_cmp_mag.c | 35 + src/crypto/tfm/fp_count_bits.c | 36 + src/crypto/tfm/fp_div.c | 157 + src/crypto/tfm/fp_div_2.c | 53 + src/crypto/tfm/fp_div_2d.c | 79 + src/crypto/tfm/fp_exptmod.c | 276 ++ src/crypto/tfm/fp_invmod.c | 207 + src/crypto/tfm/fp_lshd.c | 38 + src/crypto/tfm/fp_mod.c | 34 + src/crypto/tfm/fp_mod_2d.c | 42 + .../tfm/fp_montgomery_calc_normalization.c | 43 + src/crypto/tfm/fp_montgomery_reduce.c | 552 +++ src/crypto/tfm/fp_montgomery_setup.c | 48 + src/crypto/tfm/fp_mul.c | 139 + src/crypto/tfm/fp_mul_2.c | 67 + src/crypto/tfm/fp_mul_2d.c | 47 + src/crypto/tfm/fp_mul_comba.c | 373 ++ src/crypto/tfm/fp_mul_comba_20.c | 180 + src/crypto/tfm/fp_mul_comba_24.c | 212 + src/crypto/tfm/fp_mul_comba_28.c | 244 ++ src/crypto/tfm/fp_mul_comba_32.c | 290 ++ src/crypto/tfm/fp_mul_comba_small_set.c | 1233 ++++++ src/crypto/tfm/fp_mul_d.c | 40 + src/crypto/tfm/fp_mulmod.c | 22 + src/crypto/tfm/fp_read_unsigned_bin.c | 70 + src/crypto/tfm/fp_reverse.c | 31 + src/crypto/tfm/fp_rshd.c | 40 + src/crypto/tfm/fp_set.c | 21 + src/crypto/tfm/fp_sqr.c | 126 + src/crypto/tfm/fp_sqr_comba.c | 677 ++++ src/crypto/tfm/fp_sqr_comba_20.c | 224 ++ src/crypto/tfm/fp_sqr_comba_24.c | 264 ++ src/crypto/tfm/fp_sqr_comba_28.c | 304 ++ src/crypto/tfm/fp_sqr_comba_32.c | 344 ++ src/crypto/tfm/fp_sqr_comba_generic.c | 98 + src/crypto/tfm/fp_sqr_comba_small_set.c | 1523 ++++++++ src/crypto/tfm/fp_sub.c | 50 + src/crypto/tfm/fp_to_unsigned_bin.c | 29 + src/crypto/tfm/fp_unsigned_bin_size.c | 20 + src/crypto/tfm/s_fp_add.c | 42 + src/crypto/tfm/s_fp_sub.c | 40 + src/crypto/tfm/tfm.h | 498 +++ src/crypto/tfm/tfm_private.h | 125 + 51 files changed, 9143 insertions(+), 3646 deletions(-) delete mode 100644 src/crypto/bignum.c delete mode 100644 src/crypto/libtommath.c create mode 100644 src/crypto/tfm/fp_2expt.c create mode 100644 src/crypto/tfm/fp_add.c create mode 100644 src/crypto/tfm/fp_cmp.c create mode 100644 src/crypto/tfm/fp_cmp_d.c create mode 100644 src/crypto/tfm/fp_cmp_mag.c create mode 100644 src/crypto/tfm/fp_count_bits.c create mode 100644 src/crypto/tfm/fp_div.c create mode 100644 src/crypto/tfm/fp_div_2.c create mode 100644 src/crypto/tfm/fp_div_2d.c create mode 100644 src/crypto/tfm/fp_exptmod.c create mode 100644 src/crypto/tfm/fp_invmod.c create mode 100644 src/crypto/tfm/fp_lshd.c create mode 100644 src/crypto/tfm/fp_mod.c create mode 100644 src/crypto/tfm/fp_mod_2d.c create mode 100644 src/crypto/tfm/fp_montgomery_calc_normalization.c create mode 100644 src/crypto/tfm/fp_montgomery_reduce.c create mode 100644 src/crypto/tfm/fp_montgomery_setup.c create mode 100644 src/crypto/tfm/fp_mul.c create mode 100644 src/crypto/tfm/fp_mul_2.c create mode 100644 src/crypto/tfm/fp_mul_2d.c create mode 100644 src/crypto/tfm/fp_mul_comba.c create mode 100644 src/crypto/tfm/fp_mul_comba_20.c create mode 100644 src/crypto/tfm/fp_mul_comba_24.c create mode 100644 src/crypto/tfm/fp_mul_comba_28.c create mode 100644 src/crypto/tfm/fp_mul_comba_32.c create mode 100644 src/crypto/tfm/fp_mul_comba_small_set.c create mode 100644 src/crypto/tfm/fp_mul_d.c create mode 100644 src/crypto/tfm/fp_mulmod.c create mode 100644 src/crypto/tfm/fp_read_unsigned_bin.c create mode 100644 src/crypto/tfm/fp_reverse.c create mode 100644 src/crypto/tfm/fp_rshd.c create mode 100644 src/crypto/tfm/fp_set.c create mode 100644 src/crypto/tfm/fp_sqr.c create mode 100644 src/crypto/tfm/fp_sqr_comba.c create mode 100644 src/crypto/tfm/fp_sqr_comba_20.c create mode 100644 src/crypto/tfm/fp_sqr_comba_24.c create mode 100644 src/crypto/tfm/fp_sqr_comba_28.c create mode 100644 src/crypto/tfm/fp_sqr_comba_32.c create mode 100644 src/crypto/tfm/fp_sqr_comba_generic.c create mode 100644 src/crypto/tfm/fp_sqr_comba_small_set.c create mode 100644 src/crypto/tfm/fp_sub.c create mode 100644 src/crypto/tfm/fp_to_unsigned_bin.c create mode 100644 src/crypto/tfm/fp_unsigned_bin_size.c create mode 100644 src/crypto/tfm/s_fp_add.c create mode 100644 src/crypto/tfm/s_fp_sub.c create mode 100644 src/crypto/tfm/tfm.h create mode 100644 src/crypto/tfm/tfm_private.h diff --git a/Makefile b/Makefile index d099c51..73d4894 100644 --- a/Makefile +++ b/Makefile @@ -18,6 +18,9 @@ INTFLAGS += -DUSE_OPENSSL endif TARGET = pixiewps +TFMSRC = $(sort $(wildcard $(SRCDIR)/crypto/tfm/*.c)) +TFMOBJS = $(TFMSRC:.c=.o) + SOURCE = $(SRCDIR)/pixiewps.c -include config.mak @@ -26,8 +29,11 @@ SOURCE = $(SRCDIR)/pixiewps.c all: $(TARGET) -$(TARGET): $(SOURCE) $(HDRS) - $(CC) $(INTFLAGS) $(CFLAGS) $(CPPFLAGS) -o $(TARGET) $(SOURCE) $(LIBS) $(LDFLAGS) +$(TARGET): $(SOURCE) $(HDRS) $(TFMOBJS) + $(CC) $(INTFLAGS) $(CFLAGS) $(CPPFLAGS) -o $(TARGET) $(SOURCE) $(LIBS) $(LDFLAGS) $(TFMOBJS) + +$(SRCDIR)/crypto/tfm/%.o: $(SRCDIR)/crypto/tfm/%.c + $(CC) $(CFLAGS) $(CPPFLAGS) -I$(SRCDIR)/crypto/tfm -c -o $@ $< install: install-bin install-man @@ -43,4 +49,4 @@ strip: $(TARGET) strip $(TARGET) clean: - rm -f $(TARGET) + rm -f $(TARGET) $(TFMOBJS) diff --git a/src/crypto/bignum.c b/src/crypto/bignum.c deleted file mode 100644 index 1cca885..0000000 --- a/src/crypto/bignum.c +++ /dev/null @@ -1,207 +0,0 @@ -/* - * Big number math - * Copyright (c) 2006, Jouni Malinen - * - * This software may be distributed under the terms of the BSD license. - * See README for more details. - */ - -#include "libtommath.c" - -/* - * The current version is just a wrapper for LibTomMath library, so - * struct bignum is just typecast to mp_int. - */ - -/** - * bignum_init - Allocate memory for bignum - * Returns: Pointer to allocated bignum or %NULL on failure - */ -struct bignum *bignum_init(void) -{ - struct bignum *n = calloc(sizeof(mp_int), 1); - if (n == NULL) - return NULL; - if (mp_init((mp_int *) n) != MP_OKAY) { - free(n); - n = NULL; - } - return n; -} - - -/** - * bignum_deinit - Free bignum - * @n: Bignum from bignum_init() - */ -void bignum_deinit(struct bignum *n) -{ - if (n) { - mp_clear((mp_int *) n); - free(n); - } -} - - -/** - * bignum_get_unsigned_bin - Get length of bignum as an unsigned binary buffer - * @n: Bignum from bignum_init() - * Returns: Length of n if written to a binary buffer - */ -size_t bignum_get_unsigned_bin_len(struct bignum *n) -{ - return mp_unsigned_bin_size((mp_int *) n); -} - - -/** - * bignum_get_unsigned_bin - Set binary buffer to unsigned bignum - * @n: Bignum from bignum_init() - * @buf: Buffer for the binary number - * @len: Length of the buffer, can be %NULL if buffer is known to be long - * enough. Set to used buffer length on success if not %NULL. - * Returns: 0 on success, -1 on failure - */ -int bignum_get_unsigned_bin(const struct bignum *n, u8 *buf, size_t *len) -{ - size_t need = mp_unsigned_bin_size((mp_int *) n); - if (len && need > *len) { - *len = need; - return -1; - } - if (mp_to_unsigned_bin((mp_int *) n, buf) != MP_OKAY) { - return -1; - } - if (len) - *len = need; - return 0; -} - - -/** - * bignum_set_unsigned_bin - Set bignum based on unsigned binary buffer - * @n: Bignum from bignum_init(); to be set to the given value - * @buf: Buffer with unsigned binary value - * @len: Length of buf in octets - * Returns: 0 on success, -1 on failure - */ -int bignum_set_unsigned_bin(struct bignum *n, const u8 *buf, size_t len) -{ - if (mp_read_unsigned_bin((mp_int *) n, (u8 *) buf, len) != MP_OKAY) { - return -1; - } - return 0; -} - - -/** - * bignum_cmp - Signed comparison - * @a: Bignum from bignum_init() - * @b: Bignum from bignum_init() - * Returns: 0 on success, -1 on failure - */ -int bignum_cmp(const struct bignum *a, const struct bignum *b) -{ - return mp_cmp((mp_int *) a, (mp_int *) b); -} - - -/** - * bignum_cmd_d - Compare bignum to standard integer - * @a: Bignum from bignum_init() - * @b: Small integer - * Returns: 0 on success, -1 on failure - */ -int bignum_cmp_d(const struct bignum *a, unsigned long b) -{ - return mp_cmp_d((mp_int *) a, b); -} - - -/** - * bignum_add - c = a + b - * @a: Bignum from bignum_init() - * @b: Bignum from bignum_init() - * @c: Bignum from bignum_init(); used to store the result of a + b - * Returns: 0 on success, -1 on failure - */ -int bignum_add(const struct bignum *a, const struct bignum *b, - struct bignum *c) -{ - if (mp_add((mp_int *) a, (mp_int *) b, (mp_int *) c) != MP_OKAY) { - return -1; - } - return 0; -} - - -/** - * bignum_sub - c = a - b - * @a: Bignum from bignum_init() - * @b: Bignum from bignum_init() - * @c: Bignum from bignum_init(); used to store the result of a - b - * Returns: 0 on success, -1 on failure - */ -int bignum_sub(const struct bignum *a, const struct bignum *b, - struct bignum *c) -{ - if (mp_sub((mp_int *) a, (mp_int *) b, (mp_int *) c) != MP_OKAY) { - return -1; - } - return 0; -} - - -/** - * bignum_mul - c = a * b - * @a: Bignum from bignum_init() - * @b: Bignum from bignum_init() - * @c: Bignum from bignum_init(); used to store the result of a * b - * Returns: 0 on success, -1 on failure - */ -int bignum_mul(const struct bignum *a, const struct bignum *b, - struct bignum *c) -{ - if (mp_mul((mp_int *) a, (mp_int *) b, (mp_int *) c) != MP_OKAY) { - return -1; - } - return 0; -} - - -/** - * bignum_mulmod - d = a * b (mod c) - * @a: Bignum from bignum_init() - * @b: Bignum from bignum_init() - * @c: Bignum from bignum_init(); modulus - * @d: Bignum from bignum_init(); used to store the result of a * b (mod c) - * Returns: 0 on success, -1 on failure - */ -int bignum_mulmod(const struct bignum *a, const struct bignum *b, - const struct bignum *c, struct bignum *d) -{ - if (mp_mulmod((mp_int *) a, (mp_int *) b, (mp_int *) c, (mp_int *) d) - != MP_OKAY) { - return -1; - } - return 0; -} - - -/** - * bignum_exptmod - Modular exponentiation: d = a^b (mod c) - * @a: Bignum from bignum_init(); base - * @b: Bignum from bignum_init(); exponent - * @c: Bignum from bignum_init(); modulus - * @d: Bignum from bignum_init(); used to store the result of a^b (mod c) - * Returns: 0 on success, -1 on failure - */ -int bignum_exptmod(const struct bignum *a, const struct bignum *b, - const struct bignum *c, struct bignum *d) -{ - if (mp_exptmod((mp_int *) a, (mp_int *) b, (mp_int *) c, (mp_int *) d) - != MP_OKAY) { - return -1; - } - return 0; -} diff --git a/src/crypto/crypto_internal-modexp.c b/src/crypto/crypto_internal-modexp.c index 8be1ced..1adf2fc 100644 --- a/src/crypto/crypto_internal-modexp.c +++ b/src/crypto/crypto_internal-modexp.c @@ -1,21 +1,7 @@ -/* - * Crypto wrapper for internal crypto implementation - modexp - * Copyright (c) 2006-2009, Jouni Malinen - * - * This program is free software; you can redistribute it and/or modify - * it under the terms of the GNU General Public License version 2 as - * published by the Free Software Foundation. - * - * Alternatively, this software may be distributed under the terms of BSD - * license. - * - * See README and COPYING for more details. - */ - #include #include -#include "bignum.c" +#include "tfm/tfm.h" #define u8 uint8_t @@ -24,32 +10,17 @@ int crypto_mod_exp(const u8 *base, size_t base_len, const u8 *modulus, size_t modulus_len, u8 *result, size_t *result_len) { - struct bignum *bn_base, *bn_exp, *bn_modulus, *bn_result; - int ret = -1; + fp_int bn_base, bn_exp, bn_modulus, bn_result; - bn_base = bignum_init(); - bn_exp = bignum_init(); - bn_modulus = bignum_init(); - bn_result = bignum_init(); + fp_read_unsigned_bin(&bn_base, base, base_len); + fp_read_unsigned_bin(&bn_exp, power, power_len); + fp_read_unsigned_bin(&bn_modulus, modulus, modulus_len); + fp_init(&bn_result); - if (bn_base == NULL || bn_exp == NULL || bn_modulus == NULL || - bn_result == NULL) - goto error; + fp_exptmod(&bn_base, &bn_exp, &bn_modulus, &bn_result); - if (bignum_set_unsigned_bin(bn_base, base, base_len) < 0 || - bignum_set_unsigned_bin(bn_exp, power, power_len) < 0 || - bignum_set_unsigned_bin(bn_modulus, modulus, modulus_len) < 0) - goto error; + fp_to_unsigned_bin(&bn_result, result); - if (bignum_exptmod(bn_base, bn_exp, bn_modulus, bn_result) < 0) - goto error; - - ret = bignum_get_unsigned_bin(bn_result, result, result_len); - -error: - bignum_deinit(bn_base); - bignum_deinit(bn_exp); - bignum_deinit(bn_modulus); - bignum_deinit(bn_result); - return ret; + *result_len = fp_unsigned_bin_size(&bn_result); + return 0; } diff --git a/src/crypto/libtommath.c b/src/crypto/libtommath.c deleted file mode 100644 index 05338e9..0000000 --- a/src/crypto/libtommath.c +++ /dev/null @@ -1,3397 +0,0 @@ -/* - * Minimal code for RSA support from LibTomMath 0.41 - * http://libtom.org/ - * http://libtom.org/files/ltm-0.41.tar.bz2 - * This library was released in public domain by Tom St Denis. - * - * The combination in this file may not use all of the optimized algorithms - * from LibTomMath and may be considerable slower than the LibTomMath with its - * default settings. The main purpose of having this version here is to make it - * easier to build bignum.c wrapper without having to install and build an - * external library. - */ - -#ifndef CHAR_BIT -#define CHAR_BIT 8 -#endif - -#define BN_MP_INVMOD_C -#define BN_S_MP_EXPTMOD_C /* Note: #undef in tommath_superclass.h; this would - * require BN_MP_EXPTMOD_FAST_C instead */ -#define BN_S_MP_MUL_DIGS_C -#define BN_MP_INVMOD_SLOW_C -#define BN_S_MP_SQR_C -#define BN_S_MP_MUL_HIGH_DIGS_C /* Note: #undef in tommath_superclass.h; this - * would require other than mp_reduce */ - -#ifdef LTM_FAST - -/* Use faster div at the cost of about 1 kB */ -#define BN_MP_MUL_D_C - -/* Include faster exptmod (Montgomery) at the cost of about 2.5 kB in code */ -#define BN_MP_EXPTMOD_FAST_C -#define BN_MP_MONTGOMERY_SETUP_C -#define BN_FAST_MP_MONTGOMERY_REDUCE_C -#define BN_MP_MONTGOMERY_CALC_NORMALIZATION_C -#define BN_MP_MUL_2_C - -/* Include faster sqr at the cost of about 0.5 kB in code */ -#define BN_FAST_S_MP_SQR_C - -/* About 0.25 kB of code, but ~1.7kB of stack space! */ -#define BN_FAST_S_MP_MUL_DIGS_C - -#else /* LTM_FAST */ - -#define BN_MP_DIV_SMALL -#define BN_MP_INIT_MULTI_C -#define BN_MP_CLEAR_MULTI_C -#define BN_MP_ABS_C -#endif /* LTM_FAST */ - -/* Current uses do not require support for negative exponent in exptmod, so we - * can save about 1.5 kB in leaving out invmod. */ -#define LTM_NO_NEG_EXP - -/* from tommath.h */ - -#ifndef MIN - #define MIN(x,y) ((x)<(y)?(x):(y)) -#endif - -#ifndef MAX - #define MAX(x,y) ((x)>(y)?(x):(y)) -#endif - -#define OPT_CAST(x) - -#ifdef __x86_64__ -typedef unsigned long mp_digit; -typedef unsigned long mp_word __attribute__((mode(TI))); - -#define DIGIT_BIT 60 -#define MP_64BIT -#else -typedef unsigned long mp_digit; -typedef u64 mp_word; - -#define DIGIT_BIT 28 -#define MP_28BIT -#endif - - -#define XMALLOC malloc -#define XFREE free -#define XREALLOC realloc - - -#define MP_MASK ((((mp_digit)1)<<((mp_digit)DIGIT_BIT))-((mp_digit)1)) - -#define MP_LT -1 /* less than */ -#define MP_EQ 0 /* equal to */ -#define MP_GT 1 /* greater than */ - -#define MP_ZPOS 0 /* positive integer */ -#define MP_NEG 1 /* negative */ - -#define MP_OKAY 0 /* ok result */ -#define MP_MEM -2 /* out of mem */ -#define MP_VAL -3 /* invalid input */ - -#define MP_YES 1 /* yes response */ -#define MP_NO 0 /* no response */ - -typedef int mp_err; - -/* define this to use lower memory usage routines (exptmods mostly) */ -#define MP_LOW_MEM - -/* default precision */ -#ifndef MP_PREC - #ifndef MP_LOW_MEM - #define MP_PREC 32 /* default digits of precision */ - #else - #define MP_PREC 8 /* default digits of precision */ - #endif -#endif - -/* size of comba arrays, should be at least 2 * 2**(BITS_PER_WORD - BITS_PER_DIGIT*2) */ -#define MP_WARRAY (1 << (sizeof(mp_word) * CHAR_BIT - 2 * DIGIT_BIT + 1)) - -/* the infamous mp_int structure */ -typedef struct { - int used, alloc, sign; - mp_digit *dp; -} mp_int; - - -/* ---> Basic Manipulations <--- */ -#define mp_iszero(a) (((a)->used == 0) ? MP_YES : MP_NO) -#define mp_iseven(a) (((a)->used > 0 && (((a)->dp[0] & 1) == 0)) ? MP_YES : MP_NO) -#define mp_isodd(a) (((a)->used > 0 && (((a)->dp[0] & 1) == 1)) ? MP_YES : MP_NO) - - -/* prototypes for copied functions */ -#define s_mp_mul(a, b, c) s_mp_mul_digs(a, b, c, (a)->used + (b)->used + 1) -static int s_mp_exptmod(mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode); -static int s_mp_mul_digs (mp_int * a, mp_int * b, mp_int * c, int digs); -static int s_mp_sqr(mp_int * a, mp_int * b); -static int s_mp_mul_high_digs(mp_int * a, mp_int * b, mp_int * c, int digs); - -#ifdef BN_FAST_S_MP_MUL_DIGS_C -static int fast_s_mp_mul_digs (mp_int * a, mp_int * b, mp_int * c, int digs); -#endif - -#ifdef BN_MP_INIT_MULTI_C -static int mp_init_multi(mp_int *mp, ...); -#endif -#ifdef BN_MP_CLEAR_MULTI_C -static void mp_clear_multi(mp_int *mp, ...); -#endif -static int mp_lshd(mp_int * a, int b); -static void mp_set(mp_int * a, mp_digit b); -static void mp_clamp(mp_int * a); -static void mp_exch(mp_int * a, mp_int * b); -static void mp_rshd(mp_int * a, int b); -static void mp_zero(mp_int * a); -static int mp_mod_2d(mp_int * a, int b, mp_int * c); -static int mp_div_2d(mp_int * a, int b, mp_int * c, mp_int * d); -static int mp_init_copy(mp_int * a, mp_int * b); -static int mp_mul_2d(mp_int * a, int b, mp_int * c); -#ifndef LTM_NO_NEG_EXP -static int mp_div_2(mp_int * a, mp_int * b); -static int mp_invmod(mp_int * a, mp_int * b, mp_int * c); -static int mp_invmod_slow(mp_int * a, mp_int * b, mp_int * c); -#endif /* LTM_NO_NEG_EXP */ -static int mp_copy(mp_int * a, mp_int * b); -static int mp_count_bits(mp_int * a); -static int mp_div(mp_int * a, mp_int * b, mp_int * c, mp_int * d); -static int mp_mod(mp_int * a, mp_int * b, mp_int * c); -static int mp_grow(mp_int * a, int size); -static int mp_cmp_mag(mp_int * a, mp_int * b); -#ifdef BN_MP_ABS_C -static int mp_abs(mp_int * a, mp_int * b); -#endif -static int mp_sqr(mp_int * a, mp_int * b); -static int mp_reduce_2k_l(mp_int *a, mp_int *n, mp_int *d); -static int mp_reduce_2k_setup_l(mp_int *a, mp_int *d); -static int mp_2expt(mp_int * a, int b); -static int mp_reduce_setup(mp_int * a, mp_int * b); -static int mp_reduce(mp_int * x, mp_int * m, mp_int * mu); -static int mp_init_size(mp_int * a, int size); -#ifdef BN_MP_EXPTMOD_FAST_C -static int mp_exptmod_fast (mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode); -#endif /* BN_MP_EXPTMOD_FAST_C */ -#ifdef BN_FAST_S_MP_SQR_C -static int fast_s_mp_sqr (mp_int * a, mp_int * b); -#endif /* BN_FAST_S_MP_SQR_C */ -#ifdef BN_MP_MUL_D_C -static int mp_mul_d (mp_int * a, mp_digit b, mp_int * c); -#endif /* BN_MP_MUL_D_C */ - - - -/* functions from bn_.c */ - - -/* reverse an array, used for radix code */ -static void bn_reverse (unsigned char *s, int len) -{ - int ix, iy; - unsigned char t; - - ix = 0; - iy = len - 1; - while (ix < iy) { - t = s[ix]; - s[ix] = s[iy]; - s[iy] = t; - ++ix; - --iy; - } -} - - -/* low level addition, based on HAC pp.594, Algorithm 14.7 */ -static int s_mp_add (mp_int * a, mp_int * b, mp_int * c) -{ - mp_int *x; - int olduse, res, min, max; - - /* find sizes, we let |a| <= |b| which means we have to sort - * them. "x" will point to the input with the most digits - */ - if (a->used > b->used) { - min = b->used; - max = a->used; - x = a; - } else { - min = a->used; - max = b->used; - x = b; - } - - /* init result */ - if (c->alloc < max + 1) { - if ((res = mp_grow (c, max + 1)) != MP_OKAY) { - return res; - } - } - - /* get old used digit count and set new one */ - olduse = c->used; - c->used = max + 1; - - { - register mp_digit u, *tmpa, *tmpb, *tmpc; - register int i; - - /* alias for digit pointers */ - - /* first input */ - tmpa = a->dp; - - /* second input */ - tmpb = b->dp; - - /* destination */ - tmpc = c->dp; - - /* zero the carry */ - u = 0; - for (i = 0; i < min; i++) { - /* Compute the sum at one digit, T[i] = A[i] + B[i] + U */ - *tmpc = *tmpa++ + *tmpb++ + u; - - /* U = carry bit of T[i] */ - u = *tmpc >> ((mp_digit)DIGIT_BIT); - - /* take away carry bit from T[i] */ - *tmpc++ &= MP_MASK; - } - - /* now copy higher words if any, that is in A+B - * if A or B has more digits add those in - */ - if (min != max) { - for (; i < max; i++) { - /* T[i] = X[i] + U */ - *tmpc = x->dp[i] + u; - - /* U = carry bit of T[i] */ - u = *tmpc >> ((mp_digit)DIGIT_BIT); - - /* take away carry bit from T[i] */ - *tmpc++ &= MP_MASK; - } - } - - /* add carry */ - *tmpc++ = u; - - /* clear digits above oldused */ - for (i = c->used; i < olduse; i++) { - *tmpc++ = 0; - } - } - - mp_clamp (c); - return MP_OKAY; -} - - -/* low level subtraction (assumes |a| > |b|), HAC pp.595 Algorithm 14.9 */ -static int s_mp_sub (mp_int * a, mp_int * b, mp_int * c) -{ - int olduse, res, min, max; - - /* find sizes */ - min = b->used; - max = a->used; - - /* init result */ - if (c->alloc < max) { - if ((res = mp_grow (c, max)) != MP_OKAY) { - return res; - } - } - olduse = c->used; - c->used = max; - - { - register mp_digit u, *tmpa, *tmpb, *tmpc; - register int i; - - /* alias for digit pointers */ - tmpa = a->dp; - tmpb = b->dp; - tmpc = c->dp; - - /* set carry to zero */ - u = 0; - for (i = 0; i < min; i++) { - /* T[i] = A[i] - B[i] - U */ - *tmpc = *tmpa++ - *tmpb++ - u; - - /* U = carry bit of T[i] - * Note this saves performing an AND operation since - * if a carry does occur it will propagate all the way to the - * MSB. As a result a single shift is enough to get the carry - */ - u = *tmpc >> ((mp_digit)(CHAR_BIT * sizeof (mp_digit) - 1)); - - /* Clear carry from T[i] */ - *tmpc++ &= MP_MASK; - } - - /* now copy higher words if any, e.g. if A has more digits than B */ - for (; i < max; i++) { - /* T[i] = A[i] - U */ - *tmpc = *tmpa++ - u; - - /* U = carry bit of T[i] */ - u = *tmpc >> ((mp_digit)(CHAR_BIT * sizeof (mp_digit) - 1)); - - /* Clear carry from T[i] */ - *tmpc++ &= MP_MASK; - } - - /* clear digits above used (since we may not have grown result above) */ - for (i = c->used; i < olduse; i++) { - *tmpc++ = 0; - } - } - - mp_clamp (c); - return MP_OKAY; -} - - -/* init a new mp_int */ -static int mp_init (mp_int * a) -{ - int i; - - /* allocate memory required and clear it */ - a->dp = OPT_CAST(mp_digit) XMALLOC (sizeof (mp_digit) * MP_PREC); - if (a->dp == NULL) { - return MP_MEM; - } - - /* set the digits to zero */ - for (i = 0; i < MP_PREC; i++) { - a->dp[i] = 0; - } - - /* set the used to zero, allocated digits to the default precision - * and sign to positive */ - a->used = 0; - a->alloc = MP_PREC; - a->sign = MP_ZPOS; - - return MP_OKAY; -} - - -/* clear one (frees) */ -static void mp_clear (mp_int * a) -{ - int i; - - /* only do anything if a hasn't been freed previously */ - if (a->dp != NULL) { - /* first zero the digits */ - for (i = 0; i < a->used; i++) { - a->dp[i] = 0; - } - - /* free ram */ - XFREE(a->dp); - - /* reset members to make debugging easier */ - a->dp = NULL; - a->alloc = a->used = 0; - a->sign = MP_ZPOS; - } -} - - -/* high level addition (handles signs) */ -static int mp_add (mp_int * a, mp_int * b, mp_int * c) -{ - int sa, sb, res; - - /* get sign of both inputs */ - sa = a->sign; - sb = b->sign; - - /* handle two cases, not four */ - if (sa == sb) { - /* both positive or both negative */ - /* add their magnitudes, copy the sign */ - c->sign = sa; - res = s_mp_add (a, b, c); - } else { - /* one positive, the other negative */ - /* subtract the one with the greater magnitude from */ - /* the one of the lesser magnitude. The result gets */ - /* the sign of the one with the greater magnitude. */ - if (mp_cmp_mag (a, b) == MP_LT) { - c->sign = sb; - res = s_mp_sub (b, a, c); - } else { - c->sign = sa; - res = s_mp_sub (a, b, c); - } - } - return res; -} - - -/* high level subtraction (handles signs) */ -static int mp_sub (mp_int * a, mp_int * b, mp_int * c) -{ - int sa, sb, res; - - sa = a->sign; - sb = b->sign; - - if (sa != sb) { - /* subtract a negative from a positive, OR */ - /* subtract a positive from a negative. */ - /* In either case, ADD their magnitudes, */ - /* and use the sign of the first number. */ - c->sign = sa; - res = s_mp_add (a, b, c); - } else { - /* subtract a positive from a positive, OR */ - /* subtract a negative from a negative. */ - /* First, take the difference between their */ - /* magnitudes, then... */ - if (mp_cmp_mag (a, b) != MP_LT) { - /* Copy the sign from the first */ - c->sign = sa; - /* The first has a larger or equal magnitude */ - res = s_mp_sub (a, b, c); - } else { - /* The result has the *opposite* sign from */ - /* the first number. */ - c->sign = (sa == MP_ZPOS) ? MP_NEG : MP_ZPOS; - /* The second has a larger magnitude */ - res = s_mp_sub (b, a, c); - } - } - return res; -} - - -/* high level multiplication (handles sign) */ -static int mp_mul (mp_int * a, mp_int * b, mp_int * c) -{ - int res, neg; - neg = (a->sign == b->sign) ? MP_ZPOS : MP_NEG; - - /* use Toom-Cook? */ -#ifdef BN_MP_TOOM_MUL_C - if (MIN (a->used, b->used) >= TOOM_MUL_CUTOFF) { - res = mp_toom_mul(a, b, c); - } else -#endif -#ifdef BN_MP_KARATSUBA_MUL_C - /* use Karatsuba? */ - if (MIN (a->used, b->used) >= KARATSUBA_MUL_CUTOFF) { - res = mp_karatsuba_mul (a, b, c); - } else -#endif - { - /* can we use the fast multiplier? - * - * The fast multiplier can be used if the output will - * have less than MP_WARRAY digits and the number of - * digits won't affect carry propagation - */ -#ifdef BN_FAST_S_MP_MUL_DIGS_C - int digs = a->used + b->used + 1; - - if ((digs < MP_WARRAY) && - MIN(a->used, b->used) <= - (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) { - res = fast_s_mp_mul_digs (a, b, c, digs); - } else -#endif -#ifdef BN_S_MP_MUL_DIGS_C - res = s_mp_mul (a, b, c); /* uses s_mp_mul_digs */ -#else -#error mp_mul could fail - res = MP_VAL; -#endif - - } - c->sign = (c->used > 0) ? neg : MP_ZPOS; - return res; -} - - -/* d = a * b (mod c) */ -static int mp_mulmod (mp_int * a, mp_int * b, mp_int * c, mp_int * d) -{ - int res; - mp_int t; - - if ((res = mp_init (&t)) != MP_OKAY) { - return res; - } - - if ((res = mp_mul (a, b, &t)) != MP_OKAY) { - mp_clear (&t); - return res; - } - res = mp_mod (&t, c, d); - mp_clear (&t); - return res; -} - - -/* c = a mod b, 0 <= c < b */ -static int mp_mod (mp_int * a, mp_int * b, mp_int * c) -{ - mp_int t; - int res; - - if ((res = mp_init (&t)) != MP_OKAY) { - return res; - } - - if ((res = mp_div (a, b, NULL, &t)) != MP_OKAY) { - mp_clear (&t); - return res; - } - - if (t.sign != b->sign) { - res = mp_add (b, &t, c); - } else { - res = MP_OKAY; - mp_exch (&t, c); - } - - mp_clear (&t); - return res; -} - - -/* this is a shell function that calls either the normal or Montgomery - * exptmod functions. Originally the call to the montgomery code was - * embedded in the normal function but that wasted a lot of stack space - * for nothing (since 99% of the time the Montgomery code would be called) - */ -static int mp_exptmod (mp_int * G, mp_int * X, mp_int * P, mp_int * Y) -{ - int dr; - - /* modulus P must be positive */ - if (P->sign == MP_NEG) { - return MP_VAL; - } - - /* if exponent X is negative we have to recurse */ - if (X->sign == MP_NEG) { -#ifdef LTM_NO_NEG_EXP - return MP_VAL; -#else /* LTM_NO_NEG_EXP */ -#ifdef BN_MP_INVMOD_C - mp_int tmpG, tmpX; - int err; - - /* first compute 1/G mod P */ - if ((err = mp_init(&tmpG)) != MP_OKAY) { - return err; - } - if ((err = mp_invmod(G, P, &tmpG)) != MP_OKAY) { - mp_clear(&tmpG); - return err; - } - - /* now get |X| */ - if ((err = mp_init(&tmpX)) != MP_OKAY) { - mp_clear(&tmpG); - return err; - } - if ((err = mp_abs(X, &tmpX)) != MP_OKAY) { - mp_clear_multi(&tmpG, &tmpX, NULL); - return err; - } - - /* and now compute (1/G)**|X| instead of G**X [X < 0] */ - err = mp_exptmod(&tmpG, &tmpX, P, Y); - mp_clear_multi(&tmpG, &tmpX, NULL); - return err; -#else -#error mp_exptmod would always fail - /* no invmod */ - return MP_VAL; -#endif -#endif /* LTM_NO_NEG_EXP */ - } - -/* modified diminished radix reduction */ -#if defined(BN_MP_REDUCE_IS_2K_L_C) && defined(BN_MP_REDUCE_2K_L_C) && defined(BN_S_MP_EXPTMOD_C) - if (mp_reduce_is_2k_l(P) == MP_YES) { - return s_mp_exptmod(G, X, P, Y, 1); - } -#endif - -#ifdef BN_MP_DR_IS_MODULUS_C - /* is it a DR modulus? */ - dr = mp_dr_is_modulus(P); -#else - /* default to no */ - dr = 0; -#endif - -#ifdef BN_MP_REDUCE_IS_2K_C - /* if not, is it a unrestricted DR modulus? */ - if (dr == 0) { - dr = mp_reduce_is_2k(P) << 1; - } -#endif - - /* if the modulus is odd or dr != 0 use the montgomery method */ -#ifdef BN_MP_EXPTMOD_FAST_C - if (mp_isodd (P) == 1 || dr != 0) { - return mp_exptmod_fast (G, X, P, Y, dr); - } else { -#endif -#ifdef BN_S_MP_EXPTMOD_C - /* otherwise use the generic Barrett reduction technique */ - return s_mp_exptmod (G, X, P, Y, 0); -#else -#error mp_exptmod could fail - /* no exptmod for evens */ - return MP_VAL; -#endif -#ifdef BN_MP_EXPTMOD_FAST_C - } -#endif - if (dr == 0) { - /* avoid compiler warnings about possibly unused variable */ - } -} - - -/* compare two ints (signed)*/ -static int mp_cmp (mp_int * a, mp_int * b) -{ - /* compare based on sign */ - if (a->sign != b->sign) { - if (a->sign == MP_NEG) { - return MP_LT; - } else { - return MP_GT; - } - } - - /* compare digits */ - if (a->sign == MP_NEG) { - /* if negative compare opposite direction */ - return mp_cmp_mag(b, a); - } else { - return mp_cmp_mag(a, b); - } -} - - -/* compare a digit */ -static int mp_cmp_d(mp_int * a, mp_digit b) -{ - /* compare based on sign */ - if (a->sign == MP_NEG) { - return MP_LT; - } - - /* compare based on magnitude */ - if (a->used > 1) { - return MP_GT; - } - - /* compare the only digit of a to b */ - if (a->dp[0] > b) { - return MP_GT; - } else if (a->dp[0] < b) { - return MP_LT; - } else { - return MP_EQ; - } -} - - -#ifndef LTM_NO_NEG_EXP -/* hac 14.61, pp608 */ -static int mp_invmod (mp_int * a, mp_int * b, mp_int * c) -{ - /* b cannot be negative */ - if (b->sign == MP_NEG || mp_iszero(b) == 1) { - return MP_VAL; - } - -#ifdef BN_FAST_MP_INVMOD_C - /* if the modulus is odd we can use a faster routine instead */ - if (mp_isodd (b) == 1) { - return fast_mp_invmod (a, b, c); - } -#endif - -#ifdef BN_MP_INVMOD_SLOW_C - return mp_invmod_slow(a, b, c); -#endif - -#ifndef BN_FAST_MP_INVMOD_C -#ifndef BN_MP_INVMOD_SLOW_C -#error mp_invmod would always fail -#endif -#endif - return MP_VAL; -} -#endif /* LTM_NO_NEG_EXP */ - - -/* get the size for an unsigned equivalent */ -static int mp_unsigned_bin_size (mp_int * a) -{ - int size = mp_count_bits (a); - return (size / 8 + ((size & 7) != 0 ? 1 : 0)); -} - - -#ifndef LTM_NO_NEG_EXP -/* hac 14.61, pp608 */ -static int mp_invmod_slow (mp_int * a, mp_int * b, mp_int * c) -{ - mp_int x, y, u, v, A, B, C, D; - int res; - - /* b cannot be negative */ - if (b->sign == MP_NEG || mp_iszero(b) == 1) { - return MP_VAL; - } - - /* init temps */ - if ((res = mp_init_multi(&x, &y, &u, &v, - &A, &B, &C, &D, NULL)) != MP_OKAY) { - return res; - } - - /* x = a, y = b */ - if ((res = mp_mod(a, b, &x)) != MP_OKAY) { - goto LBL_ERR; - } - if ((res = mp_copy (b, &y)) != MP_OKAY) { - goto LBL_ERR; - } - - /* 2. [modified] if x,y are both even then return an error! */ - if (mp_iseven (&x) == 1 && mp_iseven (&y) == 1) { - res = MP_VAL; - goto LBL_ERR; - } - - /* 3. u=x, v=y, A=1, B=0, C=0,D=1 */ - if ((res = mp_copy (&x, &u)) != MP_OKAY) { - goto LBL_ERR; - } - if ((res = mp_copy (&y, &v)) != MP_OKAY) { - goto LBL_ERR; - } - mp_set (&A, 1); - mp_set (&D, 1); - -top: - /* 4. while u is even do */ - while (mp_iseven (&u) == 1) { - /* 4.1 u = u/2 */ - if ((res = mp_div_2 (&u, &u)) != MP_OKAY) { - goto LBL_ERR; - } - /* 4.2 if A or B is odd then */ - if (mp_isodd (&A) == 1 || mp_isodd (&B) == 1) { - /* A = (A+y)/2, B = (B-x)/2 */ - if ((res = mp_add (&A, &y, &A)) != MP_OKAY) { - goto LBL_ERR; - } - if ((res = mp_sub (&B, &x, &B)) != MP_OKAY) { - goto LBL_ERR; - } - } - /* A = A/2, B = B/2 */ - if ((res = mp_div_2 (&A, &A)) != MP_OKAY) { - goto LBL_ERR; - } - if ((res = mp_div_2 (&B, &B)) != MP_OKAY) { - goto LBL_ERR; - } - } - - /* 5. while v is even do */ - while (mp_iseven (&v) == 1) { - /* 5.1 v = v/2 */ - if ((res = mp_div_2 (&v, &v)) != MP_OKAY) { - goto LBL_ERR; - } - /* 5.2 if C or D is odd then */ - if (mp_isodd (&C) == 1 || mp_isodd (&D) == 1) { - /* C = (C+y)/2, D = (D-x)/2 */ - if ((res = mp_add (&C, &y, &C)) != MP_OKAY) { - goto LBL_ERR; - } - if ((res = mp_sub (&D, &x, &D)) != MP_OKAY) { - goto LBL_ERR; - } - } - /* C = C/2, D = D/2 */ - if ((res = mp_div_2 (&C, &C)) != MP_OKAY) { - goto LBL_ERR; - } - if ((res = mp_div_2 (&D, &D)) != MP_OKAY) { - goto LBL_ERR; - } - } - - /* 6. if u >= v then */ - if (mp_cmp (&u, &v) != MP_LT) { - /* u = u - v, A = A - C, B = B - D */ - if ((res = mp_sub (&u, &v, &u)) != MP_OKAY) { - goto LBL_ERR; - } - - if ((res = mp_sub (&A, &C, &A)) != MP_OKAY) { - goto LBL_ERR; - } - - if ((res = mp_sub (&B, &D, &B)) != MP_OKAY) { - goto LBL_ERR; - } - } else { - /* v - v - u, C = C - A, D = D - B */ - if ((res = mp_sub (&v, &u, &v)) != MP_OKAY) { - goto LBL_ERR; - } - - if ((res = mp_sub (&C, &A, &C)) != MP_OKAY) { - goto LBL_ERR; - } - - if ((res = mp_sub (&D, &B, &D)) != MP_OKAY) { - goto LBL_ERR; - } - } - - /* if not zero goto step 4 */ - if (mp_iszero (&u) == 0) - goto top; - - /* now a = C, b = D, gcd == g*v */ - - /* if v != 1 then there is no inverse */ - if (mp_cmp_d (&v, 1) != MP_EQ) { - res = MP_VAL; - goto LBL_ERR; - } - - /* if its too low */ - while (mp_cmp_d(&C, 0) == MP_LT) { - if ((res = mp_add(&C, b, &C)) != MP_OKAY) { - goto LBL_ERR; - } - } - - /* too big */ - while (mp_cmp_mag(&C, b) != MP_LT) { - if ((res = mp_sub(&C, b, &C)) != MP_OKAY) { - goto LBL_ERR; - } - } - - /* C is now the inverse */ - mp_exch (&C, c); - res = MP_OKAY; -LBL_ERR:mp_clear_multi (&x, &y, &u, &v, &A, &B, &C, &D, NULL); - return res; -} -#endif /* LTM_NO_NEG_EXP */ - - -/* compare maginitude of two ints (unsigned) */ -static int mp_cmp_mag (mp_int * a, mp_int * b) -{ - int n; - mp_digit *tmpa, *tmpb; - - /* compare based on # of non-zero digits */ - if (a->used > b->used) { - return MP_GT; - } - - if (a->used < b->used) { - return MP_LT; - } - - /* alias for a */ - tmpa = a->dp + (a->used - 1); - - /* alias for b */ - tmpb = b->dp + (a->used - 1); - - /* compare based on digits */ - for (n = 0; n < a->used; ++n, --tmpa, --tmpb) { - if (*tmpa > *tmpb) { - return MP_GT; - } - - if (*tmpa < *tmpb) { - return MP_LT; - } - } - return MP_EQ; -} - - -/* reads a unsigned char array, assumes the msb is stored first [big endian] */ -static int mp_read_unsigned_bin (mp_int * a, const unsigned char *b, int c) -{ - int res; - - /* make sure there are at least two digits */ - if (a->alloc < 2) { - if ((res = mp_grow(a, 2)) != MP_OKAY) { - return res; - } - } - - /* zero the int */ - mp_zero (a); - - /* read the bytes in */ - while (c-- > 0) { - if ((res = mp_mul_2d (a, 8, a)) != MP_OKAY) { - return res; - } - -#ifndef MP_8BIT - a->dp[0] |= *b++; - a->used += 1; -#else - a->dp[0] = (*b & MP_MASK); - a->dp[1] |= ((*b++ >> 7U) & 1); - a->used += 2; -#endif - } - mp_clamp (a); - return MP_OKAY; -} - - -/* store in unsigned [big endian] format */ -static int mp_to_unsigned_bin (mp_int * a, unsigned char *b) -{ - int x, res; - mp_int t; - - if ((res = mp_init_copy (&t, a)) != MP_OKAY) { - return res; - } - - x = 0; - while (mp_iszero (&t) == 0) { -#ifndef MP_8BIT - b[x++] = (unsigned char) (t.dp[0] & 255); -#else - b[x++] = (unsigned char) (t.dp[0] | ((t.dp[1] & 0x01) << 7)); -#endif - if ((res = mp_div_2d (&t, 8, &t, NULL)) != MP_OKAY) { - mp_clear (&t); - return res; - } - } - bn_reverse (b, x); - mp_clear (&t); - return MP_OKAY; -} - - -/* shift right by a certain bit count (store quotient in c, optional remainder in d) */ -static int mp_div_2d (mp_int * a, int b, mp_int * c, mp_int * d) -{ - mp_digit D, r, rr; - int x, res; - mp_int t; - - - /* if the shift count is <= 0 then we do no work */ - if (b <= 0) { - res = mp_copy (a, c); - if (d != NULL) { - mp_zero (d); - } - return res; - } - - if ((res = mp_init (&t)) != MP_OKAY) { - return res; - } - - /* get the remainder */ - if (d != NULL) { - if ((res = mp_mod_2d (a, b, &t)) != MP_OKAY) { - mp_clear (&t); - return res; - } - } - - /* copy */ - if ((res = mp_copy (a, c)) != MP_OKAY) { - mp_clear (&t); - return res; - } - - /* shift by as many digits in the bit count */ - if (b >= (int)DIGIT_BIT) { - mp_rshd (c, b / DIGIT_BIT); - } - - /* shift any bit count < DIGIT_BIT */ - D = (mp_digit) (b % DIGIT_BIT); - if (D != 0) { - register mp_digit *tmpc, mask, shift; - - /* mask */ - mask = (((mp_digit)1) << D) - 1; - - /* shift for lsb */ - shift = DIGIT_BIT - D; - - /* alias */ - tmpc = c->dp + (c->used - 1); - - /* carry */ - r = 0; - for (x = c->used - 1; x >= 0; x--) { - /* get the lower bits of this word in a temp */ - rr = *tmpc & mask; - - /* shift the current word and mix in the carry bits from the previous word */ - *tmpc = (*tmpc >> D) | (r << shift); - --tmpc; - - /* set the carry to the carry bits of the current word found above */ - r = rr; - } - } - mp_clamp (c); - if (d != NULL) { - mp_exch (&t, d); - } - mp_clear (&t); - return MP_OKAY; -} - - -static int mp_init_copy (mp_int * a, mp_int * b) -{ - int res; - - if ((res = mp_init (a)) != MP_OKAY) { - return res; - } - return mp_copy (b, a); -} - - -/* set to zero */ -static void mp_zero (mp_int * a) -{ - int n; - mp_digit *tmp; - - a->sign = MP_ZPOS; - a->used = 0; - - tmp = a->dp; - for (n = 0; n < a->alloc; n++) { - *tmp++ = 0; - } -} - - -/* copy, b = a */ -static int mp_copy (mp_int * a, mp_int * b) -{ - int res, n; - - /* if dst == src do nothing */ - if (a == b) { - return MP_OKAY; - } - - /* grow dest */ - if (b->alloc < a->used) { - if ((res = mp_grow (b, a->used)) != MP_OKAY) { - return res; - } - } - - /* zero b and copy the parameters over */ - { - register mp_digit *tmpa, *tmpb; - - /* pointer aliases */ - - /* source */ - tmpa = a->dp; - - /* destination */ - tmpb = b->dp; - - /* copy all the digits */ - for (n = 0; n < a->used; n++) { - *tmpb++ = *tmpa++; - } - - /* clear high digits */ - for (; n < b->used; n++) { - *tmpb++ = 0; - } - } - - /* copy used count and sign */ - b->used = a->used; - b->sign = a->sign; - return MP_OKAY; -} - - -/* shift right a certain amount of digits */ -static void mp_rshd (mp_int * a, int b) -{ - int x; - - /* if b <= 0 then ignore it */ - if (b <= 0) { - return; - } - - /* if b > used then simply zero it and return */ - if (a->used <= b) { - mp_zero (a); - return; - } - - { - register mp_digit *bottom, *top; - - /* shift the digits down */ - - /* bottom */ - bottom = a->dp; - - /* top [offset into digits] */ - top = a->dp + b; - - /* this is implemented as a sliding window where - * the window is b-digits long and digits from - * the top of the window are copied to the bottom - * - * e.g. - - b-2 | b-1 | b0 | b1 | b2 | ... | bb | ----> - /\ | ----> - \-------------------/ ----> - */ - for (x = 0; x < (a->used - b); x++) { - *bottom++ = *top++; - } - - /* zero the top digits */ - for (; x < a->used; x++) { - *bottom++ = 0; - } - } - - /* remove excess digits */ - a->used -= b; -} - - -/* swap the elements of two integers, for cases where you can't simply swap the - * mp_int pointers around - */ -static void mp_exch (mp_int * a, mp_int * b) -{ - mp_int t; - - t = *a; - *a = *b; - *b = t; -} - - -/* trim unused digits - * - * This is used to ensure that leading zero digits are - * trimed and the leading "used" digit will be non-zero - * Typically very fast. Also fixes the sign if there - * are no more leading digits - */ -static void mp_clamp (mp_int * a) -{ - /* decrease used while the most significant digit is - * zero. - */ - while (a->used > 0 && a->dp[a->used - 1] == 0) { - --(a->used); - } - - /* reset the sign flag if used == 0 */ - if (a->used == 0) { - a->sign = MP_ZPOS; - } -} - - -/* grow as required */ -static int mp_grow (mp_int * a, int size) -{ - int i; - mp_digit *tmp; - - /* if the alloc size is smaller alloc more ram */ - if (a->alloc < size) { - /* ensure there are always at least MP_PREC digits extra on top */ - size += (MP_PREC * 2) - (size % MP_PREC); - - /* reallocate the array a->dp - * - * We store the return in a temporary variable - * in case the operation failed we don't want - * to overwrite the dp member of a. - */ - tmp = OPT_CAST(mp_digit) XREALLOC (a->dp, sizeof (mp_digit) * size); - if (tmp == NULL) { - /* reallocation failed but "a" is still valid [can be freed] */ - return MP_MEM; - } - - /* reallocation succeeded so set a->dp */ - a->dp = tmp; - - /* zero excess digits */ - i = a->alloc; - a->alloc = size; - for (; i < a->alloc; i++) { - a->dp[i] = 0; - } - } - return MP_OKAY; -} - - -#ifdef BN_MP_ABS_C -/* b = |a| - * - * Simple function copies the input and fixes the sign to positive - */ -static int mp_abs (mp_int * a, mp_int * b) -{ - int res; - - /* copy a to b */ - if (a != b) { - if ((res = mp_copy (a, b)) != MP_OKAY) { - return res; - } - } - - /* force the sign of b to positive */ - b->sign = MP_ZPOS; - - return MP_OKAY; -} -#endif - - -/* set to a digit */ -static void mp_set (mp_int * a, mp_digit b) -{ - mp_zero (a); - a->dp[0] = b & MP_MASK; - a->used = (a->dp[0] != 0) ? 1 : 0; -} - - -#ifndef LTM_NO_NEG_EXP -/* b = a/2 */ -static int mp_div_2(mp_int * a, mp_int * b) -{ - int x, res, oldused; - - /* copy */ - if (b->alloc < a->used) { - if ((res = mp_grow (b, a->used)) != MP_OKAY) { - return res; - } - } - - oldused = b->used; - b->used = a->used; - { - register mp_digit r, rr, *tmpa, *tmpb; - - /* source alias */ - tmpa = a->dp + b->used - 1; - - /* dest alias */ - tmpb = b->dp + b->used - 1; - - /* carry */ - r = 0; - for (x = b->used - 1; x >= 0; x--) { - /* get the carry for the next iteration */ - rr = *tmpa & 1; - - /* shift the current digit, add in carry and store */ - *tmpb-- = (*tmpa-- >> 1) | (r << (DIGIT_BIT - 1)); - - /* forward carry to next iteration */ - r = rr; - } - - /* zero excess digits */ - tmpb = b->dp + b->used; - for (x = b->used; x < oldused; x++) { - *tmpb++ = 0; - } - } - b->sign = a->sign; - mp_clamp (b); - return MP_OKAY; -} -#endif /* LTM_NO_NEG_EXP */ - - -/* shift left by a certain bit count */ -static int mp_mul_2d (mp_int * a, int b, mp_int * c) -{ - mp_digit d; - int res; - - /* copy */ - if (a != c) { - if ((res = mp_copy (a, c)) != MP_OKAY) { - return res; - } - } - - if (c->alloc < (int)(c->used + b/DIGIT_BIT + 1)) { - if ((res = mp_grow (c, c->used + b / DIGIT_BIT + 1)) != MP_OKAY) { - return res; - } - } - - /* shift by as many digits in the bit count */ - if (b >= (int)DIGIT_BIT) { - if ((res = mp_lshd (c, b / DIGIT_BIT)) != MP_OKAY) { - return res; - } - } - - /* shift any bit count < DIGIT_BIT */ - d = (mp_digit) (b % DIGIT_BIT); - if (d != 0) { - register mp_digit *tmpc, shift, mask, r, rr; - register int x; - - /* bitmask for carries */ - mask = (((mp_digit)1) << d) - 1; - - /* shift for msbs */ - shift = DIGIT_BIT - d; - - /* alias */ - tmpc = c->dp; - - /* carry */ - r = 0; - for (x = 0; x < c->used; x++) { - /* get the higher bits of the current word */ - rr = (*tmpc >> shift) & mask; - - /* shift the current word and OR in the carry */ - *tmpc = ((*tmpc << d) | r) & MP_MASK; - ++tmpc; - - /* set the carry to the carry bits of the current word */ - r = rr; - } - - /* set final carry */ - if (r != 0) { - c->dp[(c->used)++] = r; - } - } - mp_clamp (c); - return MP_OKAY; -} - - -#ifdef BN_MP_INIT_MULTI_C -static int mp_init_multi(mp_int *mp, ...) -{ - mp_err res = MP_OKAY; /* Assume ok until proven otherwise */ - int n = 0; /* Number of ok inits */ - mp_int* cur_arg = mp; - va_list args; - - va_start(args, mp); /* init args to next argument from caller */ - while (cur_arg != NULL) { - if (mp_init(cur_arg) != MP_OKAY) { - /* Oops - error! Back-track and mp_clear what we already - succeeded in init-ing, then return error. - */ - va_list clean_args; - - /* end the current list */ - va_end(args); - - /* now start cleaning up */ - cur_arg = mp; - va_start(clean_args, mp); - while (n--) { - mp_clear(cur_arg); - cur_arg = va_arg(clean_args, mp_int*); - } - va_end(clean_args); - return MP_MEM; - } - n++; - cur_arg = va_arg(args, mp_int*); - } - va_end(args); - return res; /* Assumed ok, if error flagged above. */ -} -#endif - - -#ifdef BN_MP_CLEAR_MULTI_C -static void mp_clear_multi(mp_int *mp, ...) -{ - mp_int* next_mp = mp; - va_list args; - va_start(args, mp); - while (next_mp != NULL) { - mp_clear(next_mp); - next_mp = va_arg(args, mp_int*); - } - va_end(args); -} -#endif - - -/* shift left a certain amount of digits */ -static int mp_lshd (mp_int * a, int b) -{ - int x, res; - - /* if its less than zero return */ - if (b <= 0) { - return MP_OKAY; - } - - /* grow to fit the new digits */ - if (a->alloc < a->used + b) { - if ((res = mp_grow (a, a->used + b)) != MP_OKAY) { - return res; - } - } - - { - register mp_digit *top, *bottom; - - /* increment the used by the shift amount then copy upwards */ - a->used += b; - - /* top */ - top = a->dp + a->used - 1; - - /* base */ - bottom = a->dp + a->used - 1 - b; - - /* much like mp_rshd this is implemented using a sliding window - * except the window goes the otherway around. Copying from - * the bottom to the top. see bn_mp_rshd.c for more info. - */ - for (x = a->used - 1; x >= b; x--) { - *top-- = *bottom--; - } - - /* zero the lower digits */ - top = a->dp; - for (x = 0; x < b; x++) { - *top++ = 0; - } - } - return MP_OKAY; -} - - -/* returns the number of bits in an int */ -static int mp_count_bits (mp_int * a) -{ - int r; - mp_digit q; - - /* shortcut */ - if (a->used == 0) { - return 0; - } - - /* get number of digits and add that */ - r = (a->used - 1) * DIGIT_BIT; - - /* take the last digit and count the bits in it */ - q = a->dp[a->used - 1]; - while (q > ((mp_digit) 0)) { - ++r; - q >>= ((mp_digit) 1); - } - return r; -} - - -/* calc a value mod 2**b */ -static int mp_mod_2d (mp_int * a, int b, mp_int * c) -{ - int x, res; - - /* if b is <= 0 then zero the int */ - if (b <= 0) { - mp_zero (c); - return MP_OKAY; - } - - /* if the modulus is larger than the value than return */ - if (b >= (int) (a->used * DIGIT_BIT)) { - res = mp_copy (a, c); - return res; - } - - /* copy */ - if ((res = mp_copy (a, c)) != MP_OKAY) { - return res; - } - - /* zero digits above the last digit of the modulus */ - for (x = (b / DIGIT_BIT) + ((b % DIGIT_BIT) == 0 ? 0 : 1); x < c->used; x++) { - c->dp[x] = 0; - } - /* clear the digit that is not completely outside/inside the modulus */ - c->dp[b / DIGIT_BIT] &= - (mp_digit) ((((mp_digit) 1) << (((mp_digit) b) % DIGIT_BIT)) - ((mp_digit) 1)); - mp_clamp (c); - return MP_OKAY; -} - - -#ifdef BN_MP_DIV_SMALL - -/* slower bit-bang division... also smaller */ -static int mp_div(mp_int * a, mp_int * b, mp_int * c, mp_int * d) -{ - mp_int ta, tb, tq, q; - int res, n, n2; - - /* is divisor zero ? */ - if (mp_iszero (b) == 1) { - return MP_VAL; - } - - /* if a < b then q=0, r = a */ - if (mp_cmp_mag (a, b) == MP_LT) { - if (d != NULL) { - res = mp_copy (a, d); - } else { - res = MP_OKAY; - } - if (c != NULL) { - mp_zero (c); - } - return res; - } - - /* init our temps */ - if ((res = mp_init_multi(&ta, &tb, &tq, &q, NULL)) != MP_OKAY) { - return res; - } - - - mp_set(&tq, 1); - n = mp_count_bits(a) - mp_count_bits(b); - if (((res = mp_abs(a, &ta)) != MP_OKAY) || - ((res = mp_abs(b, &tb)) != MP_OKAY) || - ((res = mp_mul_2d(&tb, n, &tb)) != MP_OKAY) || - ((res = mp_mul_2d(&tq, n, &tq)) != MP_OKAY)) { - goto LBL_ERR; - } - - while (n-- >= 0) { - if (mp_cmp(&tb, &ta) != MP_GT) { - if (((res = mp_sub(&ta, &tb, &ta)) != MP_OKAY) || - ((res = mp_add(&q, &tq, &q)) != MP_OKAY)) { - goto LBL_ERR; - } - } - if (((res = mp_div_2d(&tb, 1, &tb, NULL)) != MP_OKAY) || - ((res = mp_div_2d(&tq, 1, &tq, NULL)) != MP_OKAY)) { - goto LBL_ERR; - } - } - - /* now q == quotient and ta == remainder */ - n = a->sign; - n2 = (a->sign == b->sign ? MP_ZPOS : MP_NEG); - if (c != NULL) { - mp_exch(c, &q); - c->sign = (mp_iszero(c) == MP_YES) ? MP_ZPOS : n2; - } - if (d != NULL) { - mp_exch(d, &ta); - d->sign = (mp_iszero(d) == MP_YES) ? MP_ZPOS : n; - } -LBL_ERR: - mp_clear_multi(&ta, &tb, &tq, &q, NULL); - return res; -} - -#else - -/* integer signed division. - * c*b + d == a [e.g. a/b, c=quotient, d=remainder] - * HAC pp.598 Algorithm 14.20 - * - * Note that the description in HAC is horribly - * incomplete. For example, it doesn't consider - * the case where digits are removed from 'x' in - * the inner loop. It also doesn't consider the - * case that y has fewer than three digits, etc.. - * - * The overall algorithm is as described as - * 14.20 from HAC but fixed to treat these cases. -*/ -static int mp_div (mp_int * a, mp_int * b, mp_int * c, mp_int * d) -{ - mp_int q, x, y, t1, t2; - int res, n, t, i, norm, neg; - - /* is divisor zero ? */ - if (mp_iszero (b) == 1) { - return MP_VAL; - } - - /* if a < b then q=0, r = a */ - if (mp_cmp_mag (a, b) == MP_LT) { - if (d != NULL) { - res = mp_copy (a, d); - } else { - res = MP_OKAY; - } - if (c != NULL) { - mp_zero (c); - } - return res; - } - - if ((res = mp_init_size (&q, a->used + 2)) != MP_OKAY) { - return res; - } - q.used = a->used + 2; - - if ((res = mp_init (&t1)) != MP_OKAY) { - goto LBL_Q; - } - - if ((res = mp_init (&t2)) != MP_OKAY) { - goto LBL_T1; - } - - if ((res = mp_init_copy (&x, a)) != MP_OKAY) { - goto LBL_T2; - } - - if ((res = mp_init_copy (&y, b)) != MP_OKAY) { - goto LBL_X; - } - - /* fix the sign */ - neg = (a->sign == b->sign) ? MP_ZPOS : MP_NEG; - x.sign = y.sign = MP_ZPOS; - - /* normalize both x and y, ensure that y >= b/2, [b == 2**DIGIT_BIT] */ - norm = mp_count_bits(&y) % DIGIT_BIT; - if (norm < (int)(DIGIT_BIT-1)) { - norm = (DIGIT_BIT-1) - norm; - if ((res = mp_mul_2d (&x, norm, &x)) != MP_OKAY) { - goto LBL_Y; - } - if ((res = mp_mul_2d (&y, norm, &y)) != MP_OKAY) { - goto LBL_Y; - } - } else { - norm = 0; - } - - /* note hac does 0 based, so if used==5 then its 0,1,2,3,4, e.g. use 4 */ - n = x.used - 1; - t = y.used - 1; - - /* while (x >= y*b**n-t) do { q[n-t] += 1; x -= y*b**{n-t} } */ - if ((res = mp_lshd (&y, n - t)) != MP_OKAY) { /* y = y*b**{n-t} */ - goto LBL_Y; - } - - while (mp_cmp (&x, &y) != MP_LT) { - ++(q.dp[n - t]); - if ((res = mp_sub (&x, &y, &x)) != MP_OKAY) { - goto LBL_Y; - } - } - - /* reset y by shifting it back down */ - mp_rshd (&y, n - t); - - /* step 3. for i from n down to (t + 1) */ - for (i = n; i >= (t + 1); i--) { - if (i > x.used) { - continue; - } - - /* step 3.1 if xi == yt then set q{i-t-1} to b-1, - * otherwise set q{i-t-1} to (xi*b + x{i-1})/yt */ - if (x.dp[i] == y.dp[t]) { - q.dp[i - t - 1] = ((((mp_digit)1) << DIGIT_BIT) - 1); - } else { - mp_word tmp; - tmp = ((mp_word) x.dp[i]) << ((mp_word) DIGIT_BIT); - tmp |= ((mp_word) x.dp[i - 1]); - tmp /= ((mp_word) y.dp[t]); - if (tmp > (mp_word) MP_MASK) - tmp = MP_MASK; - q.dp[i - t - 1] = (mp_digit) (tmp & (mp_word) (MP_MASK)); - } - - /* while (q{i-t-1} * (yt * b + y{t-1})) > - xi * b**2 + xi-1 * b + xi-2 - - do q{i-t-1} -= 1; - */ - q.dp[i - t - 1] = (q.dp[i - t - 1] + 1) & MP_MASK; - do { - q.dp[i - t - 1] = (q.dp[i - t - 1] - 1) & MP_MASK; - - /* find left hand */ - mp_zero (&t1); - t1.dp[0] = (t - 1 < 0) ? 0 : y.dp[t - 1]; - t1.dp[1] = y.dp[t]; - t1.used = 2; - if ((res = mp_mul_d (&t1, q.dp[i - t - 1], &t1)) != MP_OKAY) { - goto LBL_Y; - } - - /* find right hand */ - t2.dp[0] = (i - 2 < 0) ? 0 : x.dp[i - 2]; - t2.dp[1] = (i - 1 < 0) ? 0 : x.dp[i - 1]; - t2.dp[2] = x.dp[i]; - t2.used = 3; - } while (mp_cmp_mag(&t1, &t2) == MP_GT); - - /* step 3.3 x = x - q{i-t-1} * y * b**{i-t-1} */ - if ((res = mp_mul_d (&y, q.dp[i - t - 1], &t1)) != MP_OKAY) { - goto LBL_Y; - } - - if ((res = mp_lshd (&t1, i - t - 1)) != MP_OKAY) { - goto LBL_Y; - } - - if ((res = mp_sub (&x, &t1, &x)) != MP_OKAY) { - goto LBL_Y; - } - - /* if x < 0 then { x = x + y*b**{i-t-1}; q{i-t-1} -= 1; } */ - if (x.sign == MP_NEG) { - if ((res = mp_copy (&y, &t1)) != MP_OKAY) { - goto LBL_Y; - } - if ((res = mp_lshd (&t1, i - t - 1)) != MP_OKAY) { - goto LBL_Y; - } - if ((res = mp_add (&x, &t1, &x)) != MP_OKAY) { - goto LBL_Y; - } - - q.dp[i - t - 1] = (q.dp[i - t - 1] - 1UL) & MP_MASK; - } - } - - /* now q is the quotient and x is the remainder - * [which we have to normalize] - */ - - /* get sign before writing to c */ - x.sign = x.used == 0 ? MP_ZPOS : a->sign; - - if (c != NULL) { - mp_clamp (&q); - mp_exch (&q, c); - c->sign = neg; - } - - if (d != NULL) { - mp_div_2d (&x, norm, &x, NULL); - mp_exch (&x, d); - } - - res = MP_OKAY; - -LBL_Y:mp_clear (&y); -LBL_X:mp_clear (&x); -LBL_T2:mp_clear (&t2); -LBL_T1:mp_clear (&t1); -LBL_Q:mp_clear (&q); - return res; -} - -#endif - - -#ifdef MP_LOW_MEM - #define TAB_SIZE 32 -#else - #define TAB_SIZE 256 -#endif - -static int s_mp_exptmod (mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode) -{ - mp_int M[TAB_SIZE], res, mu; - mp_digit buf; - int err, bitbuf, bitcpy, bitcnt, mode, digidx, x, y, winsize; - int (*redux)(mp_int*,mp_int*,mp_int*); - - /* find window size */ - x = mp_count_bits (X); - if (x <= 7) { - winsize = 2; - } else if (x <= 36) { - winsize = 3; - } else if (x <= 140) { - winsize = 4; - } else if (x <= 450) { - winsize = 5; - } else if (x <= 1303) { - winsize = 6; - } else if (x <= 3529) { - winsize = 7; - } else { - winsize = 8; - } - -#ifdef MP_LOW_MEM - if (winsize > 5) { - winsize = 5; - } -#endif - - /* init M array */ - /* init first cell */ - if ((err = mp_init(&M[1])) != MP_OKAY) { - return err; - } - - /* now init the second half of the array */ - for (x = 1<<(winsize-1); x < (1 << winsize); x++) { - if ((err = mp_init(&M[x])) != MP_OKAY) { - for (y = 1<<(winsize-1); y < x; y++) { - mp_clear (&M[y]); - } - mp_clear(&M[1]); - return err; - } - } - - /* create mu, used for Barrett reduction */ - if ((err = mp_init (&mu)) != MP_OKAY) { - goto LBL_M; - } - - if (redmode == 0) { - if ((err = mp_reduce_setup (&mu, P)) != MP_OKAY) { - goto LBL_MU; - } - redux = mp_reduce; - } else { - if ((err = mp_reduce_2k_setup_l (P, &mu)) != MP_OKAY) { - goto LBL_MU; - } - redux = mp_reduce_2k_l; - } - - /* create M table - * - * The M table contains powers of the base, - * e.g. M[x] = G**x mod P - * - * The first half of the table is not - * computed though accept for M[0] and M[1] - */ - if ((err = mp_mod (G, P, &M[1])) != MP_OKAY) { - goto LBL_MU; - } - - /* compute the value at M[1<<(winsize-1)] by squaring - * M[1] (winsize-1) times - */ - if ((err = mp_copy (&M[1], &M[1 << (winsize - 1)])) != MP_OKAY) { - goto LBL_MU; - } - - for (x = 0; x < (winsize - 1); x++) { - /* square it */ - if ((err = mp_sqr (&M[1 << (winsize - 1)], - &M[1 << (winsize - 1)])) != MP_OKAY) { - goto LBL_MU; - } - - /* reduce modulo P */ - if ((err = redux (&M[1 << (winsize - 1)], P, &mu)) != MP_OKAY) { - goto LBL_MU; - } - } - - /* create upper table, that is M[x] = M[x-1] * M[1] (mod P) - * for x = (2**(winsize - 1) + 1) to (2**winsize - 1) - */ - for (x = (1 << (winsize - 1)) + 1; x < (1 << winsize); x++) { - if ((err = mp_mul (&M[x - 1], &M[1], &M[x])) != MP_OKAY) { - goto LBL_MU; - } - if ((err = redux (&M[x], P, &mu)) != MP_OKAY) { - goto LBL_MU; - } - } - - /* setup result */ - if ((err = mp_init (&res)) != MP_OKAY) { - goto LBL_MU; - } - mp_set (&res, 1); - - /* set initial mode and bit cnt */ - mode = 0; - bitcnt = 1; - buf = 0; - digidx = X->used - 1; - bitcpy = 0; - bitbuf = 0; - - for (;;) { - /* grab next digit as required */ - if (--bitcnt == 0) { - /* if digidx == -1 we are out of digits */ - if (digidx == -1) { - break; - } - /* read next digit and reset the bitcnt */ - buf = X->dp[digidx--]; - bitcnt = (int) DIGIT_BIT; - } - - /* grab the next msb from the exponent */ - y = (buf >> (mp_digit)(DIGIT_BIT - 1)) & 1; - buf <<= (mp_digit)1; - - /* if the bit is zero and mode == 0 then we ignore it - * These represent the leading zero bits before the first 1 bit - * in the exponent. Technically this opt is not required but it - * does lower the # of trivial squaring/reductions used - */ - if (mode == 0 && y == 0) { - continue; - } - - /* if the bit is zero and mode == 1 then we square */ - if (mode == 1 && y == 0) { - if ((err = mp_sqr (&res, &res)) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&res, P, &mu)) != MP_OKAY) { - goto LBL_RES; - } - continue; - } - - /* else we add it to the window */ - bitbuf |= (y << (winsize - ++bitcpy)); - mode = 2; - - if (bitcpy == winsize) { - /* ok window is filled so square as required and multiply */ - /* square first */ - for (x = 0; x < winsize; x++) { - if ((err = mp_sqr (&res, &res)) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&res, P, &mu)) != MP_OKAY) { - goto LBL_RES; - } - } - - /* then multiply */ - if ((err = mp_mul (&res, &M[bitbuf], &res)) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&res, P, &mu)) != MP_OKAY) { - goto LBL_RES; - } - - /* empty window and reset */ - bitcpy = 0; - bitbuf = 0; - mode = 1; - } - } - - /* if bits remain then square/multiply */ - if (mode == 2 && bitcpy > 0) { - /* square then multiply if the bit is set */ - for (x = 0; x < bitcpy; x++) { - if ((err = mp_sqr (&res, &res)) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&res, P, &mu)) != MP_OKAY) { - goto LBL_RES; - } - - bitbuf <<= 1; - if ((bitbuf & (1 << winsize)) != 0) { - /* then multiply */ - if ((err = mp_mul (&res, &M[1], &res)) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&res, P, &mu)) != MP_OKAY) { - goto LBL_RES; - } - } - } - } - - mp_exch (&res, Y); - err = MP_OKAY; -LBL_RES:mp_clear (&res); -LBL_MU:mp_clear (&mu); -LBL_M: - mp_clear(&M[1]); - for (x = 1<<(winsize-1); x < (1 << winsize); x++) { - mp_clear (&M[x]); - } - return err; -} - - -/* computes b = a*a */ -static int mp_sqr (mp_int * a, mp_int * b) -{ - int res; - -#ifdef BN_MP_TOOM_SQR_C - /* use Toom-Cook? */ - if (a->used >= TOOM_SQR_CUTOFF) { - res = mp_toom_sqr(a, b); - /* Karatsuba? */ - } else -#endif -#ifdef BN_MP_KARATSUBA_SQR_C -if (a->used >= KARATSUBA_SQR_CUTOFF) { - res = mp_karatsuba_sqr (a, b); - } else -#endif - { -#ifdef BN_FAST_S_MP_SQR_C - /* can we use the fast comba multiplier? */ - if ((a->used * 2 + 1) < MP_WARRAY && - a->used < - (1 << (sizeof(mp_word) * CHAR_BIT - 2*DIGIT_BIT - 1))) { - res = fast_s_mp_sqr (a, b); - } else -#endif -#ifdef BN_S_MP_SQR_C - res = s_mp_sqr (a, b); -#else -#error mp_sqr could fail - res = MP_VAL; -#endif - } - b->sign = MP_ZPOS; - return res; -} - - -/* reduces a modulo n where n is of the form 2**p - d - This differs from reduce_2k since "d" can be larger - than a single digit. -*/ -static int mp_reduce_2k_l(mp_int *a, mp_int *n, mp_int *d) -{ - mp_int q; - int p, res; - - if ((res = mp_init(&q)) != MP_OKAY) { - return res; - } - - p = mp_count_bits(n); -top: - /* q = a/2**p, a = a mod 2**p */ - if ((res = mp_div_2d(a, p, &q, a)) != MP_OKAY) { - goto ERR; - } - - /* q = q * d */ - if ((res = mp_mul(&q, d, &q)) != MP_OKAY) { - goto ERR; - } - - /* a = a + q */ - if ((res = s_mp_add(a, &q, a)) != MP_OKAY) { - goto ERR; - } - - if (mp_cmp_mag(a, n) != MP_LT) { - s_mp_sub(a, n, a); - goto top; - } - -ERR: - mp_clear(&q); - return res; -} - - -/* determines the setup value */ -static int mp_reduce_2k_setup_l(mp_int *a, mp_int *d) -{ - int res; - mp_int tmp; - - if ((res = mp_init(&tmp)) != MP_OKAY) { - return res; - } - - if ((res = mp_2expt(&tmp, mp_count_bits(a))) != MP_OKAY) { - goto ERR; - } - - if ((res = s_mp_sub(&tmp, a, d)) != MP_OKAY) { - goto ERR; - } - -ERR: - mp_clear(&tmp); - return res; -} - - -/* computes a = 2**b - * - * Simple algorithm which zeroes the int, grows it then just sets one bit - * as required. - */ -static int mp_2expt (mp_int * a, int b) -{ - int res; - - /* zero a as per default */ - mp_zero (a); - - /* grow a to accommodate the single bit */ - if ((res = mp_grow (a, b / DIGIT_BIT + 1)) != MP_OKAY) { - return res; - } - - /* set the used count of where the bit will go */ - a->used = b / DIGIT_BIT + 1; - - /* put the single bit in its place */ - a->dp[b / DIGIT_BIT] = ((mp_digit)1) << (b % DIGIT_BIT); - - return MP_OKAY; -} - - -/* pre-calculate the value required for Barrett reduction - * For a given modulus "b" it calulates the value required in "a" - */ -static int mp_reduce_setup (mp_int * a, mp_int * b) -{ - int res; - - if ((res = mp_2expt (a, b->used * 2 * DIGIT_BIT)) != MP_OKAY) { - return res; - } - return mp_div (a, b, a, NULL); -} - - -/* reduces x mod m, assumes 0 < x < m**2, mu is - * precomputed via mp_reduce_setup. - * From HAC pp.604 Algorithm 14.42 - */ -static int mp_reduce (mp_int * x, mp_int * m, mp_int * mu) -{ - mp_int q; - int res, um = m->used; - - /* q = x */ - if ((res = mp_init_copy (&q, x)) != MP_OKAY) { - return res; - } - - /* q1 = x / b**(k-1) */ - mp_rshd (&q, um - 1); - - /* according to HAC this optimization is ok */ - if (((unsigned long) um) > (((mp_digit)1) << (DIGIT_BIT - 1))) { - if ((res = mp_mul (&q, mu, &q)) != MP_OKAY) { - goto CLEANUP; - } - } else { -#ifdef BN_S_MP_MUL_HIGH_DIGS_C - if ((res = s_mp_mul_high_digs (&q, mu, &q, um)) != MP_OKAY) { - goto CLEANUP; - } -#elif defined(BN_FAST_S_MP_MUL_HIGH_DIGS_C) - if ((res = fast_s_mp_mul_high_digs (&q, mu, &q, um)) != MP_OKAY) { - goto CLEANUP; - } -#else - { -#error mp_reduce would always fail - res = MP_VAL; - goto CLEANUP; - } -#endif - } - - /* q3 = q2 / b**(k+1) */ - mp_rshd (&q, um + 1); - - /* x = x mod b**(k+1), quick (no division) */ - if ((res = mp_mod_2d (x, DIGIT_BIT * (um + 1), x)) != MP_OKAY) { - goto CLEANUP; - } - - /* q = q * m mod b**(k+1), quick (no division) */ - if ((res = s_mp_mul_digs (&q, m, &q, um + 1)) != MP_OKAY) { - goto CLEANUP; - } - - /* x = x - q */ - if ((res = mp_sub (x, &q, x)) != MP_OKAY) { - goto CLEANUP; - } - - /* If x < 0, add b**(k+1) to it */ - if (mp_cmp_d (x, 0) == MP_LT) { - mp_set (&q, 1); - if ((res = mp_lshd (&q, um + 1)) != MP_OKAY) { - goto CLEANUP; - } - if ((res = mp_add (x, &q, x)) != MP_OKAY) { - goto CLEANUP; - } - } - - /* Back off if it's too big */ - while (mp_cmp (x, m) != MP_LT) { - if ((res = s_mp_sub (x, m, x)) != MP_OKAY) { - goto CLEANUP; - } - } - -CLEANUP: - mp_clear (&q); - - return res; -} - - -/* multiplies |a| * |b| and only computes up to digs digits of result - * HAC pp. 595, Algorithm 14.12 Modified so you can control how - * many digits of output are created. - */ -static int s_mp_mul_digs (mp_int * a, mp_int * b, mp_int * c, int digs) -{ - mp_int t; - int res, pa, pb, ix, iy; - mp_digit u; - mp_word r; - mp_digit tmpx, *tmpt, *tmpy; - -#ifdef BN_FAST_S_MP_MUL_DIGS_C - /* can we use the fast multiplier? */ - if (((digs) < MP_WARRAY) && - MIN (a->used, b->used) < - (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) { - return fast_s_mp_mul_digs (a, b, c, digs); - } -#endif - - if ((res = mp_init_size (&t, digs)) != MP_OKAY) { - return res; - } - t.used = digs; - - /* compute the digits of the product directly */ - pa = a->used; - for (ix = 0; ix < pa; ix++) { - /* set the carry to zero */ - u = 0; - - /* limit ourselves to making digs digits of output */ - pb = MIN (b->used, digs - ix); - - /* setup some aliases */ - /* copy of the digit from a used within the nested loop */ - tmpx = a->dp[ix]; - - /* an alias for the destination shifted ix places */ - tmpt = t.dp + ix; - - /* an alias for the digits of b */ - tmpy = b->dp; - - /* compute the columns of the output and propagate the carry */ - for (iy = 0; iy < pb; iy++) { - /* compute the column as a mp_word */ - r = ((mp_word)*tmpt) + - ((mp_word)tmpx) * ((mp_word)*tmpy++) + - ((mp_word) u); - - /* the new column is the lower part of the result */ - *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK)); - - /* get the carry word from the result */ - u = (mp_digit) (r >> ((mp_word) DIGIT_BIT)); - } - /* set carry if it is placed below digs */ - if (ix + iy < digs) { - *tmpt = u; - } - } - - mp_clamp (&t); - mp_exch (&t, c); - - mp_clear (&t); - return MP_OKAY; -} - - -#ifdef BN_FAST_S_MP_MUL_DIGS_C -/* Fast (comba) multiplier - * - * This is the fast column-array [comba] multiplier. It is - * designed to compute the columns of the product first - * then handle the carries afterwards. This has the effect - * of making the nested loops that compute the columns very - * simple and schedulable on super-scalar processors. - * - * This has been modified to produce a variable number of - * digits of output so if say only a half-product is required - * you don't have to compute the upper half (a feature - * required for fast Barrett reduction). - * - * Based on Algorithm 14.12 on pp.595 of HAC. - * - */ -static int fast_s_mp_mul_digs (mp_int * a, mp_int * b, mp_int * c, int digs) -{ - int olduse, res, pa, ix, iz; - mp_digit W[MP_WARRAY]; - register mp_word _W; - - /* grow the destination as required */ - if (c->alloc < digs) { - if ((res = mp_grow (c, digs)) != MP_OKAY) { - return res; - } - } - - /* number of output digits to produce */ - pa = MIN(digs, a->used + b->used); - - /* clear the carry */ - _W = 0; - for (ix = 0; ix < pa; ix++) { - int tx, ty; - int iy; - mp_digit *tmpx, *tmpy; - - /* get offsets into the two bignums */ - ty = MIN(b->used-1, ix); - tx = ix - ty; - - /* setup temp aliases */ - tmpx = a->dp + tx; - tmpy = b->dp + ty; - - /* this is the number of times the loop will iterrate, essentially - while (tx++ < a->used && ty-- >= 0) { ... } - */ - iy = MIN(a->used-tx, ty+1); - - /* execute loop */ - for (iz = 0; iz < iy; ++iz) { - _W += ((mp_word)*tmpx++)*((mp_word)*tmpy--); - - } - - /* store term */ - W[ix] = ((mp_digit)_W) & MP_MASK; - - /* make next carry */ - _W = _W >> ((mp_word)DIGIT_BIT); - } - - /* setup dest */ - olduse = c->used; - c->used = pa; - - { - register mp_digit *tmpc; - tmpc = c->dp; - for (ix = 0; ix < pa+1; ix++) { - /* now extract the previous digit [below the carry] */ - *tmpc++ = W[ix]; - } - - /* clear unused digits [that existed in the old copy of c] */ - for (; ix < olduse; ix++) { - *tmpc++ = 0; - } - } - mp_clamp (c); - return MP_OKAY; -} -#endif /* BN_FAST_S_MP_MUL_DIGS_C */ - - -/* init an mp_init for a given size */ -static int mp_init_size (mp_int * a, int size) -{ - int x; - - /* pad size so there are always extra digits */ - size += (MP_PREC * 2) - (size % MP_PREC); - - /* alloc mem */ - a->dp = OPT_CAST(mp_digit) XMALLOC (sizeof (mp_digit) * size); - if (a->dp == NULL) { - return MP_MEM; - } - - /* set the members */ - a->used = 0; - a->alloc = size; - a->sign = MP_ZPOS; - - /* zero the digits */ - for (x = 0; x < size; x++) { - a->dp[x] = 0; - } - - return MP_OKAY; -} - - -/* low level squaring, b = a*a, HAC pp.596-597, Algorithm 14.16 */ -static int s_mp_sqr (mp_int * a, mp_int * b) -{ - mp_int t; - int res, ix, iy, pa; - mp_word r; - mp_digit u, tmpx, *tmpt; - - pa = a->used; - if ((res = mp_init_size (&t, 2*pa + 1)) != MP_OKAY) { - return res; - } - - /* default used is maximum possible size */ - t.used = 2*pa + 1; - - for (ix = 0; ix < pa; ix++) { - /* first calculate the digit at 2*ix */ - /* calculate double precision result */ - r = ((mp_word) t.dp[2*ix]) + - ((mp_word)a->dp[ix])*((mp_word)a->dp[ix]); - - /* store lower part in result */ - t.dp[ix+ix] = (mp_digit) (r & ((mp_word) MP_MASK)); - - /* get the carry */ - u = (mp_digit)(r >> ((mp_word) DIGIT_BIT)); - - /* left hand side of A[ix] * A[iy] */ - tmpx = a->dp[ix]; - - /* alias for where to store the results */ - tmpt = t.dp + (2*ix + 1); - - for (iy = ix + 1; iy < pa; iy++) { - /* first calculate the product */ - r = ((mp_word)tmpx) * ((mp_word)a->dp[iy]); - - /* now calculate the double precision result, note we use - * addition instead of *2 since it's easier to optimize - */ - r = ((mp_word) *tmpt) + r + r + ((mp_word) u); - - /* store lower part */ - *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK)); - - /* get carry */ - u = (mp_digit)(r >> ((mp_word) DIGIT_BIT)); - } - /* propagate upwards */ - while (u != ((mp_digit) 0)) { - r = ((mp_word) *tmpt) + ((mp_word) u); - *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK)); - u = (mp_digit)(r >> ((mp_word) DIGIT_BIT)); - } - } - - mp_clamp (&t); - mp_exch (&t, b); - mp_clear (&t); - return MP_OKAY; -} - - -/* multiplies |a| * |b| and does not compute the lower digs digits - * [meant to get the higher part of the product] - */ -static int s_mp_mul_high_digs (mp_int * a, mp_int * b, mp_int * c, int digs) -{ - mp_int t; - int res, pa, pb, ix, iy; - mp_digit u; - mp_word r; - mp_digit tmpx, *tmpt, *tmpy; - - /* can we use the fast multiplier? */ -#ifdef BN_FAST_S_MP_MUL_HIGH_DIGS_C - if (((a->used + b->used + 1) < MP_WARRAY) - && MIN (a->used, b->used) < (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) { - return fast_s_mp_mul_high_digs (a, b, c, digs); - } -#endif - - if ((res = mp_init_size (&t, a->used + b->used + 1)) != MP_OKAY) { - return res; - } - t.used = a->used + b->used + 1; - - pa = a->used; - pb = b->used; - for (ix = 0; ix < pa; ix++) { - /* clear the carry */ - u = 0; - - /* left hand side of A[ix] * B[iy] */ - tmpx = a->dp[ix]; - - /* alias to the address of where the digits will be stored */ - tmpt = &(t.dp[digs]); - - /* alias for where to read the right hand side from */ - tmpy = b->dp + (digs - ix); - - for (iy = digs - ix; iy < pb; iy++) { - /* calculate the double precision result */ - r = ((mp_word)*tmpt) + - ((mp_word)tmpx) * ((mp_word)*tmpy++) + - ((mp_word) u); - - /* get the lower part */ - *tmpt++ = (mp_digit) (r & ((mp_word) MP_MASK)); - - /* carry the carry */ - u = (mp_digit) (r >> ((mp_word) DIGIT_BIT)); - } - *tmpt = u; - } - mp_clamp (&t); - mp_exch (&t, c); - mp_clear (&t); - return MP_OKAY; -} - - -#ifdef BN_MP_MONTGOMERY_SETUP_C -/* setups the montgomery reduction stuff */ -static int -mp_montgomery_setup (mp_int * n, mp_digit * rho) -{ - mp_digit x, b; - -/* fast inversion mod 2**k - * - * Based on the fact that - * - * XA = 1 (mod 2**n) => (X(2-XA)) A = 1 (mod 2**2n) - * => 2*X*A - X*X*A*A = 1 - * => 2*(1) - (1) = 1 - */ - b = n->dp[0]; - - if ((b & 1) == 0) { - return MP_VAL; - } - - x = (((b + 2) & 4) << 1) + b; /* here x*a==1 mod 2**4 */ - x *= 2 - b * x; /* here x*a==1 mod 2**8 */ -#if !defined(MP_8BIT) - x *= 2 - b * x; /* here x*a==1 mod 2**16 */ -#endif -#if defined(MP_64BIT) || !(defined(MP_8BIT) || defined(MP_16BIT)) - x *= 2 - b * x; /* here x*a==1 mod 2**32 */ -#endif -#ifdef MP_64BIT - x *= 2 - b * x; /* here x*a==1 mod 2**64 */ -#endif - - /* rho = -1/m mod b */ - *rho = (unsigned long)(((mp_word)1 << ((mp_word) DIGIT_BIT)) - x) & MP_MASK; - - return MP_OKAY; -} -#endif - - -#ifdef BN_FAST_MP_MONTGOMERY_REDUCE_C -/* computes xR**-1 == x (mod N) via Montgomery Reduction - * - * This is an optimized implementation of montgomery_reduce - * which uses the comba method to quickly calculate the columns of the - * reduction. - * - * Based on Algorithm 14.32 on pp.601 of HAC. -*/ -static int fast_mp_montgomery_reduce (mp_int * x, mp_int * n, mp_digit rho) -{ - int ix, res, olduse; - mp_word W[MP_WARRAY]; - - /* get old used count */ - olduse = x->used; - - /* grow a as required */ - if (x->alloc < n->used + 1) { - if ((res = mp_grow (x, n->used + 1)) != MP_OKAY) { - return res; - } - } - - /* first we have to get the digits of the input into - * an array of double precision words W[...] - */ - { - register mp_word *_W; - register mp_digit *tmpx; - - /* alias for the W[] array */ - _W = W; - - /* alias for the digits of x*/ - tmpx = x->dp; - - /* copy the digits of a into W[0..a->used-1] */ - for (ix = 0; ix < x->used; ix++) { - *_W++ = *tmpx++; - } - - /* zero the high words of W[a->used..m->used*2] */ - for (; ix < n->used * 2 + 1; ix++) { - *_W++ = 0; - } - } - - /* now we proceed to zero successive digits - * from the least significant upwards - */ - for (ix = 0; ix < n->used; ix++) { - /* mu = ai * m' mod b - * - * We avoid a double precision multiplication (which isn't required) - * by casting the value down to a mp_digit. Note this requires - * that W[ix-1] have the carry cleared (see after the inner loop) - */ - register mp_digit mu; - mu = (mp_digit) (((W[ix] & MP_MASK) * rho) & MP_MASK); - - /* a = a + mu * m * b**i - * - * This is computed in place and on the fly. The multiplication - * by b**i is handled by offseting which columns the results - * are added to. - * - * Note the comba method normally doesn't handle carries in the - * inner loop In this case we fix the carry from the previous - * column since the Montgomery reduction requires digits of the - * result (so far) [see above] to work. This is - * handled by fixing up one carry after the inner loop. The - * carry fixups are done in order so after these loops the - * first m->used words of W[] have the carries fixed - */ - { - register int iy; - register mp_digit *tmpn; - register mp_word *_W; - - /* alias for the digits of the modulus */ - tmpn = n->dp; - - /* Alias for the columns set by an offset of ix */ - _W = W + ix; - - /* inner loop */ - for (iy = 0; iy < n->used; iy++) { - *_W++ += ((mp_word)mu) * ((mp_word)*tmpn++); - } - } - - /* now fix carry for next digit, W[ix+1] */ - W[ix + 1] += W[ix] >> ((mp_word) DIGIT_BIT); - } - - /* now we have to propagate the carries and - * shift the words downward [all those least - * significant digits we zeroed]. - */ - { - register mp_digit *tmpx; - register mp_word *_W, *_W1; - - /* nox fix rest of carries */ - - /* alias for current word */ - _W1 = W + ix; - - /* alias for next word, where the carry goes */ - _W = W + ++ix; - - for (; ix <= n->used * 2 + 1; ix++) { - *_W++ += *_W1++ >> ((mp_word) DIGIT_BIT); - } - - /* copy out, A = A/b**n - * - * The result is A/b**n but instead of converting from an - * array of mp_word to mp_digit than calling mp_rshd - * we just copy them in the right order - */ - - /* alias for destination word */ - tmpx = x->dp; - - /* alias for shifted double precision result */ - _W = W + n->used; - - for (ix = 0; ix < n->used + 1; ix++) { - *tmpx++ = (mp_digit)(*_W++ & ((mp_word) MP_MASK)); - } - - /* zero oldused digits, if the input a was larger than - * m->used+1 we'll have to clear the digits - */ - for (; ix < olduse; ix++) { - *tmpx++ = 0; - } - } - - /* set the max used and clamp */ - x->used = n->used + 1; - mp_clamp (x); - - /* if A >= m then A = A - m */ - if (mp_cmp_mag (x, n) != MP_LT) { - return s_mp_sub (x, n, x); - } - return MP_OKAY; -} -#endif - - -#ifdef BN_MP_MUL_2_C -/* b = a*2 */ -static int mp_mul_2(mp_int * a, mp_int * b) -{ - int x, res, oldused; - - /* grow to accommodate result */ - if (b->alloc < a->used + 1) { - if ((res = mp_grow (b, a->used + 1)) != MP_OKAY) { - return res; - } - } - - oldused = b->used; - b->used = a->used; - - { - register mp_digit r, rr, *tmpa, *tmpb; - - /* alias for source */ - tmpa = a->dp; - - /* alias for dest */ - tmpb = b->dp; - - /* carry */ - r = 0; - for (x = 0; x < a->used; x++) { - - /* get what will be the *next* carry bit from the - * MSB of the current digit - */ - rr = *tmpa >> ((mp_digit)(DIGIT_BIT - 1)); - - /* now shift up this digit, add in the carry [from the previous] */ - *tmpb++ = ((*tmpa++ << ((mp_digit)1)) | r) & MP_MASK; - - /* copy the carry that would be from the source - * digit into the next iteration - */ - r = rr; - } - - /* new leading digit? */ - if (r != 0) { - /* add a MSB which is always 1 at this point */ - *tmpb = 1; - ++(b->used); - } - - /* now zero any excess digits on the destination - * that we didn't write to - */ - tmpb = b->dp + b->used; - for (x = b->used; x < oldused; x++) { - *tmpb++ = 0; - } - } - b->sign = a->sign; - return MP_OKAY; -} -#endif - - -#ifdef BN_MP_MONTGOMERY_CALC_NORMALIZATION_C -/* - * shifts with subtractions when the result is greater than b. - * - * The method is slightly modified to shift B unconditionally up to just under - * the leading bit of b. This saves a lot of multiple precision shifting. - */ -static int mp_montgomery_calc_normalization (mp_int * a, mp_int * b) -{ - int x, bits, res; - - /* how many bits of last digit does b use */ - bits = mp_count_bits (b) % DIGIT_BIT; - - if (b->used > 1) { - if ((res = mp_2expt (a, (b->used - 1) * DIGIT_BIT + bits - 1)) != MP_OKAY) { - return res; - } - } else { - mp_set(a, 1); - bits = 1; - } - - - /* now compute C = A * B mod b */ - for (x = bits - 1; x < (int)DIGIT_BIT; x++) { - if ((res = mp_mul_2 (a, a)) != MP_OKAY) { - return res; - } - if (mp_cmp_mag (a, b) != MP_LT) { - if ((res = s_mp_sub (a, b, a)) != MP_OKAY) { - return res; - } - } - } - - return MP_OKAY; -} -#endif - - -#ifdef BN_MP_EXPTMOD_FAST_C -/* computes Y == G**X mod P, HAC pp.616, Algorithm 14.85 - * - * Uses a left-to-right k-ary sliding window to compute the modular exponentiation. - * The value of k changes based on the size of the exponent. - * - * Uses Montgomery or Diminished Radix reduction [whichever appropriate] - */ - -static int mp_exptmod_fast (mp_int * G, mp_int * X, mp_int * P, mp_int * Y, int redmode) -{ - mp_int M[TAB_SIZE], res; - mp_digit buf, mp; - int err, bitbuf, bitcpy, bitcnt, mode, digidx, x, y, winsize; - - /* use a pointer to the reduction algorithm. This allows us to use - * one of many reduction algorithms without modding the guts of - * the code with if statements everywhere. - */ - int (*redux)(mp_int*,mp_int*,mp_digit); - - /* find window size */ - x = mp_count_bits (X); - if (x <= 7) { - winsize = 2; - } else if (x <= 36) { - winsize = 3; - } else if (x <= 140) { - winsize = 4; - } else if (x <= 450) { - winsize = 5; - } else if (x <= 1303) { - winsize = 6; - } else if (x <= 3529) { - winsize = 7; - } else { - winsize = 8; - } - -#ifdef MP_LOW_MEM - if (winsize > 5) { - winsize = 5; - } -#endif - - /* init M array */ - /* init first cell */ - if ((err = mp_init(&M[1])) != MP_OKAY) { - return err; - } - - /* now init the second half of the array */ - for (x = 1<<(winsize-1); x < (1 << winsize); x++) { - if ((err = mp_init(&M[x])) != MP_OKAY) { - for (y = 1<<(winsize-1); y < x; y++) { - mp_clear (&M[y]); - } - mp_clear(&M[1]); - return err; - } - } - - /* determine and setup reduction code */ - if (redmode == 0) { -#ifdef BN_MP_MONTGOMERY_SETUP_C - /* now setup montgomery */ - if ((err = mp_montgomery_setup (P, &mp)) != MP_OKAY) { - goto LBL_M; - } -#else - err = MP_VAL; - goto LBL_M; -#endif - - /* automatically pick the comba one if available (saves quite a few calls/ifs) */ -#ifdef BN_FAST_MP_MONTGOMERY_REDUCE_C - if (((P->used * 2 + 1) < MP_WARRAY) && - P->used < (1 << ((CHAR_BIT * sizeof (mp_word)) - (2 * DIGIT_BIT)))) { - redux = fast_mp_montgomery_reduce; - } else -#endif - { -#ifdef BN_MP_MONTGOMERY_REDUCE_C - /* use slower baseline Montgomery method */ - redux = mp_montgomery_reduce; -#else - err = MP_VAL; - goto LBL_M; -#endif - } - } else if (redmode == 1) { -#if defined(BN_MP_DR_SETUP_C) && defined(BN_MP_DR_REDUCE_C) - /* setup DR reduction for moduli of the form B**k - b */ - mp_dr_setup(P, &mp); - redux = mp_dr_reduce; -#else - err = MP_VAL; - goto LBL_M; -#endif - } else { -#if defined(BN_MP_REDUCE_2K_SETUP_C) && defined(BN_MP_REDUCE_2K_C) - /* setup DR reduction for moduli of the form 2**k - b */ - if ((err = mp_reduce_2k_setup(P, &mp)) != MP_OKAY) { - goto LBL_M; - } - redux = mp_reduce_2k; -#else - err = MP_VAL; - goto LBL_M; -#endif - } - - /* setup result */ - if ((err = mp_init (&res)) != MP_OKAY) { - goto LBL_M; - } - - /* create M table - * - - * - * The first half of the table is not computed though accept for M[0] and M[1] - */ - - if (redmode == 0) { -#ifdef BN_MP_MONTGOMERY_CALC_NORMALIZATION_C - /* now we need R mod m */ - if ((err = mp_montgomery_calc_normalization (&res, P)) != MP_OKAY) { - goto LBL_RES; - } -#else - err = MP_VAL; - goto LBL_RES; -#endif - - /* now set M[1] to G * R mod m */ - if ((err = mp_mulmod (G, &res, P, &M[1])) != MP_OKAY) { - goto LBL_RES; - } - } else { - mp_set(&res, 1); - if ((err = mp_mod(G, P, &M[1])) != MP_OKAY) { - goto LBL_RES; - } - } - - /* compute the value at M[1<<(winsize-1)] by squaring M[1] (winsize-1) times */ - if ((err = mp_copy (&M[1], &M[1 << (winsize - 1)])) != MP_OKAY) { - goto LBL_RES; - } - - for (x = 0; x < (winsize - 1); x++) { - if ((err = mp_sqr (&M[1 << (winsize - 1)], &M[1 << (winsize - 1)])) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&M[1 << (winsize - 1)], P, mp)) != MP_OKAY) { - goto LBL_RES; - } - } - - /* create upper table */ - for (x = (1 << (winsize - 1)) + 1; x < (1 << winsize); x++) { - if ((err = mp_mul (&M[x - 1], &M[1], &M[x])) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&M[x], P, mp)) != MP_OKAY) { - goto LBL_RES; - } - } - - /* set initial mode and bit cnt */ - mode = 0; - bitcnt = 1; - buf = 0; - digidx = X->used - 1; - bitcpy = 0; - bitbuf = 0; - - for (;;) { - /* grab next digit as required */ - if (--bitcnt == 0) { - /* if digidx == -1 we are out of digits so break */ - if (digidx == -1) { - break; - } - /* read next digit and reset bitcnt */ - buf = X->dp[digidx--]; - bitcnt = (int)DIGIT_BIT; - } - - /* grab the next msb from the exponent */ - y = (mp_digit)(buf >> (DIGIT_BIT - 1)) & 1; - buf <<= (mp_digit)1; - - /* if the bit is zero and mode == 0 then we ignore it - * These represent the leading zero bits before the first 1 bit - * in the exponent. Technically this opt is not required but it - * does lower the # of trivial squaring/reductions used - */ - if (mode == 0 && y == 0) { - continue; - } - - /* if the bit is zero and mode == 1 then we square */ - if (mode == 1 && y == 0) { - if ((err = mp_sqr (&res, &res)) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&res, P, mp)) != MP_OKAY) { - goto LBL_RES; - } - continue; - } - - /* else we add it to the window */ - bitbuf |= (y << (winsize - ++bitcpy)); - mode = 2; - - if (bitcpy == winsize) { - /* ok window is filled so square as required and multiply */ - /* square first */ - for (x = 0; x < winsize; x++) { - if ((err = mp_sqr (&res, &res)) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&res, P, mp)) != MP_OKAY) { - goto LBL_RES; - } - } - - /* then multiply */ - if ((err = mp_mul (&res, &M[bitbuf], &res)) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&res, P, mp)) != MP_OKAY) { - goto LBL_RES; - } - - /* empty window and reset */ - bitcpy = 0; - bitbuf = 0; - mode = 1; - } - } - - /* if bits remain then square/multiply */ - if (mode == 2 && bitcpy > 0) { - /* square then multiply if the bit is set */ - for (x = 0; x < bitcpy; x++) { - if ((err = mp_sqr (&res, &res)) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&res, P, mp)) != MP_OKAY) { - goto LBL_RES; - } - - /* get next bit of the window */ - bitbuf <<= 1; - if ((bitbuf & (1 << winsize)) != 0) { - /* then multiply */ - if ((err = mp_mul (&res, &M[1], &res)) != MP_OKAY) { - goto LBL_RES; - } - if ((err = redux (&res, P, mp)) != MP_OKAY) { - goto LBL_RES; - } - } - } - } - - if (redmode == 0) { - /* fixup result if Montgomery reduction is used - * recall that any value in a Montgomery system is - * actually multiplied by R mod n. So we have - * to reduce one more time to cancel out the factor - * of R. - */ - if ((err = redux(&res, P, mp)) != MP_OKAY) { - goto LBL_RES; - } - } - - /* swap res with Y */ - mp_exch (&res, Y); - err = MP_OKAY; -LBL_RES:mp_clear (&res); -LBL_M: - mp_clear(&M[1]); - for (x = 1<<(winsize-1); x < (1 << winsize); x++) { - mp_clear (&M[x]); - } - return err; -} -#endif - - -#ifdef BN_FAST_S_MP_SQR_C -/* the jist of squaring... - * you do like mult except the offset of the tmpx [one that - * starts closer to zero] can't equal the offset of tmpy. - * So basically you set up iy like before then you min it with - * (ty-tx) so that it never happens. You double all those - * you add in the inner loop - -After that loop you do the squares and add them in. -*/ - -static int fast_s_mp_sqr (mp_int * a, mp_int * b) -{ - int olduse, res, pa, ix, iz; - mp_digit W[MP_WARRAY], *tmpx; - mp_word W1; - - /* grow the destination as required */ - pa = a->used + a->used; - if (b->alloc < pa) { - if ((res = mp_grow (b, pa)) != MP_OKAY) { - return res; - } - } - - /* number of output digits to produce */ - W1 = 0; - for (ix = 0; ix < pa; ix++) { - int tx, ty, iy; - mp_word _W; - mp_digit *tmpy; - - /* clear counter */ - _W = 0; - - /* get offsets into the two bignums */ - ty = MIN(a->used-1, ix); - tx = ix - ty; - - /* setup temp aliases */ - tmpx = a->dp + tx; - tmpy = a->dp + ty; - - /* this is the number of times the loop will iterrate, essentially - while (tx++ < a->used && ty-- >= 0) { ... } - */ - iy = MIN(a->used-tx, ty+1); - - /* now for squaring tx can never equal ty - * we halve the distance since they approach at a rate of 2x - * and we have to round because odd cases need to be executed - */ - iy = MIN(iy, (ty-tx+1)>>1); - - /* execute loop */ - for (iz = 0; iz < iy; iz++) { - _W += ((mp_word)*tmpx++)*((mp_word)*tmpy--); - } - - /* double the inner product and add carry */ - _W = _W + _W + W1; - - /* even columns have the square term in them */ - if ((ix&1) == 0) { - _W += ((mp_word)a->dp[ix>>1])*((mp_word)a->dp[ix>>1]); - } - - /* store it */ - W[ix] = (mp_digit)(_W & MP_MASK); - - /* make next carry */ - W1 = _W >> ((mp_word)DIGIT_BIT); - } - - /* setup dest */ - olduse = b->used; - b->used = a->used+a->used; - - { - mp_digit *tmpb; - tmpb = b->dp; - for (ix = 0; ix < pa; ix++) { - *tmpb++ = W[ix] & MP_MASK; - } - - /* clear unused digits [that existed in the old copy of c] */ - for (; ix < olduse; ix++) { - *tmpb++ = 0; - } - } - mp_clamp (b); - return MP_OKAY; -} -#endif - - -#ifdef BN_MP_MUL_D_C -/* multiply by a digit */ -static int -mp_mul_d (mp_int * a, mp_digit b, mp_int * c) -{ - mp_digit u, *tmpa, *tmpc; - mp_word r; - int ix, res, olduse; - - /* make sure c is big enough to hold a*b */ - if (c->alloc < a->used + 1) { - if ((res = mp_grow (c, a->used + 1)) != MP_OKAY) { - return res; - } - } - - /* get the original destinations used count */ - olduse = c->used; - - /* set the sign */ - c->sign = a->sign; - - /* alias for a->dp [source] */ - tmpa = a->dp; - - /* alias for c->dp [dest] */ - tmpc = c->dp; - - /* zero carry */ - u = 0; - - /* compute columns */ - for (ix = 0; ix < a->used; ix++) { - /* compute product and carry sum for this term */ - r = ((mp_word) u) + ((mp_word)*tmpa++) * ((mp_word)b); - - /* mask off higher bits to get a single digit */ - *tmpc++ = (mp_digit) (r & ((mp_word) MP_MASK)); - - /* send carry into next iteration */ - u = (mp_digit) (r >> ((mp_word) DIGIT_BIT)); - } - - /* store final carry [if any] and increment ix offset */ - *tmpc++ = u; - ++ix; - - /* now zero digits above the top */ - while (ix++ < olduse) { - *tmpc++ = 0; - } - - /* set used count */ - c->used = a->used + 1; - mp_clamp(c); - - return MP_OKAY; -} -#endif diff --git a/src/crypto/tfm/fp_2expt.c b/src/crypto/tfm/fp_2expt.c new file mode 100644 index 0000000..a6ed603 --- /dev/null +++ b/src/crypto/tfm/fp_2expt.c @@ -0,0 +1,39 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* computes a = 2**b */ +void fp_2expt(fp_int *a, int b) +{ + int z; + + /* zero a as per default */ + fp_zero (a); + + if (b < 0) { + return; + } + + z = b / DIGIT_BIT; + if (z >= FP_SIZE) { + return; + } + + /* set the used count of where the bit will go */ + a->used = z + 1; + + /* put the single bit in its place */ + a->dp[z] = ((fp_digit)1) << (b % DIGIT_BIT); +} + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_add.c b/src/crypto/tfm/fp_add.c new file mode 100644 index 0000000..9d1cf45 --- /dev/null +++ b/src/crypto/tfm/fp_add.c @@ -0,0 +1,43 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +void fp_add(fp_int *a, fp_int *b, fp_int *c) +{ + int sa, sb; + + /* get sign of both inputs */ + sa = a->sign; + sb = b->sign; + + /* handle two cases, not four */ + if (sa == sb) { + /* both positive or both negative */ + /* add their magnitudes, copy the sign */ + c->sign = sa; + s_fp_add (a, b, c); + } else { + /* one positive, the other negative */ + /* subtract the one with the greater magnitude from */ + /* the one of the lesser magnitude. The result gets */ + /* the sign of the one with the greater magnitude. */ + if (fp_cmp_mag (a, b) == FP_LT) { + c->sign = sb; + s_fp_sub (b, a, c); + } else { + c->sign = sa; + s_fp_sub (a, b, c); + } + } +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_cmp.c b/src/crypto/tfm/fp_cmp.c new file mode 100644 index 0000000..23f27e8 --- /dev/null +++ b/src/crypto/tfm/fp_cmp.c @@ -0,0 +1,31 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +int fp_cmp(fp_int *a, fp_int *b) +{ + if (a->sign == FP_NEG && b->sign == FP_ZPOS) { + return FP_LT; + } else if (a->sign == FP_ZPOS && b->sign == FP_NEG) { + return FP_GT; + } else { + /* compare digits */ + if (a->sign == FP_NEG) { + /* if negative compare opposite direction */ + return fp_cmp_mag(b, a); + } else { + return fp_cmp_mag(a, b); + } + } +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_cmp_d.c b/src/crypto/tfm/fp_cmp_d.c new file mode 100644 index 0000000..c2fb2ec --- /dev/null +++ b/src/crypto/tfm/fp_cmp_d.c @@ -0,0 +1,38 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* compare against a single digit */ +int fp_cmp_d(fp_int *a, fp_digit b) +{ + /* compare based on sign */ + if ((b && a->used == 0) || a->sign == FP_NEG) { + return FP_LT; + } + + /* compare based on magnitude */ + if (a->used > 1) { + return FP_GT; + } + + /* compare the only digit of a to b */ + if (a->dp[0] > b) { + return FP_GT; + } else if (a->dp[0] < b) { + return FP_LT; + } else { + return FP_EQ; + } + +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_cmp_mag.c b/src/crypto/tfm/fp_cmp_mag.c new file mode 100644 index 0000000..bdce60b --- /dev/null +++ b/src/crypto/tfm/fp_cmp_mag.c @@ -0,0 +1,35 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +int fp_cmp_mag(fp_int *a, fp_int *b) +{ + int x; + + if (a->used > b->used) { + return FP_GT; + } else if (a->used < b->used) { + return FP_LT; + } else { + for (x = a->used - 1; x >= 0; x--) { + if (a->dp[x] > b->dp[x]) { + return FP_GT; + } else if (a->dp[x] < b->dp[x]) { + return FP_LT; + } + } + } + return FP_EQ; +} + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_count_bits.c b/src/crypto/tfm/fp_count_bits.c new file mode 100644 index 0000000..4968a0f --- /dev/null +++ b/src/crypto/tfm/fp_count_bits.c @@ -0,0 +1,36 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +int fp_count_bits (fp_int * a) +{ + int r; + fp_digit q; + + /* shortcut */ + if (a->used == 0) { + return 0; + } + + /* get number of digits and add that */ + r = (a->used - 1) * DIGIT_BIT; + + /* take the last digit and count the bits in it */ + q = a->dp[a->used - 1]; + while (q > ((fp_digit) 0)) { + ++r; + q >>= ((fp_digit) 1); + } + return r; +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_div.c b/src/crypto/tfm/fp_div.c new file mode 100644 index 0000000..4f6dd47 --- /dev/null +++ b/src/crypto/tfm/fp_div.c @@ -0,0 +1,157 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* a/b => cb + d == a */ +int fp_div(fp_int *a, fp_int *b, fp_int *c, fp_int *d) +{ + fp_int q, x, y, t1, t2; + int n, t, i, norm, neg; + + /* is divisor zero ? */ + if (fp_iszero (b) == 1) { + return FP_VAL; + } + + /* if a < b then q=0, r = a */ + if (fp_cmp_mag (a, b) == FP_LT) { + if (d != NULL) { + fp_copy (a, d); + } + if (c != NULL) { + fp_zero (c); + } + return FP_OKAY; + } + + fp_init(&q); + q.used = a->used + 2; + + fp_init(&t1); + fp_init(&t2); + fp_init_copy(&x, a); + fp_init_copy(&y, b); + + /* fix the sign */ + neg = (a->sign == b->sign) ? FP_ZPOS : FP_NEG; + x.sign = y.sign = FP_ZPOS; + + /* normalize both x and y, ensure that y >= b/2, [b == 2**DIGIT_BIT] */ + norm = fp_count_bits(&y) % DIGIT_BIT; + if (norm < (int)(DIGIT_BIT-1)) { + norm = (DIGIT_BIT-1) - norm; + fp_mul_2d (&x, norm, &x); + fp_mul_2d (&y, norm, &y); + } else { + norm = 0; + } + + /* note hac does 0 based, so if used==5 then its 0,1,2,3,4, e.g. use 4 */ + n = x.used - 1; + t = y.used - 1; + + /* while (x >= y*b**n-t) do { q[n-t] += 1; x -= y*b**{n-t} } */ + fp_lshd (&y, n - t); /* y = y*b**{n-t} */ + + while (fp_cmp (&x, &y) != FP_LT) { + ++(q.dp[n - t]); + fp_sub (&x, &y, &x); + } + + /* reset y by shifting it back down */ + fp_rshd (&y, n - t); + + /* step 3. for i from n down to (t + 1) */ + for (i = n; i >= (t + 1); i--) { + if (i > x.used) { + continue; + } + + /* step 3.1 if xi == yt then set q{i-t-1} to b-1, + * otherwise set q{i-t-1} to (xi*b + x{i-1})/yt */ + if (x.dp[i] == y.dp[t]) { + q.dp[i - t - 1] = ((((fp_word)1) << DIGIT_BIT) - 1); + } else { + fp_word tmp; + tmp = ((fp_word) x.dp[i]) << ((fp_word) DIGIT_BIT); + tmp |= ((fp_word) x.dp[i - 1]); + tmp /= ((fp_word) y.dp[t]); + q.dp[i - t - 1] = (fp_digit) (tmp); + } + + /* while (q{i-t-1} * (yt * b + y{t-1})) > + xi * b**2 + xi-1 * b + xi-2 + + do q{i-t-1} -= 1; + */ + q.dp[i - t - 1] = (q.dp[i - t - 1] + 1); + do { + q.dp[i - t - 1] = (q.dp[i - t - 1] - 1); + + /* find left hand */ + fp_zero (&t1); + t1.dp[0] = (t - 1 < 0) ? 0 : y.dp[t - 1]; + t1.dp[1] = y.dp[t]; + t1.used = 2; + fp_mul_d (&t1, q.dp[i - t - 1], &t1); + + /* find right hand */ + t2.dp[0] = (i - 2 < 0) ? 0 : x.dp[i - 2]; + t2.dp[1] = (i - 1 < 0) ? 0 : x.dp[i - 1]; + t2.dp[2] = x.dp[i]; + t2.used = 3; + } while (fp_cmp_mag(&t1, &t2) == FP_GT); + + /* step 3.3 x = x - q{i-t-1} * y * b**{i-t-1} */ + fp_mul_d (&y, q.dp[i - t - 1], &t1); + fp_lshd (&t1, i - t - 1); + fp_sub (&x, &t1, &x); + + /* if x < 0 then { x = x + y*b**{i-t-1}; q{i-t-1} -= 1; } */ + if (x.sign == FP_NEG) { + fp_copy (&y, &t1); + fp_lshd (&t1, i - t - 1); + fp_add (&x, &t1, &x); + q.dp[i - t - 1] = q.dp[i - t - 1] - 1; + } + } + + /* now q is the quotient and x is the remainder + * [which we have to normalize] + */ + + /* get sign before writing to c */ + x.sign = x.used == 0 ? FP_ZPOS : a->sign; + + if (c != NULL) { + fp_clamp (&q); + fp_copy (&q, c); + c->sign = neg; + } + + if (d != NULL) { + fp_div_2d (&x, norm, &x, NULL); + +/* the following is a kludge, essentially we were seeing the right remainder but + with excess digits that should have been zero + */ + for (i = b->used; i < x.used; i++) { + x.dp[i] = 0; + } + fp_clamp(&x); + fp_copy (&x, d); + } + + return FP_OKAY; +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_div_2.c b/src/crypto/tfm/fp_div_2.c new file mode 100644 index 0000000..4ddc4b5 --- /dev/null +++ b/src/crypto/tfm/fp_div_2.c @@ -0,0 +1,53 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* b = a/2 */ +void fp_div_2(fp_int * a, fp_int * b) +{ + int x, oldused; + + oldused = b->used; + b->used = a->used; + { + register fp_digit r, rr, *tmpa, *tmpb; + + /* source alias */ + tmpa = a->dp + b->used - 1; + + /* dest alias */ + tmpb = b->dp + b->used - 1; + + /* carry */ + r = 0; + for (x = b->used - 1; x >= 0; x--) { + /* get the carry for the next iteration */ + rr = *tmpa & 1; + + /* shift the current digit, add in carry and store */ + *tmpb-- = (*tmpa-- >> 1) | (r << (DIGIT_BIT - 1)); + + /* forward carry to next iteration */ + r = rr; + } + + /* zero excess digits */ + tmpb = b->dp + b->used; + for (x = b->used; x < oldused; x++) { + *tmpb++ = 0; + } + } + b->sign = a->sign; + fp_clamp (b); +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_div_2d.c b/src/crypto/tfm/fp_div_2d.c new file mode 100644 index 0000000..e4a3edd --- /dev/null +++ b/src/crypto/tfm/fp_div_2d.c @@ -0,0 +1,79 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* c = a / 2**b */ +void fp_div_2d(fp_int *a, int b, fp_int *c, fp_int *d) +{ + fp_digit D, r, rr; + int x; + fp_int t; + + /* if the shift count is <= 0 then we do no work */ + if (b <= 0) { + fp_copy (a, c); + if (d != NULL) { + fp_zero (d); + } + return; + } + + fp_init(&t); + + /* get the remainder */ + if (d != NULL) { + fp_mod_2d (a, b, &t); + } + + /* copy */ + fp_copy(a, c); + + /* shift by as many digits in the bit count */ + if (b >= (int)DIGIT_BIT) { + fp_rshd (c, b / DIGIT_BIT); + } + + /* shift any bit count < DIGIT_BIT */ + D = (fp_digit) (b % DIGIT_BIT); + if (D != 0) { + register fp_digit *tmpc, mask, shift; + + /* mask */ + mask = (((fp_digit)1) << D) - 1; + + /* shift for lsb */ + shift = DIGIT_BIT - D; + + /* alias */ + tmpc = c->dp + (c->used - 1); + + /* carry */ + r = 0; + for (x = c->used - 1; x >= 0; x--) { + /* get the lower bits of this word in a temp */ + rr = *tmpc & mask; + + /* shift the current word and mix in the carry bits from the previous word */ + *tmpc = (*tmpc >> D) | (r << shift); + --tmpc; + + /* set the carry to the carry bits of the current word found above */ + r = rr; + } + } + fp_clamp (c); + if (d != NULL) { + fp_copy (&t, d); + } +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_exptmod.c b/src/crypto/tfm/fp_exptmod.c new file mode 100644 index 0000000..418a931 --- /dev/null +++ b/src/crypto/tfm/fp_exptmod.c @@ -0,0 +1,276 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +#ifdef TFM_TIMING_RESISTANT + +/* timing resistant montgomery ladder based exptmod + + Based on work by Marc Joye, Sung-Ming Yen, "The Montgomery Powering Ladder", Cryptographic Hardware and Embedded Systems, CHES 2002 +*/ +static int _fp_exptmod(fp_int * G, fp_int * X, fp_int * P, fp_int * Y) +{ + fp_int R[2]; + fp_digit buf, mp; + int err, bitcnt, digidx, y; + + /* now setup montgomery */ + if ((err = fp_montgomery_setup (P, &mp)) != FP_OKAY) { + return err; + } + + fp_init(&R[0]); + fp_init(&R[1]); + + /* now we need R mod m */ + fp_montgomery_calc_normalization (&R[0], P); + + /* now set R[0][1] to G * R mod m */ + if (fp_cmp_mag(P, G) != FP_GT) { + /* G > P so we reduce it first */ + fp_mod(G, P, &R[1]); + } else { + fp_copy(G, &R[1]); + } + fp_mulmod (&R[1], &R[0], P, &R[1]); + + /* for j = t-1 downto 0 do + r_!k = R0*R1; r_k = r_k^2 + */ + + /* set initial mode and bit cnt */ + bitcnt = 1; + buf = 0; + digidx = X->used - 1; + + for (;;) { + /* grab next digit as required */ + if (--bitcnt == 0) { + /* if digidx == -1 we are out of digits so break */ + if (digidx == -1) { + break; + } + /* read next digit and reset bitcnt */ + buf = X->dp[digidx--]; + bitcnt = (int)DIGIT_BIT; + } + + /* grab the next msb from the exponent */ + y = (fp_digit)(buf >> (DIGIT_BIT - 1)) & 1; + buf <<= (fp_digit)1; + + /* do ops */ + fp_mul(&R[0], &R[1], &R[y^1]); fp_montgomery_reduce(&R[y^1], P, mp); + fp_sqr(&R[y], &R[y]); fp_montgomery_reduce(&R[y], P, mp); + } + + fp_montgomery_reduce(&R[0], P, mp); + fp_copy(&R[0], Y); + return FP_OKAY; +} + +#else + +/* y = g**x (mod b) + * Some restrictions... x must be positive and < b + */ +static int _fp_exptmod(fp_int * G, fp_int * X, fp_int * P, fp_int * Y) +{ + fp_int M[64], res; + fp_digit buf, mp; + int err, bitbuf, bitcpy, bitcnt, mode, digidx, x, y, winsize; + + /* find window size */ + x = fp_count_bits (X); + if (x <= 21) { + winsize = 1; + } else if (x <= 36) { + winsize = 3; + } else if (x <= 140) { + winsize = 4; + } else if (x <= 450) { + winsize = 5; + } else { + winsize = 6; + } + + /* init M array */ + memset(M, 0, sizeof(M)); + + /* now setup montgomery */ + if ((err = fp_montgomery_setup (P, &mp)) != FP_OKAY) { + return err; + } + + /* setup result */ + fp_init(&res); + + /* create M table + * + * The M table contains powers of the input base, e.g. M[x] = G^x mod P + * + * The first half of the table is not computed though accept for M[0] and M[1] + */ + + /* now we need R mod m */ + fp_montgomery_calc_normalization (&res, P); + + /* now set M[1] to G * R mod m */ + if (fp_cmp_mag(P, G) != FP_GT) { + /* G > P so we reduce it first */ + fp_mod(G, P, &M[1]); + } else { + fp_copy(G, &M[1]); + } + fp_mulmod (&M[1], &res, P, &M[1]); + + /* compute the value at M[1<<(winsize-1)] by squaring M[1] (winsize-1) times */ + fp_copy (&M[1], &M[1 << (winsize - 1)]); + for (x = 0; x < (winsize - 1); x++) { + fp_sqr (&M[1 << (winsize - 1)], &M[1 << (winsize - 1)]); + fp_montgomery_reduce (&M[1 << (winsize - 1)], P, mp); + } + + /* create upper table */ + for (x = (1 << (winsize - 1)) + 1; x < (1 << winsize); x++) { + fp_mul(&M[x - 1], &M[1], &M[x]); + fp_montgomery_reduce(&M[x], P, mp); + } + + /* set initial mode and bit cnt */ + mode = 0; + bitcnt = 1; + buf = 0; + digidx = X->used - 1; + bitcpy = 0; + bitbuf = 0; + + for (;;) { + /* grab next digit as required */ + if (--bitcnt == 0) { + /* if digidx == -1 we are out of digits so break */ + if (digidx == -1) { + break; + } + /* read next digit and reset bitcnt */ + buf = X->dp[digidx--]; + bitcnt = (int)DIGIT_BIT; + } + + /* grab the next msb from the exponent */ + y = (fp_digit)(buf >> (DIGIT_BIT - 1)) & 1; + buf <<= (fp_digit)1; + + /* if the bit is zero and mode == 0 then we ignore it + * These represent the leading zero bits before the first 1 bit + * in the exponent. Technically this opt is not required but it + * does lower the # of trivial squaring/reductions used + */ + if (mode == 0 && y == 0) { + continue; + } + + /* if the bit is zero and mode == 1 then we square */ + if (mode == 1 && y == 0) { + fp_sqr(&res, &res); + fp_montgomery_reduce(&res, P, mp); + continue; + } + + /* else we add it to the window */ + bitbuf |= (y << (winsize - ++bitcpy)); + mode = 2; + + if (bitcpy == winsize) { + /* ok window is filled so square as required and multiply */ + /* square first */ + for (x = 0; x < winsize; x++) { + fp_sqr(&res, &res); + fp_montgomery_reduce(&res, P, mp); + } + + /* then multiply */ + fp_mul(&res, &M[bitbuf], &res); + fp_montgomery_reduce(&res, P, mp); + + /* empty window and reset */ + bitcpy = 0; + bitbuf = 0; + mode = 1; + } + } + + /* if bits remain then square/multiply */ + if (mode == 2 && bitcpy > 0) { + /* square then multiply if the bit is set */ + for (x = 0; x < bitcpy; x++) { + fp_sqr(&res, &res); + fp_montgomery_reduce(&res, P, mp); + + /* get next bit of the window */ + bitbuf <<= 1; + if ((bitbuf & (1 << winsize)) != 0) { + /* then multiply */ + fp_mul(&res, &M[1], &res); + fp_montgomery_reduce(&res, P, mp); + } + } + } + + /* fixup result if Montgomery reduction is used + * recall that any value in a Montgomery system is + * actually multiplied by R mod n. So we have + * to reduce one more time to cancel out the factor + * of R. + */ + fp_montgomery_reduce(&res, P, mp); + + /* swap res with Y */ + fp_copy (&res, Y); + return FP_OKAY; +} + +#endif + + +int fp_exptmod(fp_int * G, fp_int * X, fp_int * P, fp_int * Y) +{ + fp_int tmp; + int err; + +#ifdef TFM_CHECK + /* prevent overflows */ + if (P->used > (FP_SIZE/2)) { + return FP_VAL; + } +#endif + + /* is X negative? */ + if (X->sign == FP_NEG) { + /* yes, copy G and invmod it */ + fp_copy(G, &tmp); + if ((err = fp_invmod(&tmp, P, &tmp)) != FP_OKAY) { + return err; + } + X->sign = FP_ZPOS; + err = _fp_exptmod(&tmp, X, P, Y); + if (X != Y) { + X->sign = FP_NEG; + } + return err; + } else { + /* Positive exponent so just exptmod */ + return _fp_exptmod(G, X, P, Y); + } +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_invmod.c b/src/crypto/tfm/fp_invmod.c new file mode 100644 index 0000000..9a2329e --- /dev/null +++ b/src/crypto/tfm/fp_invmod.c @@ -0,0 +1,207 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +static int fp_invmod_slow (fp_int * a, fp_int * b, fp_int * c) +{ + fp_int x, y, u, v, A, B, C, D; + int res; + + /* b cannot be negative */ + if (b->sign == FP_NEG || fp_iszero(b) == 1) { + return FP_VAL; + } + + /* init temps */ + fp_init(&x); fp_init(&y); + fp_init(&u); fp_init(&v); + fp_init(&A); fp_init(&B); + fp_init(&C); fp_init(&D); + + /* x = a, y = b */ + if ((res = fp_mod(a, b, &x)) != FP_OKAY) { + return res; + } + fp_copy(b, &y); + + /* 2. [modified] if x,y are both even then return an error! */ + if (fp_iseven (&x) == 1 && fp_iseven (&y) == 1) { + return FP_VAL; + } + + /* 3. u=x, v=y, A=1, B=0, C=0,D=1 */ + fp_copy (&x, &u); + fp_copy (&y, &v); + fp_set (&A, 1); + fp_set (&D, 1); + +top: + /* 4. while u is even do */ + while (fp_iseven (&u) == 1) { + /* 4.1 u = u/2 */ + fp_div_2 (&u, &u); + + /* 4.2 if A or B is odd then */ + if (fp_isodd (&A) == 1 || fp_isodd (&B) == 1) { + /* A = (A+y)/2, B = (B-x)/2 */ + fp_add (&A, &y, &A); + fp_sub (&B, &x, &B); + } + /* A = A/2, B = B/2 */ + fp_div_2 (&A, &A); + fp_div_2 (&B, &B); + } + + /* 5. while v is even do */ + while (fp_iseven (&v) == 1) { + /* 5.1 v = v/2 */ + fp_div_2 (&v, &v); + + /* 5.2 if C or D is odd then */ + if (fp_isodd (&C) == 1 || fp_isodd (&D) == 1) { + /* C = (C+y)/2, D = (D-x)/2 */ + fp_add (&C, &y, &C); + fp_sub (&D, &x, &D); + } + /* C = C/2, D = D/2 */ + fp_div_2 (&C, &C); + fp_div_2 (&D, &D); + } + + /* 6. if u >= v then */ + if (fp_cmp (&u, &v) != FP_LT) { + /* u = u - v, A = A - C, B = B - D */ + fp_sub (&u, &v, &u); + fp_sub (&A, &C, &A); + fp_sub (&B, &D, &B); + } else { + /* v - v - u, C = C - A, D = D - B */ + fp_sub (&v, &u, &v); + fp_sub (&C, &A, &C); + fp_sub (&D, &B, &D); + } + + /* if not zero goto step 4 */ + if (fp_iszero (&u) == 0) + goto top; + + /* now a = C, b = D, gcd == g*v */ + + /* if v != 1 then there is no inverse */ + if (fp_cmp_d (&v, 1) != FP_EQ) { + return FP_VAL; + } + + /* if its too low */ + while (fp_cmp_d(&C, 0) == FP_LT) { + fp_add(&C, b, &C); + } + + /* too big */ + while (fp_cmp_mag(&C, b) != FP_LT) { + fp_sub(&C, b, &C); + } + + /* C is now the inverse */ + fp_copy(&C, c); + return FP_OKAY; +} + +/* c = 1/a (mod b) for odd b only */ +int fp_invmod(fp_int *a, fp_int *b, fp_int *c) +{ + fp_int x, y, u, v, B, D; + int neg; + + /* 2. [modified] b must be odd */ + if (fp_iseven (b) == FP_YES) { + return fp_invmod_slow(a,b,c); + } + + /* init all our temps */ + fp_init(&x); fp_init(&y); + fp_init(&u); fp_init(&v); + fp_init(&B); fp_init(&D); + + /* x == modulus, y == value to invert */ + fp_copy(b, &x); + + /* we need y = |a| */ + fp_abs(a, &y); + + /* 3. u=x, v=y, A=1, B=0, C=0,D=1 */ + fp_copy(&x, &u); + fp_copy(&y, &v); + fp_set (&D, 1); + +top: + /* 4. while u is even do */ + while (fp_iseven (&u) == FP_YES) { + /* 4.1 u = u/2 */ + fp_div_2 (&u, &u); + + /* 4.2 if B is odd then */ + if (fp_isodd (&B) == FP_YES) { + fp_sub (&B, &x, &B); + } + /* B = B/2 */ + fp_div_2 (&B, &B); + } + + /* 5. while v is even do */ + while (fp_iseven (&v) == FP_YES) { + /* 5.1 v = v/2 */ + fp_div_2 (&v, &v); + + /* 5.2 if D is odd then */ + if (fp_isodd (&D) == FP_YES) { + /* D = (D-x)/2 */ + fp_sub (&D, &x, &D); + } + /* D = D/2 */ + fp_div_2 (&D, &D); + } + + /* 6. if u >= v then */ + if (fp_cmp (&u, &v) != FP_LT) { + /* u = u - v, B = B - D */ + fp_sub (&u, &v, &u); + fp_sub (&B, &D, &B); + } else { + /* v - v - u, D = D - B */ + fp_sub (&v, &u, &v); + fp_sub (&D, &B, &D); + } + + /* if not zero goto step 4 */ + if (fp_iszero (&u) == FP_NO) { + goto top; + } + + /* now a = C, b = D, gcd == g*v */ + + /* if v != 1 then there is no inverse */ + if (fp_cmp_d (&v, 1) != FP_EQ) { + return FP_VAL; + } + + /* b is now the inverse */ + neg = a->sign; + while (D.sign == FP_NEG) { + fp_add (&D, b, &D); + } + fp_copy (&D, c); + c->sign = neg; + return FP_OKAY; +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_lshd.c b/src/crypto/tfm/fp_lshd.c new file mode 100644 index 0000000..0933655 --- /dev/null +++ b/src/crypto/tfm/fp_lshd.c @@ -0,0 +1,38 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +void fp_lshd(fp_int *a, int x) +{ + int y; + + /* move up and truncate as required */ + y = MIN(a->used + x - 1, (int)(FP_SIZE-1)); + + /* store new size */ + a->used = y + 1; + + /* move digits */ + for (; y >= x; y--) { + a->dp[y] = a->dp[y-x]; + } + + /* zero lower digits */ + for (; y >= 0; y--) { + a->dp[y] = 0; + } + + /* clamp digits */ + fp_clamp(a); +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mod.c b/src/crypto/tfm/fp_mod.c new file mode 100644 index 0000000..147c0d7 --- /dev/null +++ b/src/crypto/tfm/fp_mod.c @@ -0,0 +1,34 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* c = a mod b, 0 <= c < b */ +int fp_mod(fp_int *a, fp_int *b, fp_int *c) +{ + fp_int t; + int err; + + fp_zero(&t); + if ((err = fp_div(a, b, NULL, &t)) != FP_OKAY) { + return err; + } + if (t.sign != b->sign) { + fp_add(&t, b, c); + } else { + fp_copy(&t, c); + } + return FP_OKAY; +} + + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mod_2d.c b/src/crypto/tfm/fp_mod_2d.c new file mode 100644 index 0000000..d2818f0 --- /dev/null +++ b/src/crypto/tfm/fp_mod_2d.c @@ -0,0 +1,42 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* c = a mod 2**d */ +void fp_mod_2d(fp_int *a, int b, fp_int *c) +{ + int x; + + /* zero if count less than or equal to zero */ + if (b <= 0) { + fp_zero(c); + return; + } + + /* get copy of input */ + fp_copy(a, c); + + /* if 2**d is larger than we just return */ + if (b >= (DIGIT_BIT * a->used)) { + return; + } + + /* zero digits above the last digit of the modulus */ + for (x = (b / DIGIT_BIT) + ((b % DIGIT_BIT) == 0 ? 0 : 1); x < c->used; x++) { + c->dp[x] = 0; + } + /* clear the digit that is not completely outside/inside the modulus */ + c->dp[b / DIGIT_BIT] &= ~((fp_digit)0) >> (DIGIT_BIT - b); + fp_clamp (c); +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_montgomery_calc_normalization.c b/src/crypto/tfm/fp_montgomery_calc_normalization.c new file mode 100644 index 0000000..c21c8d0 --- /dev/null +++ b/src/crypto/tfm/fp_montgomery_calc_normalization.c @@ -0,0 +1,43 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* computes a = B**n mod b without division or multiplication useful for + * normalizing numbers in a Montgomery system. + */ +void fp_montgomery_calc_normalization(fp_int *a, fp_int *b) +{ + int x, bits; + + /* how many bits of last digit does b use */ + bits = fp_count_bits (b) % DIGIT_BIT; + if (!bits) bits = DIGIT_BIT; + + /* compute A = B^(n-1) * 2^(bits-1) */ + if (b->used > 1) { + fp_2expt (a, (b->used - 1) * DIGIT_BIT + bits - 1); + } else { + fp_set(a, 1); + bits = 1; + } + + /* now compute C = A * B mod b */ + for (x = bits - 1; x < (int)DIGIT_BIT; x++) { + fp_mul_2 (a, a); + if (fp_cmp_mag (a, b) != FP_LT) { + s_fp_sub (a, b, a); + } + } +} + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_montgomery_reduce.c b/src/crypto/tfm/fp_montgomery_reduce.c new file mode 100644 index 0000000..7222e44 --- /dev/null +++ b/src/crypto/tfm/fp_montgomery_reduce.c @@ -0,0 +1,552 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/******************************************************************/ +#if defined(TFM_X86) && !defined(TFM_SSE2) +/* x86-32 code */ + +#define MONT_START +#define MONT_FINI +#define LOOP_END +#define LOOP_START \ + mu = c[x] * mp + +#define INNERMUL \ +asm( \ + "movl %5,%%eax \n\t" \ + "mull %4 \n\t" \ + "addl %1,%%eax \n\t" \ + "adcl $0,%%edx \n\t" \ + "addl %%eax,%0 \n\t" \ + "adcl $0,%%edx \n\t" \ + "movl %%edx,%1 \n\t" \ +:"=g"(_c[LO]), "=r"(cy) \ +:"0"(_c[LO]), "1"(cy), "r"(mu), "r"(*tmpm++) \ +: "%eax", "%edx", "cc") + +#define PROPCARRY \ +asm( \ + "addl %1,%0 \n\t" \ + "setb %%al \n\t" \ + "movzbl %%al,%1 \n\t" \ +:"=g"(_c[LO]), "=r"(cy) \ +:"0"(_c[LO]), "1"(cy) \ +: "%eax", "cc") + +/******************************************************************/ +#elif defined(TFM_X86_64) +/* x86-64 code */ + +#define MONT_START +#define MONT_FINI +#define LOOP_END +#define LOOP_START \ + mu = c[x] * mp + +#define INNERMUL \ +asm( \ + "movq %5,%%rax \n\t" \ + "mulq %4 \n\t" \ + "addq %1,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "addq %%rax,%0 \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq %%rdx,%1 \n\t" \ +:"=g"(_c[LO]), "=r"(cy) \ +:"0"(_c[LO]), "1"(cy), "r"(mu), "r"(*tmpm++) \ +: "%rax", "%rdx", "cc") + +#define INNERMUL8 \ + asm( \ + "movq 0(%5),%%rax \n\t" \ + "movq 0(%2),%%r10 \n\t" \ + "movq 0x8(%5),%%r11 \n\t" \ + "mulq %4 \n\t" \ + "addq %%r10,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq 0x8(%2),%%r10 \n\t" \ + "addq %3,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq %%rax,0(%0) \n\t" \ + "movq %%rdx,%1 \n\t" \ + \ + "movq %%r11,%%rax \n\t" \ + "movq 0x10(%5),%%r11 \n\t" \ + "mulq %4 \n\t" \ + "addq %%r10,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq 0x10(%2),%%r10 \n\t" \ + "addq %3,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq %%rax,0x8(%0) \n\t" \ + "movq %%rdx,%1 \n\t" \ + \ + "movq %%r11,%%rax \n\t" \ + "movq 0x18(%5),%%r11 \n\t" \ + "mulq %4 \n\t" \ + "addq %%r10,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq 0x18(%2),%%r10 \n\t" \ + "addq %3,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq %%rax,0x10(%0) \n\t" \ + "movq %%rdx,%1 \n\t" \ + \ + "movq %%r11,%%rax \n\t" \ + "movq 0x20(%5),%%r11 \n\t" \ + "mulq %4 \n\t" \ + "addq %%r10,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq 0x20(%2),%%r10 \n\t" \ + "addq %3,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq %%rax,0x18(%0) \n\t" \ + "movq %%rdx,%1 \n\t" \ + \ + "movq %%r11,%%rax \n\t" \ + "movq 0x28(%5),%%r11 \n\t" \ + "mulq %4 \n\t" \ + "addq %%r10,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq 0x28(%2),%%r10 \n\t" \ + "addq %3,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq %%rax,0x20(%0) \n\t" \ + "movq %%rdx,%1 \n\t" \ + \ + "movq %%r11,%%rax \n\t" \ + "movq 0x30(%5),%%r11 \n\t" \ + "mulq %4 \n\t" \ + "addq %%r10,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq 0x30(%2),%%r10 \n\t" \ + "addq %3,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq %%rax,0x28(%0) \n\t" \ + "movq %%rdx,%1 \n\t" \ + \ + "movq %%r11,%%rax \n\t" \ + "movq 0x38(%5),%%r11 \n\t" \ + "mulq %4 \n\t" \ + "addq %%r10,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq 0x38(%2),%%r10 \n\t" \ + "addq %3,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq %%rax,0x30(%0) \n\t" \ + "movq %%rdx,%1 \n\t" \ + \ + "movq %%r11,%%rax \n\t" \ + "mulq %4 \n\t" \ + "addq %%r10,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "addq %3,%%rax \n\t" \ + "adcq $0,%%rdx \n\t" \ + "movq %%rax,0x38(%0) \n\t" \ + "movq %%rdx,%1 \n\t" \ + \ +:"=r"(_c), "=r"(cy) \ +: "0"(_c), "1"(cy), "g"(mu), "r"(tmpm)\ +: "%rax", "%rdx", "%r10", "%r11", "cc") + + +#define PROPCARRY \ +asm( \ + "addq %1,%0 \n\t" \ + "setb %%al \n\t" \ + "movzbq %%al,%1 \n\t" \ +:"=g"(_c[LO]), "=r"(cy) \ +:"0"(_c[LO]), "1"(cy) \ +: "%rax", "cc") + +/******************************************************************/ +#elif defined(TFM_SSE2) +/* SSE2 code (assumes 32-bit fp_digits) */ +/* XMM register assignments: + * xmm0 *tmpm++, then Mu * (*tmpm++) + * xmm1 c[x], then Mu + * xmm2 mp + * xmm3 cy + * xmm4 _c[LO] + */ + +#define MONT_START \ + asm("movd %0,%%mm2"::"g"(mp)) + +#define MONT_FINI \ + asm("emms") + +#define LOOP_START \ +asm( \ +"movd %0,%%mm1 \n\t" \ +"pxor %%mm3,%%mm3 \n\t" \ +"pmuludq %%mm2,%%mm1 \n\t" \ +:: "g"(c[x])) + +/* pmuludq on mmx registers does a 32x32->64 multiply. */ +#define INNERMUL \ +asm( \ + "movd %1,%%mm4 \n\t" \ + "movd %2,%%mm0 \n\t" \ + "paddq %%mm4,%%mm3 \n\t" \ + "pmuludq %%mm1,%%mm0 \n\t" \ + "paddq %%mm0,%%mm3 \n\t" \ + "movd %%mm3,%0 \n\t" \ + "psrlq $32, %%mm3 \n\t" \ +:"=g"(_c[LO]) : "0"(_c[LO]), "g"(*tmpm++) ); + +#define INNERMUL8 \ +asm( \ + "movd 0(%1),%%mm4 \n\t" \ + "movd 0(%2),%%mm0 \n\t" \ + "paddq %%mm4,%%mm3 \n\t" \ + "pmuludq %%mm1,%%mm0 \n\t" \ + "movd 4(%2),%%mm5 \n\t" \ + "paddq %%mm0,%%mm3 \n\t" \ + "movd 4(%1),%%mm6 \n\t" \ + "movd %%mm3,0(%0) \n\t" \ + "psrlq $32, %%mm3 \n\t" \ +\ + "paddq %%mm6,%%mm3 \n\t" \ + "pmuludq %%mm1,%%mm5 \n\t" \ + "movd 8(%2),%%mm6 \n\t" \ + "paddq %%mm5,%%mm3 \n\t" \ + "movd 8(%1),%%mm7 \n\t" \ + "movd %%mm3,4(%0) \n\t" \ + "psrlq $32, %%mm3 \n\t" \ +\ + "paddq %%mm7,%%mm3 \n\t" \ + "pmuludq %%mm1,%%mm6 \n\t" \ + "movd 12(%2),%%mm7 \n\t" \ + "paddq %%mm6,%%mm3 \n\t" \ + "movd 12(%1),%%mm5 \n\t" \ + "movd %%mm3,8(%0) \n\t" \ + "psrlq $32, %%mm3 \n\t" \ +\ + "paddq %%mm5,%%mm3 \n\t" \ + "pmuludq %%mm1,%%mm7 \n\t" \ + "movd 16(%2),%%mm5 \n\t" \ + "paddq %%mm7,%%mm3 \n\t" \ + "movd 16(%1),%%mm6 \n\t" \ + "movd %%mm3,12(%0) \n\t" \ + "psrlq $32, %%mm3 \n\t" \ +\ + "paddq %%mm6,%%mm3 \n\t" \ + "pmuludq %%mm1,%%mm5 \n\t" \ + "movd 20(%2),%%mm6 \n\t" \ + "paddq %%mm5,%%mm3 \n\t" \ + "movd 20(%1),%%mm7 \n\t" \ + "movd %%mm3,16(%0) \n\t" \ + "psrlq $32, %%mm3 \n\t" \ +\ + "paddq %%mm7,%%mm3 \n\t" \ + "pmuludq %%mm1,%%mm6 \n\t" \ + "movd 24(%2),%%mm7 \n\t" \ + "paddq %%mm6,%%mm3 \n\t" \ + "movd 24(%1),%%mm5 \n\t" \ + "movd %%mm3,20(%0) \n\t" \ + "psrlq $32, %%mm3 \n\t" \ +\ + "paddq %%mm5,%%mm3 \n\t" \ + "pmuludq %%mm1,%%mm7 \n\t" \ + "movd 28(%2),%%mm5 \n\t" \ + "paddq %%mm7,%%mm3 \n\t" \ + "movd 28(%1),%%mm6 \n\t" \ + "movd %%mm3,24(%0) \n\t" \ + "psrlq $32, %%mm3 \n\t" \ +\ + "paddq %%mm6,%%mm3 \n\t" \ + "pmuludq %%mm1,%%mm5 \n\t" \ + "paddq %%mm5,%%mm3 \n\t" \ + "movd %%mm3,28(%0) \n\t" \ + "psrlq $32, %%mm3 \n\t" \ +:"=r"(_c) : "0"(_c), "g"(tmpm) ); + +#define LOOP_END \ +asm( "movd %%mm3,%0 \n" :"=r"(cy)) + +#define PROPCARRY \ +asm( \ + "addl %1,%0 \n\t" \ + "setb %%al \n\t" \ + "movzbl %%al,%1 \n\t" \ +:"=g"(_c[LO]), "=r"(cy) \ +:"0"(_c[LO]), "1"(cy) \ +: "%eax", "cc") + +/******************************************************************/ +#elif defined(TFM_ARM) + /* ARMv4 code */ + +#define MONT_START +#define MONT_FINI +#define LOOP_END +#define LOOP_START \ + mu = c[x] * mp + +#define INNERMUL \ +asm( \ + " LDR r0,%1 \n\t" \ + " ADDS r0,r0,%0 \n\t" \ + " MOVCS %0,#1 \n\t" \ + " MOVCC %0,#0 \n\t" \ + " UMLAL r0,%0,%3,%4 \n\t" \ + " STR r0,%1 \n\t" \ +:"=r"(cy),"=m"(_c[0]):"0"(cy),"r"(mu),"r"(*tmpm++),"1"(_c[0]):"r0","cc"); + +#define PROPCARRY \ +asm( \ + " LDR r0,%1 \n\t" \ + " ADDS r0,r0,%0 \n\t" \ + " STR r0,%1 \n\t" \ + " MOVCS %0,#1 \n\t" \ + " MOVCC %0,#0 \n\t" \ +:"=r"(cy),"=m"(_c[0]):"0"(cy),"1"(_c[0]):"r0","cc"); + +/******************************************************************/ +#elif defined(TFM_PPC32) + +/* PPC32 */ +#define MONT_START +#define MONT_FINI +#define LOOP_END +#define LOOP_START \ + mu = c[x] * mp + +#define INNERMUL \ +asm( \ + " mullw 16,%3,%4 \n\t" \ + " mulhwu 17,%3,%4 \n\t" \ + " addc 16,16,%2 \n\t" \ + " addze 17,17 \n\t" \ + " addc %1,16,%5 \n\t" \ + " addze %0,17 \n\t" \ +:"=r"(cy),"=r"(_c[0]):"0"(cy),"r"(mu),"r"(tmpm[0]),"1"(_c[0]):"16", "17", "cc"); ++tmpm; + +#define PROPCARRY \ +asm( \ + " addc %1,%3,%2 \n\t" \ + " xor %0,%2,%2 \n\t" \ + " addze %0,%2 \n\t" \ +:"=r"(cy),"=r"(_c[0]):"0"(cy),"1"(_c[0]):"cc"); + +/******************************************************************/ +#elif defined(TFM_PPC64) + +/* PPC64 */ +#define MONT_START +#define MONT_FINI +#define LOOP_END +#define LOOP_START \ + mu = c[x] * mp + +#define INNERMUL \ +asm( \ + " mulld r16,%3,%4 \n\t" \ + " mulhdu r17,%3,%4 \n\t" \ + " addc r16,16,%0 \n\t" \ + " addze r17,r17 \n\t" \ + " ldx r18,0,%1 \n\t" \ + " addc r16,r16,r18 \n\t" \ + " addze %0,r17 \n\t" \ + " sdx r16,0,%1 \n\t" \ +:"=r"(cy),"=m"(_c[0]):"0"(cy),"r"(mu),"r"(tmpm[0]),"1"(_c[0]):"r16", "r17", "r18","cc"); ++tmpm; + +#define PROPCARRY \ +asm( \ + " ldx r16,0,%1 \n\t" \ + " addc r16,r16,%0 \n\t" \ + " sdx r16,0,%1 \n\t" \ + " xor %0,%0,%0 \n\t" \ + " addze %0,%0 \n\t" \ +:"=r"(cy),"=m"(_c[0]):"0"(cy),"1"(_c[0]):"r16","cc"); + +/******************************************************************/ +#elif defined(TFM_AVR32) + +/* AVR32 */ +#define MONT_START +#define MONT_FINI +#define LOOP_END +#define LOOP_START \ + mu = c[x] * mp + +#define INNERMUL \ +asm( \ + " ld.w r2,%1 \n\t" \ + " add r2,%0 \n\t" \ + " eor r3,r3 \n\t" \ + " acr r3 \n\t" \ + " macu.d r2,%3,%4 \n\t" \ + " st.w %1,r2 \n\t" \ + " mov %0,r3 \n\t" \ +:"=r"(cy),"=r"(_c):"0"(cy),"r"(mu),"r"(*tmpm++),"1"(_c):"r2","r3"); + +#define PROPCARRY \ +asm( \ + " ld.w r2,%1 \n\t" \ + " add r2,%0 \n\t" \ + " st.w %1,r2 \n\t" \ + " eor %0,%0 \n\t" \ + " acr %0 \n\t" \ +:"=r"(cy),"=r"(&_c[0]):"0"(cy),"1"(&_c[0]):"r2","cc"); + +/******************************************************************/ +#elif defined(TFM_MIPS) + +/* MIPS */ +#define MONT_START +#define MONT_FINI +#define LOOP_END +#define LOOP_START \ + mu = c[x] * mp + +#define INNERMUL \ +asm( \ + " multu %3,%4 \n\t" \ + " mflo $12 \n\t" \ + " mfhi $13 \n\t" \ + " addu $12,$12,%0 \n\t" \ + " sltu $10,$12,%0 \n\t" \ + " addu $13,$13,$10 \n\t" \ + " lw $10,%1 \n\t" \ + " addu $12,$12,$10 \n\t" \ + " sltu $10,$12,$10 \n\t" \ + " addu %0,$13,$10 \n\t" \ + " sw $12,%1 \n\t" \ +:"=r"(cy),"=m"(_c[0]):"0"(cy),"r"(mu),"r"(tmpm[0]),"1"(_c[0]):"$10","$12","$13"); ++tmpm; + +#define PROPCARRY \ +asm( \ + " lw $10,%1 \n\t" \ + " addu $10,$10,%0 \n\t" \ + " sw $10,%1 \n\t" \ + " sltu %0,$10,%0 \n\t" \ +:"=r"(cy),"=m"(_c[0]):"0"(cy),"1"(_c[0]):"$10"); + +/******************************************************************/ +#else + +/* ISO C code */ +#define MONT_START +#define MONT_FINI +#define LOOP_END +#define LOOP_START \ + mu = c[x] * mp + +#define INNERMUL \ + do { fp_word t; \ + _c[0] = t = ((fp_word)_c[0] + (fp_word)cy) + \ + (((fp_word)mu) * ((fp_word)*tmpm++)); \ + cy = (t >> DIGIT_BIT); \ + } while (0) + +#define PROPCARRY \ + do { fp_digit t = _c[0] += cy; cy = (t < cy); } while (0) + +#endif +/******************************************************************/ + + +#define LO 0 + +#ifdef TFM_SMALL_MONT_SET +#include "fp_mont_small.i" +#endif + +/* computes x/R == x (mod N) via Montgomery Reduction */ +void fp_montgomery_reduce(fp_int *a, fp_int *m, fp_digit mp) +{ + fp_digit c[FP_SIZE], *_c, *tmpm, mu; + int oldused, x, y, pa; + + /* bail if too large */ + if (m->used > (FP_SIZE/2)) { + return; + } + +#ifdef TFM_SMALL_MONT_SET + if (m->used <= 16) { + fp_montgomery_reduce_small(a, m, mp); + return; + } +#endif + +#if defined(USE_MEMSET) + /* now zero the buff */ + memset(c, 0, sizeof c); +#endif + pa = m->used; + + /* copy the input */ + oldused = a->used; + for (x = 0; x < oldused; x++) { + c[x] = a->dp[x]; + } +#if !defined(USE_MEMSET) + for (; x < 2*pa+1; x++) { + c[x] = 0; + } +#endif + MONT_START; + + for (x = 0; x < pa; x++) { + fp_digit cy = 0; + /* get Mu for this round */ + LOOP_START; + _c = c + x; + tmpm = m->dp; + y = 0; + #if defined(INNERMUL8) + for (; y < (pa & ~7); y += 8) { + INNERMUL8; + _c += 8; + tmpm += 8; + } + #endif + + for (; y < pa; y++) { + INNERMUL; + ++_c; + } + LOOP_END; + while (cy) { + PROPCARRY; + ++_c; + } + } + + /* now copy out */ + _c = c + pa; + tmpm = a->dp; + for (x = 0; x < pa+1; x++) { + *tmpm++ = *_c++; + } + + for (; x < oldused; x++) { + *tmpm++ = 0; + } + + MONT_FINI; + + a->used = pa+1; + fp_clamp(a); + + /* if A >= m then A = A - m */ + if (fp_cmp_mag (a, m) != FP_LT) { + s_fp_sub (a, m, a); + } +} + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_montgomery_setup.c b/src/crypto/tfm/fp_montgomery_setup.c new file mode 100644 index 0000000..55f16dc --- /dev/null +++ b/src/crypto/tfm/fp_montgomery_setup.c @@ -0,0 +1,48 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* setups the montgomery reduction */ +int fp_montgomery_setup(fp_int *a, fp_digit *rho) +{ + fp_digit x, b; + +/* fast inversion mod 2**k + * + * Based on the fact that + * + * XA = 1 (mod 2**n) => (X(2-XA)) A = 1 (mod 2**2n) + * => 2*X*A - X*X*A*A = 1 + * => 2*(1) - (1) = 1 + */ + b = a->dp[0]; + + if ((b & 1) == 0) { + return FP_VAL; + } + + x = (((b + 2) & 4) << 1) + b; /* here x*a==1 mod 2**4 */ + x *= 2 - b * x; /* here x*a==1 mod 2**8 */ + x *= 2 - b * x; /* here x*a==1 mod 2**16 */ + x *= 2 - b * x; /* here x*a==1 mod 2**32 */ +#ifdef FP_64BIT + x *= 2 - b * x; /* here x*a==1 mod 2**64 */ +#endif + + /* rho = -1/m mod b */ + *rho = (((fp_word) 1 << ((fp_word) DIGIT_BIT)) - ((fp_word)x)); + + return FP_OKAY; +} + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mul.c b/src/crypto/tfm/fp_mul.c new file mode 100644 index 0000000..e97c6cf --- /dev/null +++ b/src/crypto/tfm/fp_mul.c @@ -0,0 +1,139 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* c = a * b */ +void fp_mul(fp_int *A, fp_int *B, fp_int *C) +{ + int y, old_used; +#if FP_SIZE >= 48 + int yy; +#endif + + old_used = C->used; + + /* call generic if we're out of range */ + if (A->used + B->used > FP_SIZE) { + fp_mul_comba(A, B, C); + goto clean; + } + + y = MAX(A->used, B->used); +#if FP_SIZE >= 48 + yy = MIN(A->used, B->used); +#endif + /* pick a comba (unrolled 4/8/16/32 x or rolled) based on the size + of the largest input. We also want to avoid doing excess mults if the + inputs are not close to the next power of two. That is, for example, + if say y=17 then we would do (32-17)^2 = 225 unneeded multiplications + */ + +#if defined(TFM_MUL3) && FP_SIZE >= 6 + if (y <= 3) { + fp_mul_comba3(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL4) && FP_SIZE >= 8 + if (y == 4) { + fp_mul_comba4(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL6) && FP_SIZE >= 12 + if (y <= 6) { + fp_mul_comba6(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL7) && FP_SIZE >= 14 + if (y == 7) { + fp_mul_comba7(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL8) && FP_SIZE >= 16 + if (y == 8) { + fp_mul_comba8(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL9) && FP_SIZE >= 18 + if (y == 9) { + fp_mul_comba9(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL12) && FP_SIZE >= 24 + if (y <= 12) { + fp_mul_comba12(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL17) && FP_SIZE >= 34 + if (y <= 17) { + fp_mul_comba17(A,B,C); + goto clean; + } +#endif + +#if defined(TFM_SMALL_SET) && FP_SIZE >= 32 + if (y <= 16) { + fp_mul_comba_small(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL20) && FP_SIZE >= 40 + if (y <= 20) { + fp_mul_comba20(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL24) && FP_SIZE >= 48 + if (yy >= 16 && y <= 24) { + fp_mul_comba24(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL28) && FP_SIZE >= 56 + if (yy >= 20 && y <= 28) { + fp_mul_comba28(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL32) && FP_SIZE >= 64 + if (yy >= 24 && y <= 32) { + fp_mul_comba32(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL48) && FP_SIZE >= 96 + if (yy >= 40 && y <= 48) { + fp_mul_comba48(A,B,C); + goto clean; + } +#endif +#if defined(TFM_MUL64) && FP_SIZE >= 128 + if (yy >= 56 && y <= 64) { + fp_mul_comba64(A,B,C); + goto clean; + } +#endif + fp_mul_comba(A,B,C); +clean: + for (y = C->used; y < old_used; y++) { + C->dp[y] = 0; + } +} + + +/* $Source: /cvs/libtom/tomsfastmath/src/mul/fp_mul.c,v $ */ +/* $Revision: 1.1 $ */ +/* $Date: 2006/12/31 21:25:53 $ */ diff --git a/src/crypto/tfm/fp_mul_2.c b/src/crypto/tfm/fp_mul_2.c new file mode 100644 index 0000000..bd40163 --- /dev/null +++ b/src/crypto/tfm/fp_mul_2.c @@ -0,0 +1,67 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +void fp_mul_2(fp_int * a, fp_int * b) +{ + int x, oldused; + + oldused = b->used; + b->used = a->used; + + { + register fp_digit r, rr, *tmpa, *tmpb; + + /* alias for source */ + tmpa = a->dp; + + /* alias for dest */ + tmpb = b->dp; + + /* carry */ + r = 0; + for (x = 0; x < a->used; x++) { + + /* get what will be the *next* carry bit from the + * MSB of the current digit + */ + rr = *tmpa >> ((fp_digit)(DIGIT_BIT - 1)); + + /* now shift up this digit, add in the carry [from the previous] */ + *tmpb++ = ((*tmpa++ << ((fp_digit)1)) | r); + + /* copy the carry that would be from the source + * digit into the next iteration + */ + r = rr; + } + + /* new leading digit? */ + if (r != 0 && b->used != (FP_SIZE-1)) { + /* add a MSB which is always 1 at this point */ + *tmpb = 1; + ++(b->used); + } + + /* now zero any excess digits on the destination + * that we didn't write to + */ + tmpb = b->dp + b->used; + for (x = b->used; x < oldused; x++) { + *tmpb++ = 0; + } + } + b->sign = a->sign; +} + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mul_2d.c b/src/crypto/tfm/fp_mul_2d.c new file mode 100644 index 0000000..dd81d41 --- /dev/null +++ b/src/crypto/tfm/fp_mul_2d.c @@ -0,0 +1,47 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* c = a * 2**d */ +void fp_mul_2d(fp_int *a, int b, fp_int *c) +{ + fp_digit carry, carrytmp, shift; + int x; + + /* copy it */ + fp_copy(a, c); + + /* handle whole digits */ + if (b >= DIGIT_BIT) { + fp_lshd(c, b/DIGIT_BIT); + } + b %= DIGIT_BIT; + + /* shift the digits */ + if (b != 0) { + carry = 0; + shift = DIGIT_BIT - b; + for (x = 0; x < c->used; x++) { + carrytmp = c->dp[x] >> shift; + c->dp[x] = (c->dp[x] << b) + carry; + carry = carrytmp; + } + /* store last carry if room */ + if (carry && x < FP_SIZE) { + c->dp[c->used++] = carry; + } + } + fp_clamp(c); +} + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mul_comba.c b/src/crypto/tfm/fp_mul_comba.c new file mode 100644 index 0000000..a4253fb --- /dev/null +++ b/src/crypto/tfm/fp_mul_comba.c @@ -0,0 +1,373 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ + +/* About this file... + +*/ + +#include + +#if defined(TFM_PRESCOTT) && defined(TFM_SSE2) + #undef TFM_SSE2 + #define TFM_X86 +#endif + +/* these are the combas. Worship them. */ +#if defined(TFM_X86) +/* Generic x86 optimized code */ + +/* anything you need at the start */ +#define COMBA_START + +/* clear the chaining variables */ +#define COMBA_CLEAR \ + c0 = c1 = c2 = 0; + +/* forward the carry to the next digit */ +#define COMBA_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +/* store the first sum */ +#define COMBA_STORE(x) \ + x = c0; + +/* store the second sum [carry] */ +#define COMBA_STORE2(x) \ + x = c1; + +/* anything you need at the end */ +#define COMBA_FINI + +/* this should multiply i and j */ +#define MULADD(i, j) \ +asm( \ + "movl %6,%%eax \n\t" \ + "mull %7 \n\t" \ + "addl %%eax,%0 \n\t" \ + "adcl %%edx,%1 \n\t" \ + "adcl $0,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(i), "m"(j) :"%eax","%edx","cc"); + +#elif defined(TFM_X86_64) +/* x86-64 optimized */ + +/* anything you need at the start */ +#define COMBA_START + +/* clear the chaining variables */ +#define COMBA_CLEAR \ + c0 = c1 = c2 = 0; + +/* forward the carry to the next digit */ +#define COMBA_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +/* store the first sum */ +#define COMBA_STORE(x) \ + x = c0; + +/* store the second sum [carry] */ +#define COMBA_STORE2(x) \ + x = c1; + +/* anything you need at the end */ +#define COMBA_FINI + +/* this should multiply i and j */ +#define MULADD(i, j) \ +asm ( \ + "movq %6,%%rax \n\t" \ + "mulq %7 \n\t" \ + "addq %%rax,%0 \n\t" \ + "adcq %%rdx,%1 \n\t" \ + "adcq $0,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "g"(i), "g"(j) :"%rax","%rdx","cc"); + +#elif defined(TFM_SSE2) +/* use SSE2 optimizations */ + +/* anything you need at the start */ +#define COMBA_START + +/* clear the chaining variables */ +#define COMBA_CLEAR \ + c0 = c1 = c2 = 0; + +/* forward the carry to the next digit */ +#define COMBA_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +/* store the first sum */ +#define COMBA_STORE(x) \ + x = c0; + +/* store the second sum [carry] */ +#define COMBA_STORE2(x) \ + x = c1; + +/* anything you need at the end */ +#define COMBA_FINI \ + asm("emms"); + +/* this should multiply i and j */ +#define MULADD(i, j) \ +asm( \ + "movd %6,%%mm0 \n\t" \ + "movd %7,%%mm1 \n\t" \ + "pmuludq %%mm1,%%mm0\n\t" \ + "movd %%mm0,%%eax \n\t" \ + "psrlq $32,%%mm0 \n\t" \ + "addl %%eax,%0 \n\t" \ + "movd %%mm0,%%eax \n\t" \ + "adcl %%eax,%1 \n\t" \ + "adcl $0,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(i), "m"(j) :"%eax","cc"); + +#elif defined(TFM_ARM) +/* ARM code */ + +#define COMBA_START + +#define COMBA_CLEAR \ + c0 = c1 = c2 = 0; + +#define COMBA_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define COMBA_FINI + +#define MULADD(i, j) \ +asm( \ +" UMULL r0,r1,%6,%7 \n\t" \ +" ADDS %0,%0,r0 \n\t" \ +" ADCS %1,%1,r1 \n\t" \ +" ADC %2,%2,#0 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2) : "0"(c0), "1"(c1), "2"(c2), "r"(i), "r"(j) : "r0", "r1", "cc"); + +#elif defined(TFM_PPC32) +/* For 32-bit PPC */ + +#define COMBA_START + +#define COMBA_CLEAR \ + c0 = c1 = c2 = 0; + +#define COMBA_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define COMBA_FINI + +/* untested: will mulhwu change the flags? Docs say no */ +#define MULADD(i, j) \ +asm( \ + " mullw 16,%6,%7 \n\t" \ + " addc %0,%0,16 \n\t" \ + " mulhwu 16,%6,%7 \n\t" \ + " adde %1,%1,16 \n\t" \ + " addze %2,%2 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i), "r"(j):"16"); + +#elif defined(TFM_PPC64) +/* For 64-bit PPC */ + +#define COMBA_START + +#define COMBA_CLEAR \ + c0 = c1 = c2 = 0; + +#define COMBA_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define COMBA_FINI + +/* untested: will mulhdu change the flags? Docs say no */ +#define MULADD(i, j) \ +asm( \ + " mulld r16,%6,%7 \n\t" \ + " addc %0,%0,16 \n\t" \ + " mulhdu r16,%6,%7 \n\t" \ + " adde %1,%1,16 \n\t" \ + " addze %2,%2 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i), "r"(j):"r16"); + +#elif defined(TFM_AVR32) + +/* ISO C code */ + +#define COMBA_START + +#define COMBA_CLEAR \ + c0 = c1 = c2 = 0; + +#define COMBA_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define COMBA_FINI + +#define MULADD(i, j) \ +asm( \ + " mulu.d r2,%6,%7 \n\t"\ + " add %0,r2 \n\t"\ + " adc %1,%1,r3 \n\t"\ + " acr %2 \n\t"\ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i), "r"(j):"r2","r3"); + +#elif defined(TFM_MIPS) + +#define COMBA_START + +#define COMBA_CLEAR \ + c0 = c1 = c2 = 0; + +#define COMBA_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define COMBA_FINI + +#define MULADD(i, j) \ +asm( \ + " multu %6,%7 \n\t" \ + " mflo $12 \n\t" \ + " mfhi $13 \n\t" \ + " addu %0,%0,$12 \n\t" \ + " sltu $12,%0,$12 \n\t" \ + " addu %1,%1,$13 \n\t" \ + " sltu $13,%1,$13 \n\t" \ + " addu %1,%1,$12 \n\t" \ + " sltu $12,%1,$12 \n\t" \ + " addu %2,%2,$13 \n\t" \ + " addu %2,%2,$12 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i), "r"(j):"$12","$13"); + +#else +/* ISO C code */ + +#define COMBA_START + +#define COMBA_CLEAR \ + c0 = c1 = c2 = 0; + +#define COMBA_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define COMBA_FINI + +#define MULADD(i, j) \ + do { fp_word t; \ + t = (fp_word)c0 + ((fp_word)i) * ((fp_word)j); \ + c0 = t; \ + t = (fp_word)c1 + (t >> DIGIT_BIT); \ + c1 = t; \ + c2 += t >> DIGIT_BIT; \ + } while (0); + +#endif + +#ifndef TFM_DEFINES + +/* generic PxQ multiplier */ +void fp_mul_comba(fp_int *A, fp_int *B, fp_int *C) +{ + int ix, iy, iz, tx, ty, pa; + fp_digit c0, c1, c2, *tmpx, *tmpy; + fp_int tmp, *dst; + + COMBA_START; + COMBA_CLEAR; + + /* get size of output and trim */ + pa = A->used + B->used; + if (pa >= FP_SIZE) { + pa = FP_SIZE-1; + } + + if (A == C || B == C) { + fp_zero(&tmp); + dst = &tmp; + } else { + fp_zero(C); + dst = C; + } + + for (ix = 0; ix < pa; ix++) { + /* get offsets into the two bignums */ + ty = MIN(ix, B->used-1); + tx = ix - ty; + + /* setup temp aliases */ + tmpx = A->dp + tx; + tmpy = B->dp + ty; + + /* this is the number of times the loop will iterrate, essentially its + while (tx++ < a->used && ty-- >= 0) { ... } + */ + iy = MIN(A->used-tx, ty+1); + + /* execute loop */ + COMBA_FORWARD; + for (iz = 0; iz < iy; ++iz) { + fp_digit _tmpx = *tmpx++; + fp_digit _tmpy = *tmpy--; + MULADD(_tmpx, _tmpy); + } + + /* store term */ + COMBA_STORE(dst->dp[ix]); + } + COMBA_FINI; + + dst->used = pa; + dst->sign = A->sign ^ B->sign; + fp_clamp(dst); + fp_copy(dst, C); +} + +#endif + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ + diff --git a/src/crypto/tfm/fp_mul_comba_20.c b/src/crypto/tfm/fp_mul_comba_20.c new file mode 100644 index 0000000..905741d --- /dev/null +++ b/src/crypto/tfm/fp_mul_comba_20.c @@ -0,0 +1,180 @@ +#define TFM_DEFINES +#include "fp_mul_comba.c" + +#if defined(TFM_MUL20) && FP_SIZE >= 40 +void fp_mul_comba20(fp_int *A, fp_int *B, fp_int *C) +{ + fp_digit c0, c1, c2, at[40]; + + memcpy(at, A->dp, 20 * sizeof(fp_digit)); + memcpy(at+20, B->dp, 20 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[20]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[21]); MULADD(at[1], at[20]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[22]); MULADD(at[1], at[21]); MULADD(at[2], at[20]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[23]); MULADD(at[1], at[22]); MULADD(at[2], at[21]); MULADD(at[3], at[20]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[24]); MULADD(at[1], at[23]); MULADD(at[2], at[22]); MULADD(at[3], at[21]); MULADD(at[4], at[20]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[25]); MULADD(at[1], at[24]); MULADD(at[2], at[23]); MULADD(at[3], at[22]); MULADD(at[4], at[21]); MULADD(at[5], at[20]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[26]); MULADD(at[1], at[25]); MULADD(at[2], at[24]); MULADD(at[3], at[23]); MULADD(at[4], at[22]); MULADD(at[5], at[21]); MULADD(at[6], at[20]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[27]); MULADD(at[1], at[26]); MULADD(at[2], at[25]); MULADD(at[3], at[24]); MULADD(at[4], at[23]); MULADD(at[5], at[22]); MULADD(at[6], at[21]); MULADD(at[7], at[20]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[28]); MULADD(at[1], at[27]); MULADD(at[2], at[26]); MULADD(at[3], at[25]); MULADD(at[4], at[24]); MULADD(at[5], at[23]); MULADD(at[6], at[22]); MULADD(at[7], at[21]); MULADD(at[8], at[20]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[29]); MULADD(at[1], at[28]); MULADD(at[2], at[27]); MULADD(at[3], at[26]); MULADD(at[4], at[25]); MULADD(at[5], at[24]); MULADD(at[6], at[23]); MULADD(at[7], at[22]); MULADD(at[8], at[21]); MULADD(at[9], at[20]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[0], at[30]); MULADD(at[1], at[29]); MULADD(at[2], at[28]); MULADD(at[3], at[27]); MULADD(at[4], at[26]); MULADD(at[5], at[25]); MULADD(at[6], at[24]); MULADD(at[7], at[23]); MULADD(at[8], at[22]); MULADD(at[9], at[21]); MULADD(at[10], at[20]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[0], at[31]); MULADD(at[1], at[30]); MULADD(at[2], at[29]); MULADD(at[3], at[28]); MULADD(at[4], at[27]); MULADD(at[5], at[26]); MULADD(at[6], at[25]); MULADD(at[7], at[24]); MULADD(at[8], at[23]); MULADD(at[9], at[22]); MULADD(at[10], at[21]); MULADD(at[11], at[20]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[0], at[32]); MULADD(at[1], at[31]); MULADD(at[2], at[30]); MULADD(at[3], at[29]); MULADD(at[4], at[28]); MULADD(at[5], at[27]); MULADD(at[6], at[26]); MULADD(at[7], at[25]); MULADD(at[8], at[24]); MULADD(at[9], at[23]); MULADD(at[10], at[22]); MULADD(at[11], at[21]); MULADD(at[12], at[20]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[0], at[33]); MULADD(at[1], at[32]); MULADD(at[2], at[31]); MULADD(at[3], at[30]); MULADD(at[4], at[29]); MULADD(at[5], at[28]); MULADD(at[6], at[27]); MULADD(at[7], at[26]); MULADD(at[8], at[25]); MULADD(at[9], at[24]); MULADD(at[10], at[23]); MULADD(at[11], at[22]); MULADD(at[12], at[21]); MULADD(at[13], at[20]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[0], at[34]); MULADD(at[1], at[33]); MULADD(at[2], at[32]); MULADD(at[3], at[31]); MULADD(at[4], at[30]); MULADD(at[5], at[29]); MULADD(at[6], at[28]); MULADD(at[7], at[27]); MULADD(at[8], at[26]); MULADD(at[9], at[25]); MULADD(at[10], at[24]); MULADD(at[11], at[23]); MULADD(at[12], at[22]); MULADD(at[13], at[21]); MULADD(at[14], at[20]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[0], at[35]); MULADD(at[1], at[34]); MULADD(at[2], at[33]); MULADD(at[3], at[32]); MULADD(at[4], at[31]); MULADD(at[5], at[30]); MULADD(at[6], at[29]); MULADD(at[7], at[28]); MULADD(at[8], at[27]); MULADD(at[9], at[26]); MULADD(at[10], at[25]); MULADD(at[11], at[24]); MULADD(at[12], at[23]); MULADD(at[13], at[22]); MULADD(at[14], at[21]); MULADD(at[15], at[20]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[0], at[36]); MULADD(at[1], at[35]); MULADD(at[2], at[34]); MULADD(at[3], at[33]); MULADD(at[4], at[32]); MULADD(at[5], at[31]); MULADD(at[6], at[30]); MULADD(at[7], at[29]); MULADD(at[8], at[28]); MULADD(at[9], at[27]); MULADD(at[10], at[26]); MULADD(at[11], at[25]); MULADD(at[12], at[24]); MULADD(at[13], at[23]); MULADD(at[14], at[22]); MULADD(at[15], at[21]); MULADD(at[16], at[20]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[0], at[37]); MULADD(at[1], at[36]); MULADD(at[2], at[35]); MULADD(at[3], at[34]); MULADD(at[4], at[33]); MULADD(at[5], at[32]); MULADD(at[6], at[31]); MULADD(at[7], at[30]); MULADD(at[8], at[29]); MULADD(at[9], at[28]); MULADD(at[10], at[27]); MULADD(at[11], at[26]); MULADD(at[12], at[25]); MULADD(at[13], at[24]); MULADD(at[14], at[23]); MULADD(at[15], at[22]); MULADD(at[16], at[21]); MULADD(at[17], at[20]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[0], at[38]); MULADD(at[1], at[37]); MULADD(at[2], at[36]); MULADD(at[3], at[35]); MULADD(at[4], at[34]); MULADD(at[5], at[33]); MULADD(at[6], at[32]); MULADD(at[7], at[31]); MULADD(at[8], at[30]); MULADD(at[9], at[29]); MULADD(at[10], at[28]); MULADD(at[11], at[27]); MULADD(at[12], at[26]); MULADD(at[13], at[25]); MULADD(at[14], at[24]); MULADD(at[15], at[23]); MULADD(at[16], at[22]); MULADD(at[17], at[21]); MULADD(at[18], at[20]); + COMBA_STORE(C->dp[18]); + /* 19 */ + COMBA_FORWARD; + MULADD(at[0], at[39]); MULADD(at[1], at[38]); MULADD(at[2], at[37]); MULADD(at[3], at[36]); MULADD(at[4], at[35]); MULADD(at[5], at[34]); MULADD(at[6], at[33]); MULADD(at[7], at[32]); MULADD(at[8], at[31]); MULADD(at[9], at[30]); MULADD(at[10], at[29]); MULADD(at[11], at[28]); MULADD(at[12], at[27]); MULADD(at[13], at[26]); MULADD(at[14], at[25]); MULADD(at[15], at[24]); MULADD(at[16], at[23]); MULADD(at[17], at[22]); MULADD(at[18], at[21]); MULADD(at[19], at[20]); + COMBA_STORE(C->dp[19]); + /* 20 */ + COMBA_FORWARD; + MULADD(at[1], at[39]); MULADD(at[2], at[38]); MULADD(at[3], at[37]); MULADD(at[4], at[36]); MULADD(at[5], at[35]); MULADD(at[6], at[34]); MULADD(at[7], at[33]); MULADD(at[8], at[32]); MULADD(at[9], at[31]); MULADD(at[10], at[30]); MULADD(at[11], at[29]); MULADD(at[12], at[28]); MULADD(at[13], at[27]); MULADD(at[14], at[26]); MULADD(at[15], at[25]); MULADD(at[16], at[24]); MULADD(at[17], at[23]); MULADD(at[18], at[22]); MULADD(at[19], at[21]); + COMBA_STORE(C->dp[20]); + /* 21 */ + COMBA_FORWARD; + MULADD(at[2], at[39]); MULADD(at[3], at[38]); MULADD(at[4], at[37]); MULADD(at[5], at[36]); MULADD(at[6], at[35]); MULADD(at[7], at[34]); MULADD(at[8], at[33]); MULADD(at[9], at[32]); MULADD(at[10], at[31]); MULADD(at[11], at[30]); MULADD(at[12], at[29]); MULADD(at[13], at[28]); MULADD(at[14], at[27]); MULADD(at[15], at[26]); MULADD(at[16], at[25]); MULADD(at[17], at[24]); MULADD(at[18], at[23]); MULADD(at[19], at[22]); + COMBA_STORE(C->dp[21]); + /* 22 */ + COMBA_FORWARD; + MULADD(at[3], at[39]); MULADD(at[4], at[38]); MULADD(at[5], at[37]); MULADD(at[6], at[36]); MULADD(at[7], at[35]); MULADD(at[8], at[34]); MULADD(at[9], at[33]); MULADD(at[10], at[32]); MULADD(at[11], at[31]); MULADD(at[12], at[30]); MULADD(at[13], at[29]); MULADD(at[14], at[28]); MULADD(at[15], at[27]); MULADD(at[16], at[26]); MULADD(at[17], at[25]); MULADD(at[18], at[24]); MULADD(at[19], at[23]); + COMBA_STORE(C->dp[22]); + /* 23 */ + COMBA_FORWARD; + MULADD(at[4], at[39]); MULADD(at[5], at[38]); MULADD(at[6], at[37]); MULADD(at[7], at[36]); MULADD(at[8], at[35]); MULADD(at[9], at[34]); MULADD(at[10], at[33]); MULADD(at[11], at[32]); MULADD(at[12], at[31]); MULADD(at[13], at[30]); MULADD(at[14], at[29]); MULADD(at[15], at[28]); MULADD(at[16], at[27]); MULADD(at[17], at[26]); MULADD(at[18], at[25]); MULADD(at[19], at[24]); + COMBA_STORE(C->dp[23]); + /* 24 */ + COMBA_FORWARD; + MULADD(at[5], at[39]); MULADD(at[6], at[38]); MULADD(at[7], at[37]); MULADD(at[8], at[36]); MULADD(at[9], at[35]); MULADD(at[10], at[34]); MULADD(at[11], at[33]); MULADD(at[12], at[32]); MULADD(at[13], at[31]); MULADD(at[14], at[30]); MULADD(at[15], at[29]); MULADD(at[16], at[28]); MULADD(at[17], at[27]); MULADD(at[18], at[26]); MULADD(at[19], at[25]); + COMBA_STORE(C->dp[24]); + /* 25 */ + COMBA_FORWARD; + MULADD(at[6], at[39]); MULADD(at[7], at[38]); MULADD(at[8], at[37]); MULADD(at[9], at[36]); MULADD(at[10], at[35]); MULADD(at[11], at[34]); MULADD(at[12], at[33]); MULADD(at[13], at[32]); MULADD(at[14], at[31]); MULADD(at[15], at[30]); MULADD(at[16], at[29]); MULADD(at[17], at[28]); MULADD(at[18], at[27]); MULADD(at[19], at[26]); + COMBA_STORE(C->dp[25]); + /* 26 */ + COMBA_FORWARD; + MULADD(at[7], at[39]); MULADD(at[8], at[38]); MULADD(at[9], at[37]); MULADD(at[10], at[36]); MULADD(at[11], at[35]); MULADD(at[12], at[34]); MULADD(at[13], at[33]); MULADD(at[14], at[32]); MULADD(at[15], at[31]); MULADD(at[16], at[30]); MULADD(at[17], at[29]); MULADD(at[18], at[28]); MULADD(at[19], at[27]); + COMBA_STORE(C->dp[26]); + /* 27 */ + COMBA_FORWARD; + MULADD(at[8], at[39]); MULADD(at[9], at[38]); MULADD(at[10], at[37]); MULADD(at[11], at[36]); MULADD(at[12], at[35]); MULADD(at[13], at[34]); MULADD(at[14], at[33]); MULADD(at[15], at[32]); MULADD(at[16], at[31]); MULADD(at[17], at[30]); MULADD(at[18], at[29]); MULADD(at[19], at[28]); + COMBA_STORE(C->dp[27]); + /* 28 */ + COMBA_FORWARD; + MULADD(at[9], at[39]); MULADD(at[10], at[38]); MULADD(at[11], at[37]); MULADD(at[12], at[36]); MULADD(at[13], at[35]); MULADD(at[14], at[34]); MULADD(at[15], at[33]); MULADD(at[16], at[32]); MULADD(at[17], at[31]); MULADD(at[18], at[30]); MULADD(at[19], at[29]); + COMBA_STORE(C->dp[28]); + /* 29 */ + COMBA_FORWARD; + MULADD(at[10], at[39]); MULADD(at[11], at[38]); MULADD(at[12], at[37]); MULADD(at[13], at[36]); MULADD(at[14], at[35]); MULADD(at[15], at[34]); MULADD(at[16], at[33]); MULADD(at[17], at[32]); MULADD(at[18], at[31]); MULADD(at[19], at[30]); + COMBA_STORE(C->dp[29]); + /* 30 */ + COMBA_FORWARD; + MULADD(at[11], at[39]); MULADD(at[12], at[38]); MULADD(at[13], at[37]); MULADD(at[14], at[36]); MULADD(at[15], at[35]); MULADD(at[16], at[34]); MULADD(at[17], at[33]); MULADD(at[18], at[32]); MULADD(at[19], at[31]); + COMBA_STORE(C->dp[30]); + /* 31 */ + COMBA_FORWARD; + MULADD(at[12], at[39]); MULADD(at[13], at[38]); MULADD(at[14], at[37]); MULADD(at[15], at[36]); MULADD(at[16], at[35]); MULADD(at[17], at[34]); MULADD(at[18], at[33]); MULADD(at[19], at[32]); + COMBA_STORE(C->dp[31]); + /* 32 */ + COMBA_FORWARD; + MULADD(at[13], at[39]); MULADD(at[14], at[38]); MULADD(at[15], at[37]); MULADD(at[16], at[36]); MULADD(at[17], at[35]); MULADD(at[18], at[34]); MULADD(at[19], at[33]); + COMBA_STORE(C->dp[32]); + /* 33 */ + COMBA_FORWARD; + MULADD(at[14], at[39]); MULADD(at[15], at[38]); MULADD(at[16], at[37]); MULADD(at[17], at[36]); MULADD(at[18], at[35]); MULADD(at[19], at[34]); + COMBA_STORE(C->dp[33]); + /* 34 */ + COMBA_FORWARD; + MULADD(at[15], at[39]); MULADD(at[16], at[38]); MULADD(at[17], at[37]); MULADD(at[18], at[36]); MULADD(at[19], at[35]); + COMBA_STORE(C->dp[34]); + /* 35 */ + COMBA_FORWARD; + MULADD(at[16], at[39]); MULADD(at[17], at[38]); MULADD(at[18], at[37]); MULADD(at[19], at[36]); + COMBA_STORE(C->dp[35]); + /* 36 */ + COMBA_FORWARD; + MULADD(at[17], at[39]); MULADD(at[18], at[38]); MULADD(at[19], at[37]); + COMBA_STORE(C->dp[36]); + /* 37 */ + COMBA_FORWARD; + MULADD(at[18], at[39]); MULADD(at[19], at[38]); + COMBA_STORE(C->dp[37]); + /* 38 */ + COMBA_FORWARD; + MULADD(at[19], at[39]); + COMBA_STORE(C->dp[38]); + COMBA_STORE2(C->dp[39]); + C->used = 40; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; +} +#endif + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mul_comba_24.c b/src/crypto/tfm/fp_mul_comba_24.c new file mode 100644 index 0000000..0aaccad --- /dev/null +++ b/src/crypto/tfm/fp_mul_comba_24.c @@ -0,0 +1,212 @@ +#define TFM_DEFINES +#include "fp_mul_comba.c" + +#if defined(TFM_MUL24) && FP_SIZE >= 48 +void fp_mul_comba24(fp_int *A, fp_int *B, fp_int *C) +{ + fp_digit c0, c1, c2, at[48]; + + memcpy(at, A->dp, 24 * sizeof(fp_digit)); + memcpy(at+24, B->dp, 24 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[24]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[25]); MULADD(at[1], at[24]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[26]); MULADD(at[1], at[25]); MULADD(at[2], at[24]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[27]); MULADD(at[1], at[26]); MULADD(at[2], at[25]); MULADD(at[3], at[24]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[28]); MULADD(at[1], at[27]); MULADD(at[2], at[26]); MULADD(at[3], at[25]); MULADD(at[4], at[24]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[29]); MULADD(at[1], at[28]); MULADD(at[2], at[27]); MULADD(at[3], at[26]); MULADD(at[4], at[25]); MULADD(at[5], at[24]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[30]); MULADD(at[1], at[29]); MULADD(at[2], at[28]); MULADD(at[3], at[27]); MULADD(at[4], at[26]); MULADD(at[5], at[25]); MULADD(at[6], at[24]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[31]); MULADD(at[1], at[30]); MULADD(at[2], at[29]); MULADD(at[3], at[28]); MULADD(at[4], at[27]); MULADD(at[5], at[26]); MULADD(at[6], at[25]); MULADD(at[7], at[24]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[32]); MULADD(at[1], at[31]); MULADD(at[2], at[30]); MULADD(at[3], at[29]); MULADD(at[4], at[28]); MULADD(at[5], at[27]); MULADD(at[6], at[26]); MULADD(at[7], at[25]); MULADD(at[8], at[24]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[33]); MULADD(at[1], at[32]); MULADD(at[2], at[31]); MULADD(at[3], at[30]); MULADD(at[4], at[29]); MULADD(at[5], at[28]); MULADD(at[6], at[27]); MULADD(at[7], at[26]); MULADD(at[8], at[25]); MULADD(at[9], at[24]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[0], at[34]); MULADD(at[1], at[33]); MULADD(at[2], at[32]); MULADD(at[3], at[31]); MULADD(at[4], at[30]); MULADD(at[5], at[29]); MULADD(at[6], at[28]); MULADD(at[7], at[27]); MULADD(at[8], at[26]); MULADD(at[9], at[25]); MULADD(at[10], at[24]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[0], at[35]); MULADD(at[1], at[34]); MULADD(at[2], at[33]); MULADD(at[3], at[32]); MULADD(at[4], at[31]); MULADD(at[5], at[30]); MULADD(at[6], at[29]); MULADD(at[7], at[28]); MULADD(at[8], at[27]); MULADD(at[9], at[26]); MULADD(at[10], at[25]); MULADD(at[11], at[24]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[0], at[36]); MULADD(at[1], at[35]); MULADD(at[2], at[34]); MULADD(at[3], at[33]); MULADD(at[4], at[32]); MULADD(at[5], at[31]); MULADD(at[6], at[30]); MULADD(at[7], at[29]); MULADD(at[8], at[28]); MULADD(at[9], at[27]); MULADD(at[10], at[26]); MULADD(at[11], at[25]); MULADD(at[12], at[24]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[0], at[37]); MULADD(at[1], at[36]); MULADD(at[2], at[35]); MULADD(at[3], at[34]); MULADD(at[4], at[33]); MULADD(at[5], at[32]); MULADD(at[6], at[31]); MULADD(at[7], at[30]); MULADD(at[8], at[29]); MULADD(at[9], at[28]); MULADD(at[10], at[27]); MULADD(at[11], at[26]); MULADD(at[12], at[25]); MULADD(at[13], at[24]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[0], at[38]); MULADD(at[1], at[37]); MULADD(at[2], at[36]); MULADD(at[3], at[35]); MULADD(at[4], at[34]); MULADD(at[5], at[33]); MULADD(at[6], at[32]); MULADD(at[7], at[31]); MULADD(at[8], at[30]); MULADD(at[9], at[29]); MULADD(at[10], at[28]); MULADD(at[11], at[27]); MULADD(at[12], at[26]); MULADD(at[13], at[25]); MULADD(at[14], at[24]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[0], at[39]); MULADD(at[1], at[38]); MULADD(at[2], at[37]); MULADD(at[3], at[36]); MULADD(at[4], at[35]); MULADD(at[5], at[34]); MULADD(at[6], at[33]); MULADD(at[7], at[32]); MULADD(at[8], at[31]); MULADD(at[9], at[30]); MULADD(at[10], at[29]); MULADD(at[11], at[28]); MULADD(at[12], at[27]); MULADD(at[13], at[26]); MULADD(at[14], at[25]); MULADD(at[15], at[24]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[0], at[40]); MULADD(at[1], at[39]); MULADD(at[2], at[38]); MULADD(at[3], at[37]); MULADD(at[4], at[36]); MULADD(at[5], at[35]); MULADD(at[6], at[34]); MULADD(at[7], at[33]); MULADD(at[8], at[32]); MULADD(at[9], at[31]); MULADD(at[10], at[30]); MULADD(at[11], at[29]); MULADD(at[12], at[28]); MULADD(at[13], at[27]); MULADD(at[14], at[26]); MULADD(at[15], at[25]); MULADD(at[16], at[24]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[0], at[41]); MULADD(at[1], at[40]); MULADD(at[2], at[39]); MULADD(at[3], at[38]); MULADD(at[4], at[37]); MULADD(at[5], at[36]); MULADD(at[6], at[35]); MULADD(at[7], at[34]); MULADD(at[8], at[33]); MULADD(at[9], at[32]); MULADD(at[10], at[31]); MULADD(at[11], at[30]); MULADD(at[12], at[29]); MULADD(at[13], at[28]); MULADD(at[14], at[27]); MULADD(at[15], at[26]); MULADD(at[16], at[25]); MULADD(at[17], at[24]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[0], at[42]); MULADD(at[1], at[41]); MULADD(at[2], at[40]); MULADD(at[3], at[39]); MULADD(at[4], at[38]); MULADD(at[5], at[37]); MULADD(at[6], at[36]); MULADD(at[7], at[35]); MULADD(at[8], at[34]); MULADD(at[9], at[33]); MULADD(at[10], at[32]); MULADD(at[11], at[31]); MULADD(at[12], at[30]); MULADD(at[13], at[29]); MULADD(at[14], at[28]); MULADD(at[15], at[27]); MULADD(at[16], at[26]); MULADD(at[17], at[25]); MULADD(at[18], at[24]); + COMBA_STORE(C->dp[18]); + /* 19 */ + COMBA_FORWARD; + MULADD(at[0], at[43]); MULADD(at[1], at[42]); MULADD(at[2], at[41]); MULADD(at[3], at[40]); MULADD(at[4], at[39]); MULADD(at[5], at[38]); MULADD(at[6], at[37]); MULADD(at[7], at[36]); MULADD(at[8], at[35]); MULADD(at[9], at[34]); MULADD(at[10], at[33]); MULADD(at[11], at[32]); MULADD(at[12], at[31]); MULADD(at[13], at[30]); MULADD(at[14], at[29]); MULADD(at[15], at[28]); MULADD(at[16], at[27]); MULADD(at[17], at[26]); MULADD(at[18], at[25]); MULADD(at[19], at[24]); + COMBA_STORE(C->dp[19]); + /* 20 */ + COMBA_FORWARD; + MULADD(at[0], at[44]); MULADD(at[1], at[43]); MULADD(at[2], at[42]); MULADD(at[3], at[41]); MULADD(at[4], at[40]); MULADD(at[5], at[39]); MULADD(at[6], at[38]); MULADD(at[7], at[37]); MULADD(at[8], at[36]); MULADD(at[9], at[35]); MULADD(at[10], at[34]); MULADD(at[11], at[33]); MULADD(at[12], at[32]); MULADD(at[13], at[31]); MULADD(at[14], at[30]); MULADD(at[15], at[29]); MULADD(at[16], at[28]); MULADD(at[17], at[27]); MULADD(at[18], at[26]); MULADD(at[19], at[25]); MULADD(at[20], at[24]); + COMBA_STORE(C->dp[20]); + /* 21 */ + COMBA_FORWARD; + MULADD(at[0], at[45]); MULADD(at[1], at[44]); MULADD(at[2], at[43]); MULADD(at[3], at[42]); MULADD(at[4], at[41]); MULADD(at[5], at[40]); MULADD(at[6], at[39]); MULADD(at[7], at[38]); MULADD(at[8], at[37]); MULADD(at[9], at[36]); MULADD(at[10], at[35]); MULADD(at[11], at[34]); MULADD(at[12], at[33]); MULADD(at[13], at[32]); MULADD(at[14], at[31]); MULADD(at[15], at[30]); MULADD(at[16], at[29]); MULADD(at[17], at[28]); MULADD(at[18], at[27]); MULADD(at[19], at[26]); MULADD(at[20], at[25]); MULADD(at[21], at[24]); + COMBA_STORE(C->dp[21]); + /* 22 */ + COMBA_FORWARD; + MULADD(at[0], at[46]); MULADD(at[1], at[45]); MULADD(at[2], at[44]); MULADD(at[3], at[43]); MULADD(at[4], at[42]); MULADD(at[5], at[41]); MULADD(at[6], at[40]); MULADD(at[7], at[39]); MULADD(at[8], at[38]); MULADD(at[9], at[37]); MULADD(at[10], at[36]); MULADD(at[11], at[35]); MULADD(at[12], at[34]); MULADD(at[13], at[33]); MULADD(at[14], at[32]); MULADD(at[15], at[31]); MULADD(at[16], at[30]); MULADD(at[17], at[29]); MULADD(at[18], at[28]); MULADD(at[19], at[27]); MULADD(at[20], at[26]); MULADD(at[21], at[25]); MULADD(at[22], at[24]); + COMBA_STORE(C->dp[22]); + /* 23 */ + COMBA_FORWARD; + MULADD(at[0], at[47]); MULADD(at[1], at[46]); MULADD(at[2], at[45]); MULADD(at[3], at[44]); MULADD(at[4], at[43]); MULADD(at[5], at[42]); MULADD(at[6], at[41]); MULADD(at[7], at[40]); MULADD(at[8], at[39]); MULADD(at[9], at[38]); MULADD(at[10], at[37]); MULADD(at[11], at[36]); MULADD(at[12], at[35]); MULADD(at[13], at[34]); MULADD(at[14], at[33]); MULADD(at[15], at[32]); MULADD(at[16], at[31]); MULADD(at[17], at[30]); MULADD(at[18], at[29]); MULADD(at[19], at[28]); MULADD(at[20], at[27]); MULADD(at[21], at[26]); MULADD(at[22], at[25]); MULADD(at[23], at[24]); + COMBA_STORE(C->dp[23]); + /* 24 */ + COMBA_FORWARD; + MULADD(at[1], at[47]); MULADD(at[2], at[46]); MULADD(at[3], at[45]); MULADD(at[4], at[44]); MULADD(at[5], at[43]); MULADD(at[6], at[42]); MULADD(at[7], at[41]); MULADD(at[8], at[40]); MULADD(at[9], at[39]); MULADD(at[10], at[38]); MULADD(at[11], at[37]); MULADD(at[12], at[36]); MULADD(at[13], at[35]); MULADD(at[14], at[34]); MULADD(at[15], at[33]); MULADD(at[16], at[32]); MULADD(at[17], at[31]); MULADD(at[18], at[30]); MULADD(at[19], at[29]); MULADD(at[20], at[28]); MULADD(at[21], at[27]); MULADD(at[22], at[26]); MULADD(at[23], at[25]); + COMBA_STORE(C->dp[24]); + /* 25 */ + COMBA_FORWARD; + MULADD(at[2], at[47]); MULADD(at[3], at[46]); MULADD(at[4], at[45]); MULADD(at[5], at[44]); MULADD(at[6], at[43]); MULADD(at[7], at[42]); MULADD(at[8], at[41]); MULADD(at[9], at[40]); MULADD(at[10], at[39]); MULADD(at[11], at[38]); MULADD(at[12], at[37]); MULADD(at[13], at[36]); MULADD(at[14], at[35]); MULADD(at[15], at[34]); MULADD(at[16], at[33]); MULADD(at[17], at[32]); MULADD(at[18], at[31]); MULADD(at[19], at[30]); MULADD(at[20], at[29]); MULADD(at[21], at[28]); MULADD(at[22], at[27]); MULADD(at[23], at[26]); + COMBA_STORE(C->dp[25]); + /* 26 */ + COMBA_FORWARD; + MULADD(at[3], at[47]); MULADD(at[4], at[46]); MULADD(at[5], at[45]); MULADD(at[6], at[44]); MULADD(at[7], at[43]); MULADD(at[8], at[42]); MULADD(at[9], at[41]); MULADD(at[10], at[40]); MULADD(at[11], at[39]); MULADD(at[12], at[38]); MULADD(at[13], at[37]); MULADD(at[14], at[36]); MULADD(at[15], at[35]); MULADD(at[16], at[34]); MULADD(at[17], at[33]); MULADD(at[18], at[32]); MULADD(at[19], at[31]); MULADD(at[20], at[30]); MULADD(at[21], at[29]); MULADD(at[22], at[28]); MULADD(at[23], at[27]); + COMBA_STORE(C->dp[26]); + /* 27 */ + COMBA_FORWARD; + MULADD(at[4], at[47]); MULADD(at[5], at[46]); MULADD(at[6], at[45]); MULADD(at[7], at[44]); MULADD(at[8], at[43]); MULADD(at[9], at[42]); MULADD(at[10], at[41]); MULADD(at[11], at[40]); MULADD(at[12], at[39]); MULADD(at[13], at[38]); MULADD(at[14], at[37]); MULADD(at[15], at[36]); MULADD(at[16], at[35]); MULADD(at[17], at[34]); MULADD(at[18], at[33]); MULADD(at[19], at[32]); MULADD(at[20], at[31]); MULADD(at[21], at[30]); MULADD(at[22], at[29]); MULADD(at[23], at[28]); + COMBA_STORE(C->dp[27]); + /* 28 */ + COMBA_FORWARD; + MULADD(at[5], at[47]); MULADD(at[6], at[46]); MULADD(at[7], at[45]); MULADD(at[8], at[44]); MULADD(at[9], at[43]); MULADD(at[10], at[42]); MULADD(at[11], at[41]); MULADD(at[12], at[40]); MULADD(at[13], at[39]); MULADD(at[14], at[38]); MULADD(at[15], at[37]); MULADD(at[16], at[36]); MULADD(at[17], at[35]); MULADD(at[18], at[34]); MULADD(at[19], at[33]); MULADD(at[20], at[32]); MULADD(at[21], at[31]); MULADD(at[22], at[30]); MULADD(at[23], at[29]); + COMBA_STORE(C->dp[28]); + /* 29 */ + COMBA_FORWARD; + MULADD(at[6], at[47]); MULADD(at[7], at[46]); MULADD(at[8], at[45]); MULADD(at[9], at[44]); MULADD(at[10], at[43]); MULADD(at[11], at[42]); MULADD(at[12], at[41]); MULADD(at[13], at[40]); MULADD(at[14], at[39]); MULADD(at[15], at[38]); MULADD(at[16], at[37]); MULADD(at[17], at[36]); MULADD(at[18], at[35]); MULADD(at[19], at[34]); MULADD(at[20], at[33]); MULADD(at[21], at[32]); MULADD(at[22], at[31]); MULADD(at[23], at[30]); + COMBA_STORE(C->dp[29]); + /* 30 */ + COMBA_FORWARD; + MULADD(at[7], at[47]); MULADD(at[8], at[46]); MULADD(at[9], at[45]); MULADD(at[10], at[44]); MULADD(at[11], at[43]); MULADD(at[12], at[42]); MULADD(at[13], at[41]); MULADD(at[14], at[40]); MULADD(at[15], at[39]); MULADD(at[16], at[38]); MULADD(at[17], at[37]); MULADD(at[18], at[36]); MULADD(at[19], at[35]); MULADD(at[20], at[34]); MULADD(at[21], at[33]); MULADD(at[22], at[32]); MULADD(at[23], at[31]); + COMBA_STORE(C->dp[30]); + /* 31 */ + COMBA_FORWARD; + MULADD(at[8], at[47]); MULADD(at[9], at[46]); MULADD(at[10], at[45]); MULADD(at[11], at[44]); MULADD(at[12], at[43]); MULADD(at[13], at[42]); MULADD(at[14], at[41]); MULADD(at[15], at[40]); MULADD(at[16], at[39]); MULADD(at[17], at[38]); MULADD(at[18], at[37]); MULADD(at[19], at[36]); MULADD(at[20], at[35]); MULADD(at[21], at[34]); MULADD(at[22], at[33]); MULADD(at[23], at[32]); + COMBA_STORE(C->dp[31]); + /* 32 */ + COMBA_FORWARD; + MULADD(at[9], at[47]); MULADD(at[10], at[46]); MULADD(at[11], at[45]); MULADD(at[12], at[44]); MULADD(at[13], at[43]); MULADD(at[14], at[42]); MULADD(at[15], at[41]); MULADD(at[16], at[40]); MULADD(at[17], at[39]); MULADD(at[18], at[38]); MULADD(at[19], at[37]); MULADD(at[20], at[36]); MULADD(at[21], at[35]); MULADD(at[22], at[34]); MULADD(at[23], at[33]); + COMBA_STORE(C->dp[32]); + /* 33 */ + COMBA_FORWARD; + MULADD(at[10], at[47]); MULADD(at[11], at[46]); MULADD(at[12], at[45]); MULADD(at[13], at[44]); MULADD(at[14], at[43]); MULADD(at[15], at[42]); MULADD(at[16], at[41]); MULADD(at[17], at[40]); MULADD(at[18], at[39]); MULADD(at[19], at[38]); MULADD(at[20], at[37]); MULADD(at[21], at[36]); MULADD(at[22], at[35]); MULADD(at[23], at[34]); + COMBA_STORE(C->dp[33]); + /* 34 */ + COMBA_FORWARD; + MULADD(at[11], at[47]); MULADD(at[12], at[46]); MULADD(at[13], at[45]); MULADD(at[14], at[44]); MULADD(at[15], at[43]); MULADD(at[16], at[42]); MULADD(at[17], at[41]); MULADD(at[18], at[40]); MULADD(at[19], at[39]); MULADD(at[20], at[38]); MULADD(at[21], at[37]); MULADD(at[22], at[36]); MULADD(at[23], at[35]); + COMBA_STORE(C->dp[34]); + /* 35 */ + COMBA_FORWARD; + MULADD(at[12], at[47]); MULADD(at[13], at[46]); MULADD(at[14], at[45]); MULADD(at[15], at[44]); MULADD(at[16], at[43]); MULADD(at[17], at[42]); MULADD(at[18], at[41]); MULADD(at[19], at[40]); MULADD(at[20], at[39]); MULADD(at[21], at[38]); MULADD(at[22], at[37]); MULADD(at[23], at[36]); + COMBA_STORE(C->dp[35]); + /* 36 */ + COMBA_FORWARD; + MULADD(at[13], at[47]); MULADD(at[14], at[46]); MULADD(at[15], at[45]); MULADD(at[16], at[44]); MULADD(at[17], at[43]); MULADD(at[18], at[42]); MULADD(at[19], at[41]); MULADD(at[20], at[40]); MULADD(at[21], at[39]); MULADD(at[22], at[38]); MULADD(at[23], at[37]); + COMBA_STORE(C->dp[36]); + /* 37 */ + COMBA_FORWARD; + MULADD(at[14], at[47]); MULADD(at[15], at[46]); MULADD(at[16], at[45]); MULADD(at[17], at[44]); MULADD(at[18], at[43]); MULADD(at[19], at[42]); MULADD(at[20], at[41]); MULADD(at[21], at[40]); MULADD(at[22], at[39]); MULADD(at[23], at[38]); + COMBA_STORE(C->dp[37]); + /* 38 */ + COMBA_FORWARD; + MULADD(at[15], at[47]); MULADD(at[16], at[46]); MULADD(at[17], at[45]); MULADD(at[18], at[44]); MULADD(at[19], at[43]); MULADD(at[20], at[42]); MULADD(at[21], at[41]); MULADD(at[22], at[40]); MULADD(at[23], at[39]); + COMBA_STORE(C->dp[38]); + /* 39 */ + COMBA_FORWARD; + MULADD(at[16], at[47]); MULADD(at[17], at[46]); MULADD(at[18], at[45]); MULADD(at[19], at[44]); MULADD(at[20], at[43]); MULADD(at[21], at[42]); MULADD(at[22], at[41]); MULADD(at[23], at[40]); + COMBA_STORE(C->dp[39]); + /* 40 */ + COMBA_FORWARD; + MULADD(at[17], at[47]); MULADD(at[18], at[46]); MULADD(at[19], at[45]); MULADD(at[20], at[44]); MULADD(at[21], at[43]); MULADD(at[22], at[42]); MULADD(at[23], at[41]); + COMBA_STORE(C->dp[40]); + /* 41 */ + COMBA_FORWARD; + MULADD(at[18], at[47]); MULADD(at[19], at[46]); MULADD(at[20], at[45]); MULADD(at[21], at[44]); MULADD(at[22], at[43]); MULADD(at[23], at[42]); + COMBA_STORE(C->dp[41]); + /* 42 */ + COMBA_FORWARD; + MULADD(at[19], at[47]); MULADD(at[20], at[46]); MULADD(at[21], at[45]); MULADD(at[22], at[44]); MULADD(at[23], at[43]); + COMBA_STORE(C->dp[42]); + /* 43 */ + COMBA_FORWARD; + MULADD(at[20], at[47]); MULADD(at[21], at[46]); MULADD(at[22], at[45]); MULADD(at[23], at[44]); + COMBA_STORE(C->dp[43]); + /* 44 */ + COMBA_FORWARD; + MULADD(at[21], at[47]); MULADD(at[22], at[46]); MULADD(at[23], at[45]); + COMBA_STORE(C->dp[44]); + /* 45 */ + COMBA_FORWARD; + MULADD(at[22], at[47]); MULADD(at[23], at[46]); + COMBA_STORE(C->dp[45]); + /* 46 */ + COMBA_FORWARD; + MULADD(at[23], at[47]); + COMBA_STORE(C->dp[46]); + COMBA_STORE2(C->dp[47]); + C->used = 48; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; +} +#endif + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mul_comba_28.c b/src/crypto/tfm/fp_mul_comba_28.c new file mode 100644 index 0000000..d72c08e --- /dev/null +++ b/src/crypto/tfm/fp_mul_comba_28.c @@ -0,0 +1,244 @@ +#define TFM_DEFINES +#include "fp_mul_comba.c" + +#if defined(TFM_MUL28) && FP_SIZE >= 56 +void fp_mul_comba28(fp_int *A, fp_int *B, fp_int *C) +{ + fp_digit c0, c1, c2, at[56]; + + memcpy(at, A->dp, 28 * sizeof(fp_digit)); + memcpy(at+28, B->dp, 28 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[28]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[29]); MULADD(at[1], at[28]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[30]); MULADD(at[1], at[29]); MULADD(at[2], at[28]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[31]); MULADD(at[1], at[30]); MULADD(at[2], at[29]); MULADD(at[3], at[28]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[32]); MULADD(at[1], at[31]); MULADD(at[2], at[30]); MULADD(at[3], at[29]); MULADD(at[4], at[28]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[33]); MULADD(at[1], at[32]); MULADD(at[2], at[31]); MULADD(at[3], at[30]); MULADD(at[4], at[29]); MULADD(at[5], at[28]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[34]); MULADD(at[1], at[33]); MULADD(at[2], at[32]); MULADD(at[3], at[31]); MULADD(at[4], at[30]); MULADD(at[5], at[29]); MULADD(at[6], at[28]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[35]); MULADD(at[1], at[34]); MULADD(at[2], at[33]); MULADD(at[3], at[32]); MULADD(at[4], at[31]); MULADD(at[5], at[30]); MULADD(at[6], at[29]); MULADD(at[7], at[28]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[36]); MULADD(at[1], at[35]); MULADD(at[2], at[34]); MULADD(at[3], at[33]); MULADD(at[4], at[32]); MULADD(at[5], at[31]); MULADD(at[6], at[30]); MULADD(at[7], at[29]); MULADD(at[8], at[28]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[37]); MULADD(at[1], at[36]); MULADD(at[2], at[35]); MULADD(at[3], at[34]); MULADD(at[4], at[33]); MULADD(at[5], at[32]); MULADD(at[6], at[31]); MULADD(at[7], at[30]); MULADD(at[8], at[29]); MULADD(at[9], at[28]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[0], at[38]); MULADD(at[1], at[37]); MULADD(at[2], at[36]); MULADD(at[3], at[35]); MULADD(at[4], at[34]); MULADD(at[5], at[33]); MULADD(at[6], at[32]); MULADD(at[7], at[31]); MULADD(at[8], at[30]); MULADD(at[9], at[29]); MULADD(at[10], at[28]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[0], at[39]); MULADD(at[1], at[38]); MULADD(at[2], at[37]); MULADD(at[3], at[36]); MULADD(at[4], at[35]); MULADD(at[5], at[34]); MULADD(at[6], at[33]); MULADD(at[7], at[32]); MULADD(at[8], at[31]); MULADD(at[9], at[30]); MULADD(at[10], at[29]); MULADD(at[11], at[28]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[0], at[40]); MULADD(at[1], at[39]); MULADD(at[2], at[38]); MULADD(at[3], at[37]); MULADD(at[4], at[36]); MULADD(at[5], at[35]); MULADD(at[6], at[34]); MULADD(at[7], at[33]); MULADD(at[8], at[32]); MULADD(at[9], at[31]); MULADD(at[10], at[30]); MULADD(at[11], at[29]); MULADD(at[12], at[28]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[0], at[41]); MULADD(at[1], at[40]); MULADD(at[2], at[39]); MULADD(at[3], at[38]); MULADD(at[4], at[37]); MULADD(at[5], at[36]); MULADD(at[6], at[35]); MULADD(at[7], at[34]); MULADD(at[8], at[33]); MULADD(at[9], at[32]); MULADD(at[10], at[31]); MULADD(at[11], at[30]); MULADD(at[12], at[29]); MULADD(at[13], at[28]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[0], at[42]); MULADD(at[1], at[41]); MULADD(at[2], at[40]); MULADD(at[3], at[39]); MULADD(at[4], at[38]); MULADD(at[5], at[37]); MULADD(at[6], at[36]); MULADD(at[7], at[35]); MULADD(at[8], at[34]); MULADD(at[9], at[33]); MULADD(at[10], at[32]); MULADD(at[11], at[31]); MULADD(at[12], at[30]); MULADD(at[13], at[29]); MULADD(at[14], at[28]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[0], at[43]); MULADD(at[1], at[42]); MULADD(at[2], at[41]); MULADD(at[3], at[40]); MULADD(at[4], at[39]); MULADD(at[5], at[38]); MULADD(at[6], at[37]); MULADD(at[7], at[36]); MULADD(at[8], at[35]); MULADD(at[9], at[34]); MULADD(at[10], at[33]); MULADD(at[11], at[32]); MULADD(at[12], at[31]); MULADD(at[13], at[30]); MULADD(at[14], at[29]); MULADD(at[15], at[28]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[0], at[44]); MULADD(at[1], at[43]); MULADD(at[2], at[42]); MULADD(at[3], at[41]); MULADD(at[4], at[40]); MULADD(at[5], at[39]); MULADD(at[6], at[38]); MULADD(at[7], at[37]); MULADD(at[8], at[36]); MULADD(at[9], at[35]); MULADD(at[10], at[34]); MULADD(at[11], at[33]); MULADD(at[12], at[32]); MULADD(at[13], at[31]); MULADD(at[14], at[30]); MULADD(at[15], at[29]); MULADD(at[16], at[28]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[0], at[45]); MULADD(at[1], at[44]); MULADD(at[2], at[43]); MULADD(at[3], at[42]); MULADD(at[4], at[41]); MULADD(at[5], at[40]); MULADD(at[6], at[39]); MULADD(at[7], at[38]); MULADD(at[8], at[37]); MULADD(at[9], at[36]); MULADD(at[10], at[35]); MULADD(at[11], at[34]); MULADD(at[12], at[33]); MULADD(at[13], at[32]); MULADD(at[14], at[31]); MULADD(at[15], at[30]); MULADD(at[16], at[29]); MULADD(at[17], at[28]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[0], at[46]); MULADD(at[1], at[45]); MULADD(at[2], at[44]); MULADD(at[3], at[43]); MULADD(at[4], at[42]); MULADD(at[5], at[41]); MULADD(at[6], at[40]); MULADD(at[7], at[39]); MULADD(at[8], at[38]); MULADD(at[9], at[37]); MULADD(at[10], at[36]); MULADD(at[11], at[35]); MULADD(at[12], at[34]); MULADD(at[13], at[33]); MULADD(at[14], at[32]); MULADD(at[15], at[31]); MULADD(at[16], at[30]); MULADD(at[17], at[29]); MULADD(at[18], at[28]); + COMBA_STORE(C->dp[18]); + /* 19 */ + COMBA_FORWARD; + MULADD(at[0], at[47]); MULADD(at[1], at[46]); MULADD(at[2], at[45]); MULADD(at[3], at[44]); MULADD(at[4], at[43]); MULADD(at[5], at[42]); MULADD(at[6], at[41]); MULADD(at[7], at[40]); MULADD(at[8], at[39]); MULADD(at[9], at[38]); MULADD(at[10], at[37]); MULADD(at[11], at[36]); MULADD(at[12], at[35]); MULADD(at[13], at[34]); MULADD(at[14], at[33]); MULADD(at[15], at[32]); MULADD(at[16], at[31]); MULADD(at[17], at[30]); MULADD(at[18], at[29]); MULADD(at[19], at[28]); + COMBA_STORE(C->dp[19]); + /* 20 */ + COMBA_FORWARD; + MULADD(at[0], at[48]); MULADD(at[1], at[47]); MULADD(at[2], at[46]); MULADD(at[3], at[45]); MULADD(at[4], at[44]); MULADD(at[5], at[43]); MULADD(at[6], at[42]); MULADD(at[7], at[41]); MULADD(at[8], at[40]); MULADD(at[9], at[39]); MULADD(at[10], at[38]); MULADD(at[11], at[37]); MULADD(at[12], at[36]); MULADD(at[13], at[35]); MULADD(at[14], at[34]); MULADD(at[15], at[33]); MULADD(at[16], at[32]); MULADD(at[17], at[31]); MULADD(at[18], at[30]); MULADD(at[19], at[29]); MULADD(at[20], at[28]); + COMBA_STORE(C->dp[20]); + /* 21 */ + COMBA_FORWARD; + MULADD(at[0], at[49]); MULADD(at[1], at[48]); MULADD(at[2], at[47]); MULADD(at[3], at[46]); MULADD(at[4], at[45]); MULADD(at[5], at[44]); MULADD(at[6], at[43]); MULADD(at[7], at[42]); MULADD(at[8], at[41]); MULADD(at[9], at[40]); MULADD(at[10], at[39]); MULADD(at[11], at[38]); MULADD(at[12], at[37]); MULADD(at[13], at[36]); MULADD(at[14], at[35]); MULADD(at[15], at[34]); MULADD(at[16], at[33]); MULADD(at[17], at[32]); MULADD(at[18], at[31]); MULADD(at[19], at[30]); MULADD(at[20], at[29]); MULADD(at[21], at[28]); + COMBA_STORE(C->dp[21]); + /* 22 */ + COMBA_FORWARD; + MULADD(at[0], at[50]); MULADD(at[1], at[49]); MULADD(at[2], at[48]); MULADD(at[3], at[47]); MULADD(at[4], at[46]); MULADD(at[5], at[45]); MULADD(at[6], at[44]); MULADD(at[7], at[43]); MULADD(at[8], at[42]); MULADD(at[9], at[41]); MULADD(at[10], at[40]); MULADD(at[11], at[39]); MULADD(at[12], at[38]); MULADD(at[13], at[37]); MULADD(at[14], at[36]); MULADD(at[15], at[35]); MULADD(at[16], at[34]); MULADD(at[17], at[33]); MULADD(at[18], at[32]); MULADD(at[19], at[31]); MULADD(at[20], at[30]); MULADD(at[21], at[29]); MULADD(at[22], at[28]); + COMBA_STORE(C->dp[22]); + /* 23 */ + COMBA_FORWARD; + MULADD(at[0], at[51]); MULADD(at[1], at[50]); MULADD(at[2], at[49]); MULADD(at[3], at[48]); MULADD(at[4], at[47]); MULADD(at[5], at[46]); MULADD(at[6], at[45]); MULADD(at[7], at[44]); MULADD(at[8], at[43]); MULADD(at[9], at[42]); MULADD(at[10], at[41]); MULADD(at[11], at[40]); MULADD(at[12], at[39]); MULADD(at[13], at[38]); MULADD(at[14], at[37]); MULADD(at[15], at[36]); MULADD(at[16], at[35]); MULADD(at[17], at[34]); MULADD(at[18], at[33]); MULADD(at[19], at[32]); MULADD(at[20], at[31]); MULADD(at[21], at[30]); MULADD(at[22], at[29]); MULADD(at[23], at[28]); + COMBA_STORE(C->dp[23]); + /* 24 */ + COMBA_FORWARD; + MULADD(at[0], at[52]); MULADD(at[1], at[51]); MULADD(at[2], at[50]); MULADD(at[3], at[49]); MULADD(at[4], at[48]); MULADD(at[5], at[47]); MULADD(at[6], at[46]); MULADD(at[7], at[45]); MULADD(at[8], at[44]); MULADD(at[9], at[43]); MULADD(at[10], at[42]); MULADD(at[11], at[41]); MULADD(at[12], at[40]); MULADD(at[13], at[39]); MULADD(at[14], at[38]); MULADD(at[15], at[37]); MULADD(at[16], at[36]); MULADD(at[17], at[35]); MULADD(at[18], at[34]); MULADD(at[19], at[33]); MULADD(at[20], at[32]); MULADD(at[21], at[31]); MULADD(at[22], at[30]); MULADD(at[23], at[29]); MULADD(at[24], at[28]); + COMBA_STORE(C->dp[24]); + /* 25 */ + COMBA_FORWARD; + MULADD(at[0], at[53]); MULADD(at[1], at[52]); MULADD(at[2], at[51]); MULADD(at[3], at[50]); MULADD(at[4], at[49]); MULADD(at[5], at[48]); MULADD(at[6], at[47]); MULADD(at[7], at[46]); MULADD(at[8], at[45]); MULADD(at[9], at[44]); MULADD(at[10], at[43]); MULADD(at[11], at[42]); MULADD(at[12], at[41]); MULADD(at[13], at[40]); MULADD(at[14], at[39]); MULADD(at[15], at[38]); MULADD(at[16], at[37]); MULADD(at[17], at[36]); MULADD(at[18], at[35]); MULADD(at[19], at[34]); MULADD(at[20], at[33]); MULADD(at[21], at[32]); MULADD(at[22], at[31]); MULADD(at[23], at[30]); MULADD(at[24], at[29]); MULADD(at[25], at[28]); + COMBA_STORE(C->dp[25]); + /* 26 */ + COMBA_FORWARD; + MULADD(at[0], at[54]); MULADD(at[1], at[53]); MULADD(at[2], at[52]); MULADD(at[3], at[51]); MULADD(at[4], at[50]); MULADD(at[5], at[49]); MULADD(at[6], at[48]); MULADD(at[7], at[47]); MULADD(at[8], at[46]); MULADD(at[9], at[45]); MULADD(at[10], at[44]); MULADD(at[11], at[43]); MULADD(at[12], at[42]); MULADD(at[13], at[41]); MULADD(at[14], at[40]); MULADD(at[15], at[39]); MULADD(at[16], at[38]); MULADD(at[17], at[37]); MULADD(at[18], at[36]); MULADD(at[19], at[35]); MULADD(at[20], at[34]); MULADD(at[21], at[33]); MULADD(at[22], at[32]); MULADD(at[23], at[31]); MULADD(at[24], at[30]); MULADD(at[25], at[29]); MULADD(at[26], at[28]); + COMBA_STORE(C->dp[26]); + /* 27 */ + COMBA_FORWARD; + MULADD(at[0], at[55]); MULADD(at[1], at[54]); MULADD(at[2], at[53]); MULADD(at[3], at[52]); MULADD(at[4], at[51]); MULADD(at[5], at[50]); MULADD(at[6], at[49]); MULADD(at[7], at[48]); MULADD(at[8], at[47]); MULADD(at[9], at[46]); MULADD(at[10], at[45]); MULADD(at[11], at[44]); MULADD(at[12], at[43]); MULADD(at[13], at[42]); MULADD(at[14], at[41]); MULADD(at[15], at[40]); MULADD(at[16], at[39]); MULADD(at[17], at[38]); MULADD(at[18], at[37]); MULADD(at[19], at[36]); MULADD(at[20], at[35]); MULADD(at[21], at[34]); MULADD(at[22], at[33]); MULADD(at[23], at[32]); MULADD(at[24], at[31]); MULADD(at[25], at[30]); MULADD(at[26], at[29]); MULADD(at[27], at[28]); + COMBA_STORE(C->dp[27]); + /* 28 */ + COMBA_FORWARD; + MULADD(at[1], at[55]); MULADD(at[2], at[54]); MULADD(at[3], at[53]); MULADD(at[4], at[52]); MULADD(at[5], at[51]); MULADD(at[6], at[50]); MULADD(at[7], at[49]); MULADD(at[8], at[48]); MULADD(at[9], at[47]); MULADD(at[10], at[46]); MULADD(at[11], at[45]); MULADD(at[12], at[44]); MULADD(at[13], at[43]); MULADD(at[14], at[42]); MULADD(at[15], at[41]); MULADD(at[16], at[40]); MULADD(at[17], at[39]); MULADD(at[18], at[38]); MULADD(at[19], at[37]); MULADD(at[20], at[36]); MULADD(at[21], at[35]); MULADD(at[22], at[34]); MULADD(at[23], at[33]); MULADD(at[24], at[32]); MULADD(at[25], at[31]); MULADD(at[26], at[30]); MULADD(at[27], at[29]); + COMBA_STORE(C->dp[28]); + /* 29 */ + COMBA_FORWARD; + MULADD(at[2], at[55]); MULADD(at[3], at[54]); MULADD(at[4], at[53]); MULADD(at[5], at[52]); MULADD(at[6], at[51]); MULADD(at[7], at[50]); MULADD(at[8], at[49]); MULADD(at[9], at[48]); MULADD(at[10], at[47]); MULADD(at[11], at[46]); MULADD(at[12], at[45]); MULADD(at[13], at[44]); MULADD(at[14], at[43]); MULADD(at[15], at[42]); MULADD(at[16], at[41]); MULADD(at[17], at[40]); MULADD(at[18], at[39]); MULADD(at[19], at[38]); MULADD(at[20], at[37]); MULADD(at[21], at[36]); MULADD(at[22], at[35]); MULADD(at[23], at[34]); MULADD(at[24], at[33]); MULADD(at[25], at[32]); MULADD(at[26], at[31]); MULADD(at[27], at[30]); + COMBA_STORE(C->dp[29]); + /* 30 */ + COMBA_FORWARD; + MULADD(at[3], at[55]); MULADD(at[4], at[54]); MULADD(at[5], at[53]); MULADD(at[6], at[52]); MULADD(at[7], at[51]); MULADD(at[8], at[50]); MULADD(at[9], at[49]); MULADD(at[10], at[48]); MULADD(at[11], at[47]); MULADD(at[12], at[46]); MULADD(at[13], at[45]); MULADD(at[14], at[44]); MULADD(at[15], at[43]); MULADD(at[16], at[42]); MULADD(at[17], at[41]); MULADD(at[18], at[40]); MULADD(at[19], at[39]); MULADD(at[20], at[38]); MULADD(at[21], at[37]); MULADD(at[22], at[36]); MULADD(at[23], at[35]); MULADD(at[24], at[34]); MULADD(at[25], at[33]); MULADD(at[26], at[32]); MULADD(at[27], at[31]); + COMBA_STORE(C->dp[30]); + /* 31 */ + COMBA_FORWARD; + MULADD(at[4], at[55]); MULADD(at[5], at[54]); MULADD(at[6], at[53]); MULADD(at[7], at[52]); MULADD(at[8], at[51]); MULADD(at[9], at[50]); MULADD(at[10], at[49]); MULADD(at[11], at[48]); MULADD(at[12], at[47]); MULADD(at[13], at[46]); MULADD(at[14], at[45]); MULADD(at[15], at[44]); MULADD(at[16], at[43]); MULADD(at[17], at[42]); MULADD(at[18], at[41]); MULADD(at[19], at[40]); MULADD(at[20], at[39]); MULADD(at[21], at[38]); MULADD(at[22], at[37]); MULADD(at[23], at[36]); MULADD(at[24], at[35]); MULADD(at[25], at[34]); MULADD(at[26], at[33]); MULADD(at[27], at[32]); + COMBA_STORE(C->dp[31]); + /* 32 */ + COMBA_FORWARD; + MULADD(at[5], at[55]); MULADD(at[6], at[54]); MULADD(at[7], at[53]); MULADD(at[8], at[52]); MULADD(at[9], at[51]); MULADD(at[10], at[50]); MULADD(at[11], at[49]); MULADD(at[12], at[48]); MULADD(at[13], at[47]); MULADD(at[14], at[46]); MULADD(at[15], at[45]); MULADD(at[16], at[44]); MULADD(at[17], at[43]); MULADD(at[18], at[42]); MULADD(at[19], at[41]); MULADD(at[20], at[40]); MULADD(at[21], at[39]); MULADD(at[22], at[38]); MULADD(at[23], at[37]); MULADD(at[24], at[36]); MULADD(at[25], at[35]); MULADD(at[26], at[34]); MULADD(at[27], at[33]); + COMBA_STORE(C->dp[32]); + /* 33 */ + COMBA_FORWARD; + MULADD(at[6], at[55]); MULADD(at[7], at[54]); MULADD(at[8], at[53]); MULADD(at[9], at[52]); MULADD(at[10], at[51]); MULADD(at[11], at[50]); MULADD(at[12], at[49]); MULADD(at[13], at[48]); MULADD(at[14], at[47]); MULADD(at[15], at[46]); MULADD(at[16], at[45]); MULADD(at[17], at[44]); MULADD(at[18], at[43]); MULADD(at[19], at[42]); MULADD(at[20], at[41]); MULADD(at[21], at[40]); MULADD(at[22], at[39]); MULADD(at[23], at[38]); MULADD(at[24], at[37]); MULADD(at[25], at[36]); MULADD(at[26], at[35]); MULADD(at[27], at[34]); + COMBA_STORE(C->dp[33]); + /* 34 */ + COMBA_FORWARD; + MULADD(at[7], at[55]); MULADD(at[8], at[54]); MULADD(at[9], at[53]); MULADD(at[10], at[52]); MULADD(at[11], at[51]); MULADD(at[12], at[50]); MULADD(at[13], at[49]); MULADD(at[14], at[48]); MULADD(at[15], at[47]); MULADD(at[16], at[46]); MULADD(at[17], at[45]); MULADD(at[18], at[44]); MULADD(at[19], at[43]); MULADD(at[20], at[42]); MULADD(at[21], at[41]); MULADD(at[22], at[40]); MULADD(at[23], at[39]); MULADD(at[24], at[38]); MULADD(at[25], at[37]); MULADD(at[26], at[36]); MULADD(at[27], at[35]); + COMBA_STORE(C->dp[34]); + /* 35 */ + COMBA_FORWARD; + MULADD(at[8], at[55]); MULADD(at[9], at[54]); MULADD(at[10], at[53]); MULADD(at[11], at[52]); MULADD(at[12], at[51]); MULADD(at[13], at[50]); MULADD(at[14], at[49]); MULADD(at[15], at[48]); MULADD(at[16], at[47]); MULADD(at[17], at[46]); MULADD(at[18], at[45]); MULADD(at[19], at[44]); MULADD(at[20], at[43]); MULADD(at[21], at[42]); MULADD(at[22], at[41]); MULADD(at[23], at[40]); MULADD(at[24], at[39]); MULADD(at[25], at[38]); MULADD(at[26], at[37]); MULADD(at[27], at[36]); + COMBA_STORE(C->dp[35]); + /* 36 */ + COMBA_FORWARD; + MULADD(at[9], at[55]); MULADD(at[10], at[54]); MULADD(at[11], at[53]); MULADD(at[12], at[52]); MULADD(at[13], at[51]); MULADD(at[14], at[50]); MULADD(at[15], at[49]); MULADD(at[16], at[48]); MULADD(at[17], at[47]); MULADD(at[18], at[46]); MULADD(at[19], at[45]); MULADD(at[20], at[44]); MULADD(at[21], at[43]); MULADD(at[22], at[42]); MULADD(at[23], at[41]); MULADD(at[24], at[40]); MULADD(at[25], at[39]); MULADD(at[26], at[38]); MULADD(at[27], at[37]); + COMBA_STORE(C->dp[36]); + /* 37 */ + COMBA_FORWARD; + MULADD(at[10], at[55]); MULADD(at[11], at[54]); MULADD(at[12], at[53]); MULADD(at[13], at[52]); MULADD(at[14], at[51]); MULADD(at[15], at[50]); MULADD(at[16], at[49]); MULADD(at[17], at[48]); MULADD(at[18], at[47]); MULADD(at[19], at[46]); MULADD(at[20], at[45]); MULADD(at[21], at[44]); MULADD(at[22], at[43]); MULADD(at[23], at[42]); MULADD(at[24], at[41]); MULADD(at[25], at[40]); MULADD(at[26], at[39]); MULADD(at[27], at[38]); + COMBA_STORE(C->dp[37]); + /* 38 */ + COMBA_FORWARD; + MULADD(at[11], at[55]); MULADD(at[12], at[54]); MULADD(at[13], at[53]); MULADD(at[14], at[52]); MULADD(at[15], at[51]); MULADD(at[16], at[50]); MULADD(at[17], at[49]); MULADD(at[18], at[48]); MULADD(at[19], at[47]); MULADD(at[20], at[46]); MULADD(at[21], at[45]); MULADD(at[22], at[44]); MULADD(at[23], at[43]); MULADD(at[24], at[42]); MULADD(at[25], at[41]); MULADD(at[26], at[40]); MULADD(at[27], at[39]); + COMBA_STORE(C->dp[38]); + /* 39 */ + COMBA_FORWARD; + MULADD(at[12], at[55]); MULADD(at[13], at[54]); MULADD(at[14], at[53]); MULADD(at[15], at[52]); MULADD(at[16], at[51]); MULADD(at[17], at[50]); MULADD(at[18], at[49]); MULADD(at[19], at[48]); MULADD(at[20], at[47]); MULADD(at[21], at[46]); MULADD(at[22], at[45]); MULADD(at[23], at[44]); MULADD(at[24], at[43]); MULADD(at[25], at[42]); MULADD(at[26], at[41]); MULADD(at[27], at[40]); + COMBA_STORE(C->dp[39]); + /* 40 */ + COMBA_FORWARD; + MULADD(at[13], at[55]); MULADD(at[14], at[54]); MULADD(at[15], at[53]); MULADD(at[16], at[52]); MULADD(at[17], at[51]); MULADD(at[18], at[50]); MULADD(at[19], at[49]); MULADD(at[20], at[48]); MULADD(at[21], at[47]); MULADD(at[22], at[46]); MULADD(at[23], at[45]); MULADD(at[24], at[44]); MULADD(at[25], at[43]); MULADD(at[26], at[42]); MULADD(at[27], at[41]); + COMBA_STORE(C->dp[40]); + /* 41 */ + COMBA_FORWARD; + MULADD(at[14], at[55]); MULADD(at[15], at[54]); MULADD(at[16], at[53]); MULADD(at[17], at[52]); MULADD(at[18], at[51]); MULADD(at[19], at[50]); MULADD(at[20], at[49]); MULADD(at[21], at[48]); MULADD(at[22], at[47]); MULADD(at[23], at[46]); MULADD(at[24], at[45]); MULADD(at[25], at[44]); MULADD(at[26], at[43]); MULADD(at[27], at[42]); + COMBA_STORE(C->dp[41]); + /* 42 */ + COMBA_FORWARD; + MULADD(at[15], at[55]); MULADD(at[16], at[54]); MULADD(at[17], at[53]); MULADD(at[18], at[52]); MULADD(at[19], at[51]); MULADD(at[20], at[50]); MULADD(at[21], at[49]); MULADD(at[22], at[48]); MULADD(at[23], at[47]); MULADD(at[24], at[46]); MULADD(at[25], at[45]); MULADD(at[26], at[44]); MULADD(at[27], at[43]); + COMBA_STORE(C->dp[42]); + /* 43 */ + COMBA_FORWARD; + MULADD(at[16], at[55]); MULADD(at[17], at[54]); MULADD(at[18], at[53]); MULADD(at[19], at[52]); MULADD(at[20], at[51]); MULADD(at[21], at[50]); MULADD(at[22], at[49]); MULADD(at[23], at[48]); MULADD(at[24], at[47]); MULADD(at[25], at[46]); MULADD(at[26], at[45]); MULADD(at[27], at[44]); + COMBA_STORE(C->dp[43]); + /* 44 */ + COMBA_FORWARD; + MULADD(at[17], at[55]); MULADD(at[18], at[54]); MULADD(at[19], at[53]); MULADD(at[20], at[52]); MULADD(at[21], at[51]); MULADD(at[22], at[50]); MULADD(at[23], at[49]); MULADD(at[24], at[48]); MULADD(at[25], at[47]); MULADD(at[26], at[46]); MULADD(at[27], at[45]); + COMBA_STORE(C->dp[44]); + /* 45 */ + COMBA_FORWARD; + MULADD(at[18], at[55]); MULADD(at[19], at[54]); MULADD(at[20], at[53]); MULADD(at[21], at[52]); MULADD(at[22], at[51]); MULADD(at[23], at[50]); MULADD(at[24], at[49]); MULADD(at[25], at[48]); MULADD(at[26], at[47]); MULADD(at[27], at[46]); + COMBA_STORE(C->dp[45]); + /* 46 */ + COMBA_FORWARD; + MULADD(at[19], at[55]); MULADD(at[20], at[54]); MULADD(at[21], at[53]); MULADD(at[22], at[52]); MULADD(at[23], at[51]); MULADD(at[24], at[50]); MULADD(at[25], at[49]); MULADD(at[26], at[48]); MULADD(at[27], at[47]); + COMBA_STORE(C->dp[46]); + /* 47 */ + COMBA_FORWARD; + MULADD(at[20], at[55]); MULADD(at[21], at[54]); MULADD(at[22], at[53]); MULADD(at[23], at[52]); MULADD(at[24], at[51]); MULADD(at[25], at[50]); MULADD(at[26], at[49]); MULADD(at[27], at[48]); + COMBA_STORE(C->dp[47]); + /* 48 */ + COMBA_FORWARD; + MULADD(at[21], at[55]); MULADD(at[22], at[54]); MULADD(at[23], at[53]); MULADD(at[24], at[52]); MULADD(at[25], at[51]); MULADD(at[26], at[50]); MULADD(at[27], at[49]); + COMBA_STORE(C->dp[48]); + /* 49 */ + COMBA_FORWARD; + MULADD(at[22], at[55]); MULADD(at[23], at[54]); MULADD(at[24], at[53]); MULADD(at[25], at[52]); MULADD(at[26], at[51]); MULADD(at[27], at[50]); + COMBA_STORE(C->dp[49]); + /* 50 */ + COMBA_FORWARD; + MULADD(at[23], at[55]); MULADD(at[24], at[54]); MULADD(at[25], at[53]); MULADD(at[26], at[52]); MULADD(at[27], at[51]); + COMBA_STORE(C->dp[50]); + /* 51 */ + COMBA_FORWARD; + MULADD(at[24], at[55]); MULADD(at[25], at[54]); MULADD(at[26], at[53]); MULADD(at[27], at[52]); + COMBA_STORE(C->dp[51]); + /* 52 */ + COMBA_FORWARD; + MULADD(at[25], at[55]); MULADD(at[26], at[54]); MULADD(at[27], at[53]); + COMBA_STORE(C->dp[52]); + /* 53 */ + COMBA_FORWARD; + MULADD(at[26], at[55]); MULADD(at[27], at[54]); + COMBA_STORE(C->dp[53]); + /* 54 */ + COMBA_FORWARD; + MULADD(at[27], at[55]); + COMBA_STORE(C->dp[54]); + COMBA_STORE2(C->dp[55]); + C->used = 56; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; +} +#endif + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mul_comba_32.c b/src/crypto/tfm/fp_mul_comba_32.c new file mode 100644 index 0000000..f8c304e --- /dev/null +++ b/src/crypto/tfm/fp_mul_comba_32.c @@ -0,0 +1,290 @@ +#define TFM_DEFINES +#include "fp_mul_comba.c" + +#if defined(TFM_MUL32) && FP_SIZE >= 64 +void fp_mul_comba32(fp_int *A, fp_int *B, fp_int *C) +{ + fp_digit c0, c1, c2, at[64]; + int out_size; + + out_size = A->used + B->used; + memcpy(at, A->dp, 32 * sizeof(fp_digit)); + memcpy(at+32, B->dp, 32 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[32]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[33]); MULADD(at[1], at[32]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[34]); MULADD(at[1], at[33]); MULADD(at[2], at[32]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[35]); MULADD(at[1], at[34]); MULADD(at[2], at[33]); MULADD(at[3], at[32]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[36]); MULADD(at[1], at[35]); MULADD(at[2], at[34]); MULADD(at[3], at[33]); MULADD(at[4], at[32]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[37]); MULADD(at[1], at[36]); MULADD(at[2], at[35]); MULADD(at[3], at[34]); MULADD(at[4], at[33]); MULADD(at[5], at[32]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[38]); MULADD(at[1], at[37]); MULADD(at[2], at[36]); MULADD(at[3], at[35]); MULADD(at[4], at[34]); MULADD(at[5], at[33]); MULADD(at[6], at[32]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[39]); MULADD(at[1], at[38]); MULADD(at[2], at[37]); MULADD(at[3], at[36]); MULADD(at[4], at[35]); MULADD(at[5], at[34]); MULADD(at[6], at[33]); MULADD(at[7], at[32]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[40]); MULADD(at[1], at[39]); MULADD(at[2], at[38]); MULADD(at[3], at[37]); MULADD(at[4], at[36]); MULADD(at[5], at[35]); MULADD(at[6], at[34]); MULADD(at[7], at[33]); MULADD(at[8], at[32]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[41]); MULADD(at[1], at[40]); MULADD(at[2], at[39]); MULADD(at[3], at[38]); MULADD(at[4], at[37]); MULADD(at[5], at[36]); MULADD(at[6], at[35]); MULADD(at[7], at[34]); MULADD(at[8], at[33]); MULADD(at[9], at[32]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[0], at[42]); MULADD(at[1], at[41]); MULADD(at[2], at[40]); MULADD(at[3], at[39]); MULADD(at[4], at[38]); MULADD(at[5], at[37]); MULADD(at[6], at[36]); MULADD(at[7], at[35]); MULADD(at[8], at[34]); MULADD(at[9], at[33]); MULADD(at[10], at[32]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[0], at[43]); MULADD(at[1], at[42]); MULADD(at[2], at[41]); MULADD(at[3], at[40]); MULADD(at[4], at[39]); MULADD(at[5], at[38]); MULADD(at[6], at[37]); MULADD(at[7], at[36]); MULADD(at[8], at[35]); MULADD(at[9], at[34]); MULADD(at[10], at[33]); MULADD(at[11], at[32]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[0], at[44]); MULADD(at[1], at[43]); MULADD(at[2], at[42]); MULADD(at[3], at[41]); MULADD(at[4], at[40]); MULADD(at[5], at[39]); MULADD(at[6], at[38]); MULADD(at[7], at[37]); MULADD(at[8], at[36]); MULADD(at[9], at[35]); MULADD(at[10], at[34]); MULADD(at[11], at[33]); MULADD(at[12], at[32]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[0], at[45]); MULADD(at[1], at[44]); MULADD(at[2], at[43]); MULADD(at[3], at[42]); MULADD(at[4], at[41]); MULADD(at[5], at[40]); MULADD(at[6], at[39]); MULADD(at[7], at[38]); MULADD(at[8], at[37]); MULADD(at[9], at[36]); MULADD(at[10], at[35]); MULADD(at[11], at[34]); MULADD(at[12], at[33]); MULADD(at[13], at[32]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[0], at[46]); MULADD(at[1], at[45]); MULADD(at[2], at[44]); MULADD(at[3], at[43]); MULADD(at[4], at[42]); MULADD(at[5], at[41]); MULADD(at[6], at[40]); MULADD(at[7], at[39]); MULADD(at[8], at[38]); MULADD(at[9], at[37]); MULADD(at[10], at[36]); MULADD(at[11], at[35]); MULADD(at[12], at[34]); MULADD(at[13], at[33]); MULADD(at[14], at[32]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[0], at[47]); MULADD(at[1], at[46]); MULADD(at[2], at[45]); MULADD(at[3], at[44]); MULADD(at[4], at[43]); MULADD(at[5], at[42]); MULADD(at[6], at[41]); MULADD(at[7], at[40]); MULADD(at[8], at[39]); MULADD(at[9], at[38]); MULADD(at[10], at[37]); MULADD(at[11], at[36]); MULADD(at[12], at[35]); MULADD(at[13], at[34]); MULADD(at[14], at[33]); MULADD(at[15], at[32]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[0], at[48]); MULADD(at[1], at[47]); MULADD(at[2], at[46]); MULADD(at[3], at[45]); MULADD(at[4], at[44]); MULADD(at[5], at[43]); MULADD(at[6], at[42]); MULADD(at[7], at[41]); MULADD(at[8], at[40]); MULADD(at[9], at[39]); MULADD(at[10], at[38]); MULADD(at[11], at[37]); MULADD(at[12], at[36]); MULADD(at[13], at[35]); MULADD(at[14], at[34]); MULADD(at[15], at[33]); MULADD(at[16], at[32]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[0], at[49]); MULADD(at[1], at[48]); MULADD(at[2], at[47]); MULADD(at[3], at[46]); MULADD(at[4], at[45]); MULADD(at[5], at[44]); MULADD(at[6], at[43]); MULADD(at[7], at[42]); MULADD(at[8], at[41]); MULADD(at[9], at[40]); MULADD(at[10], at[39]); MULADD(at[11], at[38]); MULADD(at[12], at[37]); MULADD(at[13], at[36]); MULADD(at[14], at[35]); MULADD(at[15], at[34]); MULADD(at[16], at[33]); MULADD(at[17], at[32]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[0], at[50]); MULADD(at[1], at[49]); MULADD(at[2], at[48]); MULADD(at[3], at[47]); MULADD(at[4], at[46]); MULADD(at[5], at[45]); MULADD(at[6], at[44]); MULADD(at[7], at[43]); MULADD(at[8], at[42]); MULADD(at[9], at[41]); MULADD(at[10], at[40]); MULADD(at[11], at[39]); MULADD(at[12], at[38]); MULADD(at[13], at[37]); MULADD(at[14], at[36]); MULADD(at[15], at[35]); MULADD(at[16], at[34]); MULADD(at[17], at[33]); MULADD(at[18], at[32]); + COMBA_STORE(C->dp[18]); + /* 19 */ + COMBA_FORWARD; + MULADD(at[0], at[51]); MULADD(at[1], at[50]); MULADD(at[2], at[49]); MULADD(at[3], at[48]); MULADD(at[4], at[47]); MULADD(at[5], at[46]); MULADD(at[6], at[45]); MULADD(at[7], at[44]); MULADD(at[8], at[43]); MULADD(at[9], at[42]); MULADD(at[10], at[41]); MULADD(at[11], at[40]); MULADD(at[12], at[39]); MULADD(at[13], at[38]); MULADD(at[14], at[37]); MULADD(at[15], at[36]); MULADD(at[16], at[35]); MULADD(at[17], at[34]); MULADD(at[18], at[33]); MULADD(at[19], at[32]); + COMBA_STORE(C->dp[19]); + /* 20 */ + COMBA_FORWARD; + MULADD(at[0], at[52]); MULADD(at[1], at[51]); MULADD(at[2], at[50]); MULADD(at[3], at[49]); MULADD(at[4], at[48]); MULADD(at[5], at[47]); MULADD(at[6], at[46]); MULADD(at[7], at[45]); MULADD(at[8], at[44]); MULADD(at[9], at[43]); MULADD(at[10], at[42]); MULADD(at[11], at[41]); MULADD(at[12], at[40]); MULADD(at[13], at[39]); MULADD(at[14], at[38]); MULADD(at[15], at[37]); MULADD(at[16], at[36]); MULADD(at[17], at[35]); MULADD(at[18], at[34]); MULADD(at[19], at[33]); MULADD(at[20], at[32]); + COMBA_STORE(C->dp[20]); + /* 21 */ + COMBA_FORWARD; + MULADD(at[0], at[53]); MULADD(at[1], at[52]); MULADD(at[2], at[51]); MULADD(at[3], at[50]); MULADD(at[4], at[49]); MULADD(at[5], at[48]); MULADD(at[6], at[47]); MULADD(at[7], at[46]); MULADD(at[8], at[45]); MULADD(at[9], at[44]); MULADD(at[10], at[43]); MULADD(at[11], at[42]); MULADD(at[12], at[41]); MULADD(at[13], at[40]); MULADD(at[14], at[39]); MULADD(at[15], at[38]); MULADD(at[16], at[37]); MULADD(at[17], at[36]); MULADD(at[18], at[35]); MULADD(at[19], at[34]); MULADD(at[20], at[33]); MULADD(at[21], at[32]); + COMBA_STORE(C->dp[21]); + /* 22 */ + COMBA_FORWARD; + MULADD(at[0], at[54]); MULADD(at[1], at[53]); MULADD(at[2], at[52]); MULADD(at[3], at[51]); MULADD(at[4], at[50]); MULADD(at[5], at[49]); MULADD(at[6], at[48]); MULADD(at[7], at[47]); MULADD(at[8], at[46]); MULADD(at[9], at[45]); MULADD(at[10], at[44]); MULADD(at[11], at[43]); MULADD(at[12], at[42]); MULADD(at[13], at[41]); MULADD(at[14], at[40]); MULADD(at[15], at[39]); MULADD(at[16], at[38]); MULADD(at[17], at[37]); MULADD(at[18], at[36]); MULADD(at[19], at[35]); MULADD(at[20], at[34]); MULADD(at[21], at[33]); MULADD(at[22], at[32]); + COMBA_STORE(C->dp[22]); + /* 23 */ + COMBA_FORWARD; + MULADD(at[0], at[55]); MULADD(at[1], at[54]); MULADD(at[2], at[53]); MULADD(at[3], at[52]); MULADD(at[4], at[51]); MULADD(at[5], at[50]); MULADD(at[6], at[49]); MULADD(at[7], at[48]); MULADD(at[8], at[47]); MULADD(at[9], at[46]); MULADD(at[10], at[45]); MULADD(at[11], at[44]); MULADD(at[12], at[43]); MULADD(at[13], at[42]); MULADD(at[14], at[41]); MULADD(at[15], at[40]); MULADD(at[16], at[39]); MULADD(at[17], at[38]); MULADD(at[18], at[37]); MULADD(at[19], at[36]); MULADD(at[20], at[35]); MULADD(at[21], at[34]); MULADD(at[22], at[33]); MULADD(at[23], at[32]); + COMBA_STORE(C->dp[23]); + /* 24 */ + COMBA_FORWARD; + MULADD(at[0], at[56]); MULADD(at[1], at[55]); MULADD(at[2], at[54]); MULADD(at[3], at[53]); MULADD(at[4], at[52]); MULADD(at[5], at[51]); MULADD(at[6], at[50]); MULADD(at[7], at[49]); MULADD(at[8], at[48]); MULADD(at[9], at[47]); MULADD(at[10], at[46]); MULADD(at[11], at[45]); MULADD(at[12], at[44]); MULADD(at[13], at[43]); MULADD(at[14], at[42]); MULADD(at[15], at[41]); MULADD(at[16], at[40]); MULADD(at[17], at[39]); MULADD(at[18], at[38]); MULADD(at[19], at[37]); MULADD(at[20], at[36]); MULADD(at[21], at[35]); MULADD(at[22], at[34]); MULADD(at[23], at[33]); MULADD(at[24], at[32]); + COMBA_STORE(C->dp[24]); + /* 25 */ + COMBA_FORWARD; + MULADD(at[0], at[57]); MULADD(at[1], at[56]); MULADD(at[2], at[55]); MULADD(at[3], at[54]); MULADD(at[4], at[53]); MULADD(at[5], at[52]); MULADD(at[6], at[51]); MULADD(at[7], at[50]); MULADD(at[8], at[49]); MULADD(at[9], at[48]); MULADD(at[10], at[47]); MULADD(at[11], at[46]); MULADD(at[12], at[45]); MULADD(at[13], at[44]); MULADD(at[14], at[43]); MULADD(at[15], at[42]); MULADD(at[16], at[41]); MULADD(at[17], at[40]); MULADD(at[18], at[39]); MULADD(at[19], at[38]); MULADD(at[20], at[37]); MULADD(at[21], at[36]); MULADD(at[22], at[35]); MULADD(at[23], at[34]); MULADD(at[24], at[33]); MULADD(at[25], at[32]); + COMBA_STORE(C->dp[25]); + /* 26 */ + COMBA_FORWARD; + MULADD(at[0], at[58]); MULADD(at[1], at[57]); MULADD(at[2], at[56]); MULADD(at[3], at[55]); MULADD(at[4], at[54]); MULADD(at[5], at[53]); MULADD(at[6], at[52]); MULADD(at[7], at[51]); MULADD(at[8], at[50]); MULADD(at[9], at[49]); MULADD(at[10], at[48]); MULADD(at[11], at[47]); MULADD(at[12], at[46]); MULADD(at[13], at[45]); MULADD(at[14], at[44]); MULADD(at[15], at[43]); MULADD(at[16], at[42]); MULADD(at[17], at[41]); MULADD(at[18], at[40]); MULADD(at[19], at[39]); MULADD(at[20], at[38]); MULADD(at[21], at[37]); MULADD(at[22], at[36]); MULADD(at[23], at[35]); MULADD(at[24], at[34]); MULADD(at[25], at[33]); MULADD(at[26], at[32]); + COMBA_STORE(C->dp[26]); + /* 27 */ + COMBA_FORWARD; + MULADD(at[0], at[59]); MULADD(at[1], at[58]); MULADD(at[2], at[57]); MULADD(at[3], at[56]); MULADD(at[4], at[55]); MULADD(at[5], at[54]); MULADD(at[6], at[53]); MULADD(at[7], at[52]); MULADD(at[8], at[51]); MULADD(at[9], at[50]); MULADD(at[10], at[49]); MULADD(at[11], at[48]); MULADD(at[12], at[47]); MULADD(at[13], at[46]); MULADD(at[14], at[45]); MULADD(at[15], at[44]); MULADD(at[16], at[43]); MULADD(at[17], at[42]); MULADD(at[18], at[41]); MULADD(at[19], at[40]); MULADD(at[20], at[39]); MULADD(at[21], at[38]); MULADD(at[22], at[37]); MULADD(at[23], at[36]); MULADD(at[24], at[35]); MULADD(at[25], at[34]); MULADD(at[26], at[33]); MULADD(at[27], at[32]); + COMBA_STORE(C->dp[27]); + /* 28 */ + COMBA_FORWARD; + MULADD(at[0], at[60]); MULADD(at[1], at[59]); MULADD(at[2], at[58]); MULADD(at[3], at[57]); MULADD(at[4], at[56]); MULADD(at[5], at[55]); MULADD(at[6], at[54]); MULADD(at[7], at[53]); MULADD(at[8], at[52]); MULADD(at[9], at[51]); MULADD(at[10], at[50]); MULADD(at[11], at[49]); MULADD(at[12], at[48]); MULADD(at[13], at[47]); MULADD(at[14], at[46]); MULADD(at[15], at[45]); MULADD(at[16], at[44]); MULADD(at[17], at[43]); MULADD(at[18], at[42]); MULADD(at[19], at[41]); MULADD(at[20], at[40]); MULADD(at[21], at[39]); MULADD(at[22], at[38]); MULADD(at[23], at[37]); MULADD(at[24], at[36]); MULADD(at[25], at[35]); MULADD(at[26], at[34]); MULADD(at[27], at[33]); MULADD(at[28], at[32]); + COMBA_STORE(C->dp[28]); + /* 29 */ + COMBA_FORWARD; + MULADD(at[0], at[61]); MULADD(at[1], at[60]); MULADD(at[2], at[59]); MULADD(at[3], at[58]); MULADD(at[4], at[57]); MULADD(at[5], at[56]); MULADD(at[6], at[55]); MULADD(at[7], at[54]); MULADD(at[8], at[53]); MULADD(at[9], at[52]); MULADD(at[10], at[51]); MULADD(at[11], at[50]); MULADD(at[12], at[49]); MULADD(at[13], at[48]); MULADD(at[14], at[47]); MULADD(at[15], at[46]); MULADD(at[16], at[45]); MULADD(at[17], at[44]); MULADD(at[18], at[43]); MULADD(at[19], at[42]); MULADD(at[20], at[41]); MULADD(at[21], at[40]); MULADD(at[22], at[39]); MULADD(at[23], at[38]); MULADD(at[24], at[37]); MULADD(at[25], at[36]); MULADD(at[26], at[35]); MULADD(at[27], at[34]); MULADD(at[28], at[33]); MULADD(at[29], at[32]); + COMBA_STORE(C->dp[29]); + /* 30 */ + COMBA_FORWARD; + MULADD(at[0], at[62]); MULADD(at[1], at[61]); MULADD(at[2], at[60]); MULADD(at[3], at[59]); MULADD(at[4], at[58]); MULADD(at[5], at[57]); MULADD(at[6], at[56]); MULADD(at[7], at[55]); MULADD(at[8], at[54]); MULADD(at[9], at[53]); MULADD(at[10], at[52]); MULADD(at[11], at[51]); MULADD(at[12], at[50]); MULADD(at[13], at[49]); MULADD(at[14], at[48]); MULADD(at[15], at[47]); MULADD(at[16], at[46]); MULADD(at[17], at[45]); MULADD(at[18], at[44]); MULADD(at[19], at[43]); MULADD(at[20], at[42]); MULADD(at[21], at[41]); MULADD(at[22], at[40]); MULADD(at[23], at[39]); MULADD(at[24], at[38]); MULADD(at[25], at[37]); MULADD(at[26], at[36]); MULADD(at[27], at[35]); MULADD(at[28], at[34]); MULADD(at[29], at[33]); MULADD(at[30], at[32]); + COMBA_STORE(C->dp[30]); + /* 31 */ + COMBA_FORWARD; + MULADD(at[0], at[63]); MULADD(at[1], at[62]); MULADD(at[2], at[61]); MULADD(at[3], at[60]); MULADD(at[4], at[59]); MULADD(at[5], at[58]); MULADD(at[6], at[57]); MULADD(at[7], at[56]); MULADD(at[8], at[55]); MULADD(at[9], at[54]); MULADD(at[10], at[53]); MULADD(at[11], at[52]); MULADD(at[12], at[51]); MULADD(at[13], at[50]); MULADD(at[14], at[49]); MULADD(at[15], at[48]); MULADD(at[16], at[47]); MULADD(at[17], at[46]); MULADD(at[18], at[45]); MULADD(at[19], at[44]); MULADD(at[20], at[43]); MULADD(at[21], at[42]); MULADD(at[22], at[41]); MULADD(at[23], at[40]); MULADD(at[24], at[39]); MULADD(at[25], at[38]); MULADD(at[26], at[37]); MULADD(at[27], at[36]); MULADD(at[28], at[35]); MULADD(at[29], at[34]); MULADD(at[30], at[33]); MULADD(at[31], at[32]); + COMBA_STORE(C->dp[31]); + /* 32 */ + COMBA_FORWARD; + MULADD(at[1], at[63]); MULADD(at[2], at[62]); MULADD(at[3], at[61]); MULADD(at[4], at[60]); MULADD(at[5], at[59]); MULADD(at[6], at[58]); MULADD(at[7], at[57]); MULADD(at[8], at[56]); MULADD(at[9], at[55]); MULADD(at[10], at[54]); MULADD(at[11], at[53]); MULADD(at[12], at[52]); MULADD(at[13], at[51]); MULADD(at[14], at[50]); MULADD(at[15], at[49]); MULADD(at[16], at[48]); MULADD(at[17], at[47]); MULADD(at[18], at[46]); MULADD(at[19], at[45]); MULADD(at[20], at[44]); MULADD(at[21], at[43]); MULADD(at[22], at[42]); MULADD(at[23], at[41]); MULADD(at[24], at[40]); MULADD(at[25], at[39]); MULADD(at[26], at[38]); MULADD(at[27], at[37]); MULADD(at[28], at[36]); MULADD(at[29], at[35]); MULADD(at[30], at[34]); MULADD(at[31], at[33]); + COMBA_STORE(C->dp[32]); + /* 33 */ + COMBA_FORWARD; + MULADD(at[2], at[63]); MULADD(at[3], at[62]); MULADD(at[4], at[61]); MULADD(at[5], at[60]); MULADD(at[6], at[59]); MULADD(at[7], at[58]); MULADD(at[8], at[57]); MULADD(at[9], at[56]); MULADD(at[10], at[55]); MULADD(at[11], at[54]); MULADD(at[12], at[53]); MULADD(at[13], at[52]); MULADD(at[14], at[51]); MULADD(at[15], at[50]); MULADD(at[16], at[49]); MULADD(at[17], at[48]); MULADD(at[18], at[47]); MULADD(at[19], at[46]); MULADD(at[20], at[45]); MULADD(at[21], at[44]); MULADD(at[22], at[43]); MULADD(at[23], at[42]); MULADD(at[24], at[41]); MULADD(at[25], at[40]); MULADD(at[26], at[39]); MULADD(at[27], at[38]); MULADD(at[28], at[37]); MULADD(at[29], at[36]); MULADD(at[30], at[35]); MULADD(at[31], at[34]); + COMBA_STORE(C->dp[33]); + /* 34 */ + COMBA_FORWARD; + MULADD(at[3], at[63]); MULADD(at[4], at[62]); MULADD(at[5], at[61]); MULADD(at[6], at[60]); MULADD(at[7], at[59]); MULADD(at[8], at[58]); MULADD(at[9], at[57]); MULADD(at[10], at[56]); MULADD(at[11], at[55]); MULADD(at[12], at[54]); MULADD(at[13], at[53]); MULADD(at[14], at[52]); MULADD(at[15], at[51]); MULADD(at[16], at[50]); MULADD(at[17], at[49]); MULADD(at[18], at[48]); MULADD(at[19], at[47]); MULADD(at[20], at[46]); MULADD(at[21], at[45]); MULADD(at[22], at[44]); MULADD(at[23], at[43]); MULADD(at[24], at[42]); MULADD(at[25], at[41]); MULADD(at[26], at[40]); MULADD(at[27], at[39]); MULADD(at[28], at[38]); MULADD(at[29], at[37]); MULADD(at[30], at[36]); MULADD(at[31], at[35]); + COMBA_STORE(C->dp[34]); + /* 35 */ + COMBA_FORWARD; + MULADD(at[4], at[63]); MULADD(at[5], at[62]); MULADD(at[6], at[61]); MULADD(at[7], at[60]); MULADD(at[8], at[59]); MULADD(at[9], at[58]); MULADD(at[10], at[57]); MULADD(at[11], at[56]); MULADD(at[12], at[55]); MULADD(at[13], at[54]); MULADD(at[14], at[53]); MULADD(at[15], at[52]); MULADD(at[16], at[51]); MULADD(at[17], at[50]); MULADD(at[18], at[49]); MULADD(at[19], at[48]); MULADD(at[20], at[47]); MULADD(at[21], at[46]); MULADD(at[22], at[45]); MULADD(at[23], at[44]); MULADD(at[24], at[43]); MULADD(at[25], at[42]); MULADD(at[26], at[41]); MULADD(at[27], at[40]); MULADD(at[28], at[39]); MULADD(at[29], at[38]); MULADD(at[30], at[37]); MULADD(at[31], at[36]); + COMBA_STORE(C->dp[35]); + /* 36 */ + COMBA_FORWARD; + MULADD(at[5], at[63]); MULADD(at[6], at[62]); MULADD(at[7], at[61]); MULADD(at[8], at[60]); MULADD(at[9], at[59]); MULADD(at[10], at[58]); MULADD(at[11], at[57]); MULADD(at[12], at[56]); MULADD(at[13], at[55]); MULADD(at[14], at[54]); MULADD(at[15], at[53]); MULADD(at[16], at[52]); MULADD(at[17], at[51]); MULADD(at[18], at[50]); MULADD(at[19], at[49]); MULADD(at[20], at[48]); MULADD(at[21], at[47]); MULADD(at[22], at[46]); MULADD(at[23], at[45]); MULADD(at[24], at[44]); MULADD(at[25], at[43]); MULADD(at[26], at[42]); MULADD(at[27], at[41]); MULADD(at[28], at[40]); MULADD(at[29], at[39]); MULADD(at[30], at[38]); MULADD(at[31], at[37]); + COMBA_STORE(C->dp[36]); + /* 37 */ + COMBA_FORWARD; + MULADD(at[6], at[63]); MULADD(at[7], at[62]); MULADD(at[8], at[61]); MULADD(at[9], at[60]); MULADD(at[10], at[59]); MULADD(at[11], at[58]); MULADD(at[12], at[57]); MULADD(at[13], at[56]); MULADD(at[14], at[55]); MULADD(at[15], at[54]); MULADD(at[16], at[53]); MULADD(at[17], at[52]); MULADD(at[18], at[51]); MULADD(at[19], at[50]); MULADD(at[20], at[49]); MULADD(at[21], at[48]); MULADD(at[22], at[47]); MULADD(at[23], at[46]); MULADD(at[24], at[45]); MULADD(at[25], at[44]); MULADD(at[26], at[43]); MULADD(at[27], at[42]); MULADD(at[28], at[41]); MULADD(at[29], at[40]); MULADD(at[30], at[39]); MULADD(at[31], at[38]); + COMBA_STORE(C->dp[37]); + /* 38 */ + COMBA_FORWARD; + MULADD(at[7], at[63]); MULADD(at[8], at[62]); MULADD(at[9], at[61]); MULADD(at[10], at[60]); MULADD(at[11], at[59]); MULADD(at[12], at[58]); MULADD(at[13], at[57]); MULADD(at[14], at[56]); MULADD(at[15], at[55]); MULADD(at[16], at[54]); MULADD(at[17], at[53]); MULADD(at[18], at[52]); MULADD(at[19], at[51]); MULADD(at[20], at[50]); MULADD(at[21], at[49]); MULADD(at[22], at[48]); MULADD(at[23], at[47]); MULADD(at[24], at[46]); MULADD(at[25], at[45]); MULADD(at[26], at[44]); MULADD(at[27], at[43]); MULADD(at[28], at[42]); MULADD(at[29], at[41]); MULADD(at[30], at[40]); MULADD(at[31], at[39]); + COMBA_STORE(C->dp[38]); + + /* early out at 40 digits, 40*32==1280, or two 640 bit operands */ + if (out_size <= 40) { COMBA_STORE2(C->dp[39]); C->used = 40; C->sign = A->sign ^ B->sign; fp_clamp(C); COMBA_FINI; return; } + + /* 39 */ + COMBA_FORWARD; + MULADD(at[8], at[63]); MULADD(at[9], at[62]); MULADD(at[10], at[61]); MULADD(at[11], at[60]); MULADD(at[12], at[59]); MULADD(at[13], at[58]); MULADD(at[14], at[57]); MULADD(at[15], at[56]); MULADD(at[16], at[55]); MULADD(at[17], at[54]); MULADD(at[18], at[53]); MULADD(at[19], at[52]); MULADD(at[20], at[51]); MULADD(at[21], at[50]); MULADD(at[22], at[49]); MULADD(at[23], at[48]); MULADD(at[24], at[47]); MULADD(at[25], at[46]); MULADD(at[26], at[45]); MULADD(at[27], at[44]); MULADD(at[28], at[43]); MULADD(at[29], at[42]); MULADD(at[30], at[41]); MULADD(at[31], at[40]); + COMBA_STORE(C->dp[39]); + /* 40 */ + COMBA_FORWARD; + MULADD(at[9], at[63]); MULADD(at[10], at[62]); MULADD(at[11], at[61]); MULADD(at[12], at[60]); MULADD(at[13], at[59]); MULADD(at[14], at[58]); MULADD(at[15], at[57]); MULADD(at[16], at[56]); MULADD(at[17], at[55]); MULADD(at[18], at[54]); MULADD(at[19], at[53]); MULADD(at[20], at[52]); MULADD(at[21], at[51]); MULADD(at[22], at[50]); MULADD(at[23], at[49]); MULADD(at[24], at[48]); MULADD(at[25], at[47]); MULADD(at[26], at[46]); MULADD(at[27], at[45]); MULADD(at[28], at[44]); MULADD(at[29], at[43]); MULADD(at[30], at[42]); MULADD(at[31], at[41]); + COMBA_STORE(C->dp[40]); + /* 41 */ + COMBA_FORWARD; + MULADD(at[10], at[63]); MULADD(at[11], at[62]); MULADD(at[12], at[61]); MULADD(at[13], at[60]); MULADD(at[14], at[59]); MULADD(at[15], at[58]); MULADD(at[16], at[57]); MULADD(at[17], at[56]); MULADD(at[18], at[55]); MULADD(at[19], at[54]); MULADD(at[20], at[53]); MULADD(at[21], at[52]); MULADD(at[22], at[51]); MULADD(at[23], at[50]); MULADD(at[24], at[49]); MULADD(at[25], at[48]); MULADD(at[26], at[47]); MULADD(at[27], at[46]); MULADD(at[28], at[45]); MULADD(at[29], at[44]); MULADD(at[30], at[43]); MULADD(at[31], at[42]); + COMBA_STORE(C->dp[41]); + /* 42 */ + COMBA_FORWARD; + MULADD(at[11], at[63]); MULADD(at[12], at[62]); MULADD(at[13], at[61]); MULADD(at[14], at[60]); MULADD(at[15], at[59]); MULADD(at[16], at[58]); MULADD(at[17], at[57]); MULADD(at[18], at[56]); MULADD(at[19], at[55]); MULADD(at[20], at[54]); MULADD(at[21], at[53]); MULADD(at[22], at[52]); MULADD(at[23], at[51]); MULADD(at[24], at[50]); MULADD(at[25], at[49]); MULADD(at[26], at[48]); MULADD(at[27], at[47]); MULADD(at[28], at[46]); MULADD(at[29], at[45]); MULADD(at[30], at[44]); MULADD(at[31], at[43]); + COMBA_STORE(C->dp[42]); + /* 43 */ + COMBA_FORWARD; + MULADD(at[12], at[63]); MULADD(at[13], at[62]); MULADD(at[14], at[61]); MULADD(at[15], at[60]); MULADD(at[16], at[59]); MULADD(at[17], at[58]); MULADD(at[18], at[57]); MULADD(at[19], at[56]); MULADD(at[20], at[55]); MULADD(at[21], at[54]); MULADD(at[22], at[53]); MULADD(at[23], at[52]); MULADD(at[24], at[51]); MULADD(at[25], at[50]); MULADD(at[26], at[49]); MULADD(at[27], at[48]); MULADD(at[28], at[47]); MULADD(at[29], at[46]); MULADD(at[30], at[45]); MULADD(at[31], at[44]); + COMBA_STORE(C->dp[43]); + /* 44 */ + COMBA_FORWARD; + MULADD(at[13], at[63]); MULADD(at[14], at[62]); MULADD(at[15], at[61]); MULADD(at[16], at[60]); MULADD(at[17], at[59]); MULADD(at[18], at[58]); MULADD(at[19], at[57]); MULADD(at[20], at[56]); MULADD(at[21], at[55]); MULADD(at[22], at[54]); MULADD(at[23], at[53]); MULADD(at[24], at[52]); MULADD(at[25], at[51]); MULADD(at[26], at[50]); MULADD(at[27], at[49]); MULADD(at[28], at[48]); MULADD(at[29], at[47]); MULADD(at[30], at[46]); MULADD(at[31], at[45]); + COMBA_STORE(C->dp[44]); + /* 45 */ + COMBA_FORWARD; + MULADD(at[14], at[63]); MULADD(at[15], at[62]); MULADD(at[16], at[61]); MULADD(at[17], at[60]); MULADD(at[18], at[59]); MULADD(at[19], at[58]); MULADD(at[20], at[57]); MULADD(at[21], at[56]); MULADD(at[22], at[55]); MULADD(at[23], at[54]); MULADD(at[24], at[53]); MULADD(at[25], at[52]); MULADD(at[26], at[51]); MULADD(at[27], at[50]); MULADD(at[28], at[49]); MULADD(at[29], at[48]); MULADD(at[30], at[47]); MULADD(at[31], at[46]); + COMBA_STORE(C->dp[45]); + /* 46 */ + COMBA_FORWARD; + MULADD(at[15], at[63]); MULADD(at[16], at[62]); MULADD(at[17], at[61]); MULADD(at[18], at[60]); MULADD(at[19], at[59]); MULADD(at[20], at[58]); MULADD(at[21], at[57]); MULADD(at[22], at[56]); MULADD(at[23], at[55]); MULADD(at[24], at[54]); MULADD(at[25], at[53]); MULADD(at[26], at[52]); MULADD(at[27], at[51]); MULADD(at[28], at[50]); MULADD(at[29], at[49]); MULADD(at[30], at[48]); MULADD(at[31], at[47]); + COMBA_STORE(C->dp[46]); + + /* early out at 48 digits, 48*32==1536, or two 768 bit operands */ + if (out_size <= 48) { COMBA_STORE2(C->dp[47]); C->used = 48; C->sign = A->sign ^ B->sign; fp_clamp(C); COMBA_FINI; return; } + + /* 47 */ + COMBA_FORWARD; + MULADD(at[16], at[63]); MULADD(at[17], at[62]); MULADD(at[18], at[61]); MULADD(at[19], at[60]); MULADD(at[20], at[59]); MULADD(at[21], at[58]); MULADD(at[22], at[57]); MULADD(at[23], at[56]); MULADD(at[24], at[55]); MULADD(at[25], at[54]); MULADD(at[26], at[53]); MULADD(at[27], at[52]); MULADD(at[28], at[51]); MULADD(at[29], at[50]); MULADD(at[30], at[49]); MULADD(at[31], at[48]); + COMBA_STORE(C->dp[47]); + /* 48 */ + COMBA_FORWARD; + MULADD(at[17], at[63]); MULADD(at[18], at[62]); MULADD(at[19], at[61]); MULADD(at[20], at[60]); MULADD(at[21], at[59]); MULADD(at[22], at[58]); MULADD(at[23], at[57]); MULADD(at[24], at[56]); MULADD(at[25], at[55]); MULADD(at[26], at[54]); MULADD(at[27], at[53]); MULADD(at[28], at[52]); MULADD(at[29], at[51]); MULADD(at[30], at[50]); MULADD(at[31], at[49]); + COMBA_STORE(C->dp[48]); + /* 49 */ + COMBA_FORWARD; + MULADD(at[18], at[63]); MULADD(at[19], at[62]); MULADD(at[20], at[61]); MULADD(at[21], at[60]); MULADD(at[22], at[59]); MULADD(at[23], at[58]); MULADD(at[24], at[57]); MULADD(at[25], at[56]); MULADD(at[26], at[55]); MULADD(at[27], at[54]); MULADD(at[28], at[53]); MULADD(at[29], at[52]); MULADD(at[30], at[51]); MULADD(at[31], at[50]); + COMBA_STORE(C->dp[49]); + /* 50 */ + COMBA_FORWARD; + MULADD(at[19], at[63]); MULADD(at[20], at[62]); MULADD(at[21], at[61]); MULADD(at[22], at[60]); MULADD(at[23], at[59]); MULADD(at[24], at[58]); MULADD(at[25], at[57]); MULADD(at[26], at[56]); MULADD(at[27], at[55]); MULADD(at[28], at[54]); MULADD(at[29], at[53]); MULADD(at[30], at[52]); MULADD(at[31], at[51]); + COMBA_STORE(C->dp[50]); + /* 51 */ + COMBA_FORWARD; + MULADD(at[20], at[63]); MULADD(at[21], at[62]); MULADD(at[22], at[61]); MULADD(at[23], at[60]); MULADD(at[24], at[59]); MULADD(at[25], at[58]); MULADD(at[26], at[57]); MULADD(at[27], at[56]); MULADD(at[28], at[55]); MULADD(at[29], at[54]); MULADD(at[30], at[53]); MULADD(at[31], at[52]); + COMBA_STORE(C->dp[51]); + /* 52 */ + COMBA_FORWARD; + MULADD(at[21], at[63]); MULADD(at[22], at[62]); MULADD(at[23], at[61]); MULADD(at[24], at[60]); MULADD(at[25], at[59]); MULADD(at[26], at[58]); MULADD(at[27], at[57]); MULADD(at[28], at[56]); MULADD(at[29], at[55]); MULADD(at[30], at[54]); MULADD(at[31], at[53]); + COMBA_STORE(C->dp[52]); + /* 53 */ + COMBA_FORWARD; + MULADD(at[22], at[63]); MULADD(at[23], at[62]); MULADD(at[24], at[61]); MULADD(at[25], at[60]); MULADD(at[26], at[59]); MULADD(at[27], at[58]); MULADD(at[28], at[57]); MULADD(at[29], at[56]); MULADD(at[30], at[55]); MULADD(at[31], at[54]); + COMBA_STORE(C->dp[53]); + /* 54 */ + COMBA_FORWARD; + MULADD(at[23], at[63]); MULADD(at[24], at[62]); MULADD(at[25], at[61]); MULADD(at[26], at[60]); MULADD(at[27], at[59]); MULADD(at[28], at[58]); MULADD(at[29], at[57]); MULADD(at[30], at[56]); MULADD(at[31], at[55]); + COMBA_STORE(C->dp[54]); + + /* early out at 56 digits, 56*32==1792, or two 896 bit operands */ + if (out_size <= 56) { COMBA_STORE2(C->dp[55]); C->used = 56; C->sign = A->sign ^ B->sign; fp_clamp(C); COMBA_FINI; return; } + + /* 55 */ + COMBA_FORWARD; + MULADD(at[24], at[63]); MULADD(at[25], at[62]); MULADD(at[26], at[61]); MULADD(at[27], at[60]); MULADD(at[28], at[59]); MULADD(at[29], at[58]); MULADD(at[30], at[57]); MULADD(at[31], at[56]); + COMBA_STORE(C->dp[55]); + /* 56 */ + COMBA_FORWARD; + MULADD(at[25], at[63]); MULADD(at[26], at[62]); MULADD(at[27], at[61]); MULADD(at[28], at[60]); MULADD(at[29], at[59]); MULADD(at[30], at[58]); MULADD(at[31], at[57]); + COMBA_STORE(C->dp[56]); + /* 57 */ + COMBA_FORWARD; + MULADD(at[26], at[63]); MULADD(at[27], at[62]); MULADD(at[28], at[61]); MULADD(at[29], at[60]); MULADD(at[30], at[59]); MULADD(at[31], at[58]); + COMBA_STORE(C->dp[57]); + /* 58 */ + COMBA_FORWARD; + MULADD(at[27], at[63]); MULADD(at[28], at[62]); MULADD(at[29], at[61]); MULADD(at[30], at[60]); MULADD(at[31], at[59]); + COMBA_STORE(C->dp[58]); + /* 59 */ + COMBA_FORWARD; + MULADD(at[28], at[63]); MULADD(at[29], at[62]); MULADD(at[30], at[61]); MULADD(at[31], at[60]); + COMBA_STORE(C->dp[59]); + /* 60 */ + COMBA_FORWARD; + MULADD(at[29], at[63]); MULADD(at[30], at[62]); MULADD(at[31], at[61]); + COMBA_STORE(C->dp[60]); + /* 61 */ + COMBA_FORWARD; + MULADD(at[30], at[63]); MULADD(at[31], at[62]); + COMBA_STORE(C->dp[61]); + /* 62 */ + COMBA_FORWARD; + MULADD(at[31], at[63]); + COMBA_STORE(C->dp[62]); + COMBA_STORE2(C->dp[63]); + C->used = 64; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; +} +#endif + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mul_comba_small_set.c b/src/crypto/tfm/fp_mul_comba_small_set.c new file mode 100644 index 0000000..c4b9413 --- /dev/null +++ b/src/crypto/tfm/fp_mul_comba_small_set.c @@ -0,0 +1,1233 @@ +#define TFM_DEFINES +#include "fp_mul_comba.c" + +#if defined(TFM_SMALL_SET) +void fp_mul_comba_small(fp_int *A, fp_int *B, fp_int *C) +{ + fp_digit c0, c1, c2, at[32]; + switch (MAX(A->used, B->used)) { + + case 1: + memcpy(at, A->dp, 1 * sizeof(fp_digit)); + memcpy(at+1, B->dp, 1 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[1]); + COMBA_STORE(C->dp[0]); + COMBA_STORE2(C->dp[1]); + C->used = 2; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 2: + memcpy(at, A->dp, 2 * sizeof(fp_digit)); + memcpy(at+2, B->dp, 2 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[2]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[3]); MULADD(at[1], at[2]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[1], at[3]); + COMBA_STORE(C->dp[2]); + COMBA_STORE2(C->dp[3]); + C->used = 4; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 3: + memcpy(at, A->dp, 3 * sizeof(fp_digit)); + memcpy(at+3, B->dp, 3 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[3]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[4]); MULADD(at[1], at[3]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[5]); MULADD(at[1], at[4]); MULADD(at[2], at[3]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[1], at[5]); MULADD(at[2], at[4]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[2], at[5]); + COMBA_STORE(C->dp[4]); + COMBA_STORE2(C->dp[5]); + C->used = 6; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 4: + memcpy(at, A->dp, 4 * sizeof(fp_digit)); + memcpy(at+4, B->dp, 4 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[4]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[5]); MULADD(at[1], at[4]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[6]); MULADD(at[1], at[5]); MULADD(at[2], at[4]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[7]); MULADD(at[1], at[6]); MULADD(at[2], at[5]); MULADD(at[3], at[4]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[1], at[7]); MULADD(at[2], at[6]); MULADD(at[3], at[5]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[2], at[7]); MULADD(at[3], at[6]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[3], at[7]); + COMBA_STORE(C->dp[6]); + COMBA_STORE2(C->dp[7]); + C->used = 8; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 5: + memcpy(at, A->dp, 5 * sizeof(fp_digit)); + memcpy(at+5, B->dp, 5 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[5]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[6]); MULADD(at[1], at[5]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[7]); MULADD(at[1], at[6]); MULADD(at[2], at[5]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[8]); MULADD(at[1], at[7]); MULADD(at[2], at[6]); MULADD(at[3], at[5]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[9]); MULADD(at[1], at[8]); MULADD(at[2], at[7]); MULADD(at[3], at[6]); MULADD(at[4], at[5]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[1], at[9]); MULADD(at[2], at[8]); MULADD(at[3], at[7]); MULADD(at[4], at[6]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[2], at[9]); MULADD(at[3], at[8]); MULADD(at[4], at[7]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[3], at[9]); MULADD(at[4], at[8]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[4], at[9]); + COMBA_STORE(C->dp[8]); + COMBA_STORE2(C->dp[9]); + C->used = 10; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 6: + memcpy(at, A->dp, 6 * sizeof(fp_digit)); + memcpy(at+6, B->dp, 6 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[6]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[7]); MULADD(at[1], at[6]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[8]); MULADD(at[1], at[7]); MULADD(at[2], at[6]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[9]); MULADD(at[1], at[8]); MULADD(at[2], at[7]); MULADD(at[3], at[6]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[10]); MULADD(at[1], at[9]); MULADD(at[2], at[8]); MULADD(at[3], at[7]); MULADD(at[4], at[6]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[11]); MULADD(at[1], at[10]); MULADD(at[2], at[9]); MULADD(at[3], at[8]); MULADD(at[4], at[7]); MULADD(at[5], at[6]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[1], at[11]); MULADD(at[2], at[10]); MULADD(at[3], at[9]); MULADD(at[4], at[8]); MULADD(at[5], at[7]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[2], at[11]); MULADD(at[3], at[10]); MULADD(at[4], at[9]); MULADD(at[5], at[8]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[3], at[11]); MULADD(at[4], at[10]); MULADD(at[5], at[9]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[4], at[11]); MULADD(at[5], at[10]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[5], at[11]); + COMBA_STORE(C->dp[10]); + COMBA_STORE2(C->dp[11]); + C->used = 12; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 7: + memcpy(at, A->dp, 7 * sizeof(fp_digit)); + memcpy(at+7, B->dp, 7 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[7]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[8]); MULADD(at[1], at[7]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[9]); MULADD(at[1], at[8]); MULADD(at[2], at[7]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[10]); MULADD(at[1], at[9]); MULADD(at[2], at[8]); MULADD(at[3], at[7]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[11]); MULADD(at[1], at[10]); MULADD(at[2], at[9]); MULADD(at[3], at[8]); MULADD(at[4], at[7]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[12]); MULADD(at[1], at[11]); MULADD(at[2], at[10]); MULADD(at[3], at[9]); MULADD(at[4], at[8]); MULADD(at[5], at[7]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[13]); MULADD(at[1], at[12]); MULADD(at[2], at[11]); MULADD(at[3], at[10]); MULADD(at[4], at[9]); MULADD(at[5], at[8]); MULADD(at[6], at[7]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[1], at[13]); MULADD(at[2], at[12]); MULADD(at[3], at[11]); MULADD(at[4], at[10]); MULADD(at[5], at[9]); MULADD(at[6], at[8]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[2], at[13]); MULADD(at[3], at[12]); MULADD(at[4], at[11]); MULADD(at[5], at[10]); MULADD(at[6], at[9]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[3], at[13]); MULADD(at[4], at[12]); MULADD(at[5], at[11]); MULADD(at[6], at[10]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[4], at[13]); MULADD(at[5], at[12]); MULADD(at[6], at[11]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[5], at[13]); MULADD(at[6], at[12]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[6], at[13]); + COMBA_STORE(C->dp[12]); + COMBA_STORE2(C->dp[13]); + C->used = 14; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 8: + memcpy(at, A->dp, 8 * sizeof(fp_digit)); + memcpy(at+8, B->dp, 8 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[8]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[9]); MULADD(at[1], at[8]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[10]); MULADD(at[1], at[9]); MULADD(at[2], at[8]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[11]); MULADD(at[1], at[10]); MULADD(at[2], at[9]); MULADD(at[3], at[8]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[12]); MULADD(at[1], at[11]); MULADD(at[2], at[10]); MULADD(at[3], at[9]); MULADD(at[4], at[8]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[13]); MULADD(at[1], at[12]); MULADD(at[2], at[11]); MULADD(at[3], at[10]); MULADD(at[4], at[9]); MULADD(at[5], at[8]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[14]); MULADD(at[1], at[13]); MULADD(at[2], at[12]); MULADD(at[3], at[11]); MULADD(at[4], at[10]); MULADD(at[5], at[9]); MULADD(at[6], at[8]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[15]); MULADD(at[1], at[14]); MULADD(at[2], at[13]); MULADD(at[3], at[12]); MULADD(at[4], at[11]); MULADD(at[5], at[10]); MULADD(at[6], at[9]); MULADD(at[7], at[8]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[1], at[15]); MULADD(at[2], at[14]); MULADD(at[3], at[13]); MULADD(at[4], at[12]); MULADD(at[5], at[11]); MULADD(at[6], at[10]); MULADD(at[7], at[9]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[2], at[15]); MULADD(at[3], at[14]); MULADD(at[4], at[13]); MULADD(at[5], at[12]); MULADD(at[6], at[11]); MULADD(at[7], at[10]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[3], at[15]); MULADD(at[4], at[14]); MULADD(at[5], at[13]); MULADD(at[6], at[12]); MULADD(at[7], at[11]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[4], at[15]); MULADD(at[5], at[14]); MULADD(at[6], at[13]); MULADD(at[7], at[12]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[5], at[15]); MULADD(at[6], at[14]); MULADD(at[7], at[13]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[6], at[15]); MULADD(at[7], at[14]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[7], at[15]); + COMBA_STORE(C->dp[14]); + COMBA_STORE2(C->dp[15]); + C->used = 16; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 9: + memcpy(at, A->dp, 9 * sizeof(fp_digit)); + memcpy(at+9, B->dp, 9 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[9]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[10]); MULADD(at[1], at[9]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[11]); MULADD(at[1], at[10]); MULADD(at[2], at[9]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[12]); MULADD(at[1], at[11]); MULADD(at[2], at[10]); MULADD(at[3], at[9]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[13]); MULADD(at[1], at[12]); MULADD(at[2], at[11]); MULADD(at[3], at[10]); MULADD(at[4], at[9]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[14]); MULADD(at[1], at[13]); MULADD(at[2], at[12]); MULADD(at[3], at[11]); MULADD(at[4], at[10]); MULADD(at[5], at[9]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[15]); MULADD(at[1], at[14]); MULADD(at[2], at[13]); MULADD(at[3], at[12]); MULADD(at[4], at[11]); MULADD(at[5], at[10]); MULADD(at[6], at[9]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[16]); MULADD(at[1], at[15]); MULADD(at[2], at[14]); MULADD(at[3], at[13]); MULADD(at[4], at[12]); MULADD(at[5], at[11]); MULADD(at[6], at[10]); MULADD(at[7], at[9]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[17]); MULADD(at[1], at[16]); MULADD(at[2], at[15]); MULADD(at[3], at[14]); MULADD(at[4], at[13]); MULADD(at[5], at[12]); MULADD(at[6], at[11]); MULADD(at[7], at[10]); MULADD(at[8], at[9]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[1], at[17]); MULADD(at[2], at[16]); MULADD(at[3], at[15]); MULADD(at[4], at[14]); MULADD(at[5], at[13]); MULADD(at[6], at[12]); MULADD(at[7], at[11]); MULADD(at[8], at[10]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[2], at[17]); MULADD(at[3], at[16]); MULADD(at[4], at[15]); MULADD(at[5], at[14]); MULADD(at[6], at[13]); MULADD(at[7], at[12]); MULADD(at[8], at[11]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[3], at[17]); MULADD(at[4], at[16]); MULADD(at[5], at[15]); MULADD(at[6], at[14]); MULADD(at[7], at[13]); MULADD(at[8], at[12]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[4], at[17]); MULADD(at[5], at[16]); MULADD(at[6], at[15]); MULADD(at[7], at[14]); MULADD(at[8], at[13]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[5], at[17]); MULADD(at[6], at[16]); MULADD(at[7], at[15]); MULADD(at[8], at[14]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[6], at[17]); MULADD(at[7], at[16]); MULADD(at[8], at[15]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[7], at[17]); MULADD(at[8], at[16]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[8], at[17]); + COMBA_STORE(C->dp[16]); + COMBA_STORE2(C->dp[17]); + C->used = 18; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 10: + memcpy(at, A->dp, 10 * sizeof(fp_digit)); + memcpy(at+10, B->dp, 10 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[10]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[11]); MULADD(at[1], at[10]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[12]); MULADD(at[1], at[11]); MULADD(at[2], at[10]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[13]); MULADD(at[1], at[12]); MULADD(at[2], at[11]); MULADD(at[3], at[10]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[14]); MULADD(at[1], at[13]); MULADD(at[2], at[12]); MULADD(at[3], at[11]); MULADD(at[4], at[10]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[15]); MULADD(at[1], at[14]); MULADD(at[2], at[13]); MULADD(at[3], at[12]); MULADD(at[4], at[11]); MULADD(at[5], at[10]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[16]); MULADD(at[1], at[15]); MULADD(at[2], at[14]); MULADD(at[3], at[13]); MULADD(at[4], at[12]); MULADD(at[5], at[11]); MULADD(at[6], at[10]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[17]); MULADD(at[1], at[16]); MULADD(at[2], at[15]); MULADD(at[3], at[14]); MULADD(at[4], at[13]); MULADD(at[5], at[12]); MULADD(at[6], at[11]); MULADD(at[7], at[10]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[18]); MULADD(at[1], at[17]); MULADD(at[2], at[16]); MULADD(at[3], at[15]); MULADD(at[4], at[14]); MULADD(at[5], at[13]); MULADD(at[6], at[12]); MULADD(at[7], at[11]); MULADD(at[8], at[10]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[19]); MULADD(at[1], at[18]); MULADD(at[2], at[17]); MULADD(at[3], at[16]); MULADD(at[4], at[15]); MULADD(at[5], at[14]); MULADD(at[6], at[13]); MULADD(at[7], at[12]); MULADD(at[8], at[11]); MULADD(at[9], at[10]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[1], at[19]); MULADD(at[2], at[18]); MULADD(at[3], at[17]); MULADD(at[4], at[16]); MULADD(at[5], at[15]); MULADD(at[6], at[14]); MULADD(at[7], at[13]); MULADD(at[8], at[12]); MULADD(at[9], at[11]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[2], at[19]); MULADD(at[3], at[18]); MULADD(at[4], at[17]); MULADD(at[5], at[16]); MULADD(at[6], at[15]); MULADD(at[7], at[14]); MULADD(at[8], at[13]); MULADD(at[9], at[12]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[3], at[19]); MULADD(at[4], at[18]); MULADD(at[5], at[17]); MULADD(at[6], at[16]); MULADD(at[7], at[15]); MULADD(at[8], at[14]); MULADD(at[9], at[13]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[4], at[19]); MULADD(at[5], at[18]); MULADD(at[6], at[17]); MULADD(at[7], at[16]); MULADD(at[8], at[15]); MULADD(at[9], at[14]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[5], at[19]); MULADD(at[6], at[18]); MULADD(at[7], at[17]); MULADD(at[8], at[16]); MULADD(at[9], at[15]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[6], at[19]); MULADD(at[7], at[18]); MULADD(at[8], at[17]); MULADD(at[9], at[16]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[7], at[19]); MULADD(at[8], at[18]); MULADD(at[9], at[17]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[8], at[19]); MULADD(at[9], at[18]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[9], at[19]); + COMBA_STORE(C->dp[18]); + COMBA_STORE2(C->dp[19]); + C->used = 20; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 11: + memcpy(at, A->dp, 11 * sizeof(fp_digit)); + memcpy(at+11, B->dp, 11 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[11]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[12]); MULADD(at[1], at[11]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[13]); MULADD(at[1], at[12]); MULADD(at[2], at[11]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[14]); MULADD(at[1], at[13]); MULADD(at[2], at[12]); MULADD(at[3], at[11]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[15]); MULADD(at[1], at[14]); MULADD(at[2], at[13]); MULADD(at[3], at[12]); MULADD(at[4], at[11]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[16]); MULADD(at[1], at[15]); MULADD(at[2], at[14]); MULADD(at[3], at[13]); MULADD(at[4], at[12]); MULADD(at[5], at[11]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[17]); MULADD(at[1], at[16]); MULADD(at[2], at[15]); MULADD(at[3], at[14]); MULADD(at[4], at[13]); MULADD(at[5], at[12]); MULADD(at[6], at[11]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[18]); MULADD(at[1], at[17]); MULADD(at[2], at[16]); MULADD(at[3], at[15]); MULADD(at[4], at[14]); MULADD(at[5], at[13]); MULADD(at[6], at[12]); MULADD(at[7], at[11]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[19]); MULADD(at[1], at[18]); MULADD(at[2], at[17]); MULADD(at[3], at[16]); MULADD(at[4], at[15]); MULADD(at[5], at[14]); MULADD(at[6], at[13]); MULADD(at[7], at[12]); MULADD(at[8], at[11]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[20]); MULADD(at[1], at[19]); MULADD(at[2], at[18]); MULADD(at[3], at[17]); MULADD(at[4], at[16]); MULADD(at[5], at[15]); MULADD(at[6], at[14]); MULADD(at[7], at[13]); MULADD(at[8], at[12]); MULADD(at[9], at[11]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[0], at[21]); MULADD(at[1], at[20]); MULADD(at[2], at[19]); MULADD(at[3], at[18]); MULADD(at[4], at[17]); MULADD(at[5], at[16]); MULADD(at[6], at[15]); MULADD(at[7], at[14]); MULADD(at[8], at[13]); MULADD(at[9], at[12]); MULADD(at[10], at[11]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[1], at[21]); MULADD(at[2], at[20]); MULADD(at[3], at[19]); MULADD(at[4], at[18]); MULADD(at[5], at[17]); MULADD(at[6], at[16]); MULADD(at[7], at[15]); MULADD(at[8], at[14]); MULADD(at[9], at[13]); MULADD(at[10], at[12]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[2], at[21]); MULADD(at[3], at[20]); MULADD(at[4], at[19]); MULADD(at[5], at[18]); MULADD(at[6], at[17]); MULADD(at[7], at[16]); MULADD(at[8], at[15]); MULADD(at[9], at[14]); MULADD(at[10], at[13]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[3], at[21]); MULADD(at[4], at[20]); MULADD(at[5], at[19]); MULADD(at[6], at[18]); MULADD(at[7], at[17]); MULADD(at[8], at[16]); MULADD(at[9], at[15]); MULADD(at[10], at[14]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[4], at[21]); MULADD(at[5], at[20]); MULADD(at[6], at[19]); MULADD(at[7], at[18]); MULADD(at[8], at[17]); MULADD(at[9], at[16]); MULADD(at[10], at[15]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[5], at[21]); MULADD(at[6], at[20]); MULADD(at[7], at[19]); MULADD(at[8], at[18]); MULADD(at[9], at[17]); MULADD(at[10], at[16]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[6], at[21]); MULADD(at[7], at[20]); MULADD(at[8], at[19]); MULADD(at[9], at[18]); MULADD(at[10], at[17]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[7], at[21]); MULADD(at[8], at[20]); MULADD(at[9], at[19]); MULADD(at[10], at[18]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[8], at[21]); MULADD(at[9], at[20]); MULADD(at[10], at[19]); + COMBA_STORE(C->dp[18]); + /* 19 */ + COMBA_FORWARD; + MULADD(at[9], at[21]); MULADD(at[10], at[20]); + COMBA_STORE(C->dp[19]); + /* 20 */ + COMBA_FORWARD; + MULADD(at[10], at[21]); + COMBA_STORE(C->dp[20]); + COMBA_STORE2(C->dp[21]); + C->used = 22; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 12: + memcpy(at, A->dp, 12 * sizeof(fp_digit)); + memcpy(at+12, B->dp, 12 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[12]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[13]); MULADD(at[1], at[12]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[14]); MULADD(at[1], at[13]); MULADD(at[2], at[12]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[15]); MULADD(at[1], at[14]); MULADD(at[2], at[13]); MULADD(at[3], at[12]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[16]); MULADD(at[1], at[15]); MULADD(at[2], at[14]); MULADD(at[3], at[13]); MULADD(at[4], at[12]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[17]); MULADD(at[1], at[16]); MULADD(at[2], at[15]); MULADD(at[3], at[14]); MULADD(at[4], at[13]); MULADD(at[5], at[12]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[18]); MULADD(at[1], at[17]); MULADD(at[2], at[16]); MULADD(at[3], at[15]); MULADD(at[4], at[14]); MULADD(at[5], at[13]); MULADD(at[6], at[12]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[19]); MULADD(at[1], at[18]); MULADD(at[2], at[17]); MULADD(at[3], at[16]); MULADD(at[4], at[15]); MULADD(at[5], at[14]); MULADD(at[6], at[13]); MULADD(at[7], at[12]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[20]); MULADD(at[1], at[19]); MULADD(at[2], at[18]); MULADD(at[3], at[17]); MULADD(at[4], at[16]); MULADD(at[5], at[15]); MULADD(at[6], at[14]); MULADD(at[7], at[13]); MULADD(at[8], at[12]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[21]); MULADD(at[1], at[20]); MULADD(at[2], at[19]); MULADD(at[3], at[18]); MULADD(at[4], at[17]); MULADD(at[5], at[16]); MULADD(at[6], at[15]); MULADD(at[7], at[14]); MULADD(at[8], at[13]); MULADD(at[9], at[12]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[0], at[22]); MULADD(at[1], at[21]); MULADD(at[2], at[20]); MULADD(at[3], at[19]); MULADD(at[4], at[18]); MULADD(at[5], at[17]); MULADD(at[6], at[16]); MULADD(at[7], at[15]); MULADD(at[8], at[14]); MULADD(at[9], at[13]); MULADD(at[10], at[12]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[0], at[23]); MULADD(at[1], at[22]); MULADD(at[2], at[21]); MULADD(at[3], at[20]); MULADD(at[4], at[19]); MULADD(at[5], at[18]); MULADD(at[6], at[17]); MULADD(at[7], at[16]); MULADD(at[8], at[15]); MULADD(at[9], at[14]); MULADD(at[10], at[13]); MULADD(at[11], at[12]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[1], at[23]); MULADD(at[2], at[22]); MULADD(at[3], at[21]); MULADD(at[4], at[20]); MULADD(at[5], at[19]); MULADD(at[6], at[18]); MULADD(at[7], at[17]); MULADD(at[8], at[16]); MULADD(at[9], at[15]); MULADD(at[10], at[14]); MULADD(at[11], at[13]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[2], at[23]); MULADD(at[3], at[22]); MULADD(at[4], at[21]); MULADD(at[5], at[20]); MULADD(at[6], at[19]); MULADD(at[7], at[18]); MULADD(at[8], at[17]); MULADD(at[9], at[16]); MULADD(at[10], at[15]); MULADD(at[11], at[14]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[3], at[23]); MULADD(at[4], at[22]); MULADD(at[5], at[21]); MULADD(at[6], at[20]); MULADD(at[7], at[19]); MULADD(at[8], at[18]); MULADD(at[9], at[17]); MULADD(at[10], at[16]); MULADD(at[11], at[15]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[4], at[23]); MULADD(at[5], at[22]); MULADD(at[6], at[21]); MULADD(at[7], at[20]); MULADD(at[8], at[19]); MULADD(at[9], at[18]); MULADD(at[10], at[17]); MULADD(at[11], at[16]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[5], at[23]); MULADD(at[6], at[22]); MULADD(at[7], at[21]); MULADD(at[8], at[20]); MULADD(at[9], at[19]); MULADD(at[10], at[18]); MULADD(at[11], at[17]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[6], at[23]); MULADD(at[7], at[22]); MULADD(at[8], at[21]); MULADD(at[9], at[20]); MULADD(at[10], at[19]); MULADD(at[11], at[18]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[7], at[23]); MULADD(at[8], at[22]); MULADD(at[9], at[21]); MULADD(at[10], at[20]); MULADD(at[11], at[19]); + COMBA_STORE(C->dp[18]); + /* 19 */ + COMBA_FORWARD; + MULADD(at[8], at[23]); MULADD(at[9], at[22]); MULADD(at[10], at[21]); MULADD(at[11], at[20]); + COMBA_STORE(C->dp[19]); + /* 20 */ + COMBA_FORWARD; + MULADD(at[9], at[23]); MULADD(at[10], at[22]); MULADD(at[11], at[21]); + COMBA_STORE(C->dp[20]); + /* 21 */ + COMBA_FORWARD; + MULADD(at[10], at[23]); MULADD(at[11], at[22]); + COMBA_STORE(C->dp[21]); + /* 22 */ + COMBA_FORWARD; + MULADD(at[11], at[23]); + COMBA_STORE(C->dp[22]); + COMBA_STORE2(C->dp[23]); + C->used = 24; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 13: + memcpy(at, A->dp, 13 * sizeof(fp_digit)); + memcpy(at+13, B->dp, 13 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[13]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[14]); MULADD(at[1], at[13]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[15]); MULADD(at[1], at[14]); MULADD(at[2], at[13]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[16]); MULADD(at[1], at[15]); MULADD(at[2], at[14]); MULADD(at[3], at[13]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[17]); MULADD(at[1], at[16]); MULADD(at[2], at[15]); MULADD(at[3], at[14]); MULADD(at[4], at[13]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[18]); MULADD(at[1], at[17]); MULADD(at[2], at[16]); MULADD(at[3], at[15]); MULADD(at[4], at[14]); MULADD(at[5], at[13]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[19]); MULADD(at[1], at[18]); MULADD(at[2], at[17]); MULADD(at[3], at[16]); MULADD(at[4], at[15]); MULADD(at[5], at[14]); MULADD(at[6], at[13]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[20]); MULADD(at[1], at[19]); MULADD(at[2], at[18]); MULADD(at[3], at[17]); MULADD(at[4], at[16]); MULADD(at[5], at[15]); MULADD(at[6], at[14]); MULADD(at[7], at[13]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[21]); MULADD(at[1], at[20]); MULADD(at[2], at[19]); MULADD(at[3], at[18]); MULADD(at[4], at[17]); MULADD(at[5], at[16]); MULADD(at[6], at[15]); MULADD(at[7], at[14]); MULADD(at[8], at[13]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[22]); MULADD(at[1], at[21]); MULADD(at[2], at[20]); MULADD(at[3], at[19]); MULADD(at[4], at[18]); MULADD(at[5], at[17]); MULADD(at[6], at[16]); MULADD(at[7], at[15]); MULADD(at[8], at[14]); MULADD(at[9], at[13]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[0], at[23]); MULADD(at[1], at[22]); MULADD(at[2], at[21]); MULADD(at[3], at[20]); MULADD(at[4], at[19]); MULADD(at[5], at[18]); MULADD(at[6], at[17]); MULADD(at[7], at[16]); MULADD(at[8], at[15]); MULADD(at[9], at[14]); MULADD(at[10], at[13]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[0], at[24]); MULADD(at[1], at[23]); MULADD(at[2], at[22]); MULADD(at[3], at[21]); MULADD(at[4], at[20]); MULADD(at[5], at[19]); MULADD(at[6], at[18]); MULADD(at[7], at[17]); MULADD(at[8], at[16]); MULADD(at[9], at[15]); MULADD(at[10], at[14]); MULADD(at[11], at[13]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[0], at[25]); MULADD(at[1], at[24]); MULADD(at[2], at[23]); MULADD(at[3], at[22]); MULADD(at[4], at[21]); MULADD(at[5], at[20]); MULADD(at[6], at[19]); MULADD(at[7], at[18]); MULADD(at[8], at[17]); MULADD(at[9], at[16]); MULADD(at[10], at[15]); MULADD(at[11], at[14]); MULADD(at[12], at[13]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[1], at[25]); MULADD(at[2], at[24]); MULADD(at[3], at[23]); MULADD(at[4], at[22]); MULADD(at[5], at[21]); MULADD(at[6], at[20]); MULADD(at[7], at[19]); MULADD(at[8], at[18]); MULADD(at[9], at[17]); MULADD(at[10], at[16]); MULADD(at[11], at[15]); MULADD(at[12], at[14]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[2], at[25]); MULADD(at[3], at[24]); MULADD(at[4], at[23]); MULADD(at[5], at[22]); MULADD(at[6], at[21]); MULADD(at[7], at[20]); MULADD(at[8], at[19]); MULADD(at[9], at[18]); MULADD(at[10], at[17]); MULADD(at[11], at[16]); MULADD(at[12], at[15]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[3], at[25]); MULADD(at[4], at[24]); MULADD(at[5], at[23]); MULADD(at[6], at[22]); MULADD(at[7], at[21]); MULADD(at[8], at[20]); MULADD(at[9], at[19]); MULADD(at[10], at[18]); MULADD(at[11], at[17]); MULADD(at[12], at[16]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[4], at[25]); MULADD(at[5], at[24]); MULADD(at[6], at[23]); MULADD(at[7], at[22]); MULADD(at[8], at[21]); MULADD(at[9], at[20]); MULADD(at[10], at[19]); MULADD(at[11], at[18]); MULADD(at[12], at[17]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[5], at[25]); MULADD(at[6], at[24]); MULADD(at[7], at[23]); MULADD(at[8], at[22]); MULADD(at[9], at[21]); MULADD(at[10], at[20]); MULADD(at[11], at[19]); MULADD(at[12], at[18]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[6], at[25]); MULADD(at[7], at[24]); MULADD(at[8], at[23]); MULADD(at[9], at[22]); MULADD(at[10], at[21]); MULADD(at[11], at[20]); MULADD(at[12], at[19]); + COMBA_STORE(C->dp[18]); + /* 19 */ + COMBA_FORWARD; + MULADD(at[7], at[25]); MULADD(at[8], at[24]); MULADD(at[9], at[23]); MULADD(at[10], at[22]); MULADD(at[11], at[21]); MULADD(at[12], at[20]); + COMBA_STORE(C->dp[19]); + /* 20 */ + COMBA_FORWARD; + MULADD(at[8], at[25]); MULADD(at[9], at[24]); MULADD(at[10], at[23]); MULADD(at[11], at[22]); MULADD(at[12], at[21]); + COMBA_STORE(C->dp[20]); + /* 21 */ + COMBA_FORWARD; + MULADD(at[9], at[25]); MULADD(at[10], at[24]); MULADD(at[11], at[23]); MULADD(at[12], at[22]); + COMBA_STORE(C->dp[21]); + /* 22 */ + COMBA_FORWARD; + MULADD(at[10], at[25]); MULADD(at[11], at[24]); MULADD(at[12], at[23]); + COMBA_STORE(C->dp[22]); + /* 23 */ + COMBA_FORWARD; + MULADD(at[11], at[25]); MULADD(at[12], at[24]); + COMBA_STORE(C->dp[23]); + /* 24 */ + COMBA_FORWARD; + MULADD(at[12], at[25]); + COMBA_STORE(C->dp[24]); + COMBA_STORE2(C->dp[25]); + C->used = 26; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 14: + memcpy(at, A->dp, 14 * sizeof(fp_digit)); + memcpy(at+14, B->dp, 14 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[14]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[15]); MULADD(at[1], at[14]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[16]); MULADD(at[1], at[15]); MULADD(at[2], at[14]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[17]); MULADD(at[1], at[16]); MULADD(at[2], at[15]); MULADD(at[3], at[14]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[18]); MULADD(at[1], at[17]); MULADD(at[2], at[16]); MULADD(at[3], at[15]); MULADD(at[4], at[14]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[19]); MULADD(at[1], at[18]); MULADD(at[2], at[17]); MULADD(at[3], at[16]); MULADD(at[4], at[15]); MULADD(at[5], at[14]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[20]); MULADD(at[1], at[19]); MULADD(at[2], at[18]); MULADD(at[3], at[17]); MULADD(at[4], at[16]); MULADD(at[5], at[15]); MULADD(at[6], at[14]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[21]); MULADD(at[1], at[20]); MULADD(at[2], at[19]); MULADD(at[3], at[18]); MULADD(at[4], at[17]); MULADD(at[5], at[16]); MULADD(at[6], at[15]); MULADD(at[7], at[14]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[22]); MULADD(at[1], at[21]); MULADD(at[2], at[20]); MULADD(at[3], at[19]); MULADD(at[4], at[18]); MULADD(at[5], at[17]); MULADD(at[6], at[16]); MULADD(at[7], at[15]); MULADD(at[8], at[14]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[23]); MULADD(at[1], at[22]); MULADD(at[2], at[21]); MULADD(at[3], at[20]); MULADD(at[4], at[19]); MULADD(at[5], at[18]); MULADD(at[6], at[17]); MULADD(at[7], at[16]); MULADD(at[8], at[15]); MULADD(at[9], at[14]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[0], at[24]); MULADD(at[1], at[23]); MULADD(at[2], at[22]); MULADD(at[3], at[21]); MULADD(at[4], at[20]); MULADD(at[5], at[19]); MULADD(at[6], at[18]); MULADD(at[7], at[17]); MULADD(at[8], at[16]); MULADD(at[9], at[15]); MULADD(at[10], at[14]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[0], at[25]); MULADD(at[1], at[24]); MULADD(at[2], at[23]); MULADD(at[3], at[22]); MULADD(at[4], at[21]); MULADD(at[5], at[20]); MULADD(at[6], at[19]); MULADD(at[7], at[18]); MULADD(at[8], at[17]); MULADD(at[9], at[16]); MULADD(at[10], at[15]); MULADD(at[11], at[14]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[0], at[26]); MULADD(at[1], at[25]); MULADD(at[2], at[24]); MULADD(at[3], at[23]); MULADD(at[4], at[22]); MULADD(at[5], at[21]); MULADD(at[6], at[20]); MULADD(at[7], at[19]); MULADD(at[8], at[18]); MULADD(at[9], at[17]); MULADD(at[10], at[16]); MULADD(at[11], at[15]); MULADD(at[12], at[14]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[0], at[27]); MULADD(at[1], at[26]); MULADD(at[2], at[25]); MULADD(at[3], at[24]); MULADD(at[4], at[23]); MULADD(at[5], at[22]); MULADD(at[6], at[21]); MULADD(at[7], at[20]); MULADD(at[8], at[19]); MULADD(at[9], at[18]); MULADD(at[10], at[17]); MULADD(at[11], at[16]); MULADD(at[12], at[15]); MULADD(at[13], at[14]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[1], at[27]); MULADD(at[2], at[26]); MULADD(at[3], at[25]); MULADD(at[4], at[24]); MULADD(at[5], at[23]); MULADD(at[6], at[22]); MULADD(at[7], at[21]); MULADD(at[8], at[20]); MULADD(at[9], at[19]); MULADD(at[10], at[18]); MULADD(at[11], at[17]); MULADD(at[12], at[16]); MULADD(at[13], at[15]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[2], at[27]); MULADD(at[3], at[26]); MULADD(at[4], at[25]); MULADD(at[5], at[24]); MULADD(at[6], at[23]); MULADD(at[7], at[22]); MULADD(at[8], at[21]); MULADD(at[9], at[20]); MULADD(at[10], at[19]); MULADD(at[11], at[18]); MULADD(at[12], at[17]); MULADD(at[13], at[16]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[3], at[27]); MULADD(at[4], at[26]); MULADD(at[5], at[25]); MULADD(at[6], at[24]); MULADD(at[7], at[23]); MULADD(at[8], at[22]); MULADD(at[9], at[21]); MULADD(at[10], at[20]); MULADD(at[11], at[19]); MULADD(at[12], at[18]); MULADD(at[13], at[17]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[4], at[27]); MULADD(at[5], at[26]); MULADD(at[6], at[25]); MULADD(at[7], at[24]); MULADD(at[8], at[23]); MULADD(at[9], at[22]); MULADD(at[10], at[21]); MULADD(at[11], at[20]); MULADD(at[12], at[19]); MULADD(at[13], at[18]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[5], at[27]); MULADD(at[6], at[26]); MULADD(at[7], at[25]); MULADD(at[8], at[24]); MULADD(at[9], at[23]); MULADD(at[10], at[22]); MULADD(at[11], at[21]); MULADD(at[12], at[20]); MULADD(at[13], at[19]); + COMBA_STORE(C->dp[18]); + /* 19 */ + COMBA_FORWARD; + MULADD(at[6], at[27]); MULADD(at[7], at[26]); MULADD(at[8], at[25]); MULADD(at[9], at[24]); MULADD(at[10], at[23]); MULADD(at[11], at[22]); MULADD(at[12], at[21]); MULADD(at[13], at[20]); + COMBA_STORE(C->dp[19]); + /* 20 */ + COMBA_FORWARD; + MULADD(at[7], at[27]); MULADD(at[8], at[26]); MULADD(at[9], at[25]); MULADD(at[10], at[24]); MULADD(at[11], at[23]); MULADD(at[12], at[22]); MULADD(at[13], at[21]); + COMBA_STORE(C->dp[20]); + /* 21 */ + COMBA_FORWARD; + MULADD(at[8], at[27]); MULADD(at[9], at[26]); MULADD(at[10], at[25]); MULADD(at[11], at[24]); MULADD(at[12], at[23]); MULADD(at[13], at[22]); + COMBA_STORE(C->dp[21]); + /* 22 */ + COMBA_FORWARD; + MULADD(at[9], at[27]); MULADD(at[10], at[26]); MULADD(at[11], at[25]); MULADD(at[12], at[24]); MULADD(at[13], at[23]); + COMBA_STORE(C->dp[22]); + /* 23 */ + COMBA_FORWARD; + MULADD(at[10], at[27]); MULADD(at[11], at[26]); MULADD(at[12], at[25]); MULADD(at[13], at[24]); + COMBA_STORE(C->dp[23]); + /* 24 */ + COMBA_FORWARD; + MULADD(at[11], at[27]); MULADD(at[12], at[26]); MULADD(at[13], at[25]); + COMBA_STORE(C->dp[24]); + /* 25 */ + COMBA_FORWARD; + MULADD(at[12], at[27]); MULADD(at[13], at[26]); + COMBA_STORE(C->dp[25]); + /* 26 */ + COMBA_FORWARD; + MULADD(at[13], at[27]); + COMBA_STORE(C->dp[26]); + COMBA_STORE2(C->dp[27]); + C->used = 28; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 15: + memcpy(at, A->dp, 15 * sizeof(fp_digit)); + memcpy(at+15, B->dp, 15 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[15]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[16]); MULADD(at[1], at[15]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[17]); MULADD(at[1], at[16]); MULADD(at[2], at[15]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[18]); MULADD(at[1], at[17]); MULADD(at[2], at[16]); MULADD(at[3], at[15]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[19]); MULADD(at[1], at[18]); MULADD(at[2], at[17]); MULADD(at[3], at[16]); MULADD(at[4], at[15]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[20]); MULADD(at[1], at[19]); MULADD(at[2], at[18]); MULADD(at[3], at[17]); MULADD(at[4], at[16]); MULADD(at[5], at[15]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[21]); MULADD(at[1], at[20]); MULADD(at[2], at[19]); MULADD(at[3], at[18]); MULADD(at[4], at[17]); MULADD(at[5], at[16]); MULADD(at[6], at[15]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[22]); MULADD(at[1], at[21]); MULADD(at[2], at[20]); MULADD(at[3], at[19]); MULADD(at[4], at[18]); MULADD(at[5], at[17]); MULADD(at[6], at[16]); MULADD(at[7], at[15]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[23]); MULADD(at[1], at[22]); MULADD(at[2], at[21]); MULADD(at[3], at[20]); MULADD(at[4], at[19]); MULADD(at[5], at[18]); MULADD(at[6], at[17]); MULADD(at[7], at[16]); MULADD(at[8], at[15]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[24]); MULADD(at[1], at[23]); MULADD(at[2], at[22]); MULADD(at[3], at[21]); MULADD(at[4], at[20]); MULADD(at[5], at[19]); MULADD(at[6], at[18]); MULADD(at[7], at[17]); MULADD(at[8], at[16]); MULADD(at[9], at[15]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[0], at[25]); MULADD(at[1], at[24]); MULADD(at[2], at[23]); MULADD(at[3], at[22]); MULADD(at[4], at[21]); MULADD(at[5], at[20]); MULADD(at[6], at[19]); MULADD(at[7], at[18]); MULADD(at[8], at[17]); MULADD(at[9], at[16]); MULADD(at[10], at[15]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[0], at[26]); MULADD(at[1], at[25]); MULADD(at[2], at[24]); MULADD(at[3], at[23]); MULADD(at[4], at[22]); MULADD(at[5], at[21]); MULADD(at[6], at[20]); MULADD(at[7], at[19]); MULADD(at[8], at[18]); MULADD(at[9], at[17]); MULADD(at[10], at[16]); MULADD(at[11], at[15]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[0], at[27]); MULADD(at[1], at[26]); MULADD(at[2], at[25]); MULADD(at[3], at[24]); MULADD(at[4], at[23]); MULADD(at[5], at[22]); MULADD(at[6], at[21]); MULADD(at[7], at[20]); MULADD(at[8], at[19]); MULADD(at[9], at[18]); MULADD(at[10], at[17]); MULADD(at[11], at[16]); MULADD(at[12], at[15]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[0], at[28]); MULADD(at[1], at[27]); MULADD(at[2], at[26]); MULADD(at[3], at[25]); MULADD(at[4], at[24]); MULADD(at[5], at[23]); MULADD(at[6], at[22]); MULADD(at[7], at[21]); MULADD(at[8], at[20]); MULADD(at[9], at[19]); MULADD(at[10], at[18]); MULADD(at[11], at[17]); MULADD(at[12], at[16]); MULADD(at[13], at[15]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[0], at[29]); MULADD(at[1], at[28]); MULADD(at[2], at[27]); MULADD(at[3], at[26]); MULADD(at[4], at[25]); MULADD(at[5], at[24]); MULADD(at[6], at[23]); MULADD(at[7], at[22]); MULADD(at[8], at[21]); MULADD(at[9], at[20]); MULADD(at[10], at[19]); MULADD(at[11], at[18]); MULADD(at[12], at[17]); MULADD(at[13], at[16]); MULADD(at[14], at[15]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[1], at[29]); MULADD(at[2], at[28]); MULADD(at[3], at[27]); MULADD(at[4], at[26]); MULADD(at[5], at[25]); MULADD(at[6], at[24]); MULADD(at[7], at[23]); MULADD(at[8], at[22]); MULADD(at[9], at[21]); MULADD(at[10], at[20]); MULADD(at[11], at[19]); MULADD(at[12], at[18]); MULADD(at[13], at[17]); MULADD(at[14], at[16]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[2], at[29]); MULADD(at[3], at[28]); MULADD(at[4], at[27]); MULADD(at[5], at[26]); MULADD(at[6], at[25]); MULADD(at[7], at[24]); MULADD(at[8], at[23]); MULADD(at[9], at[22]); MULADD(at[10], at[21]); MULADD(at[11], at[20]); MULADD(at[12], at[19]); MULADD(at[13], at[18]); MULADD(at[14], at[17]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[3], at[29]); MULADD(at[4], at[28]); MULADD(at[5], at[27]); MULADD(at[6], at[26]); MULADD(at[7], at[25]); MULADD(at[8], at[24]); MULADD(at[9], at[23]); MULADD(at[10], at[22]); MULADD(at[11], at[21]); MULADD(at[12], at[20]); MULADD(at[13], at[19]); MULADD(at[14], at[18]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[4], at[29]); MULADD(at[5], at[28]); MULADD(at[6], at[27]); MULADD(at[7], at[26]); MULADD(at[8], at[25]); MULADD(at[9], at[24]); MULADD(at[10], at[23]); MULADD(at[11], at[22]); MULADD(at[12], at[21]); MULADD(at[13], at[20]); MULADD(at[14], at[19]); + COMBA_STORE(C->dp[18]); + /* 19 */ + COMBA_FORWARD; + MULADD(at[5], at[29]); MULADD(at[6], at[28]); MULADD(at[7], at[27]); MULADD(at[8], at[26]); MULADD(at[9], at[25]); MULADD(at[10], at[24]); MULADD(at[11], at[23]); MULADD(at[12], at[22]); MULADD(at[13], at[21]); MULADD(at[14], at[20]); + COMBA_STORE(C->dp[19]); + /* 20 */ + COMBA_FORWARD; + MULADD(at[6], at[29]); MULADD(at[7], at[28]); MULADD(at[8], at[27]); MULADD(at[9], at[26]); MULADD(at[10], at[25]); MULADD(at[11], at[24]); MULADD(at[12], at[23]); MULADD(at[13], at[22]); MULADD(at[14], at[21]); + COMBA_STORE(C->dp[20]); + /* 21 */ + COMBA_FORWARD; + MULADD(at[7], at[29]); MULADD(at[8], at[28]); MULADD(at[9], at[27]); MULADD(at[10], at[26]); MULADD(at[11], at[25]); MULADD(at[12], at[24]); MULADD(at[13], at[23]); MULADD(at[14], at[22]); + COMBA_STORE(C->dp[21]); + /* 22 */ + COMBA_FORWARD; + MULADD(at[8], at[29]); MULADD(at[9], at[28]); MULADD(at[10], at[27]); MULADD(at[11], at[26]); MULADD(at[12], at[25]); MULADD(at[13], at[24]); MULADD(at[14], at[23]); + COMBA_STORE(C->dp[22]); + /* 23 */ + COMBA_FORWARD; + MULADD(at[9], at[29]); MULADD(at[10], at[28]); MULADD(at[11], at[27]); MULADD(at[12], at[26]); MULADD(at[13], at[25]); MULADD(at[14], at[24]); + COMBA_STORE(C->dp[23]); + /* 24 */ + COMBA_FORWARD; + MULADD(at[10], at[29]); MULADD(at[11], at[28]); MULADD(at[12], at[27]); MULADD(at[13], at[26]); MULADD(at[14], at[25]); + COMBA_STORE(C->dp[24]); + /* 25 */ + COMBA_FORWARD; + MULADD(at[11], at[29]); MULADD(at[12], at[28]); MULADD(at[13], at[27]); MULADD(at[14], at[26]); + COMBA_STORE(C->dp[25]); + /* 26 */ + COMBA_FORWARD; + MULADD(at[12], at[29]); MULADD(at[13], at[28]); MULADD(at[14], at[27]); + COMBA_STORE(C->dp[26]); + /* 27 */ + COMBA_FORWARD; + MULADD(at[13], at[29]); MULADD(at[14], at[28]); + COMBA_STORE(C->dp[27]); + /* 28 */ + COMBA_FORWARD; + MULADD(at[14], at[29]); + COMBA_STORE(C->dp[28]); + COMBA_STORE2(C->dp[29]); + C->used = 30; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + + case 16: + memcpy(at, A->dp, 16 * sizeof(fp_digit)); + memcpy(at+16, B->dp, 16 * sizeof(fp_digit)); + COMBA_START; + + COMBA_CLEAR; + /* 0 */ + MULADD(at[0], at[16]); + COMBA_STORE(C->dp[0]); + /* 1 */ + COMBA_FORWARD; + MULADD(at[0], at[17]); MULADD(at[1], at[16]); + COMBA_STORE(C->dp[1]); + /* 2 */ + COMBA_FORWARD; + MULADD(at[0], at[18]); MULADD(at[1], at[17]); MULADD(at[2], at[16]); + COMBA_STORE(C->dp[2]); + /* 3 */ + COMBA_FORWARD; + MULADD(at[0], at[19]); MULADD(at[1], at[18]); MULADD(at[2], at[17]); MULADD(at[3], at[16]); + COMBA_STORE(C->dp[3]); + /* 4 */ + COMBA_FORWARD; + MULADD(at[0], at[20]); MULADD(at[1], at[19]); MULADD(at[2], at[18]); MULADD(at[3], at[17]); MULADD(at[4], at[16]); + COMBA_STORE(C->dp[4]); + /* 5 */ + COMBA_FORWARD; + MULADD(at[0], at[21]); MULADD(at[1], at[20]); MULADD(at[2], at[19]); MULADD(at[3], at[18]); MULADD(at[4], at[17]); MULADD(at[5], at[16]); + COMBA_STORE(C->dp[5]); + /* 6 */ + COMBA_FORWARD; + MULADD(at[0], at[22]); MULADD(at[1], at[21]); MULADD(at[2], at[20]); MULADD(at[3], at[19]); MULADD(at[4], at[18]); MULADD(at[5], at[17]); MULADD(at[6], at[16]); + COMBA_STORE(C->dp[6]); + /* 7 */ + COMBA_FORWARD; + MULADD(at[0], at[23]); MULADD(at[1], at[22]); MULADD(at[2], at[21]); MULADD(at[3], at[20]); MULADD(at[4], at[19]); MULADD(at[5], at[18]); MULADD(at[6], at[17]); MULADD(at[7], at[16]); + COMBA_STORE(C->dp[7]); + /* 8 */ + COMBA_FORWARD; + MULADD(at[0], at[24]); MULADD(at[1], at[23]); MULADD(at[2], at[22]); MULADD(at[3], at[21]); MULADD(at[4], at[20]); MULADD(at[5], at[19]); MULADD(at[6], at[18]); MULADD(at[7], at[17]); MULADD(at[8], at[16]); + COMBA_STORE(C->dp[8]); + /* 9 */ + COMBA_FORWARD; + MULADD(at[0], at[25]); MULADD(at[1], at[24]); MULADD(at[2], at[23]); MULADD(at[3], at[22]); MULADD(at[4], at[21]); MULADD(at[5], at[20]); MULADD(at[6], at[19]); MULADD(at[7], at[18]); MULADD(at[8], at[17]); MULADD(at[9], at[16]); + COMBA_STORE(C->dp[9]); + /* 10 */ + COMBA_FORWARD; + MULADD(at[0], at[26]); MULADD(at[1], at[25]); MULADD(at[2], at[24]); MULADD(at[3], at[23]); MULADD(at[4], at[22]); MULADD(at[5], at[21]); MULADD(at[6], at[20]); MULADD(at[7], at[19]); MULADD(at[8], at[18]); MULADD(at[9], at[17]); MULADD(at[10], at[16]); + COMBA_STORE(C->dp[10]); + /* 11 */ + COMBA_FORWARD; + MULADD(at[0], at[27]); MULADD(at[1], at[26]); MULADD(at[2], at[25]); MULADD(at[3], at[24]); MULADD(at[4], at[23]); MULADD(at[5], at[22]); MULADD(at[6], at[21]); MULADD(at[7], at[20]); MULADD(at[8], at[19]); MULADD(at[9], at[18]); MULADD(at[10], at[17]); MULADD(at[11], at[16]); + COMBA_STORE(C->dp[11]); + /* 12 */ + COMBA_FORWARD; + MULADD(at[0], at[28]); MULADD(at[1], at[27]); MULADD(at[2], at[26]); MULADD(at[3], at[25]); MULADD(at[4], at[24]); MULADD(at[5], at[23]); MULADD(at[6], at[22]); MULADD(at[7], at[21]); MULADD(at[8], at[20]); MULADD(at[9], at[19]); MULADD(at[10], at[18]); MULADD(at[11], at[17]); MULADD(at[12], at[16]); + COMBA_STORE(C->dp[12]); + /* 13 */ + COMBA_FORWARD; + MULADD(at[0], at[29]); MULADD(at[1], at[28]); MULADD(at[2], at[27]); MULADD(at[3], at[26]); MULADD(at[4], at[25]); MULADD(at[5], at[24]); MULADD(at[6], at[23]); MULADD(at[7], at[22]); MULADD(at[8], at[21]); MULADD(at[9], at[20]); MULADD(at[10], at[19]); MULADD(at[11], at[18]); MULADD(at[12], at[17]); MULADD(at[13], at[16]); + COMBA_STORE(C->dp[13]); + /* 14 */ + COMBA_FORWARD; + MULADD(at[0], at[30]); MULADD(at[1], at[29]); MULADD(at[2], at[28]); MULADD(at[3], at[27]); MULADD(at[4], at[26]); MULADD(at[5], at[25]); MULADD(at[6], at[24]); MULADD(at[7], at[23]); MULADD(at[8], at[22]); MULADD(at[9], at[21]); MULADD(at[10], at[20]); MULADD(at[11], at[19]); MULADD(at[12], at[18]); MULADD(at[13], at[17]); MULADD(at[14], at[16]); + COMBA_STORE(C->dp[14]); + /* 15 */ + COMBA_FORWARD; + MULADD(at[0], at[31]); MULADD(at[1], at[30]); MULADD(at[2], at[29]); MULADD(at[3], at[28]); MULADD(at[4], at[27]); MULADD(at[5], at[26]); MULADD(at[6], at[25]); MULADD(at[7], at[24]); MULADD(at[8], at[23]); MULADD(at[9], at[22]); MULADD(at[10], at[21]); MULADD(at[11], at[20]); MULADD(at[12], at[19]); MULADD(at[13], at[18]); MULADD(at[14], at[17]); MULADD(at[15], at[16]); + COMBA_STORE(C->dp[15]); + /* 16 */ + COMBA_FORWARD; + MULADD(at[1], at[31]); MULADD(at[2], at[30]); MULADD(at[3], at[29]); MULADD(at[4], at[28]); MULADD(at[5], at[27]); MULADD(at[6], at[26]); MULADD(at[7], at[25]); MULADD(at[8], at[24]); MULADD(at[9], at[23]); MULADD(at[10], at[22]); MULADD(at[11], at[21]); MULADD(at[12], at[20]); MULADD(at[13], at[19]); MULADD(at[14], at[18]); MULADD(at[15], at[17]); + COMBA_STORE(C->dp[16]); + /* 17 */ + COMBA_FORWARD; + MULADD(at[2], at[31]); MULADD(at[3], at[30]); MULADD(at[4], at[29]); MULADD(at[5], at[28]); MULADD(at[6], at[27]); MULADD(at[7], at[26]); MULADD(at[8], at[25]); MULADD(at[9], at[24]); MULADD(at[10], at[23]); MULADD(at[11], at[22]); MULADD(at[12], at[21]); MULADD(at[13], at[20]); MULADD(at[14], at[19]); MULADD(at[15], at[18]); + COMBA_STORE(C->dp[17]); + /* 18 */ + COMBA_FORWARD; + MULADD(at[3], at[31]); MULADD(at[4], at[30]); MULADD(at[5], at[29]); MULADD(at[6], at[28]); MULADD(at[7], at[27]); MULADD(at[8], at[26]); MULADD(at[9], at[25]); MULADD(at[10], at[24]); MULADD(at[11], at[23]); MULADD(at[12], at[22]); MULADD(at[13], at[21]); MULADD(at[14], at[20]); MULADD(at[15], at[19]); + COMBA_STORE(C->dp[18]); + /* 19 */ + COMBA_FORWARD; + MULADD(at[4], at[31]); MULADD(at[5], at[30]); MULADD(at[6], at[29]); MULADD(at[7], at[28]); MULADD(at[8], at[27]); MULADD(at[9], at[26]); MULADD(at[10], at[25]); MULADD(at[11], at[24]); MULADD(at[12], at[23]); MULADD(at[13], at[22]); MULADD(at[14], at[21]); MULADD(at[15], at[20]); + COMBA_STORE(C->dp[19]); + /* 20 */ + COMBA_FORWARD; + MULADD(at[5], at[31]); MULADD(at[6], at[30]); MULADD(at[7], at[29]); MULADD(at[8], at[28]); MULADD(at[9], at[27]); MULADD(at[10], at[26]); MULADD(at[11], at[25]); MULADD(at[12], at[24]); MULADD(at[13], at[23]); MULADD(at[14], at[22]); MULADD(at[15], at[21]); + COMBA_STORE(C->dp[20]); + /* 21 */ + COMBA_FORWARD; + MULADD(at[6], at[31]); MULADD(at[7], at[30]); MULADD(at[8], at[29]); MULADD(at[9], at[28]); MULADD(at[10], at[27]); MULADD(at[11], at[26]); MULADD(at[12], at[25]); MULADD(at[13], at[24]); MULADD(at[14], at[23]); MULADD(at[15], at[22]); + COMBA_STORE(C->dp[21]); + /* 22 */ + COMBA_FORWARD; + MULADD(at[7], at[31]); MULADD(at[8], at[30]); MULADD(at[9], at[29]); MULADD(at[10], at[28]); MULADD(at[11], at[27]); MULADD(at[12], at[26]); MULADD(at[13], at[25]); MULADD(at[14], at[24]); MULADD(at[15], at[23]); + COMBA_STORE(C->dp[22]); + /* 23 */ + COMBA_FORWARD; + MULADD(at[8], at[31]); MULADD(at[9], at[30]); MULADD(at[10], at[29]); MULADD(at[11], at[28]); MULADD(at[12], at[27]); MULADD(at[13], at[26]); MULADD(at[14], at[25]); MULADD(at[15], at[24]); + COMBA_STORE(C->dp[23]); + /* 24 */ + COMBA_FORWARD; + MULADD(at[9], at[31]); MULADD(at[10], at[30]); MULADD(at[11], at[29]); MULADD(at[12], at[28]); MULADD(at[13], at[27]); MULADD(at[14], at[26]); MULADD(at[15], at[25]); + COMBA_STORE(C->dp[24]); + /* 25 */ + COMBA_FORWARD; + MULADD(at[10], at[31]); MULADD(at[11], at[30]); MULADD(at[12], at[29]); MULADD(at[13], at[28]); MULADD(at[14], at[27]); MULADD(at[15], at[26]); + COMBA_STORE(C->dp[25]); + /* 26 */ + COMBA_FORWARD; + MULADD(at[11], at[31]); MULADD(at[12], at[30]); MULADD(at[13], at[29]); MULADD(at[14], at[28]); MULADD(at[15], at[27]); + COMBA_STORE(C->dp[26]); + /* 27 */ + COMBA_FORWARD; + MULADD(at[12], at[31]); MULADD(at[13], at[30]); MULADD(at[14], at[29]); MULADD(at[15], at[28]); + COMBA_STORE(C->dp[27]); + /* 28 */ + COMBA_FORWARD; + MULADD(at[13], at[31]); MULADD(at[14], at[30]); MULADD(at[15], at[29]); + COMBA_STORE(C->dp[28]); + /* 29 */ + COMBA_FORWARD; + MULADD(at[14], at[31]); MULADD(at[15], at[30]); + COMBA_STORE(C->dp[29]); + /* 30 */ + COMBA_FORWARD; + MULADD(at[15], at[31]); + COMBA_STORE(C->dp[30]); + COMBA_STORE2(C->dp[31]); + C->used = 32; + C->sign = A->sign ^ B->sign; + fp_clamp(C); + COMBA_FINI; + break; + } +} + +#endif + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mul_d.c b/src/crypto/tfm/fp_mul_d.c new file mode 100644 index 0000000..27dbc0d --- /dev/null +++ b/src/crypto/tfm/fp_mul_d.c @@ -0,0 +1,40 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* c = a * b */ +void fp_mul_d(fp_int *a, fp_digit b, fp_int *c) +{ + fp_word w; + int x, oldused; + + oldused = c->used; + c->used = a->used; + c->sign = a->sign; + w = 0; + for (x = 0; x < a->used; x++) { + w = ((fp_word)a->dp[x]) * ((fp_word)b) + w; + c->dp[x] = (fp_digit)w; + w = w >> DIGIT_BIT; + } + if (w != 0 && (a->used != FP_SIZE)) { + c->dp[c->used++] = w; + ++x; + } + for (; x < oldused; x++) { + c->dp[x] = 0; + } + fp_clamp(c); +} + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_mulmod.c b/src/crypto/tfm/fp_mulmod.c new file mode 100644 index 0000000..3c0069c --- /dev/null +++ b/src/crypto/tfm/fp_mulmod.c @@ -0,0 +1,22 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include +/* d = a * b (mod c) */ +int fp_mulmod(fp_int *a, fp_int *b, fp_int *c, fp_int *d) +{ + fp_int tmp; + fp_zero(&tmp); + fp_mul(a, b, &tmp); + return fp_mod(&tmp, c, d); +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_read_unsigned_bin.c b/src/crypto/tfm/fp_read_unsigned_bin.c new file mode 100644 index 0000000..975fe8a --- /dev/null +++ b/src/crypto/tfm/fp_read_unsigned_bin.c @@ -0,0 +1,70 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +void fp_read_unsigned_bin(fp_int *a, const unsigned char *b, int c) +{ + /* zero the int */ + fp_zero (a); + + if ((unsigned)c > (FP_SIZE * sizeof(fp_digit))) { + int excess = c - (FP_SIZE * sizeof(fp_digit)); + c -= excess; + b += excess; + } + + /* If we know the endianness of this architecture, and we're using + 32-bit fp_digits, we can optimize this */ +#if (defined(ENDIAN_LITTLE) || defined(ENDIAN_BIG)) && !defined(FP_64BIT) + /* But not for both simultaneously */ +#if defined(ENDIAN_LITTLE) && defined(ENDIAN_BIG) +#error Both ENDIAN_LITTLE and ENDIAN_BIG defined. +#endif + { + unsigned char *pd = (unsigned char *)a->dp; + + a->used = (c + sizeof(fp_digit) - 1)/sizeof(fp_digit); + /* read the bytes in */ +#ifdef ENDIAN_BIG + { + /* Use Duff's device to unroll the loop. */ + int idx = (c - 1) & ~3; + switch (c % 4) { + case 0: do { pd[idx+0] = *b++; + case 3: pd[idx+1] = *b++; + case 2: pd[idx+2] = *b++; + case 1: pd[idx+3] = *b++; + idx -= 4; + } while ((c -= 4) > 0); + } + } +#else + for (c -= 1; c >= 0; c -= 1) { + pd[c] = *b++; + } +#endif + } +#else + /* read the bytes in */ + for (; c > 0; c--) { + fp_mul_2d (a, 8, a); + a->dp[0] |= *b++; + + if (a->used == 0) { + a->used = 1; + } + } +#endif + fp_clamp (a); +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_reverse.c b/src/crypto/tfm/fp_reverse.c new file mode 100644 index 0000000..ece1d4a --- /dev/null +++ b/src/crypto/tfm/fp_reverse.c @@ -0,0 +1,31 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* reverse an array, used for radix code */ +void fp_reverse (unsigned char *s, int len) +{ + int ix, iy; + unsigned char t; + + ix = 0; + iy = len - 1; + while (ix < iy) { + t = s[ix]; + s[ix] = s[iy]; + s[iy] = t; + ++ix; + --iy; + } +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_rshd.c b/src/crypto/tfm/fp_rshd.c new file mode 100644 index 0000000..6dc3589 --- /dev/null +++ b/src/crypto/tfm/fp_rshd.c @@ -0,0 +1,40 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +void fp_rshd(fp_int *a, int x) +{ + int y; + + /* too many digits just zero and return */ + if (x >= a->used) { + fp_zero(a); + return; + } + + /* shift */ + for (y = 0; y < a->used - x; y++) { + a->dp[y] = a->dp[y+x]; + } + + /* zero rest */ + for (; y < a->used; y++) { + a->dp[y] = 0; + } + + /* decrement count */ + a->used -= x; + fp_clamp(a); +} + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_set.c b/src/crypto/tfm/fp_set.c new file mode 100644 index 0000000..5693759 --- /dev/null +++ b/src/crypto/tfm/fp_set.c @@ -0,0 +1,21 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +void fp_set(fp_int *a, fp_digit b) +{ + fp_zero(a); + a->dp[0] = b; + a->used = a->dp[0] ? 1 : 0; +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_sqr.c b/src/crypto/tfm/fp_sqr.c new file mode 100644 index 0000000..3678c7a --- /dev/null +++ b/src/crypto/tfm/fp_sqr.c @@ -0,0 +1,126 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* b = a*a */ +void fp_sqr(fp_int *A, fp_int *B) +{ + int y, old_used; + + old_used = B->used; + + /* call generic if we're out of range */ + if (A->used + A->used > FP_SIZE) { + fp_sqr_comba(A, B); + goto clean; + } + + y = A->used; +#if defined(TFM_SQR3) && FP_SIZE >= 6 + if (y <= 3) { + fp_sqr_comba3(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR4) && FP_SIZE >= 8 + if (y == 4) { + fp_sqr_comba4(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR6) && FP_SIZE >= 12 + if (y <= 6) { + fp_sqr_comba6(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR7) && FP_SIZE >= 14 + if (y == 7) { + fp_sqr_comba7(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR8) && FP_SIZE >= 16 + if (y == 8) { + fp_sqr_comba8(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR9) && FP_SIZE >= 18 + if (y == 9) { + fp_sqr_comba9(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR12) && FP_SIZE >= 24 + if (y <= 12) { + fp_sqr_comba12(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR17) && FP_SIZE >= 34 + if (y <= 17) { + fp_sqr_comba17(A,B); + goto clean; + } +#endif +#if defined(TFM_SMALL_SET) + if (y <= 16) { + fp_sqr_comba_small(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR20) && FP_SIZE >= 40 + if (y <= 20) { + fp_sqr_comba20(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR24) && FP_SIZE >= 48 + if (y <= 24) { + fp_sqr_comba24(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR28) && FP_SIZE >= 56 + if (y <= 28) { + fp_sqr_comba28(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR32) && FP_SIZE >= 64 + if (y <= 32) { + fp_sqr_comba32(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR48) && FP_SIZE >= 96 + if (y <= 48) { + fp_sqr_comba48(A,B); + goto clean; + } +#endif +#if defined(TFM_SQR64) && FP_SIZE >= 128 + if (y <= 64) { + fp_sqr_comba64(A,B); + goto clean; + } +#endif + fp_sqr_comba(A, B); +clean: + for (y = B->used; y < old_used; y++) { + B->dp[y] = 0; + } +} + + +/* $Source: /cvs/libtom/tomsfastmath/src/sqr/fp_sqr.c,v $ */ +/* $Revision: 1.1 $ */ +/* $Date: 2006/12/31 21:25:53 $ */ diff --git a/src/crypto/tfm/fp_sqr_comba.c b/src/crypto/tfm/fp_sqr_comba.c new file mode 100644 index 0000000..1150333 --- /dev/null +++ b/src/crypto/tfm/fp_sqr_comba.c @@ -0,0 +1,677 @@ +/* + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +#if defined(TFM_PRESCOTT) && defined(TFM_SSE2) + #undef TFM_SSE2 + #define TFM_X86 +#endif + +#if defined(TFM_X86) + +/* x86-32 optimized */ + +#define COMBA_START + +#define CLEAR_CARRY \ + c0 = c1 = c2 = 0; + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define CARRY_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_FINI + +#define SQRADD(i, j) \ +asm( \ + "movl %6,%%eax \n\t" \ + "mull %%eax \n\t" \ + "addl %%eax,%0 \n\t" \ + "adcl %%edx,%1 \n\t" \ + "adcl $0,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(i) :"%eax","%edx","cc"); + +#define SQRADD2(i, j) \ +asm( \ + "movl %6,%%eax \n\t" \ + "mull %7 \n\t" \ + "addl %%eax,%0 \n\t" \ + "adcl %%edx,%1 \n\t" \ + "adcl $0,%2 \n\t" \ + "addl %%eax,%0 \n\t" \ + "adcl %%edx,%1 \n\t" \ + "adcl $0,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(i), "m"(j) :"%eax","%edx","cc"); + +#define SQRADDSC(i, j) \ +asm( \ + "movl %3,%%eax \n\t" \ + "mull %4 \n\t" \ + "movl %%eax,%0 \n\t" \ + "movl %%edx,%1 \n\t" \ + "xorl %2,%2 \n\t" \ + :"=r"(sc0), "=r"(sc1), "=r"(sc2): "g"(i), "g"(j) :"%eax","%edx","cc"); + +#define SQRADDAC(i, j) \ +asm( \ + "movl %6,%%eax \n\t" \ + "mull %7 \n\t" \ + "addl %%eax,%0 \n\t" \ + "adcl %%edx,%1 \n\t" \ + "adcl $0,%2 \n\t" \ + :"=r"(sc0), "=r"(sc1), "=r"(sc2): "0"(sc0), "1"(sc1), "2"(sc2), "g"(i), "g"(j) :"%eax","%edx","cc"); + +#define SQRADDDB \ +asm( \ + "addl %6,%0 \n\t" \ + "adcl %7,%1 \n\t" \ + "adcl %8,%2 \n\t" \ + "addl %6,%0 \n\t" \ + "adcl %7,%1 \n\t" \ + "adcl %8,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2) : "0"(c0), "1"(c1), "2"(c2), "r"(sc0), "r"(sc1), "r"(sc2) : "cc"); + +#elif defined(TFM_X86_64) +/* x86-64 optimized */ + +#define COMBA_START + +#define CLEAR_CARRY \ + c0 = c1 = c2 = 0; + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define CARRY_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_FINI + +#define SQRADD(i, j) \ +asm( \ + "movq %6,%%rax \n\t" \ + "mulq %%rax \n\t" \ + "addq %%rax,%0 \n\t" \ + "adcq %%rdx,%1 \n\t" \ + "adcq $0,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "x"(i) :"%rax","%rdx","cc"); + +#define SQRADD2(i, j) \ +asm( \ + "movq %6,%%rax \n\t" \ + "mulq %7 \n\t" \ + "addq %%rax,%0 \n\t" \ + "adcq %%rdx,%1 \n\t" \ + "adcq $0,%2 \n\t" \ + "addq %%rax,%0 \n\t" \ + "adcq %%rdx,%1 \n\t" \ + "adcq $0,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "g"(i), "g"(j) :"%rax","%rdx","cc"); + +#define SQRADDSC(i, j) \ +asm( \ + "movq %3,%%rax \n\t" \ + "mulq %4 \n\t" \ + "movq %%rax,%0 \n\t" \ + "movq %%rdx,%1 \n\t" \ + "xorq %2,%2 \n\t" \ + :"=r"(sc0), "=r"(sc1), "=r"(sc2): "g"(i), "g"(j) :"%rax","%rdx","cc"); + +#define SQRADDAC(i, j) \ +asm( \ + "movq %6,%%rax \n\t" \ + "mulq %7 \n\t" \ + "addq %%rax,%0 \n\t" \ + "adcq %%rdx,%1 \n\t" \ + "adcq $0,%2 \n\t" \ + :"=r"(sc0), "=r"(sc1), "=r"(sc2): "0"(sc0), "1"(sc1), "2"(sc2), "g"(i), "g"(j) :"%rax","%rdx","cc"); + +#define SQRADDDB \ +asm( \ + "addq %6,%0 \n\t" \ + "adcq %7,%1 \n\t" \ + "adcq %8,%2 \n\t" \ + "addq %6,%0 \n\t" \ + "adcq %7,%1 \n\t" \ + "adcq %8,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2) : "0"(c0), "1"(c1), "2"(c2), "r"(sc0), "r"(sc1), "r"(sc2) : "cc"); + +#elif defined(TFM_SSE2) + +/* SSE2 Optimized */ +#define COMBA_START + +#define CLEAR_CARRY \ + c0 = c1 = c2 = 0; + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define CARRY_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_FINI \ + asm("emms"); + +#define SQRADD(i, j) \ +asm( \ + "movd %6,%%mm0 \n\t" \ + "pmuludq %%mm0,%%mm0\n\t" \ + "movd %%mm0,%%eax \n\t" \ + "psrlq $32,%%mm0 \n\t" \ + "addl %%eax,%0 \n\t" \ + "movd %%mm0,%%eax \n\t" \ + "adcl %%eax,%1 \n\t" \ + "adcl $0,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(i) :"%eax","cc"); + +#define SQRADD2(i, j) \ +asm( \ + "movd %6,%%mm0 \n\t" \ + "movd %7,%%mm1 \n\t" \ + "pmuludq %%mm1,%%mm0\n\t" \ + "movd %%mm0,%%eax \n\t" \ + "psrlq $32,%%mm0 \n\t" \ + "movd %%mm0,%%edx \n\t" \ + "addl %%eax,%0 \n\t" \ + "adcl %%edx,%1 \n\t" \ + "adcl $0,%2 \n\t" \ + "addl %%eax,%0 \n\t" \ + "adcl %%edx,%1 \n\t" \ + "adcl $0,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2): "0"(c0), "1"(c1), "2"(c2), "m"(i), "m"(j) :"%eax","%edx","cc"); + +#define SQRADDSC(i, j) \ +asm( \ + "movd %6,%%mm0 \n\t" \ + "movd %7,%%mm1 \n\t" \ + "pmuludq %%mm1,%%mm0\n\t" \ + "movd %%mm0,%0 \n\t" \ + "psrlq $32,%%mm0 \n\t" \ + "movd %%mm0,%1 \n\t" \ + "xorl %2,%2 \n\t" \ + :"=r"(sc0), "=r"(sc1), "=r"(sc2): "0"(sc0), "1"(sc1), "2"(sc2), "m"(i), "m"(j)); + +#define SQRADDAC(i, j) \ +asm( \ + "movd %6,%%mm0 \n\t" \ + "movd %7,%%mm1 \n\t" \ + "pmuludq %%mm1,%%mm0\n\t" \ + "movd %%mm0,%%eax \n\t" \ + "psrlq $32,%%mm0 \n\t" \ + "movd %%mm0,%%edx \n\t" \ + "addl %%eax,%0 \n\t" \ + "adcl %%edx,%1 \n\t" \ + "adcl $0,%2 \n\t" \ + :"=r"(sc0), "=r"(sc1), "=r"(sc2): "0"(sc0), "1"(sc1), "2"(sc2), "m"(i), "m"(j) :"%eax","%edx","cc"); + +#define SQRADDDB \ +asm( \ + "addl %6,%0 \n\t" \ + "adcl %7,%1 \n\t" \ + "adcl %8,%2 \n\t" \ + "addl %6,%0 \n\t" \ + "adcl %7,%1 \n\t" \ + "adcl %8,%2 \n\t" \ + :"=r"(c0), "=r"(c1), "=r"(c2) : "0"(c0), "1"(c1), "2"(c2), "r"(sc0), "r"(sc1), "r"(sc2) : "cc"); + +#elif defined(TFM_ARM) + +/* ARM code */ + +#define COMBA_START + +#define CLEAR_CARRY \ + c0 = c1 = c2 = 0; + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define CARRY_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_FINI + +/* multiplies point i and j, updates carry "c1" and digit c2 */ +#define SQRADD(i, j) \ +asm( \ +" UMULL r0,r1,%6,%6 \n\t" \ +" ADDS %0,%0,r0 \n\t" \ +" ADCS %1,%1,r1 \n\t" \ +" ADC %2,%2,#0 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2) : "0"(c0), "1"(c1), "2"(c2), "r"(i) : "r0", "r1", "cc"); + +/* for squaring some of the terms are doubled... */ +#define SQRADD2(i, j) \ +asm( \ +" UMULL r0,r1,%6,%7 \n\t" \ +" ADDS %0,%0,r0 \n\t" \ +" ADCS %1,%1,r1 \n\t" \ +" ADC %2,%2,#0 \n\t" \ +" ADDS %0,%0,r0 \n\t" \ +" ADCS %1,%1,r1 \n\t" \ +" ADC %2,%2,#0 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2) : "0"(c0), "1"(c1), "2"(c2), "r"(i), "r"(j) : "r0", "r1", "cc"); + +#define SQRADDSC(i, j) \ +asm( \ +" UMULL %0,%1,%6,%7 \n\t" \ +" SUB %2,%2,%2 \n\t" \ +:"=r"(sc0), "=r"(sc1), "=r"(sc2) : "0"(sc0), "1"(sc1), "2"(sc2), "r"(i), "r"(j) : "cc"); + +#define SQRADDAC(i, j) \ +asm( \ +" UMULL r0,r1,%6,%7 \n\t" \ +" ADDS %0,%0,r0 \n\t" \ +" ADCS %1,%1,r1 \n\t" \ +" ADC %2,%2,#0 \n\t" \ +:"=r"(sc0), "=r"(sc1), "=r"(sc2) : "0"(sc0), "1"(sc1), "2"(sc2), "r"(i), "r"(j) : "r0", "r1", "cc"); + +#define SQRADDDB \ +asm( \ +" ADDS %0,%0,%3 \n\t" \ +" ADCS %1,%1,%4 \n\t" \ +" ADC %2,%2,%5 \n\t" \ +" ADDS %0,%0,%3 \n\t" \ +" ADCS %1,%1,%4 \n\t" \ +" ADC %2,%2,%5 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2) : "r"(sc0), "r"(sc1), "r"(sc2), "0"(c0), "1"(c1), "2"(c2) : "cc"); + +#elif defined(TFM_PPC32) + +/* PPC32 */ + +#define COMBA_START + +#define CLEAR_CARRY \ + c0 = c1 = c2 = 0; + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define CARRY_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_FINI + +/* multiplies point i and j, updates carry "c1" and digit c2 */ +#define SQRADD(i, j) \ +asm( \ + " mullw 16,%6,%6 \n\t" \ + " addc %0,%0,16 \n\t" \ + " mulhwu 16,%6,%6 \n\t" \ + " adde %1,%1,16 \n\t" \ + " addze %2,%2 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i):"16","cc"); + +/* for squaring some of the terms are doubled... */ +#define SQRADD2(i, j) \ +asm( \ + " mullw 16,%6,%7 \n\t" \ + " mulhwu 17,%6,%7 \n\t" \ + " addc %0,%0,16 \n\t" \ + " adde %1,%1,17 \n\t" \ + " addze %2,%2 \n\t" \ + " addc %0,%0,16 \n\t" \ + " adde %1,%1,17 \n\t" \ + " addze %2,%2 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i), "r"(j):"16", "17","cc"); + +#define SQRADDSC(i, j) \ +asm( \ + " mullw %0,%6,%7 \n\t" \ + " mulhwu %1,%6,%7 \n\t" \ + " xor %2,%2,%2 \n\t" \ +:"=r"(sc0), "=r"(sc1), "=r"(sc2):"0"(sc0), "1"(sc1), "2"(sc2), "r"(i),"r"(j) : "cc"); + +#define SQRADDAC(i, j) \ +asm( \ + " mullw 16,%6,%7 \n\t" \ + " addc %0,%0,16 \n\t" \ + " mulhwu 16,%6,%7 \n\t" \ + " adde %1,%1,16 \n\t" \ + " addze %2,%2 \n\t" \ +:"=r"(sc0), "=r"(sc1), "=r"(sc2):"0"(sc0), "1"(sc1), "2"(sc2), "r"(i), "r"(j):"16", "cc"); + +#define SQRADDDB \ +asm( \ + " addc %0,%0,%3 \n\t" \ + " adde %1,%1,%4 \n\t" \ + " adde %2,%2,%5 \n\t" \ + " addc %0,%0,%3 \n\t" \ + " adde %1,%1,%4 \n\t" \ + " adde %2,%2,%5 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2) : "r"(sc0), "r"(sc1), "r"(sc2), "0"(c0), "1"(c1), "2"(c2) : "cc"); + +#elif defined(TFM_PPC64) +/* PPC64 */ + +#define COMBA_START + +#define CLEAR_CARRY \ + c0 = c1 = c2 = 0; + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define CARRY_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_FINI + +/* multiplies point i and j, updates carry "c1" and digit c2 */ +#define SQRADD(i, j) \ +asm( \ + " mulld r16,%6,%6 \n\t" \ + " addc %0,%0,r16 \n\t" \ + " mulhdu r16,%6,%6 \n\t" \ + " adde %1,%1,r16 \n\t" \ + " addze %2,%2 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i):"r16","cc"); + +/* for squaring some of the terms are doubled... */ +#define SQRADD2(i, j) \ +asm( \ + " mulld r16,%6,%7 \n\t" \ + " mulhdu r17,%6,%7 \n\t" \ + " addc %0,%0,r16 \n\t" \ + " adde %1,%1,r17 \n\t" \ + " addze %2,%2 \n\t" \ + " addc %0,%0,r16 \n\t" \ + " adde %1,%1,r17 \n\t" \ + " addze %2,%2 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i), "r"(j):"r16", "r17","cc"); + +#define SQRADDSC(i, j) \ +asm( \ + " mulld %0,%6,%7 \n\t" \ + " mulhdu %1,%6,%7 \n\t" \ + " xor %2,%2,%2 \n\t" \ +:"=r"(sc0), "=r"(sc1), "=r"(sc2):"0"(sc0), "1"(sc1), "2"(sc2), "r"(i),"r"(j) : "cc"); + +#define SQRADDAC(i, j) \ +asm( \ + " mulld r16,%6,%7 \n\t" \ + " addc %0,%0,r16 \n\t" \ + " mulhdu r16,%6,%7 \n\t" \ + " adde %1,%1,r16 \n\t" \ + " addze %2,%2 \n\t" \ +:"=r"(sc0), "=r"(sc1), "=r"(sc2):"0"(sc0), "1"(sc1), "2"(sc2), "r"(i), "r"(j):"r16", "cc"); + +#define SQRADDDB \ +asm( \ + " addc %0,%0,%3 \n\t" \ + " adde %1,%1,%4 \n\t" \ + " adde %2,%2,%5 \n\t" \ + " addc %0,%0,%3 \n\t" \ + " adde %1,%1,%4 \n\t" \ + " adde %2,%2,%5 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2) : "r"(sc0), "r"(sc1), "r"(sc2), "0"(c0), "1"(c1), "2"(c2) : "cc"); + + +#elif defined(TFM_AVR32) + +/* AVR32 */ + +#define COMBA_START + +#define CLEAR_CARRY \ + c0 = c1 = c2 = 0; + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define CARRY_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_FINI + +/* multiplies point i and j, updates carry "c1" and digit c2 */ +#define SQRADD(i, j) \ +asm( \ + " mulu.d r2,%6,%6 \n\t" \ + " add %0,%0,r2 \n\t" \ + " adc %1,%1,r3 \n\t" \ + " acr %2 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i):"r2","r3"); + +/* for squaring some of the terms are doubled... */ +#define SQRADD2(i, j) \ +asm( \ + " mulu.d r2,%6,%7 \n\t" \ + " add %0,%0,r2 \n\t" \ + " adc %1,%1,r3 \n\t" \ + " acr %2, \n\t" \ + " add %0,%0,r2 \n\t" \ + " adc %1,%1,r3 \n\t" \ + " acr %2, \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i), "r"(j):"r2", "r3"); + +#define SQRADDSC(i, j) \ +asm( \ + " mulu.d r2,%6,%7 \n\t" \ + " mov %0,r2 \n\t" \ + " mov %1,r3 \n\t" \ + " eor %2,%2 \n\t" \ +:"=r"(sc0), "=r"(sc1), "=r"(sc2):"0"(sc0), "1"(sc1), "2"(sc2), "r"(i),"r"(j) : "r2", "r3"); + +#define SQRADDAC(i, j) \ +asm( \ + " mulu.d r2,%6,%7 \n\t" \ + " add %0,%0,r2 \n\t" \ + " adc %1,%1,r3 \n\t" \ + " acr %2 \n\t" \ +:"=r"(sc0), "=r"(sc1), "=r"(sc2):"0"(sc0), "1"(sc1), "2"(sc2), "r"(i), "r"(j):"r2", "r3"); + +#define SQRADDDB \ +asm( \ + " add %0,%0,%3 \n\t" \ + " adc %1,%1,%4 \n\t" \ + " adc %2,%2,%5 \n\t" \ + " add %0,%0,%3 \n\t" \ + " adc %1,%1,%4 \n\t" \ + " adc %2,%2,%5 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2) : "r"(sc0), "r"(sc1), "r"(sc2), "0"(c0), "1"(c1), "2"(c2) : "cc"); + +#elif defined(TFM_MIPS) + +/* MIPS */ + +#define COMBA_START + +#define CLEAR_CARRY \ + c0 = c1 = c2 = 0; + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define CARRY_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_FINI + +/* multiplies point i and j, updates carry "c1" and digit c2 */ +#define SQRADD(i, j) \ +asm( \ + " multu %6,%6 \n\t" \ + " mflo $12 \n\t" \ + " mfhi $13 \n\t" \ + " addu %0,%0,$12 \n\t" \ + " sltu $12,%0,$12 \n\t" \ + " addu %1,%1,$13 \n\t" \ + " sltu $13,%1,$13 \n\t" \ + " addu %1,%1,$12 \n\t" \ + " sltu $12,%1,$12 \n\t" \ + " addu %2,%2,$13 \n\t" \ + " addu %2,%2,$12 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i):"$12","$13"); + +/* for squaring some of the terms are doubled... */ +#define SQRADD2(i, j) \ +asm( \ + " multu %6,%7 \n\t" \ + " mflo $12 \n\t" \ + " mfhi $13 \n\t" \ + \ + " addu %0,%0,$12 \n\t" \ + " sltu $14,%0,$12 \n\t" \ + " addu %1,%1,$13 \n\t" \ + " sltu $15,%1,$13 \n\t" \ + " addu %1,%1,$14 \n\t" \ + " sltu $14,%1,$14 \n\t" \ + " addu %2,%2,$15 \n\t" \ + " addu %2,%2,$14 \n\t" \ + \ + " addu %0,%0,$12 \n\t" \ + " sltu $14,%0,$12 \n\t" \ + " addu %1,%1,$13 \n\t" \ + " sltu $15,%1,$13 \n\t" \ + " addu %1,%1,$14 \n\t" \ + " sltu $14,%1,$14 \n\t" \ + " addu %2,%2,$15 \n\t" \ + " addu %2,%2,$14 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2):"0"(c0), "1"(c1), "2"(c2), "r"(i), "r"(j):"$12", "$13", "$14", "$15"); + +#define SQRADDSC(i, j) \ +asm( \ + " multu %6,%7 \n\t" \ + " mflo %0 \n\t" \ + " mfhi %1 \n\t" \ + " xor %2,%2,%2 \n\t" \ +:"=r"(sc0), "=r"(sc1), "=r"(sc2):"0"(sc0), "1"(sc1), "2"(sc2), "r"(i),"r"(j) : "cc"); + +#define SQRADDAC(i, j) \ +asm( \ + " multu %6,%7 \n\t" \ + " mflo $12 \n\t" \ + " mfhi $13 \n\t" \ + " addu %0,%0,$12 \n\t" \ + " sltu $12,%0,$12 \n\t" \ + " addu %1,%1,$13 \n\t" \ + " sltu $13,%1,$13 \n\t" \ + " addu %1,%1,$12 \n\t" \ + " sltu $12,%1,$12 \n\t" \ + " addu %2,%2,$13 \n\t" \ + " addu %2,%2,$12 \n\t" \ +:"=r"(sc0), "=r"(sc1), "=r"(sc2):"0"(sc0), "1"(sc1), "2"(sc2), "r"(i), "r"(j):"$12", "$13", "$14"); + +#define SQRADDDB \ +asm( \ + " addu %0,%0,%3 \n\t" \ + " sltu $10,%0,%3 \n\t" \ + " addu %1,%1,$10 \n\t" \ + " sltu $10,%1,$10 \n\t" \ + " addu %1,%1,%4 \n\t" \ + " sltu $11,%1,%4 \n\t" \ + " addu %2,%2,$10 \n\t" \ + " addu %2,%2,$11 \n\t" \ + " addu %2,%2,%5 \n\t" \ + \ + " addu %0,%0,%3 \n\t" \ + " sltu $10,%0,%3 \n\t" \ + " addu %1,%1,$10 \n\t" \ + " sltu $10,%1,$10 \n\t" \ + " addu %1,%1,%4 \n\t" \ + " sltu $11,%1,%4 \n\t" \ + " addu %2,%2,$10 \n\t" \ + " addu %2,%2,$11 \n\t" \ + " addu %2,%2,%5 \n\t" \ +:"=r"(c0), "=r"(c1), "=r"(c2) : "r"(sc0), "r"(sc1), "r"(sc2), "0"(c0), "1"(c1), "2"(c2) : "$10", "$11"); + +#else + +#define TFM_ISO + +/* ISO C portable code */ + +#define COMBA_START + +#define CLEAR_CARRY \ + c0 = c1 = c2 = 0; + +#define COMBA_STORE(x) \ + x = c0; + +#define COMBA_STORE2(x) \ + x = c1; + +#define CARRY_FORWARD \ + do { c0 = c1; c1 = c2; c2 = 0; } while (0); + +#define COMBA_FINI + +/* multiplies point i and j, updates carry "c1" and digit c2 */ +#define SQRADD(i, j) \ + do { fp_word t; \ + t = c0 + ((fp_word)i) * ((fp_word)j); c0 = t; \ + t = c1 + (t >> DIGIT_BIT); c1 = t; c2 += t >> DIGIT_BIT; \ + } while (0); + + +/* for squaring some of the terms are doubled... */ +#define SQRADD2(i, j) \ + do { fp_word t; \ + t = ((fp_word)i) * ((fp_word)j); \ + tt = (fp_word)c0 + t; c0 = tt; \ + tt = (fp_word)c1 + (tt >> DIGIT_BIT); c1 = tt; c2 += tt >> DIGIT_BIT; \ + tt = (fp_word)c0 + t; c0 = tt; \ + tt = (fp_word)c1 + (tt >> DIGIT_BIT); c1 = tt; c2 += tt >> DIGIT_BIT; \ + } while (0); + +#define SQRADDSC(i, j) \ + do { fp_word t; \ + t = ((fp_word)i) * ((fp_word)j); \ + sc0 = (fp_digit)t; sc1 = (t >> DIGIT_BIT); sc2 = 0; \ + } while (0); + +#define SQRADDAC(i, j) \ + do { fp_word t; \ + t = sc0 + ((fp_word)i) * ((fp_word)j); sc0 = t; \ + t = sc1 + (t >> DIGIT_BIT); sc1 = t; sc2 += t >> DIGIT_BIT; \ + } while (0); + +#define SQRADDDB \ + do { fp_word t; \ + t = ((fp_word)sc0) + ((fp_word)sc0) + c0; c0 = t; \ + t = ((fp_word)sc1) + ((fp_word)sc1) + c1 + (t >> DIGIT_BIT); c1 = t; \ + c2 = c2 + ((fp_word)sc2) + ((fp_word)sc2) + (t >> DIGIT_BIT); \ + } while (0); + +#endif + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_sqr_comba_20.c b/src/crypto/tfm/fp_sqr_comba_20.c new file mode 100644 index 0000000..506f754 --- /dev/null +++ b/src/crypto/tfm/fp_sqr_comba_20.c @@ -0,0 +1,224 @@ +#define TFM_DEFINES +#include "fp_sqr_comba.c" + +#if defined(TFM_SQR20) && FP_SIZE >= 40 +void fp_sqr_comba20(fp_int *A, fp_int *B) +{ + fp_digit *a, b[40], c0, c1, c2, sc0, sc1, sc2; +#ifdef TFM_ISO + fp_word tt; +#endif + + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[10]); SQRADDAC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[11]); SQRADDAC(a[1], a[10]); SQRADDAC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[12]); SQRADDAC(a[1], a[11]); SQRADDAC(a[2], a[10]); SQRADDAC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[13]); SQRADDAC(a[1], a[12]); SQRADDAC(a[2], a[11]); SQRADDAC(a[3], a[10]); SQRADDAC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[14]); SQRADDAC(a[1], a[13]); SQRADDAC(a[2], a[12]); SQRADDAC(a[3], a[11]); SQRADDAC(a[4], a[10]); SQRADDAC(a[5], a[9]); SQRADDAC(a[6], a[8]); SQRADDDB; SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[15]); SQRADDAC(a[1], a[14]); SQRADDAC(a[2], a[13]); SQRADDAC(a[3], a[12]); SQRADDAC(a[4], a[11]); SQRADDAC(a[5], a[10]); SQRADDAC(a[6], a[9]); SQRADDAC(a[7], a[8]); SQRADDDB; + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[16]); SQRADDAC(a[1], a[15]); SQRADDAC(a[2], a[14]); SQRADDAC(a[3], a[13]); SQRADDAC(a[4], a[12]); SQRADDAC(a[5], a[11]); SQRADDAC(a[6], a[10]); SQRADDAC(a[7], a[9]); SQRADDDB; SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[17]); SQRADDAC(a[1], a[16]); SQRADDAC(a[2], a[15]); SQRADDAC(a[3], a[14]); SQRADDAC(a[4], a[13]); SQRADDAC(a[5], a[12]); SQRADDAC(a[6], a[11]); SQRADDAC(a[7], a[10]); SQRADDAC(a[8], a[9]); SQRADDDB; + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[18]); SQRADDAC(a[1], a[17]); SQRADDAC(a[2], a[16]); SQRADDAC(a[3], a[15]); SQRADDAC(a[4], a[14]); SQRADDAC(a[5], a[13]); SQRADDAC(a[6], a[12]); SQRADDAC(a[7], a[11]); SQRADDAC(a[8], a[10]); SQRADDDB; SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + + /* output 19 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[19]); SQRADDAC(a[1], a[18]); SQRADDAC(a[2], a[17]); SQRADDAC(a[3], a[16]); SQRADDAC(a[4], a[15]); SQRADDAC(a[5], a[14]); SQRADDAC(a[6], a[13]); SQRADDAC(a[7], a[12]); SQRADDAC(a[8], a[11]); SQRADDAC(a[9], a[10]); SQRADDDB; + COMBA_STORE(b[19]); + + /* output 20 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[19]); SQRADDAC(a[2], a[18]); SQRADDAC(a[3], a[17]); SQRADDAC(a[4], a[16]); SQRADDAC(a[5], a[15]); SQRADDAC(a[6], a[14]); SQRADDAC(a[7], a[13]); SQRADDAC(a[8], a[12]); SQRADDAC(a[9], a[11]); SQRADDDB; SQRADD(a[10], a[10]); + COMBA_STORE(b[20]); + + /* output 21 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[19]); SQRADDAC(a[3], a[18]); SQRADDAC(a[4], a[17]); SQRADDAC(a[5], a[16]); SQRADDAC(a[6], a[15]); SQRADDAC(a[7], a[14]); SQRADDAC(a[8], a[13]); SQRADDAC(a[9], a[12]); SQRADDAC(a[10], a[11]); SQRADDDB; + COMBA_STORE(b[21]); + + /* output 22 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[19]); SQRADDAC(a[4], a[18]); SQRADDAC(a[5], a[17]); SQRADDAC(a[6], a[16]); SQRADDAC(a[7], a[15]); SQRADDAC(a[8], a[14]); SQRADDAC(a[9], a[13]); SQRADDAC(a[10], a[12]); SQRADDDB; SQRADD(a[11], a[11]); + COMBA_STORE(b[22]); + + /* output 23 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[19]); SQRADDAC(a[5], a[18]); SQRADDAC(a[6], a[17]); SQRADDAC(a[7], a[16]); SQRADDAC(a[8], a[15]); SQRADDAC(a[9], a[14]); SQRADDAC(a[10], a[13]); SQRADDAC(a[11], a[12]); SQRADDDB; + COMBA_STORE(b[23]); + + /* output 24 */ + CARRY_FORWARD; + SQRADDSC(a[5], a[19]); SQRADDAC(a[6], a[18]); SQRADDAC(a[7], a[17]); SQRADDAC(a[8], a[16]); SQRADDAC(a[9], a[15]); SQRADDAC(a[10], a[14]); SQRADDAC(a[11], a[13]); SQRADDDB; SQRADD(a[12], a[12]); + COMBA_STORE(b[24]); + + /* output 25 */ + CARRY_FORWARD; + SQRADDSC(a[6], a[19]); SQRADDAC(a[7], a[18]); SQRADDAC(a[8], a[17]); SQRADDAC(a[9], a[16]); SQRADDAC(a[10], a[15]); SQRADDAC(a[11], a[14]); SQRADDAC(a[12], a[13]); SQRADDDB; + COMBA_STORE(b[25]); + + /* output 26 */ + CARRY_FORWARD; + SQRADDSC(a[7], a[19]); SQRADDAC(a[8], a[18]); SQRADDAC(a[9], a[17]); SQRADDAC(a[10], a[16]); SQRADDAC(a[11], a[15]); SQRADDAC(a[12], a[14]); SQRADDDB; SQRADD(a[13], a[13]); + COMBA_STORE(b[26]); + + /* output 27 */ + CARRY_FORWARD; + SQRADDSC(a[8], a[19]); SQRADDAC(a[9], a[18]); SQRADDAC(a[10], a[17]); SQRADDAC(a[11], a[16]); SQRADDAC(a[12], a[15]); SQRADDAC(a[13], a[14]); SQRADDDB; + COMBA_STORE(b[27]); + + /* output 28 */ + CARRY_FORWARD; + SQRADDSC(a[9], a[19]); SQRADDAC(a[10], a[18]); SQRADDAC(a[11], a[17]); SQRADDAC(a[12], a[16]); SQRADDAC(a[13], a[15]); SQRADDDB; SQRADD(a[14], a[14]); + COMBA_STORE(b[28]); + + /* output 29 */ + CARRY_FORWARD; + SQRADDSC(a[10], a[19]); SQRADDAC(a[11], a[18]); SQRADDAC(a[12], a[17]); SQRADDAC(a[13], a[16]); SQRADDAC(a[14], a[15]); SQRADDDB; + COMBA_STORE(b[29]); + + /* output 30 */ + CARRY_FORWARD; + SQRADDSC(a[11], a[19]); SQRADDAC(a[12], a[18]); SQRADDAC(a[13], a[17]); SQRADDAC(a[14], a[16]); SQRADDDB; SQRADD(a[15], a[15]); + COMBA_STORE(b[30]); + + /* output 31 */ + CARRY_FORWARD; + SQRADDSC(a[12], a[19]); SQRADDAC(a[13], a[18]); SQRADDAC(a[14], a[17]); SQRADDAC(a[15], a[16]); SQRADDDB; + COMBA_STORE(b[31]); + + /* output 32 */ + CARRY_FORWARD; + SQRADDSC(a[13], a[19]); SQRADDAC(a[14], a[18]); SQRADDAC(a[15], a[17]); SQRADDDB; SQRADD(a[16], a[16]); + COMBA_STORE(b[32]); + + /* output 33 */ + CARRY_FORWARD; + SQRADDSC(a[14], a[19]); SQRADDAC(a[15], a[18]); SQRADDAC(a[16], a[17]); SQRADDDB; + COMBA_STORE(b[33]); + + /* output 34 */ + CARRY_FORWARD; + SQRADD2(a[15], a[19]); SQRADD2(a[16], a[18]); SQRADD(a[17], a[17]); + COMBA_STORE(b[34]); + + /* output 35 */ + CARRY_FORWARD; + SQRADD2(a[16], a[19]); SQRADD2(a[17], a[18]); + COMBA_STORE(b[35]); + + /* output 36 */ + CARRY_FORWARD; + SQRADD2(a[17], a[19]); SQRADD(a[18], a[18]); + COMBA_STORE(b[36]); + + /* output 37 */ + CARRY_FORWARD; + SQRADD2(a[18], a[19]); + COMBA_STORE(b[37]); + + /* output 38 */ + CARRY_FORWARD; + SQRADD(a[19], a[19]); + COMBA_STORE(b[38]); + COMBA_STORE2(b[39]); + COMBA_FINI; + + B->used = 40; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 40 * sizeof(fp_digit)); + fp_clamp(B); +} +#endif + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_sqr_comba_24.c b/src/crypto/tfm/fp_sqr_comba_24.c new file mode 100644 index 0000000..4c9f0f8 --- /dev/null +++ b/src/crypto/tfm/fp_sqr_comba_24.c @@ -0,0 +1,264 @@ +#define TFM_DEFINES +#include "fp_sqr_comba.c" + +#if defined(TFM_SQR24) && FP_SIZE >= 48 +void fp_sqr_comba24(fp_int *A, fp_int *B) +{ + fp_digit *a, b[48], c0, c1, c2, sc0, sc1, sc2; +#ifdef TFM_ISO + fp_word tt; +#endif + + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[10]); SQRADDAC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[11]); SQRADDAC(a[1], a[10]); SQRADDAC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[12]); SQRADDAC(a[1], a[11]); SQRADDAC(a[2], a[10]); SQRADDAC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[13]); SQRADDAC(a[1], a[12]); SQRADDAC(a[2], a[11]); SQRADDAC(a[3], a[10]); SQRADDAC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[14]); SQRADDAC(a[1], a[13]); SQRADDAC(a[2], a[12]); SQRADDAC(a[3], a[11]); SQRADDAC(a[4], a[10]); SQRADDAC(a[5], a[9]); SQRADDAC(a[6], a[8]); SQRADDDB; SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[15]); SQRADDAC(a[1], a[14]); SQRADDAC(a[2], a[13]); SQRADDAC(a[3], a[12]); SQRADDAC(a[4], a[11]); SQRADDAC(a[5], a[10]); SQRADDAC(a[6], a[9]); SQRADDAC(a[7], a[8]); SQRADDDB; + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[16]); SQRADDAC(a[1], a[15]); SQRADDAC(a[2], a[14]); SQRADDAC(a[3], a[13]); SQRADDAC(a[4], a[12]); SQRADDAC(a[5], a[11]); SQRADDAC(a[6], a[10]); SQRADDAC(a[7], a[9]); SQRADDDB; SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[17]); SQRADDAC(a[1], a[16]); SQRADDAC(a[2], a[15]); SQRADDAC(a[3], a[14]); SQRADDAC(a[4], a[13]); SQRADDAC(a[5], a[12]); SQRADDAC(a[6], a[11]); SQRADDAC(a[7], a[10]); SQRADDAC(a[8], a[9]); SQRADDDB; + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[18]); SQRADDAC(a[1], a[17]); SQRADDAC(a[2], a[16]); SQRADDAC(a[3], a[15]); SQRADDAC(a[4], a[14]); SQRADDAC(a[5], a[13]); SQRADDAC(a[6], a[12]); SQRADDAC(a[7], a[11]); SQRADDAC(a[8], a[10]); SQRADDDB; SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + + /* output 19 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[19]); SQRADDAC(a[1], a[18]); SQRADDAC(a[2], a[17]); SQRADDAC(a[3], a[16]); SQRADDAC(a[4], a[15]); SQRADDAC(a[5], a[14]); SQRADDAC(a[6], a[13]); SQRADDAC(a[7], a[12]); SQRADDAC(a[8], a[11]); SQRADDAC(a[9], a[10]); SQRADDDB; + COMBA_STORE(b[19]); + + /* output 20 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[20]); SQRADDAC(a[1], a[19]); SQRADDAC(a[2], a[18]); SQRADDAC(a[3], a[17]); SQRADDAC(a[4], a[16]); SQRADDAC(a[5], a[15]); SQRADDAC(a[6], a[14]); SQRADDAC(a[7], a[13]); SQRADDAC(a[8], a[12]); SQRADDAC(a[9], a[11]); SQRADDDB; SQRADD(a[10], a[10]); + COMBA_STORE(b[20]); + + /* output 21 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[21]); SQRADDAC(a[1], a[20]); SQRADDAC(a[2], a[19]); SQRADDAC(a[3], a[18]); SQRADDAC(a[4], a[17]); SQRADDAC(a[5], a[16]); SQRADDAC(a[6], a[15]); SQRADDAC(a[7], a[14]); SQRADDAC(a[8], a[13]); SQRADDAC(a[9], a[12]); SQRADDAC(a[10], a[11]); SQRADDDB; + COMBA_STORE(b[21]); + + /* output 22 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[22]); SQRADDAC(a[1], a[21]); SQRADDAC(a[2], a[20]); SQRADDAC(a[3], a[19]); SQRADDAC(a[4], a[18]); SQRADDAC(a[5], a[17]); SQRADDAC(a[6], a[16]); SQRADDAC(a[7], a[15]); SQRADDAC(a[8], a[14]); SQRADDAC(a[9], a[13]); SQRADDAC(a[10], a[12]); SQRADDDB; SQRADD(a[11], a[11]); + COMBA_STORE(b[22]); + + /* output 23 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[23]); SQRADDAC(a[1], a[22]); SQRADDAC(a[2], a[21]); SQRADDAC(a[3], a[20]); SQRADDAC(a[4], a[19]); SQRADDAC(a[5], a[18]); SQRADDAC(a[6], a[17]); SQRADDAC(a[7], a[16]); SQRADDAC(a[8], a[15]); SQRADDAC(a[9], a[14]); SQRADDAC(a[10], a[13]); SQRADDAC(a[11], a[12]); SQRADDDB; + COMBA_STORE(b[23]); + + /* output 24 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[23]); SQRADDAC(a[2], a[22]); SQRADDAC(a[3], a[21]); SQRADDAC(a[4], a[20]); SQRADDAC(a[5], a[19]); SQRADDAC(a[6], a[18]); SQRADDAC(a[7], a[17]); SQRADDAC(a[8], a[16]); SQRADDAC(a[9], a[15]); SQRADDAC(a[10], a[14]); SQRADDAC(a[11], a[13]); SQRADDDB; SQRADD(a[12], a[12]); + COMBA_STORE(b[24]); + + /* output 25 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[23]); SQRADDAC(a[3], a[22]); SQRADDAC(a[4], a[21]); SQRADDAC(a[5], a[20]); SQRADDAC(a[6], a[19]); SQRADDAC(a[7], a[18]); SQRADDAC(a[8], a[17]); SQRADDAC(a[9], a[16]); SQRADDAC(a[10], a[15]); SQRADDAC(a[11], a[14]); SQRADDAC(a[12], a[13]); SQRADDDB; + COMBA_STORE(b[25]); + + /* output 26 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[23]); SQRADDAC(a[4], a[22]); SQRADDAC(a[5], a[21]); SQRADDAC(a[6], a[20]); SQRADDAC(a[7], a[19]); SQRADDAC(a[8], a[18]); SQRADDAC(a[9], a[17]); SQRADDAC(a[10], a[16]); SQRADDAC(a[11], a[15]); SQRADDAC(a[12], a[14]); SQRADDDB; SQRADD(a[13], a[13]); + COMBA_STORE(b[26]); + + /* output 27 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[23]); SQRADDAC(a[5], a[22]); SQRADDAC(a[6], a[21]); SQRADDAC(a[7], a[20]); SQRADDAC(a[8], a[19]); SQRADDAC(a[9], a[18]); SQRADDAC(a[10], a[17]); SQRADDAC(a[11], a[16]); SQRADDAC(a[12], a[15]); SQRADDAC(a[13], a[14]); SQRADDDB; + COMBA_STORE(b[27]); + + /* output 28 */ + CARRY_FORWARD; + SQRADDSC(a[5], a[23]); SQRADDAC(a[6], a[22]); SQRADDAC(a[7], a[21]); SQRADDAC(a[8], a[20]); SQRADDAC(a[9], a[19]); SQRADDAC(a[10], a[18]); SQRADDAC(a[11], a[17]); SQRADDAC(a[12], a[16]); SQRADDAC(a[13], a[15]); SQRADDDB; SQRADD(a[14], a[14]); + COMBA_STORE(b[28]); + + /* output 29 */ + CARRY_FORWARD; + SQRADDSC(a[6], a[23]); SQRADDAC(a[7], a[22]); SQRADDAC(a[8], a[21]); SQRADDAC(a[9], a[20]); SQRADDAC(a[10], a[19]); SQRADDAC(a[11], a[18]); SQRADDAC(a[12], a[17]); SQRADDAC(a[13], a[16]); SQRADDAC(a[14], a[15]); SQRADDDB; + COMBA_STORE(b[29]); + + /* output 30 */ + CARRY_FORWARD; + SQRADDSC(a[7], a[23]); SQRADDAC(a[8], a[22]); SQRADDAC(a[9], a[21]); SQRADDAC(a[10], a[20]); SQRADDAC(a[11], a[19]); SQRADDAC(a[12], a[18]); SQRADDAC(a[13], a[17]); SQRADDAC(a[14], a[16]); SQRADDDB; SQRADD(a[15], a[15]); + COMBA_STORE(b[30]); + + /* output 31 */ + CARRY_FORWARD; + SQRADDSC(a[8], a[23]); SQRADDAC(a[9], a[22]); SQRADDAC(a[10], a[21]); SQRADDAC(a[11], a[20]); SQRADDAC(a[12], a[19]); SQRADDAC(a[13], a[18]); SQRADDAC(a[14], a[17]); SQRADDAC(a[15], a[16]); SQRADDDB; + COMBA_STORE(b[31]); + + /* output 32 */ + CARRY_FORWARD; + SQRADDSC(a[9], a[23]); SQRADDAC(a[10], a[22]); SQRADDAC(a[11], a[21]); SQRADDAC(a[12], a[20]); SQRADDAC(a[13], a[19]); SQRADDAC(a[14], a[18]); SQRADDAC(a[15], a[17]); SQRADDDB; SQRADD(a[16], a[16]); + COMBA_STORE(b[32]); + + /* output 33 */ + CARRY_FORWARD; + SQRADDSC(a[10], a[23]); SQRADDAC(a[11], a[22]); SQRADDAC(a[12], a[21]); SQRADDAC(a[13], a[20]); SQRADDAC(a[14], a[19]); SQRADDAC(a[15], a[18]); SQRADDAC(a[16], a[17]); SQRADDDB; + COMBA_STORE(b[33]); + + /* output 34 */ + CARRY_FORWARD; + SQRADDSC(a[11], a[23]); SQRADDAC(a[12], a[22]); SQRADDAC(a[13], a[21]); SQRADDAC(a[14], a[20]); SQRADDAC(a[15], a[19]); SQRADDAC(a[16], a[18]); SQRADDDB; SQRADD(a[17], a[17]); + COMBA_STORE(b[34]); + + /* output 35 */ + CARRY_FORWARD; + SQRADDSC(a[12], a[23]); SQRADDAC(a[13], a[22]); SQRADDAC(a[14], a[21]); SQRADDAC(a[15], a[20]); SQRADDAC(a[16], a[19]); SQRADDAC(a[17], a[18]); SQRADDDB; + COMBA_STORE(b[35]); + + /* output 36 */ + CARRY_FORWARD; + SQRADDSC(a[13], a[23]); SQRADDAC(a[14], a[22]); SQRADDAC(a[15], a[21]); SQRADDAC(a[16], a[20]); SQRADDAC(a[17], a[19]); SQRADDDB; SQRADD(a[18], a[18]); + COMBA_STORE(b[36]); + + /* output 37 */ + CARRY_FORWARD; + SQRADDSC(a[14], a[23]); SQRADDAC(a[15], a[22]); SQRADDAC(a[16], a[21]); SQRADDAC(a[17], a[20]); SQRADDAC(a[18], a[19]); SQRADDDB; + COMBA_STORE(b[37]); + + /* output 38 */ + CARRY_FORWARD; + SQRADDSC(a[15], a[23]); SQRADDAC(a[16], a[22]); SQRADDAC(a[17], a[21]); SQRADDAC(a[18], a[20]); SQRADDDB; SQRADD(a[19], a[19]); + COMBA_STORE(b[38]); + + /* output 39 */ + CARRY_FORWARD; + SQRADDSC(a[16], a[23]); SQRADDAC(a[17], a[22]); SQRADDAC(a[18], a[21]); SQRADDAC(a[19], a[20]); SQRADDDB; + COMBA_STORE(b[39]); + + /* output 40 */ + CARRY_FORWARD; + SQRADDSC(a[17], a[23]); SQRADDAC(a[18], a[22]); SQRADDAC(a[19], a[21]); SQRADDDB; SQRADD(a[20], a[20]); + COMBA_STORE(b[40]); + + /* output 41 */ + CARRY_FORWARD; + SQRADDSC(a[18], a[23]); SQRADDAC(a[19], a[22]); SQRADDAC(a[20], a[21]); SQRADDDB; + COMBA_STORE(b[41]); + + /* output 42 */ + CARRY_FORWARD; + SQRADD2(a[19], a[23]); SQRADD2(a[20], a[22]); SQRADD(a[21], a[21]); + COMBA_STORE(b[42]); + + /* output 43 */ + CARRY_FORWARD; + SQRADD2(a[20], a[23]); SQRADD2(a[21], a[22]); + COMBA_STORE(b[43]); + + /* output 44 */ + CARRY_FORWARD; + SQRADD2(a[21], a[23]); SQRADD(a[22], a[22]); + COMBA_STORE(b[44]); + + /* output 45 */ + CARRY_FORWARD; + SQRADD2(a[22], a[23]); + COMBA_STORE(b[45]); + + /* output 46 */ + CARRY_FORWARD; + SQRADD(a[23], a[23]); + COMBA_STORE(b[46]); + COMBA_STORE2(b[47]); + COMBA_FINI; + + B->used = 48; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 48 * sizeof(fp_digit)); + fp_clamp(B); +} +#endif + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_sqr_comba_28.c b/src/crypto/tfm/fp_sqr_comba_28.c new file mode 100644 index 0000000..bd13a0a --- /dev/null +++ b/src/crypto/tfm/fp_sqr_comba_28.c @@ -0,0 +1,304 @@ +#define TFM_DEFINES +#include "fp_sqr_comba.c" + +#if defined(TFM_SQR28) && FP_SIZE >= 56 +void fp_sqr_comba28(fp_int *A, fp_int *B) +{ + fp_digit *a, b[56], c0, c1, c2, sc0, sc1, sc2; +#ifdef TFM_ISO + fp_word tt; +#endif + + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[10]); SQRADDAC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[11]); SQRADDAC(a[1], a[10]); SQRADDAC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[12]); SQRADDAC(a[1], a[11]); SQRADDAC(a[2], a[10]); SQRADDAC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[13]); SQRADDAC(a[1], a[12]); SQRADDAC(a[2], a[11]); SQRADDAC(a[3], a[10]); SQRADDAC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[14]); SQRADDAC(a[1], a[13]); SQRADDAC(a[2], a[12]); SQRADDAC(a[3], a[11]); SQRADDAC(a[4], a[10]); SQRADDAC(a[5], a[9]); SQRADDAC(a[6], a[8]); SQRADDDB; SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[15]); SQRADDAC(a[1], a[14]); SQRADDAC(a[2], a[13]); SQRADDAC(a[3], a[12]); SQRADDAC(a[4], a[11]); SQRADDAC(a[5], a[10]); SQRADDAC(a[6], a[9]); SQRADDAC(a[7], a[8]); SQRADDDB; + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[16]); SQRADDAC(a[1], a[15]); SQRADDAC(a[2], a[14]); SQRADDAC(a[3], a[13]); SQRADDAC(a[4], a[12]); SQRADDAC(a[5], a[11]); SQRADDAC(a[6], a[10]); SQRADDAC(a[7], a[9]); SQRADDDB; SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[17]); SQRADDAC(a[1], a[16]); SQRADDAC(a[2], a[15]); SQRADDAC(a[3], a[14]); SQRADDAC(a[4], a[13]); SQRADDAC(a[5], a[12]); SQRADDAC(a[6], a[11]); SQRADDAC(a[7], a[10]); SQRADDAC(a[8], a[9]); SQRADDDB; + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[18]); SQRADDAC(a[1], a[17]); SQRADDAC(a[2], a[16]); SQRADDAC(a[3], a[15]); SQRADDAC(a[4], a[14]); SQRADDAC(a[5], a[13]); SQRADDAC(a[6], a[12]); SQRADDAC(a[7], a[11]); SQRADDAC(a[8], a[10]); SQRADDDB; SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + + /* output 19 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[19]); SQRADDAC(a[1], a[18]); SQRADDAC(a[2], a[17]); SQRADDAC(a[3], a[16]); SQRADDAC(a[4], a[15]); SQRADDAC(a[5], a[14]); SQRADDAC(a[6], a[13]); SQRADDAC(a[7], a[12]); SQRADDAC(a[8], a[11]); SQRADDAC(a[9], a[10]); SQRADDDB; + COMBA_STORE(b[19]); + + /* output 20 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[20]); SQRADDAC(a[1], a[19]); SQRADDAC(a[2], a[18]); SQRADDAC(a[3], a[17]); SQRADDAC(a[4], a[16]); SQRADDAC(a[5], a[15]); SQRADDAC(a[6], a[14]); SQRADDAC(a[7], a[13]); SQRADDAC(a[8], a[12]); SQRADDAC(a[9], a[11]); SQRADDDB; SQRADD(a[10], a[10]); + COMBA_STORE(b[20]); + + /* output 21 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[21]); SQRADDAC(a[1], a[20]); SQRADDAC(a[2], a[19]); SQRADDAC(a[3], a[18]); SQRADDAC(a[4], a[17]); SQRADDAC(a[5], a[16]); SQRADDAC(a[6], a[15]); SQRADDAC(a[7], a[14]); SQRADDAC(a[8], a[13]); SQRADDAC(a[9], a[12]); SQRADDAC(a[10], a[11]); SQRADDDB; + COMBA_STORE(b[21]); + + /* output 22 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[22]); SQRADDAC(a[1], a[21]); SQRADDAC(a[2], a[20]); SQRADDAC(a[3], a[19]); SQRADDAC(a[4], a[18]); SQRADDAC(a[5], a[17]); SQRADDAC(a[6], a[16]); SQRADDAC(a[7], a[15]); SQRADDAC(a[8], a[14]); SQRADDAC(a[9], a[13]); SQRADDAC(a[10], a[12]); SQRADDDB; SQRADD(a[11], a[11]); + COMBA_STORE(b[22]); + + /* output 23 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[23]); SQRADDAC(a[1], a[22]); SQRADDAC(a[2], a[21]); SQRADDAC(a[3], a[20]); SQRADDAC(a[4], a[19]); SQRADDAC(a[5], a[18]); SQRADDAC(a[6], a[17]); SQRADDAC(a[7], a[16]); SQRADDAC(a[8], a[15]); SQRADDAC(a[9], a[14]); SQRADDAC(a[10], a[13]); SQRADDAC(a[11], a[12]); SQRADDDB; + COMBA_STORE(b[23]); + + /* output 24 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[24]); SQRADDAC(a[1], a[23]); SQRADDAC(a[2], a[22]); SQRADDAC(a[3], a[21]); SQRADDAC(a[4], a[20]); SQRADDAC(a[5], a[19]); SQRADDAC(a[6], a[18]); SQRADDAC(a[7], a[17]); SQRADDAC(a[8], a[16]); SQRADDAC(a[9], a[15]); SQRADDAC(a[10], a[14]); SQRADDAC(a[11], a[13]); SQRADDDB; SQRADD(a[12], a[12]); + COMBA_STORE(b[24]); + + /* output 25 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[25]); SQRADDAC(a[1], a[24]); SQRADDAC(a[2], a[23]); SQRADDAC(a[3], a[22]); SQRADDAC(a[4], a[21]); SQRADDAC(a[5], a[20]); SQRADDAC(a[6], a[19]); SQRADDAC(a[7], a[18]); SQRADDAC(a[8], a[17]); SQRADDAC(a[9], a[16]); SQRADDAC(a[10], a[15]); SQRADDAC(a[11], a[14]); SQRADDAC(a[12], a[13]); SQRADDDB; + COMBA_STORE(b[25]); + + /* output 26 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[26]); SQRADDAC(a[1], a[25]); SQRADDAC(a[2], a[24]); SQRADDAC(a[3], a[23]); SQRADDAC(a[4], a[22]); SQRADDAC(a[5], a[21]); SQRADDAC(a[6], a[20]); SQRADDAC(a[7], a[19]); SQRADDAC(a[8], a[18]); SQRADDAC(a[9], a[17]); SQRADDAC(a[10], a[16]); SQRADDAC(a[11], a[15]); SQRADDAC(a[12], a[14]); SQRADDDB; SQRADD(a[13], a[13]); + COMBA_STORE(b[26]); + + /* output 27 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[27]); SQRADDAC(a[1], a[26]); SQRADDAC(a[2], a[25]); SQRADDAC(a[3], a[24]); SQRADDAC(a[4], a[23]); SQRADDAC(a[5], a[22]); SQRADDAC(a[6], a[21]); SQRADDAC(a[7], a[20]); SQRADDAC(a[8], a[19]); SQRADDAC(a[9], a[18]); SQRADDAC(a[10], a[17]); SQRADDAC(a[11], a[16]); SQRADDAC(a[12], a[15]); SQRADDAC(a[13], a[14]); SQRADDDB; + COMBA_STORE(b[27]); + + /* output 28 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[27]); SQRADDAC(a[2], a[26]); SQRADDAC(a[3], a[25]); SQRADDAC(a[4], a[24]); SQRADDAC(a[5], a[23]); SQRADDAC(a[6], a[22]); SQRADDAC(a[7], a[21]); SQRADDAC(a[8], a[20]); SQRADDAC(a[9], a[19]); SQRADDAC(a[10], a[18]); SQRADDAC(a[11], a[17]); SQRADDAC(a[12], a[16]); SQRADDAC(a[13], a[15]); SQRADDDB; SQRADD(a[14], a[14]); + COMBA_STORE(b[28]); + + /* output 29 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[27]); SQRADDAC(a[3], a[26]); SQRADDAC(a[4], a[25]); SQRADDAC(a[5], a[24]); SQRADDAC(a[6], a[23]); SQRADDAC(a[7], a[22]); SQRADDAC(a[8], a[21]); SQRADDAC(a[9], a[20]); SQRADDAC(a[10], a[19]); SQRADDAC(a[11], a[18]); SQRADDAC(a[12], a[17]); SQRADDAC(a[13], a[16]); SQRADDAC(a[14], a[15]); SQRADDDB; + COMBA_STORE(b[29]); + + /* output 30 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[27]); SQRADDAC(a[4], a[26]); SQRADDAC(a[5], a[25]); SQRADDAC(a[6], a[24]); SQRADDAC(a[7], a[23]); SQRADDAC(a[8], a[22]); SQRADDAC(a[9], a[21]); SQRADDAC(a[10], a[20]); SQRADDAC(a[11], a[19]); SQRADDAC(a[12], a[18]); SQRADDAC(a[13], a[17]); SQRADDAC(a[14], a[16]); SQRADDDB; SQRADD(a[15], a[15]); + COMBA_STORE(b[30]); + + /* output 31 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[27]); SQRADDAC(a[5], a[26]); SQRADDAC(a[6], a[25]); SQRADDAC(a[7], a[24]); SQRADDAC(a[8], a[23]); SQRADDAC(a[9], a[22]); SQRADDAC(a[10], a[21]); SQRADDAC(a[11], a[20]); SQRADDAC(a[12], a[19]); SQRADDAC(a[13], a[18]); SQRADDAC(a[14], a[17]); SQRADDAC(a[15], a[16]); SQRADDDB; + COMBA_STORE(b[31]); + + /* output 32 */ + CARRY_FORWARD; + SQRADDSC(a[5], a[27]); SQRADDAC(a[6], a[26]); SQRADDAC(a[7], a[25]); SQRADDAC(a[8], a[24]); SQRADDAC(a[9], a[23]); SQRADDAC(a[10], a[22]); SQRADDAC(a[11], a[21]); SQRADDAC(a[12], a[20]); SQRADDAC(a[13], a[19]); SQRADDAC(a[14], a[18]); SQRADDAC(a[15], a[17]); SQRADDDB; SQRADD(a[16], a[16]); + COMBA_STORE(b[32]); + + /* output 33 */ + CARRY_FORWARD; + SQRADDSC(a[6], a[27]); SQRADDAC(a[7], a[26]); SQRADDAC(a[8], a[25]); SQRADDAC(a[9], a[24]); SQRADDAC(a[10], a[23]); SQRADDAC(a[11], a[22]); SQRADDAC(a[12], a[21]); SQRADDAC(a[13], a[20]); SQRADDAC(a[14], a[19]); SQRADDAC(a[15], a[18]); SQRADDAC(a[16], a[17]); SQRADDDB; + COMBA_STORE(b[33]); + + /* output 34 */ + CARRY_FORWARD; + SQRADDSC(a[7], a[27]); SQRADDAC(a[8], a[26]); SQRADDAC(a[9], a[25]); SQRADDAC(a[10], a[24]); SQRADDAC(a[11], a[23]); SQRADDAC(a[12], a[22]); SQRADDAC(a[13], a[21]); SQRADDAC(a[14], a[20]); SQRADDAC(a[15], a[19]); SQRADDAC(a[16], a[18]); SQRADDDB; SQRADD(a[17], a[17]); + COMBA_STORE(b[34]); + + /* output 35 */ + CARRY_FORWARD; + SQRADDSC(a[8], a[27]); SQRADDAC(a[9], a[26]); SQRADDAC(a[10], a[25]); SQRADDAC(a[11], a[24]); SQRADDAC(a[12], a[23]); SQRADDAC(a[13], a[22]); SQRADDAC(a[14], a[21]); SQRADDAC(a[15], a[20]); SQRADDAC(a[16], a[19]); SQRADDAC(a[17], a[18]); SQRADDDB; + COMBA_STORE(b[35]); + + /* output 36 */ + CARRY_FORWARD; + SQRADDSC(a[9], a[27]); SQRADDAC(a[10], a[26]); SQRADDAC(a[11], a[25]); SQRADDAC(a[12], a[24]); SQRADDAC(a[13], a[23]); SQRADDAC(a[14], a[22]); SQRADDAC(a[15], a[21]); SQRADDAC(a[16], a[20]); SQRADDAC(a[17], a[19]); SQRADDDB; SQRADD(a[18], a[18]); + COMBA_STORE(b[36]); + + /* output 37 */ + CARRY_FORWARD; + SQRADDSC(a[10], a[27]); SQRADDAC(a[11], a[26]); SQRADDAC(a[12], a[25]); SQRADDAC(a[13], a[24]); SQRADDAC(a[14], a[23]); SQRADDAC(a[15], a[22]); SQRADDAC(a[16], a[21]); SQRADDAC(a[17], a[20]); SQRADDAC(a[18], a[19]); SQRADDDB; + COMBA_STORE(b[37]); + + /* output 38 */ + CARRY_FORWARD; + SQRADDSC(a[11], a[27]); SQRADDAC(a[12], a[26]); SQRADDAC(a[13], a[25]); SQRADDAC(a[14], a[24]); SQRADDAC(a[15], a[23]); SQRADDAC(a[16], a[22]); SQRADDAC(a[17], a[21]); SQRADDAC(a[18], a[20]); SQRADDDB; SQRADD(a[19], a[19]); + COMBA_STORE(b[38]); + + /* output 39 */ + CARRY_FORWARD; + SQRADDSC(a[12], a[27]); SQRADDAC(a[13], a[26]); SQRADDAC(a[14], a[25]); SQRADDAC(a[15], a[24]); SQRADDAC(a[16], a[23]); SQRADDAC(a[17], a[22]); SQRADDAC(a[18], a[21]); SQRADDAC(a[19], a[20]); SQRADDDB; + COMBA_STORE(b[39]); + + /* output 40 */ + CARRY_FORWARD; + SQRADDSC(a[13], a[27]); SQRADDAC(a[14], a[26]); SQRADDAC(a[15], a[25]); SQRADDAC(a[16], a[24]); SQRADDAC(a[17], a[23]); SQRADDAC(a[18], a[22]); SQRADDAC(a[19], a[21]); SQRADDDB; SQRADD(a[20], a[20]); + COMBA_STORE(b[40]); + + /* output 41 */ + CARRY_FORWARD; + SQRADDSC(a[14], a[27]); SQRADDAC(a[15], a[26]); SQRADDAC(a[16], a[25]); SQRADDAC(a[17], a[24]); SQRADDAC(a[18], a[23]); SQRADDAC(a[19], a[22]); SQRADDAC(a[20], a[21]); SQRADDDB; + COMBA_STORE(b[41]); + + /* output 42 */ + CARRY_FORWARD; + SQRADDSC(a[15], a[27]); SQRADDAC(a[16], a[26]); SQRADDAC(a[17], a[25]); SQRADDAC(a[18], a[24]); SQRADDAC(a[19], a[23]); SQRADDAC(a[20], a[22]); SQRADDDB; SQRADD(a[21], a[21]); + COMBA_STORE(b[42]); + + /* output 43 */ + CARRY_FORWARD; + SQRADDSC(a[16], a[27]); SQRADDAC(a[17], a[26]); SQRADDAC(a[18], a[25]); SQRADDAC(a[19], a[24]); SQRADDAC(a[20], a[23]); SQRADDAC(a[21], a[22]); SQRADDDB; + COMBA_STORE(b[43]); + + /* output 44 */ + CARRY_FORWARD; + SQRADDSC(a[17], a[27]); SQRADDAC(a[18], a[26]); SQRADDAC(a[19], a[25]); SQRADDAC(a[20], a[24]); SQRADDAC(a[21], a[23]); SQRADDDB; SQRADD(a[22], a[22]); + COMBA_STORE(b[44]); + + /* output 45 */ + CARRY_FORWARD; + SQRADDSC(a[18], a[27]); SQRADDAC(a[19], a[26]); SQRADDAC(a[20], a[25]); SQRADDAC(a[21], a[24]); SQRADDAC(a[22], a[23]); SQRADDDB; + COMBA_STORE(b[45]); + + /* output 46 */ + CARRY_FORWARD; + SQRADDSC(a[19], a[27]); SQRADDAC(a[20], a[26]); SQRADDAC(a[21], a[25]); SQRADDAC(a[22], a[24]); SQRADDDB; SQRADD(a[23], a[23]); + COMBA_STORE(b[46]); + + /* output 47 */ + CARRY_FORWARD; + SQRADDSC(a[20], a[27]); SQRADDAC(a[21], a[26]); SQRADDAC(a[22], a[25]); SQRADDAC(a[23], a[24]); SQRADDDB; + COMBA_STORE(b[47]); + + /* output 48 */ + CARRY_FORWARD; + SQRADDSC(a[21], a[27]); SQRADDAC(a[22], a[26]); SQRADDAC(a[23], a[25]); SQRADDDB; SQRADD(a[24], a[24]); + COMBA_STORE(b[48]); + + /* output 49 */ + CARRY_FORWARD; + SQRADDSC(a[22], a[27]); SQRADDAC(a[23], a[26]); SQRADDAC(a[24], a[25]); SQRADDDB; + COMBA_STORE(b[49]); + + /* output 50 */ + CARRY_FORWARD; + SQRADD2(a[23], a[27]); SQRADD2(a[24], a[26]); SQRADD(a[25], a[25]); + COMBA_STORE(b[50]); + + /* output 51 */ + CARRY_FORWARD; + SQRADD2(a[24], a[27]); SQRADD2(a[25], a[26]); + COMBA_STORE(b[51]); + + /* output 52 */ + CARRY_FORWARD; + SQRADD2(a[25], a[27]); SQRADD(a[26], a[26]); + COMBA_STORE(b[52]); + + /* output 53 */ + CARRY_FORWARD; + SQRADD2(a[26], a[27]); + COMBA_STORE(b[53]); + + /* output 54 */ + CARRY_FORWARD; + SQRADD(a[27], a[27]); + COMBA_STORE(b[54]); + COMBA_STORE2(b[55]); + COMBA_FINI; + + B->used = 56; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 56 * sizeof(fp_digit)); + fp_clamp(B); +} +#endif + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_sqr_comba_32.c b/src/crypto/tfm/fp_sqr_comba_32.c new file mode 100644 index 0000000..5ce343d --- /dev/null +++ b/src/crypto/tfm/fp_sqr_comba_32.c @@ -0,0 +1,344 @@ +#define TFM_DEFINES +#include "fp_sqr_comba.c" + +#if defined(TFM_SQR32) && FP_SIZE >= 64 +void fp_sqr_comba32(fp_int *A, fp_int *B) +{ + fp_digit *a, b[64], c0, c1, c2, sc0, sc1, sc2; +#ifdef TFM_ISO + fp_word tt; +#endif + + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[10]); SQRADDAC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[11]); SQRADDAC(a[1], a[10]); SQRADDAC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[12]); SQRADDAC(a[1], a[11]); SQRADDAC(a[2], a[10]); SQRADDAC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[13]); SQRADDAC(a[1], a[12]); SQRADDAC(a[2], a[11]); SQRADDAC(a[3], a[10]); SQRADDAC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[14]); SQRADDAC(a[1], a[13]); SQRADDAC(a[2], a[12]); SQRADDAC(a[3], a[11]); SQRADDAC(a[4], a[10]); SQRADDAC(a[5], a[9]); SQRADDAC(a[6], a[8]); SQRADDDB; SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[15]); SQRADDAC(a[1], a[14]); SQRADDAC(a[2], a[13]); SQRADDAC(a[3], a[12]); SQRADDAC(a[4], a[11]); SQRADDAC(a[5], a[10]); SQRADDAC(a[6], a[9]); SQRADDAC(a[7], a[8]); SQRADDDB; + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[16]); SQRADDAC(a[1], a[15]); SQRADDAC(a[2], a[14]); SQRADDAC(a[3], a[13]); SQRADDAC(a[4], a[12]); SQRADDAC(a[5], a[11]); SQRADDAC(a[6], a[10]); SQRADDAC(a[7], a[9]); SQRADDDB; SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[17]); SQRADDAC(a[1], a[16]); SQRADDAC(a[2], a[15]); SQRADDAC(a[3], a[14]); SQRADDAC(a[4], a[13]); SQRADDAC(a[5], a[12]); SQRADDAC(a[6], a[11]); SQRADDAC(a[7], a[10]); SQRADDAC(a[8], a[9]); SQRADDDB; + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[18]); SQRADDAC(a[1], a[17]); SQRADDAC(a[2], a[16]); SQRADDAC(a[3], a[15]); SQRADDAC(a[4], a[14]); SQRADDAC(a[5], a[13]); SQRADDAC(a[6], a[12]); SQRADDAC(a[7], a[11]); SQRADDAC(a[8], a[10]); SQRADDDB; SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + + /* output 19 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[19]); SQRADDAC(a[1], a[18]); SQRADDAC(a[2], a[17]); SQRADDAC(a[3], a[16]); SQRADDAC(a[4], a[15]); SQRADDAC(a[5], a[14]); SQRADDAC(a[6], a[13]); SQRADDAC(a[7], a[12]); SQRADDAC(a[8], a[11]); SQRADDAC(a[9], a[10]); SQRADDDB; + COMBA_STORE(b[19]); + + /* output 20 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[20]); SQRADDAC(a[1], a[19]); SQRADDAC(a[2], a[18]); SQRADDAC(a[3], a[17]); SQRADDAC(a[4], a[16]); SQRADDAC(a[5], a[15]); SQRADDAC(a[6], a[14]); SQRADDAC(a[7], a[13]); SQRADDAC(a[8], a[12]); SQRADDAC(a[9], a[11]); SQRADDDB; SQRADD(a[10], a[10]); + COMBA_STORE(b[20]); + + /* output 21 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[21]); SQRADDAC(a[1], a[20]); SQRADDAC(a[2], a[19]); SQRADDAC(a[3], a[18]); SQRADDAC(a[4], a[17]); SQRADDAC(a[5], a[16]); SQRADDAC(a[6], a[15]); SQRADDAC(a[7], a[14]); SQRADDAC(a[8], a[13]); SQRADDAC(a[9], a[12]); SQRADDAC(a[10], a[11]); SQRADDDB; + COMBA_STORE(b[21]); + + /* output 22 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[22]); SQRADDAC(a[1], a[21]); SQRADDAC(a[2], a[20]); SQRADDAC(a[3], a[19]); SQRADDAC(a[4], a[18]); SQRADDAC(a[5], a[17]); SQRADDAC(a[6], a[16]); SQRADDAC(a[7], a[15]); SQRADDAC(a[8], a[14]); SQRADDAC(a[9], a[13]); SQRADDAC(a[10], a[12]); SQRADDDB; SQRADD(a[11], a[11]); + COMBA_STORE(b[22]); + + /* output 23 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[23]); SQRADDAC(a[1], a[22]); SQRADDAC(a[2], a[21]); SQRADDAC(a[3], a[20]); SQRADDAC(a[4], a[19]); SQRADDAC(a[5], a[18]); SQRADDAC(a[6], a[17]); SQRADDAC(a[7], a[16]); SQRADDAC(a[8], a[15]); SQRADDAC(a[9], a[14]); SQRADDAC(a[10], a[13]); SQRADDAC(a[11], a[12]); SQRADDDB; + COMBA_STORE(b[23]); + + /* output 24 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[24]); SQRADDAC(a[1], a[23]); SQRADDAC(a[2], a[22]); SQRADDAC(a[3], a[21]); SQRADDAC(a[4], a[20]); SQRADDAC(a[5], a[19]); SQRADDAC(a[6], a[18]); SQRADDAC(a[7], a[17]); SQRADDAC(a[8], a[16]); SQRADDAC(a[9], a[15]); SQRADDAC(a[10], a[14]); SQRADDAC(a[11], a[13]); SQRADDDB; SQRADD(a[12], a[12]); + COMBA_STORE(b[24]); + + /* output 25 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[25]); SQRADDAC(a[1], a[24]); SQRADDAC(a[2], a[23]); SQRADDAC(a[3], a[22]); SQRADDAC(a[4], a[21]); SQRADDAC(a[5], a[20]); SQRADDAC(a[6], a[19]); SQRADDAC(a[7], a[18]); SQRADDAC(a[8], a[17]); SQRADDAC(a[9], a[16]); SQRADDAC(a[10], a[15]); SQRADDAC(a[11], a[14]); SQRADDAC(a[12], a[13]); SQRADDDB; + COMBA_STORE(b[25]); + + /* output 26 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[26]); SQRADDAC(a[1], a[25]); SQRADDAC(a[2], a[24]); SQRADDAC(a[3], a[23]); SQRADDAC(a[4], a[22]); SQRADDAC(a[5], a[21]); SQRADDAC(a[6], a[20]); SQRADDAC(a[7], a[19]); SQRADDAC(a[8], a[18]); SQRADDAC(a[9], a[17]); SQRADDAC(a[10], a[16]); SQRADDAC(a[11], a[15]); SQRADDAC(a[12], a[14]); SQRADDDB; SQRADD(a[13], a[13]); + COMBA_STORE(b[26]); + + /* output 27 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[27]); SQRADDAC(a[1], a[26]); SQRADDAC(a[2], a[25]); SQRADDAC(a[3], a[24]); SQRADDAC(a[4], a[23]); SQRADDAC(a[5], a[22]); SQRADDAC(a[6], a[21]); SQRADDAC(a[7], a[20]); SQRADDAC(a[8], a[19]); SQRADDAC(a[9], a[18]); SQRADDAC(a[10], a[17]); SQRADDAC(a[11], a[16]); SQRADDAC(a[12], a[15]); SQRADDAC(a[13], a[14]); SQRADDDB; + COMBA_STORE(b[27]); + + /* output 28 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[28]); SQRADDAC(a[1], a[27]); SQRADDAC(a[2], a[26]); SQRADDAC(a[3], a[25]); SQRADDAC(a[4], a[24]); SQRADDAC(a[5], a[23]); SQRADDAC(a[6], a[22]); SQRADDAC(a[7], a[21]); SQRADDAC(a[8], a[20]); SQRADDAC(a[9], a[19]); SQRADDAC(a[10], a[18]); SQRADDAC(a[11], a[17]); SQRADDAC(a[12], a[16]); SQRADDAC(a[13], a[15]); SQRADDDB; SQRADD(a[14], a[14]); + COMBA_STORE(b[28]); + + /* output 29 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[29]); SQRADDAC(a[1], a[28]); SQRADDAC(a[2], a[27]); SQRADDAC(a[3], a[26]); SQRADDAC(a[4], a[25]); SQRADDAC(a[5], a[24]); SQRADDAC(a[6], a[23]); SQRADDAC(a[7], a[22]); SQRADDAC(a[8], a[21]); SQRADDAC(a[9], a[20]); SQRADDAC(a[10], a[19]); SQRADDAC(a[11], a[18]); SQRADDAC(a[12], a[17]); SQRADDAC(a[13], a[16]); SQRADDAC(a[14], a[15]); SQRADDDB; + COMBA_STORE(b[29]); + + /* output 30 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[30]); SQRADDAC(a[1], a[29]); SQRADDAC(a[2], a[28]); SQRADDAC(a[3], a[27]); SQRADDAC(a[4], a[26]); SQRADDAC(a[5], a[25]); SQRADDAC(a[6], a[24]); SQRADDAC(a[7], a[23]); SQRADDAC(a[8], a[22]); SQRADDAC(a[9], a[21]); SQRADDAC(a[10], a[20]); SQRADDAC(a[11], a[19]); SQRADDAC(a[12], a[18]); SQRADDAC(a[13], a[17]); SQRADDAC(a[14], a[16]); SQRADDDB; SQRADD(a[15], a[15]); + COMBA_STORE(b[30]); + + /* output 31 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[31]); SQRADDAC(a[1], a[30]); SQRADDAC(a[2], a[29]); SQRADDAC(a[3], a[28]); SQRADDAC(a[4], a[27]); SQRADDAC(a[5], a[26]); SQRADDAC(a[6], a[25]); SQRADDAC(a[7], a[24]); SQRADDAC(a[8], a[23]); SQRADDAC(a[9], a[22]); SQRADDAC(a[10], a[21]); SQRADDAC(a[11], a[20]); SQRADDAC(a[12], a[19]); SQRADDAC(a[13], a[18]); SQRADDAC(a[14], a[17]); SQRADDAC(a[15], a[16]); SQRADDDB; + COMBA_STORE(b[31]); + + /* output 32 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[31]); SQRADDAC(a[2], a[30]); SQRADDAC(a[3], a[29]); SQRADDAC(a[4], a[28]); SQRADDAC(a[5], a[27]); SQRADDAC(a[6], a[26]); SQRADDAC(a[7], a[25]); SQRADDAC(a[8], a[24]); SQRADDAC(a[9], a[23]); SQRADDAC(a[10], a[22]); SQRADDAC(a[11], a[21]); SQRADDAC(a[12], a[20]); SQRADDAC(a[13], a[19]); SQRADDAC(a[14], a[18]); SQRADDAC(a[15], a[17]); SQRADDDB; SQRADD(a[16], a[16]); + COMBA_STORE(b[32]); + + /* output 33 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[31]); SQRADDAC(a[3], a[30]); SQRADDAC(a[4], a[29]); SQRADDAC(a[5], a[28]); SQRADDAC(a[6], a[27]); SQRADDAC(a[7], a[26]); SQRADDAC(a[8], a[25]); SQRADDAC(a[9], a[24]); SQRADDAC(a[10], a[23]); SQRADDAC(a[11], a[22]); SQRADDAC(a[12], a[21]); SQRADDAC(a[13], a[20]); SQRADDAC(a[14], a[19]); SQRADDAC(a[15], a[18]); SQRADDAC(a[16], a[17]); SQRADDDB; + COMBA_STORE(b[33]); + + /* output 34 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[31]); SQRADDAC(a[4], a[30]); SQRADDAC(a[5], a[29]); SQRADDAC(a[6], a[28]); SQRADDAC(a[7], a[27]); SQRADDAC(a[8], a[26]); SQRADDAC(a[9], a[25]); SQRADDAC(a[10], a[24]); SQRADDAC(a[11], a[23]); SQRADDAC(a[12], a[22]); SQRADDAC(a[13], a[21]); SQRADDAC(a[14], a[20]); SQRADDAC(a[15], a[19]); SQRADDAC(a[16], a[18]); SQRADDDB; SQRADD(a[17], a[17]); + COMBA_STORE(b[34]); + + /* output 35 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[31]); SQRADDAC(a[5], a[30]); SQRADDAC(a[6], a[29]); SQRADDAC(a[7], a[28]); SQRADDAC(a[8], a[27]); SQRADDAC(a[9], a[26]); SQRADDAC(a[10], a[25]); SQRADDAC(a[11], a[24]); SQRADDAC(a[12], a[23]); SQRADDAC(a[13], a[22]); SQRADDAC(a[14], a[21]); SQRADDAC(a[15], a[20]); SQRADDAC(a[16], a[19]); SQRADDAC(a[17], a[18]); SQRADDDB; + COMBA_STORE(b[35]); + + /* output 36 */ + CARRY_FORWARD; + SQRADDSC(a[5], a[31]); SQRADDAC(a[6], a[30]); SQRADDAC(a[7], a[29]); SQRADDAC(a[8], a[28]); SQRADDAC(a[9], a[27]); SQRADDAC(a[10], a[26]); SQRADDAC(a[11], a[25]); SQRADDAC(a[12], a[24]); SQRADDAC(a[13], a[23]); SQRADDAC(a[14], a[22]); SQRADDAC(a[15], a[21]); SQRADDAC(a[16], a[20]); SQRADDAC(a[17], a[19]); SQRADDDB; SQRADD(a[18], a[18]); + COMBA_STORE(b[36]); + + /* output 37 */ + CARRY_FORWARD; + SQRADDSC(a[6], a[31]); SQRADDAC(a[7], a[30]); SQRADDAC(a[8], a[29]); SQRADDAC(a[9], a[28]); SQRADDAC(a[10], a[27]); SQRADDAC(a[11], a[26]); SQRADDAC(a[12], a[25]); SQRADDAC(a[13], a[24]); SQRADDAC(a[14], a[23]); SQRADDAC(a[15], a[22]); SQRADDAC(a[16], a[21]); SQRADDAC(a[17], a[20]); SQRADDAC(a[18], a[19]); SQRADDDB; + COMBA_STORE(b[37]); + + /* output 38 */ + CARRY_FORWARD; + SQRADDSC(a[7], a[31]); SQRADDAC(a[8], a[30]); SQRADDAC(a[9], a[29]); SQRADDAC(a[10], a[28]); SQRADDAC(a[11], a[27]); SQRADDAC(a[12], a[26]); SQRADDAC(a[13], a[25]); SQRADDAC(a[14], a[24]); SQRADDAC(a[15], a[23]); SQRADDAC(a[16], a[22]); SQRADDAC(a[17], a[21]); SQRADDAC(a[18], a[20]); SQRADDDB; SQRADD(a[19], a[19]); + COMBA_STORE(b[38]); + + /* output 39 */ + CARRY_FORWARD; + SQRADDSC(a[8], a[31]); SQRADDAC(a[9], a[30]); SQRADDAC(a[10], a[29]); SQRADDAC(a[11], a[28]); SQRADDAC(a[12], a[27]); SQRADDAC(a[13], a[26]); SQRADDAC(a[14], a[25]); SQRADDAC(a[15], a[24]); SQRADDAC(a[16], a[23]); SQRADDAC(a[17], a[22]); SQRADDAC(a[18], a[21]); SQRADDAC(a[19], a[20]); SQRADDDB; + COMBA_STORE(b[39]); + + /* output 40 */ + CARRY_FORWARD; + SQRADDSC(a[9], a[31]); SQRADDAC(a[10], a[30]); SQRADDAC(a[11], a[29]); SQRADDAC(a[12], a[28]); SQRADDAC(a[13], a[27]); SQRADDAC(a[14], a[26]); SQRADDAC(a[15], a[25]); SQRADDAC(a[16], a[24]); SQRADDAC(a[17], a[23]); SQRADDAC(a[18], a[22]); SQRADDAC(a[19], a[21]); SQRADDDB; SQRADD(a[20], a[20]); + COMBA_STORE(b[40]); + + /* output 41 */ + CARRY_FORWARD; + SQRADDSC(a[10], a[31]); SQRADDAC(a[11], a[30]); SQRADDAC(a[12], a[29]); SQRADDAC(a[13], a[28]); SQRADDAC(a[14], a[27]); SQRADDAC(a[15], a[26]); SQRADDAC(a[16], a[25]); SQRADDAC(a[17], a[24]); SQRADDAC(a[18], a[23]); SQRADDAC(a[19], a[22]); SQRADDAC(a[20], a[21]); SQRADDDB; + COMBA_STORE(b[41]); + + /* output 42 */ + CARRY_FORWARD; + SQRADDSC(a[11], a[31]); SQRADDAC(a[12], a[30]); SQRADDAC(a[13], a[29]); SQRADDAC(a[14], a[28]); SQRADDAC(a[15], a[27]); SQRADDAC(a[16], a[26]); SQRADDAC(a[17], a[25]); SQRADDAC(a[18], a[24]); SQRADDAC(a[19], a[23]); SQRADDAC(a[20], a[22]); SQRADDDB; SQRADD(a[21], a[21]); + COMBA_STORE(b[42]); + + /* output 43 */ + CARRY_FORWARD; + SQRADDSC(a[12], a[31]); SQRADDAC(a[13], a[30]); SQRADDAC(a[14], a[29]); SQRADDAC(a[15], a[28]); SQRADDAC(a[16], a[27]); SQRADDAC(a[17], a[26]); SQRADDAC(a[18], a[25]); SQRADDAC(a[19], a[24]); SQRADDAC(a[20], a[23]); SQRADDAC(a[21], a[22]); SQRADDDB; + COMBA_STORE(b[43]); + + /* output 44 */ + CARRY_FORWARD; + SQRADDSC(a[13], a[31]); SQRADDAC(a[14], a[30]); SQRADDAC(a[15], a[29]); SQRADDAC(a[16], a[28]); SQRADDAC(a[17], a[27]); SQRADDAC(a[18], a[26]); SQRADDAC(a[19], a[25]); SQRADDAC(a[20], a[24]); SQRADDAC(a[21], a[23]); SQRADDDB; SQRADD(a[22], a[22]); + COMBA_STORE(b[44]); + + /* output 45 */ + CARRY_FORWARD; + SQRADDSC(a[14], a[31]); SQRADDAC(a[15], a[30]); SQRADDAC(a[16], a[29]); SQRADDAC(a[17], a[28]); SQRADDAC(a[18], a[27]); SQRADDAC(a[19], a[26]); SQRADDAC(a[20], a[25]); SQRADDAC(a[21], a[24]); SQRADDAC(a[22], a[23]); SQRADDDB; + COMBA_STORE(b[45]); + + /* output 46 */ + CARRY_FORWARD; + SQRADDSC(a[15], a[31]); SQRADDAC(a[16], a[30]); SQRADDAC(a[17], a[29]); SQRADDAC(a[18], a[28]); SQRADDAC(a[19], a[27]); SQRADDAC(a[20], a[26]); SQRADDAC(a[21], a[25]); SQRADDAC(a[22], a[24]); SQRADDDB; SQRADD(a[23], a[23]); + COMBA_STORE(b[46]); + + /* output 47 */ + CARRY_FORWARD; + SQRADDSC(a[16], a[31]); SQRADDAC(a[17], a[30]); SQRADDAC(a[18], a[29]); SQRADDAC(a[19], a[28]); SQRADDAC(a[20], a[27]); SQRADDAC(a[21], a[26]); SQRADDAC(a[22], a[25]); SQRADDAC(a[23], a[24]); SQRADDDB; + COMBA_STORE(b[47]); + + /* output 48 */ + CARRY_FORWARD; + SQRADDSC(a[17], a[31]); SQRADDAC(a[18], a[30]); SQRADDAC(a[19], a[29]); SQRADDAC(a[20], a[28]); SQRADDAC(a[21], a[27]); SQRADDAC(a[22], a[26]); SQRADDAC(a[23], a[25]); SQRADDDB; SQRADD(a[24], a[24]); + COMBA_STORE(b[48]); + + /* output 49 */ + CARRY_FORWARD; + SQRADDSC(a[18], a[31]); SQRADDAC(a[19], a[30]); SQRADDAC(a[20], a[29]); SQRADDAC(a[21], a[28]); SQRADDAC(a[22], a[27]); SQRADDAC(a[23], a[26]); SQRADDAC(a[24], a[25]); SQRADDDB; + COMBA_STORE(b[49]); + + /* output 50 */ + CARRY_FORWARD; + SQRADDSC(a[19], a[31]); SQRADDAC(a[20], a[30]); SQRADDAC(a[21], a[29]); SQRADDAC(a[22], a[28]); SQRADDAC(a[23], a[27]); SQRADDAC(a[24], a[26]); SQRADDDB; SQRADD(a[25], a[25]); + COMBA_STORE(b[50]); + + /* output 51 */ + CARRY_FORWARD; + SQRADDSC(a[20], a[31]); SQRADDAC(a[21], a[30]); SQRADDAC(a[22], a[29]); SQRADDAC(a[23], a[28]); SQRADDAC(a[24], a[27]); SQRADDAC(a[25], a[26]); SQRADDDB; + COMBA_STORE(b[51]); + + /* output 52 */ + CARRY_FORWARD; + SQRADDSC(a[21], a[31]); SQRADDAC(a[22], a[30]); SQRADDAC(a[23], a[29]); SQRADDAC(a[24], a[28]); SQRADDAC(a[25], a[27]); SQRADDDB; SQRADD(a[26], a[26]); + COMBA_STORE(b[52]); + + /* output 53 */ + CARRY_FORWARD; + SQRADDSC(a[22], a[31]); SQRADDAC(a[23], a[30]); SQRADDAC(a[24], a[29]); SQRADDAC(a[25], a[28]); SQRADDAC(a[26], a[27]); SQRADDDB; + COMBA_STORE(b[53]); + + /* output 54 */ + CARRY_FORWARD; + SQRADDSC(a[23], a[31]); SQRADDAC(a[24], a[30]); SQRADDAC(a[25], a[29]); SQRADDAC(a[26], a[28]); SQRADDDB; SQRADD(a[27], a[27]); + COMBA_STORE(b[54]); + + /* output 55 */ + CARRY_FORWARD; + SQRADDSC(a[24], a[31]); SQRADDAC(a[25], a[30]); SQRADDAC(a[26], a[29]); SQRADDAC(a[27], a[28]); SQRADDDB; + COMBA_STORE(b[55]); + + /* output 56 */ + CARRY_FORWARD; + SQRADDSC(a[25], a[31]); SQRADDAC(a[26], a[30]); SQRADDAC(a[27], a[29]); SQRADDDB; SQRADD(a[28], a[28]); + COMBA_STORE(b[56]); + + /* output 57 */ + CARRY_FORWARD; + SQRADDSC(a[26], a[31]); SQRADDAC(a[27], a[30]); SQRADDAC(a[28], a[29]); SQRADDDB; + COMBA_STORE(b[57]); + + /* output 58 */ + CARRY_FORWARD; + SQRADD2(a[27], a[31]); SQRADD2(a[28], a[30]); SQRADD(a[29], a[29]); + COMBA_STORE(b[58]); + + /* output 59 */ + CARRY_FORWARD; + SQRADD2(a[28], a[31]); SQRADD2(a[29], a[30]); + COMBA_STORE(b[59]); + + /* output 60 */ + CARRY_FORWARD; + SQRADD2(a[29], a[31]); SQRADD(a[30], a[30]); + COMBA_STORE(b[60]); + + /* output 61 */ + CARRY_FORWARD; + SQRADD2(a[30], a[31]); + COMBA_STORE(b[61]); + + /* output 62 */ + CARRY_FORWARD; + SQRADD(a[31], a[31]); + COMBA_STORE(b[62]); + COMBA_STORE2(b[63]); + COMBA_FINI; + + B->used = 64; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 64 * sizeof(fp_digit)); + fp_clamp(B); +} +#endif + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_sqr_comba_generic.c b/src/crypto/tfm/fp_sqr_comba_generic.c new file mode 100644 index 0000000..168bd9d --- /dev/null +++ b/src/crypto/tfm/fp_sqr_comba_generic.c @@ -0,0 +1,98 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ + +#define TFM_DEFINES +#include "fp_sqr_comba.c" + +/* generic comba squarer */ +void fp_sqr_comba(fp_int *A, fp_int *B) +{ + int pa, ix, iz; + fp_digit c0, c1, c2; + fp_int tmp, *dst; +#ifdef TFM_ISO + fp_word tt; +#endif + + /* get size of output and trim */ + pa = A->used + A->used; + if (pa >= FP_SIZE) { + pa = FP_SIZE-1; + } + + /* number of output digits to produce */ + COMBA_START; + CLEAR_CARRY; + + if (A == B) { + fp_zero(&tmp); + dst = &tmp; + } else { + fp_zero(B); + dst = B; + } + + for (ix = 0; ix < pa; ix++) { + int tx, ty, iy; + fp_digit *tmpy, *tmpx; + + /* get offsets into the two bignums */ + ty = MIN(A->used-1, ix); + tx = ix - ty; + + /* setup temp aliases */ + tmpx = A->dp + tx; + tmpy = A->dp + ty; + + /* this is the number of times the loop will iterrate, + while (tx++ < a->used && ty-- >= 0) { ... } + */ + iy = MIN(A->used-tx, ty+1); + + /* now for squaring tx can never equal ty + * we halve the distance since they approach + * at a rate of 2x and we have to round because + * odd cases need to be executed + */ + iy = MIN(iy, (ty-tx+1)>>1); + + /* forward carries */ + CARRY_FORWARD; + + /* execute loop */ + for (iz = 0; iz < iy; iz++) { + fp_digit _tmpx = *tmpx++; + fp_digit _tmpy = *tmpy--; + SQRADD2(_tmpx, _tmpy); + } + + /* even columns have the square term in them */ + if ((ix&1) == 0) { + fp_digit _a_dp = A->dp[ix>>1]; + SQRADD(_a_dp, A->dp[ix>>1]); + } + + /* store it */ + COMBA_STORE(dst->dp[ix]); + } + + COMBA_FINI; + + /* setup dest */ + dst->used = pa; + fp_clamp (dst); + if (dst != B) { + fp_copy(dst, B); + } +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_sqr_comba_small_set.c b/src/crypto/tfm/fp_sqr_comba_small_set.c new file mode 100644 index 0000000..86ba037 --- /dev/null +++ b/src/crypto/tfm/fp_sqr_comba_small_set.c @@ -0,0 +1,1523 @@ +#define TFM_DEFINES +#include "fp_sqr_comba.c" + +#if defined(TFM_SMALL_SET) +void fp_sqr_comba_small(fp_int *A, fp_int *B) +{ + fp_digit *a, b[32], c0, c1, c2, sc0, sc1, sc2; +#ifdef TFM_ISO + fp_word tt; +#endif + switch (A->used) { + case 1: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + COMBA_STORE2(b[1]); + COMBA_FINI; + + B->used = 2; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 2 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 2: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + COMBA_STORE2(b[3]); + COMBA_FINI; + + B->used = 4; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 4 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 3: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + COMBA_STORE2(b[5]); + COMBA_FINI; + + B->used = 6; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 6 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 4: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADD2(a[2], a[3]); + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + COMBA_STORE2(b[7]); + COMBA_FINI; + + B->used = 8; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 8 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 5: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADD2(a[1], a[4]); SQRADD2(a[2], a[3]); + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADD2(a[2], a[4]); SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADD2(a[3], a[4]); + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + COMBA_STORE2(b[9]); + COMBA_FINI; + + B->used = 10; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 10 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 6: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADD2(a[1], a[5]); SQRADD2(a[2], a[4]); SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADD2(a[2], a[5]); SQRADD2(a[3], a[4]); + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADD2(a[3], a[5]); SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADD2(a[4], a[5]); + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + COMBA_STORE2(b[11]); + COMBA_FINI; + + B->used = 12; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 12 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 7: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADD2(a[2], a[6]); SQRADD2(a[3], a[5]); SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADD2(a[3], a[6]); SQRADD2(a[4], a[5]); + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADD2(a[4], a[6]); SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADD2(a[5], a[6]); + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + COMBA_STORE2(b[13]); + COMBA_FINI; + + B->used = 14; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 14 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 8: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADD2(a[3], a[7]); SQRADD2(a[4], a[6]); SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADD2(a[4], a[7]); SQRADD2(a[5], a[6]); + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADD2(a[5], a[7]); SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADD2(a[6], a[7]); + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + COMBA_STORE2(b[15]); + COMBA_FINI; + + B->used = 16; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 16 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 9: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADD2(a[4], a[8]); SQRADD2(a[5], a[7]); SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADD2(a[5], a[8]); SQRADD2(a[6], a[7]); + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADD2(a[6], a[8]); SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADD2(a[7], a[8]); + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + COMBA_STORE2(b[17]); + COMBA_FINI; + + B->used = 18; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 18 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 10: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADD2(a[5], a[9]); SQRADD2(a[6], a[8]); SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADD2(a[6], a[9]); SQRADD2(a[7], a[8]); + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADD2(a[7], a[9]); SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADD2(a[8], a[9]); + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + COMBA_STORE2(b[19]); + COMBA_FINI; + + B->used = 20; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 20 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 11: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[10]); SQRADDAC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[10]); SQRADDAC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[10]); SQRADDAC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[10]); SQRADDAC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[10]); SQRADDAC(a[5], a[9]); SQRADDAC(a[6], a[8]); SQRADDDB; SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADDSC(a[5], a[10]); SQRADDAC(a[6], a[9]); SQRADDAC(a[7], a[8]); SQRADDDB; + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADD2(a[6], a[10]); SQRADD2(a[7], a[9]); SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADD2(a[7], a[10]); SQRADD2(a[8], a[9]); + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADD2(a[8], a[10]); SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + + /* output 19 */ + CARRY_FORWARD; + SQRADD2(a[9], a[10]); + COMBA_STORE(b[19]); + + /* output 20 */ + CARRY_FORWARD; + SQRADD(a[10], a[10]); + COMBA_STORE(b[20]); + COMBA_STORE2(b[21]); + COMBA_FINI; + + B->used = 22; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 22 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 12: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[10]); SQRADDAC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[11]); SQRADDAC(a[1], a[10]); SQRADDAC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[11]); SQRADDAC(a[2], a[10]); SQRADDAC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[11]); SQRADDAC(a[3], a[10]); SQRADDAC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[11]); SQRADDAC(a[4], a[10]); SQRADDAC(a[5], a[9]); SQRADDAC(a[6], a[8]); SQRADDDB; SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[11]); SQRADDAC(a[5], a[10]); SQRADDAC(a[6], a[9]); SQRADDAC(a[7], a[8]); SQRADDDB; + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADDSC(a[5], a[11]); SQRADDAC(a[6], a[10]); SQRADDAC(a[7], a[9]); SQRADDDB; SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADDSC(a[6], a[11]); SQRADDAC(a[7], a[10]); SQRADDAC(a[8], a[9]); SQRADDDB; + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADD2(a[7], a[11]); SQRADD2(a[8], a[10]); SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + + /* output 19 */ + CARRY_FORWARD; + SQRADD2(a[8], a[11]); SQRADD2(a[9], a[10]); + COMBA_STORE(b[19]); + + /* output 20 */ + CARRY_FORWARD; + SQRADD2(a[9], a[11]); SQRADD(a[10], a[10]); + COMBA_STORE(b[20]); + + /* output 21 */ + CARRY_FORWARD; + SQRADD2(a[10], a[11]); + COMBA_STORE(b[21]); + + /* output 22 */ + CARRY_FORWARD; + SQRADD(a[11], a[11]); + COMBA_STORE(b[22]); + COMBA_STORE2(b[23]); + COMBA_FINI; + + B->used = 24; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 24 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 13: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[10]); SQRADDAC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[11]); SQRADDAC(a[1], a[10]); SQRADDAC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[12]); SQRADDAC(a[1], a[11]); SQRADDAC(a[2], a[10]); SQRADDAC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[12]); SQRADDAC(a[2], a[11]); SQRADDAC(a[3], a[10]); SQRADDAC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[12]); SQRADDAC(a[3], a[11]); SQRADDAC(a[4], a[10]); SQRADDAC(a[5], a[9]); SQRADDAC(a[6], a[8]); SQRADDDB; SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[12]); SQRADDAC(a[4], a[11]); SQRADDAC(a[5], a[10]); SQRADDAC(a[6], a[9]); SQRADDAC(a[7], a[8]); SQRADDDB; + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[12]); SQRADDAC(a[5], a[11]); SQRADDAC(a[6], a[10]); SQRADDAC(a[7], a[9]); SQRADDDB; SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADDSC(a[5], a[12]); SQRADDAC(a[6], a[11]); SQRADDAC(a[7], a[10]); SQRADDAC(a[8], a[9]); SQRADDDB; + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADDSC(a[6], a[12]); SQRADDAC(a[7], a[11]); SQRADDAC(a[8], a[10]); SQRADDDB; SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + + /* output 19 */ + CARRY_FORWARD; + SQRADDSC(a[7], a[12]); SQRADDAC(a[8], a[11]); SQRADDAC(a[9], a[10]); SQRADDDB; + COMBA_STORE(b[19]); + + /* output 20 */ + CARRY_FORWARD; + SQRADD2(a[8], a[12]); SQRADD2(a[9], a[11]); SQRADD(a[10], a[10]); + COMBA_STORE(b[20]); + + /* output 21 */ + CARRY_FORWARD; + SQRADD2(a[9], a[12]); SQRADD2(a[10], a[11]); + COMBA_STORE(b[21]); + + /* output 22 */ + CARRY_FORWARD; + SQRADD2(a[10], a[12]); SQRADD(a[11], a[11]); + COMBA_STORE(b[22]); + + /* output 23 */ + CARRY_FORWARD; + SQRADD2(a[11], a[12]); + COMBA_STORE(b[23]); + + /* output 24 */ + CARRY_FORWARD; + SQRADD(a[12], a[12]); + COMBA_STORE(b[24]); + COMBA_STORE2(b[25]); + COMBA_FINI; + + B->used = 26; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 26 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 14: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[10]); SQRADDAC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[11]); SQRADDAC(a[1], a[10]); SQRADDAC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[12]); SQRADDAC(a[1], a[11]); SQRADDAC(a[2], a[10]); SQRADDAC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[13]); SQRADDAC(a[1], a[12]); SQRADDAC(a[2], a[11]); SQRADDAC(a[3], a[10]); SQRADDAC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[13]); SQRADDAC(a[2], a[12]); SQRADDAC(a[3], a[11]); SQRADDAC(a[4], a[10]); SQRADDAC(a[5], a[9]); SQRADDAC(a[6], a[8]); SQRADDDB; SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[13]); SQRADDAC(a[3], a[12]); SQRADDAC(a[4], a[11]); SQRADDAC(a[5], a[10]); SQRADDAC(a[6], a[9]); SQRADDAC(a[7], a[8]); SQRADDDB; + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[13]); SQRADDAC(a[4], a[12]); SQRADDAC(a[5], a[11]); SQRADDAC(a[6], a[10]); SQRADDAC(a[7], a[9]); SQRADDDB; SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[13]); SQRADDAC(a[5], a[12]); SQRADDAC(a[6], a[11]); SQRADDAC(a[7], a[10]); SQRADDAC(a[8], a[9]); SQRADDDB; + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADDSC(a[5], a[13]); SQRADDAC(a[6], a[12]); SQRADDAC(a[7], a[11]); SQRADDAC(a[8], a[10]); SQRADDDB; SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + + /* output 19 */ + CARRY_FORWARD; + SQRADDSC(a[6], a[13]); SQRADDAC(a[7], a[12]); SQRADDAC(a[8], a[11]); SQRADDAC(a[9], a[10]); SQRADDDB; + COMBA_STORE(b[19]); + + /* output 20 */ + CARRY_FORWARD; + SQRADDSC(a[7], a[13]); SQRADDAC(a[8], a[12]); SQRADDAC(a[9], a[11]); SQRADDDB; SQRADD(a[10], a[10]); + COMBA_STORE(b[20]); + + /* output 21 */ + CARRY_FORWARD; + SQRADDSC(a[8], a[13]); SQRADDAC(a[9], a[12]); SQRADDAC(a[10], a[11]); SQRADDDB; + COMBA_STORE(b[21]); + + /* output 22 */ + CARRY_FORWARD; + SQRADD2(a[9], a[13]); SQRADD2(a[10], a[12]); SQRADD(a[11], a[11]); + COMBA_STORE(b[22]); + + /* output 23 */ + CARRY_FORWARD; + SQRADD2(a[10], a[13]); SQRADD2(a[11], a[12]); + COMBA_STORE(b[23]); + + /* output 24 */ + CARRY_FORWARD; + SQRADD2(a[11], a[13]); SQRADD(a[12], a[12]); + COMBA_STORE(b[24]); + + /* output 25 */ + CARRY_FORWARD; + SQRADD2(a[12], a[13]); + COMBA_STORE(b[25]); + + /* output 26 */ + CARRY_FORWARD; + SQRADD(a[13], a[13]); + COMBA_STORE(b[26]); + COMBA_STORE2(b[27]); + COMBA_FINI; + + B->used = 28; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 28 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 15: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[10]); SQRADDAC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[11]); SQRADDAC(a[1], a[10]); SQRADDAC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[12]); SQRADDAC(a[1], a[11]); SQRADDAC(a[2], a[10]); SQRADDAC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[13]); SQRADDAC(a[1], a[12]); SQRADDAC(a[2], a[11]); SQRADDAC(a[3], a[10]); SQRADDAC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[14]); SQRADDAC(a[1], a[13]); SQRADDAC(a[2], a[12]); SQRADDAC(a[3], a[11]); SQRADDAC(a[4], a[10]); SQRADDAC(a[5], a[9]); SQRADDAC(a[6], a[8]); SQRADDDB; SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[14]); SQRADDAC(a[2], a[13]); SQRADDAC(a[3], a[12]); SQRADDAC(a[4], a[11]); SQRADDAC(a[5], a[10]); SQRADDAC(a[6], a[9]); SQRADDAC(a[7], a[8]); SQRADDDB; + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[14]); SQRADDAC(a[3], a[13]); SQRADDAC(a[4], a[12]); SQRADDAC(a[5], a[11]); SQRADDAC(a[6], a[10]); SQRADDAC(a[7], a[9]); SQRADDDB; SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[14]); SQRADDAC(a[4], a[13]); SQRADDAC(a[5], a[12]); SQRADDAC(a[6], a[11]); SQRADDAC(a[7], a[10]); SQRADDAC(a[8], a[9]); SQRADDDB; + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[14]); SQRADDAC(a[5], a[13]); SQRADDAC(a[6], a[12]); SQRADDAC(a[7], a[11]); SQRADDAC(a[8], a[10]); SQRADDDB; SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + + /* output 19 */ + CARRY_FORWARD; + SQRADDSC(a[5], a[14]); SQRADDAC(a[6], a[13]); SQRADDAC(a[7], a[12]); SQRADDAC(a[8], a[11]); SQRADDAC(a[9], a[10]); SQRADDDB; + COMBA_STORE(b[19]); + + /* output 20 */ + CARRY_FORWARD; + SQRADDSC(a[6], a[14]); SQRADDAC(a[7], a[13]); SQRADDAC(a[8], a[12]); SQRADDAC(a[9], a[11]); SQRADDDB; SQRADD(a[10], a[10]); + COMBA_STORE(b[20]); + + /* output 21 */ + CARRY_FORWARD; + SQRADDSC(a[7], a[14]); SQRADDAC(a[8], a[13]); SQRADDAC(a[9], a[12]); SQRADDAC(a[10], a[11]); SQRADDDB; + COMBA_STORE(b[21]); + + /* output 22 */ + CARRY_FORWARD; + SQRADDSC(a[8], a[14]); SQRADDAC(a[9], a[13]); SQRADDAC(a[10], a[12]); SQRADDDB; SQRADD(a[11], a[11]); + COMBA_STORE(b[22]); + + /* output 23 */ + CARRY_FORWARD; + SQRADDSC(a[9], a[14]); SQRADDAC(a[10], a[13]); SQRADDAC(a[11], a[12]); SQRADDDB; + COMBA_STORE(b[23]); + + /* output 24 */ + CARRY_FORWARD; + SQRADD2(a[10], a[14]); SQRADD2(a[11], a[13]); SQRADD(a[12], a[12]); + COMBA_STORE(b[24]); + + /* output 25 */ + CARRY_FORWARD; + SQRADD2(a[11], a[14]); SQRADD2(a[12], a[13]); + COMBA_STORE(b[25]); + + /* output 26 */ + CARRY_FORWARD; + SQRADD2(a[12], a[14]); SQRADD(a[13], a[13]); + COMBA_STORE(b[26]); + + /* output 27 */ + CARRY_FORWARD; + SQRADD2(a[13], a[14]); + COMBA_STORE(b[27]); + + /* output 28 */ + CARRY_FORWARD; + SQRADD(a[14], a[14]); + COMBA_STORE(b[28]); + COMBA_STORE2(b[29]); + COMBA_FINI; + + B->used = 30; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 30 * sizeof(fp_digit)); + fp_clamp(B); + break; + + case 16: + a = A->dp; + COMBA_START; + + /* clear carries */ + CLEAR_CARRY; + + /* output 0 */ + SQRADD(a[0],a[0]); + COMBA_STORE(b[0]); + + /* output 1 */ + CARRY_FORWARD; + SQRADD2(a[0], a[1]); + COMBA_STORE(b[1]); + + /* output 2 */ + CARRY_FORWARD; + SQRADD2(a[0], a[2]); SQRADD(a[1], a[1]); + COMBA_STORE(b[2]); + + /* output 3 */ + CARRY_FORWARD; + SQRADD2(a[0], a[3]); SQRADD2(a[1], a[2]); + COMBA_STORE(b[3]); + + /* output 4 */ + CARRY_FORWARD; + SQRADD2(a[0], a[4]); SQRADD2(a[1], a[3]); SQRADD(a[2], a[2]); + COMBA_STORE(b[4]); + + /* output 5 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[5]); SQRADDAC(a[1], a[4]); SQRADDAC(a[2], a[3]); SQRADDDB; + COMBA_STORE(b[5]); + + /* output 6 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[6]); SQRADDAC(a[1], a[5]); SQRADDAC(a[2], a[4]); SQRADDDB; SQRADD(a[3], a[3]); + COMBA_STORE(b[6]); + + /* output 7 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[7]); SQRADDAC(a[1], a[6]); SQRADDAC(a[2], a[5]); SQRADDAC(a[3], a[4]); SQRADDDB; + COMBA_STORE(b[7]); + + /* output 8 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[8]); SQRADDAC(a[1], a[7]); SQRADDAC(a[2], a[6]); SQRADDAC(a[3], a[5]); SQRADDDB; SQRADD(a[4], a[4]); + COMBA_STORE(b[8]); + + /* output 9 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[9]); SQRADDAC(a[1], a[8]); SQRADDAC(a[2], a[7]); SQRADDAC(a[3], a[6]); SQRADDAC(a[4], a[5]); SQRADDDB; + COMBA_STORE(b[9]); + + /* output 10 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[10]); SQRADDAC(a[1], a[9]); SQRADDAC(a[2], a[8]); SQRADDAC(a[3], a[7]); SQRADDAC(a[4], a[6]); SQRADDDB; SQRADD(a[5], a[5]); + COMBA_STORE(b[10]); + + /* output 11 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[11]); SQRADDAC(a[1], a[10]); SQRADDAC(a[2], a[9]); SQRADDAC(a[3], a[8]); SQRADDAC(a[4], a[7]); SQRADDAC(a[5], a[6]); SQRADDDB; + COMBA_STORE(b[11]); + + /* output 12 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[12]); SQRADDAC(a[1], a[11]); SQRADDAC(a[2], a[10]); SQRADDAC(a[3], a[9]); SQRADDAC(a[4], a[8]); SQRADDAC(a[5], a[7]); SQRADDDB; SQRADD(a[6], a[6]); + COMBA_STORE(b[12]); + + /* output 13 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[13]); SQRADDAC(a[1], a[12]); SQRADDAC(a[2], a[11]); SQRADDAC(a[3], a[10]); SQRADDAC(a[4], a[9]); SQRADDAC(a[5], a[8]); SQRADDAC(a[6], a[7]); SQRADDDB; + COMBA_STORE(b[13]); + + /* output 14 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[14]); SQRADDAC(a[1], a[13]); SQRADDAC(a[2], a[12]); SQRADDAC(a[3], a[11]); SQRADDAC(a[4], a[10]); SQRADDAC(a[5], a[9]); SQRADDAC(a[6], a[8]); SQRADDDB; SQRADD(a[7], a[7]); + COMBA_STORE(b[14]); + + /* output 15 */ + CARRY_FORWARD; + SQRADDSC(a[0], a[15]); SQRADDAC(a[1], a[14]); SQRADDAC(a[2], a[13]); SQRADDAC(a[3], a[12]); SQRADDAC(a[4], a[11]); SQRADDAC(a[5], a[10]); SQRADDAC(a[6], a[9]); SQRADDAC(a[7], a[8]); SQRADDDB; + COMBA_STORE(b[15]); + + /* output 16 */ + CARRY_FORWARD; + SQRADDSC(a[1], a[15]); SQRADDAC(a[2], a[14]); SQRADDAC(a[3], a[13]); SQRADDAC(a[4], a[12]); SQRADDAC(a[5], a[11]); SQRADDAC(a[6], a[10]); SQRADDAC(a[7], a[9]); SQRADDDB; SQRADD(a[8], a[8]); + COMBA_STORE(b[16]); + + /* output 17 */ + CARRY_FORWARD; + SQRADDSC(a[2], a[15]); SQRADDAC(a[3], a[14]); SQRADDAC(a[4], a[13]); SQRADDAC(a[5], a[12]); SQRADDAC(a[6], a[11]); SQRADDAC(a[7], a[10]); SQRADDAC(a[8], a[9]); SQRADDDB; + COMBA_STORE(b[17]); + + /* output 18 */ + CARRY_FORWARD; + SQRADDSC(a[3], a[15]); SQRADDAC(a[4], a[14]); SQRADDAC(a[5], a[13]); SQRADDAC(a[6], a[12]); SQRADDAC(a[7], a[11]); SQRADDAC(a[8], a[10]); SQRADDDB; SQRADD(a[9], a[9]); + COMBA_STORE(b[18]); + + /* output 19 */ + CARRY_FORWARD; + SQRADDSC(a[4], a[15]); SQRADDAC(a[5], a[14]); SQRADDAC(a[6], a[13]); SQRADDAC(a[7], a[12]); SQRADDAC(a[8], a[11]); SQRADDAC(a[9], a[10]); SQRADDDB; + COMBA_STORE(b[19]); + + /* output 20 */ + CARRY_FORWARD; + SQRADDSC(a[5], a[15]); SQRADDAC(a[6], a[14]); SQRADDAC(a[7], a[13]); SQRADDAC(a[8], a[12]); SQRADDAC(a[9], a[11]); SQRADDDB; SQRADD(a[10], a[10]); + COMBA_STORE(b[20]); + + /* output 21 */ + CARRY_FORWARD; + SQRADDSC(a[6], a[15]); SQRADDAC(a[7], a[14]); SQRADDAC(a[8], a[13]); SQRADDAC(a[9], a[12]); SQRADDAC(a[10], a[11]); SQRADDDB; + COMBA_STORE(b[21]); + + /* output 22 */ + CARRY_FORWARD; + SQRADDSC(a[7], a[15]); SQRADDAC(a[8], a[14]); SQRADDAC(a[9], a[13]); SQRADDAC(a[10], a[12]); SQRADDDB; SQRADD(a[11], a[11]); + COMBA_STORE(b[22]); + + /* output 23 */ + CARRY_FORWARD; + SQRADDSC(a[8], a[15]); SQRADDAC(a[9], a[14]); SQRADDAC(a[10], a[13]); SQRADDAC(a[11], a[12]); SQRADDDB; + COMBA_STORE(b[23]); + + /* output 24 */ + CARRY_FORWARD; + SQRADDSC(a[9], a[15]); SQRADDAC(a[10], a[14]); SQRADDAC(a[11], a[13]); SQRADDDB; SQRADD(a[12], a[12]); + COMBA_STORE(b[24]); + + /* output 25 */ + CARRY_FORWARD; + SQRADDSC(a[10], a[15]); SQRADDAC(a[11], a[14]); SQRADDAC(a[12], a[13]); SQRADDDB; + COMBA_STORE(b[25]); + + /* output 26 */ + CARRY_FORWARD; + SQRADD2(a[11], a[15]); SQRADD2(a[12], a[14]); SQRADD(a[13], a[13]); + COMBA_STORE(b[26]); + + /* output 27 */ + CARRY_FORWARD; + SQRADD2(a[12], a[15]); SQRADD2(a[13], a[14]); + COMBA_STORE(b[27]); + + /* output 28 */ + CARRY_FORWARD; + SQRADD2(a[13], a[15]); SQRADD(a[14], a[14]); + COMBA_STORE(b[28]); + + /* output 29 */ + CARRY_FORWARD; + SQRADD2(a[14], a[15]); + COMBA_STORE(b[29]); + + /* output 30 */ + CARRY_FORWARD; + SQRADD(a[15], a[15]); + COMBA_STORE(b[30]); + COMBA_STORE2(b[31]); + COMBA_FINI; + + B->used = 32; + B->sign = FP_ZPOS; + memcpy(B->dp, b, 32 * sizeof(fp_digit)); + fp_clamp(B); + break; + +} +} + +#endif /* TFM_SMALL_SET */ + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_sub.c b/src/crypto/tfm/fp_sub.c new file mode 100644 index 0000000..c482a6a --- /dev/null +++ b/src/crypto/tfm/fp_sub.c @@ -0,0 +1,50 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* c = a - b */ +void fp_sub(fp_int *a, fp_int *b, fp_int *c) +{ + int sa, sb; + + sa = a->sign; + sb = b->sign; + + if (sa != sb) { + /* subtract a negative from a positive, OR */ + /* subtract a positive from a negative. */ + /* In either case, ADD their magnitudes, */ + /* and use the sign of the first number. */ + c->sign = sa; + s_fp_add (a, b, c); + } else { + /* subtract a positive from a positive, OR */ + /* subtract a negative from a negative. */ + /* First, take the difference between their */ + /* magnitudes, then... */ + if (fp_cmp_mag (a, b) != FP_LT) { + /* Copy the sign from the first */ + c->sign = sa; + /* The first has a larger or equal magnitude */ + s_fp_sub (a, b, c); + } else { + /* The result has the *opposite* sign from */ + /* the first number. */ + c->sign = (sa == FP_ZPOS) ? FP_NEG : FP_ZPOS; + /* The second has a larger magnitude */ + s_fp_sub (b, a, c); + } + } +} + + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_to_unsigned_bin.c b/src/crypto/tfm/fp_to_unsigned_bin.c new file mode 100644 index 0000000..f0eeb56 --- /dev/null +++ b/src/crypto/tfm/fp_to_unsigned_bin.c @@ -0,0 +1,29 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +void fp_to_unsigned_bin(fp_int *a, unsigned char *b) +{ + int x; + fp_int t; + + fp_init_copy(&t, a); + + x = 0; + while (fp_iszero (&t) == FP_NO) { + b[x++] = (unsigned char) (t.dp[0] & 255); + fp_div_2d (&t, 8, &t, NULL); + } + fp_reverse (b, x); +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/fp_unsigned_bin_size.c b/src/crypto/tfm/fp_unsigned_bin_size.c new file mode 100644 index 0000000..120a1a7 --- /dev/null +++ b/src/crypto/tfm/fp_unsigned_bin_size.c @@ -0,0 +1,20 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +int fp_unsigned_bin_size(fp_int *a) +{ + int size = fp_count_bits (a); + return (size / 8 + ((size & 7) != 0 ? 1 : 0)); +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/s_fp_add.c b/src/crypto/tfm/s_fp_add.c new file mode 100644 index 0000000..0633242 --- /dev/null +++ b/src/crypto/tfm/s_fp_add.c @@ -0,0 +1,42 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* unsigned addition */ +void s_fp_add(fp_int *a, fp_int *b, fp_int *c) +{ + int x, y, oldused; + register fp_word t; + + y = MAX(a->used, b->used); + oldused = MIN(c->used, FP_SIZE); + c->used = y; + + t = 0; + for (x = 0; x < y; x++) { + t += ((fp_word)a->dp[x]) + ((fp_word)b->dp[x]); + c->dp[x] = (fp_digit)t; + t >>= DIGIT_BIT; + } + if (t != 0 && x < FP_SIZE) { + c->dp[c->used++] = (fp_digit)t; + ++x; + } + + c->used = x; + for (; x < oldused; x++) { + c->dp[x] = 0; + } + fp_clamp(c); +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/s_fp_sub.c b/src/crypto/tfm/s_fp_sub.c new file mode 100644 index 0000000..dc03f0a --- /dev/null +++ b/src/crypto/tfm/s_fp_sub.c @@ -0,0 +1,40 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#include + +/* unsigned subtraction ||a|| >= ||b|| ALWAYS! */ +void s_fp_sub(fp_int *a, fp_int *b, fp_int *c) +{ + int x, oldbused, oldused; + fp_word t; + + oldused = c->used; + oldbused = b->used; + c->used = a->used; + t = 0; + for (x = 0; x < oldbused; x++) { + t = ((fp_word)a->dp[x]) - (((fp_word)b->dp[x]) + t); + c->dp[x] = (fp_digit)t; + t = (t >> DIGIT_BIT)&1; + } + for (; x < a->used; x++) { + t = ((fp_word)a->dp[x]) - t; + c->dp[x] = (fp_digit)t; + t = (t >> DIGIT_BIT)&1; + } + for (; x < oldused; x++) { + c->dp[x] = 0; + } + fp_clamp(c); +} + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/tfm.h b/src/crypto/tfm/tfm.h new file mode 100644 index 0000000..0d1dbf1 --- /dev/null +++ b/src/crypto/tfm/tfm.h @@ -0,0 +1,498 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#ifndef TFM_H_ +#define TFM_H_ + +#include +#include +#include +#include +#include + +/* 0xMaMiPaDe + * Major + * Minor + * Patch + * Development - 00=release, 01=in-development + */ +#define TFM_VERSION 0x000D0101 +#define TFM_VERSION_S "v0.13.1-next" + +#ifndef MIN + #define MIN(x,y) ((x)<(y)?(x):(y)) +#endif + +#ifndef MAX + #define MAX(x,y) ((x)>(y)?(x):(y)) +#endif + +/* externally define this symbol to ignore the default settings, useful for changing the build from the make process */ +#ifndef TFM_ALREADY_SET + +/* do we want the large set of small multiplications ? + Enable these if you are going to be doing a lot of small (<= 16 digit) multiplications say in ECC + Or if you're on a 64-bit machine doing RSA as a 1024-bit integer == 16 digits ;-) + */ +#define TFM_SMALL_SET + +/* do we want huge code + Enable these if you are doing 20, 24, 28, 32, 48, 64 digit multiplications (useful for RSA) + Less important on 64-bit machines as 32 digits == 2048 bits + */ +#if 0 +#define TFM_MUL3 +#define TFM_MUL4 +#define TFM_MUL6 +#define TFM_MUL7 +#define TFM_MUL8 +#define TFM_MUL9 +#define TFM_MUL12 +#define TFM_MUL17 +#endif +#define TFM_MUL20 +#define TFM_MUL24 +#define TFM_MUL28 +#define TFM_MUL32 +#define TFM_MUL48 +#define TFM_MUL64 + +#if 0 +#define TFM_SQR3 +#define TFM_SQR4 +#define TFM_SQR6 +#define TFM_SQR7 +#define TFM_SQR8 +#define TFM_SQR9 +#define TFM_SQR12 +#define TFM_SQR17 +#endif +#define TFM_SQR20 +#define TFM_SQR24 +#define TFM_SQR28 +#define TFM_SQR32 +#define TFM_SQR48 +#define TFM_SQR64 + +/* do we want some overflow checks + Not required if you make sure your numbers are within range (e.g. by default a modulus for fp_exptmod() can only be upto 2048 bits long) + */ +/* #define TFM_CHECK */ + +/* Is the target a P4 Prescott + */ +/* #define TFM_PRESCOTT */ + +/* Do we want timing resistant fp_exptmod() ? + * This makes it slower but also timing invariant with respect to the exponent + */ +/* #define TFM_TIMING_RESISTANT */ + +#endif + +/* Max size of any number in bits. Basically the largest size you will be multiplying + * should be half [or smaller] of FP_MAX_SIZE-four_digit + * + * You can externally define this or it defaults to 4096-bits [allowing multiplications upto 2048x2048 bits ] + */ +#ifndef FP_MAX_SIZE + #define FP_MAX_SIZE (4096+(8*DIGIT_BIT)) +#endif + +/* will this lib work? */ +#if (CHAR_BIT & 7) + #error CHAR_BIT must be a multiple of eight. +#endif +#if FP_MAX_SIZE % CHAR_BIT + #error FP_MAX_SIZE must be a multiple of CHAR_BIT +#endif + +#if __SIZEOF_LONG__ == 8 + #define FP_64BIT +#endif + +/* autodetect x86-64 and make sure we are using 64-bit digits with x86-64 asm */ +#if defined(__x86_64__) + #if defined(TFM_X86) || defined(TFM_SSE2) || defined(TFM_ARM) + #error x86-64 detected, x86-32/SSE2/ARM optimizations are not valid! + #endif + #if !defined(TFM_X86_64) && !defined(TFM_NO_ASM) + #define TFM_X86_64 + #endif +#endif +#if defined(TFM_X86_64) + #if !defined(FP_64BIT) + #define FP_64BIT + #endif +#endif + +/* try to detect x86-32 */ +#if defined(__i386__) && !defined(TFM_SSE2) + #if defined(TFM_X86_64) || defined(TFM_ARM) + #error x86-32 detected, x86-64/ARM optimizations are not valid! + #endif + #if !defined(TFM_X86) && !defined(TFM_NO_ASM) + #define TFM_X86 + #endif +#endif + +/* make sure we're 32-bit for x86-32/sse/arm/ppc32 */ +#if (defined(TFM_X86) || defined(TFM_SSE2) || defined(TFM_ARM) || defined(TFM_PPC32)) && defined(FP_64BIT) + #warning x86-32, SSE2 and ARM, PPC32 optimizations require 32-bit digits (undefining) + #undef FP_64BIT +#endif + +/* multi asms? */ +#ifdef TFM_X86 + #define TFM_ASM +#endif +#ifdef TFM_X86_64 + #ifdef TFM_ASM + #error TFM_ASM already defined! + #endif + #define TFM_ASM +#endif +#ifdef TFM_SSE2 + #ifdef TFM_ASM + #error TFM_ASM already defined! + #endif + #define TFM_ASM +#endif +#ifdef TFM_ARM + #ifdef TFM_ASM + #error TFM_ASM already defined! + #endif + #define TFM_ASM +#endif +#ifdef TFM_PPC32 + #ifdef TFM_ASM + #error TFM_ASM already defined! + #endif + #define TFM_ASM +#endif +#ifdef TFM_PPC64 + #ifdef TFM_ASM + #error TFM_ASM already defined! + #endif + #define TFM_ASM +#endif +#ifdef TFM_AVR32 + #ifdef TFM_ASM + #error TFM_ASM already defined! + #endif + #define TFM_ASM +#endif + +/* we want no asm? */ +#ifdef TFM_NO_ASM + #undef TFM_X86 + #undef TFM_X86_64 + #undef TFM_SSE2 + #undef TFM_ARM + #undef TFM_PPC32 + #undef TFM_PPC64 + #undef TFM_AVR32 + #undef TFM_ASM +#endif + +/* ECC helpers */ +#ifdef TFM_ECC192 + #ifdef FP_64BIT + #define TFM_MUL3 + #define TFM_SQR3 + #else + #define TFM_MUL6 + #define TFM_SQR6 + #endif +#endif + +#ifdef TFM_ECC224 + #ifdef FP_64BIT + #define TFM_MUL4 + #define TFM_SQR4 + #else + #define TFM_MUL7 + #define TFM_SQR7 + #endif +#endif + +#ifdef TFM_ECC256 + #ifdef FP_64BIT + #define TFM_MUL4 + #define TFM_SQR4 + #else + #define TFM_MUL8 + #define TFM_SQR8 + #endif +#endif + +#ifdef TFM_ECC384 + #ifdef FP_64BIT + #define TFM_MUL6 + #define TFM_SQR6 + #else + #define TFM_MUL12 + #define TFM_SQR12 + #endif +#endif + +#ifdef TFM_ECC521 + #ifdef FP_64BIT + #define TFM_MUL9 + #define TFM_SQR9 + #else + #define TFM_MUL17 + #define TFM_SQR17 + #endif +#endif + +/* use arc4random on platforms that support it */ +#if defined(__FreeBSD__) || defined(__OpenBSD__) || defined(__NetBSD__) || defined(__DragonFly__) + #define FP_GEN_RANDOM() arc4random() + #define FP_GEN_RANDOM_MAX 0xffffffff +#endif + +/* use rand() as fall-back if there's no better rand function */ +#ifndef FP_GEN_RANDOM + #define FP_GEN_RANDOM() rand() + #define FP_GEN_RANDOM_MAX RAND_MAX +#endif + +/* some default configurations. + */ +#if defined(FP_64BIT) + /* for GCC only on supported platforms */ +#ifndef CRYPT + typedef unsigned long long ulong64; +#endif /* CRYPT */ + + typedef ulong64 fp_digit; +#define SIZEOF_FP_DIGIT 8 + typedef unsigned long fp_word __attribute__ ((mode(TI))); + +#else + + /* this is to make porting into LibTomCrypt easier :-) */ +#ifndef CRYPT + #if defined(_MSC_VER) || defined(__BORLANDC__) + typedef unsigned __int64 ulong64; + typedef signed __int64 long64; + #else + typedef unsigned long long ulong64; + typedef signed long long long64; + #endif /* defined(_MSC_VER) ... */ +#endif /* CRYPT */ + + typedef unsigned int fp_digit; +#define SIZEOF_FP_DIGIT 4 + typedef ulong64 fp_word; +#endif /* FP_64BIT */ + +/* # of digits this is */ +#define DIGIT_BIT ((CHAR_BIT) * SIZEOF_FP_DIGIT) +#define FP_MASK (fp_digit)(-1) +#define FP_SIZE (FP_MAX_SIZE/DIGIT_BIT) + +/* signs */ +#define FP_ZPOS 0 +#define FP_NEG 1 + +/* return codes */ +#define FP_OKAY 0 +#define FP_VAL 1 +#define FP_MEM 2 + +/* equalities */ +#define FP_LT -1 /* less than */ +#define FP_EQ 0 /* equal to */ +#define FP_GT 1 /* greater than */ + +/* replies */ +#define FP_YES 1 /* yes response */ +#define FP_NO 0 /* no response */ + +/* a FP type */ +typedef struct { + fp_digit dp[FP_SIZE]; + int used, + sign; +} fp_int; + +/* functions */ + +/* returns a TFM ident string useful for debugging... */ +const char *fp_ident(void); + +/* initialize [or zero] an fp int */ +#define fp_init(a) (void)memset((a), 0, sizeof(fp_int)) +#define fp_zero(a) fp_init(a) + +/* zero/even/odd ? */ +#define fp_iszero(a) (((a)->used == 0) ? FP_YES : FP_NO) +#define fp_iseven(a) (((a)->used >= 0 && (((a)->dp[0] & 1) == 0)) ? FP_YES : FP_NO) +#define fp_isodd(a) (((a)->used > 0 && (((a)->dp[0] & 1) == 1)) ? FP_YES : FP_NO) + +/* set to a small digit */ +void fp_set(fp_int *a, fp_digit b); + +/* makes a pseudo-random int of a given size */ +void fp_rand(fp_int *a, int digits); + +/* copy from a to b */ +#define fp_copy(a, b) (void)(((a) != (b)) && memcpy((b), (a), sizeof(fp_int))) +#define fp_init_copy(a, b) fp_copy(b, a) + +/* clamp digits */ +#define fp_clamp(a) { while ((a)->used && (a)->dp[(a)->used-1] == 0) --((a)->used); (a)->sign = (a)->used ? (a)->sign : FP_ZPOS; } + +/* negate and absolute */ +#define fp_neg(a, b) { fp_copy(a, b); (b)->sign ^= 1; fp_clamp(b); } +#define fp_abs(a, b) { fp_copy(a, b); (b)->sign = 0; } + +/* right shift x digits */ +void fp_rshd(fp_int *a, int x); + +/* left shift x digits */ +void fp_lshd(fp_int *a, int x); + +/* signed comparison */ +int fp_cmp(fp_int *a, fp_int *b); + +/* unsigned comparison */ +int fp_cmp_mag(fp_int *a, fp_int *b); + +/* power of 2 operations */ +void fp_div_2d(fp_int *a, int b, fp_int *c, fp_int *d); +void fp_mod_2d(fp_int *a, int b, fp_int *c); +void fp_mul_2d(fp_int *a, int b, fp_int *c); +void fp_2expt (fp_int *a, int b); +void fp_mul_2(fp_int *a, fp_int *c); +void fp_div_2(fp_int *a, fp_int *c); + +/* Counts the number of lsbs which are zero before the first zero bit */ +int fp_cnt_lsb(fp_int *a); + +/* c = a + b */ +void fp_add(fp_int *a, fp_int *b, fp_int *c); + +/* c = a - b */ +void fp_sub(fp_int *a, fp_int *b, fp_int *c); + +/* c = a * b */ +void fp_mul(fp_int *a, fp_int *b, fp_int *c); + +/* b = a*a */ +void fp_sqr(fp_int *a, fp_int *b); + +/* a/b => cb + d == a */ +int fp_div(fp_int *a, fp_int *b, fp_int *c, fp_int *d); + +/* c = a mod b, 0 <= c < b */ +int fp_mod(fp_int *a, fp_int *b, fp_int *c); + +/* compare against a single digit */ +int fp_cmp_d(fp_int *a, fp_digit b); + +/* c = a + b */ +void fp_add_d(fp_int *a, fp_digit b, fp_int *c); + +/* c = a - b */ +void fp_sub_d(fp_int *a, fp_digit b, fp_int *c); + +/* c = a * b */ +void fp_mul_d(fp_int *a, fp_digit b, fp_int *c); + +/* a/b => cb + d == a */ +int fp_div_d(fp_int *a, fp_digit b, fp_int *c, fp_digit *d); + +/* c = a mod b, 0 <= c < b */ +int fp_mod_d(fp_int *a, fp_digit b, fp_digit *c); + +/* ---> number theory <--- */ +/* d = a + b (mod c) */ +int fp_addmod(fp_int *a, fp_int *b, fp_int *c, fp_int *d); + +/* d = a - b (mod c) */ +int fp_submod(fp_int *a, fp_int *b, fp_int *c, fp_int *d); + +/* d = a * b (mod c) */ +int fp_mulmod(fp_int *a, fp_int *b, fp_int *c, fp_int *d); + +/* c = a * a (mod b) */ +int fp_sqrmod(fp_int *a, fp_int *b, fp_int *c); + +/* c = 1/a (mod b) */ +int fp_invmod(fp_int *a, fp_int *b, fp_int *c); + +/* c = (a, b) */ +void fp_gcd(fp_int *a, fp_int *b, fp_int *c); + +/* c = [a, b] */ +void fp_lcm(fp_int *a, fp_int *b, fp_int *c); + +/* setups the montgomery reduction */ +int fp_montgomery_setup(fp_int *a, fp_digit *mp); + +/* computes a = B**n mod b without division or multiplication useful for + * normalizing numbers in a Montgomery system. + */ +void fp_montgomery_calc_normalization(fp_int *a, fp_int *b); + +/* computes x/R == x (mod N) via Montgomery Reduction */ +void fp_montgomery_reduce(fp_int *a, fp_int *m, fp_digit mp); + +/* d = a**b (mod c) */ +int fp_exptmod(fp_int *a, fp_int *b, fp_int *c, fp_int *d); + +/* primality stuff */ + +/* perform a Miller-Rabin test of a to the base b and store result in "result" */ +void fp_prime_miller_rabin (fp_int * a, fp_int * b, int *result); + +#define FP_PRIME_SIZE 256 +/* 256 trial divisions + 8 Miller-Rabins, returns FP_YES if probable prime */ +int fp_isprime(fp_int *a); +/* extended version of fp_isprime, do 't' Miller-Rabins instead of only 8 */ +int fp_isprime_ex(fp_int *a, int t); + +/* Primality generation flags */ +#define TFM_PRIME_BBS 0x0001 /* BBS style prime */ +#define TFM_PRIME_SAFE 0x0002 /* Safe prime (p-1)/2 == prime */ +#define TFM_PRIME_2MSB_OFF 0x0004 /* force 2nd MSB to 0 */ +#define TFM_PRIME_2MSB_ON 0x0008 /* force 2nd MSB to 1 */ + +/* callback for fp_prime_random, should fill dst with random bytes and return how many read [upto len] */ +typedef int tfm_prime_callback(unsigned char *dst, int len, void *dat); + +#define fp_prime_random(a, t, size, bbs, cb, dat) fp_prime_random_ex(a, t, ((size) * 8) + 1, (bbs==1)?TFM_PRIME_BBS:0, cb, dat) + +int fp_prime_random_ex(fp_int *a, int t, int size, int flags, tfm_prime_callback cb, void *dat); + +/* radix conersions */ +int fp_count_bits(fp_int *a); + +int fp_unsigned_bin_size(fp_int *a); +void fp_read_unsigned_bin(fp_int *a, const unsigned char *b, int c); +void fp_to_unsigned_bin(fp_int *a, unsigned char *b); + +int fp_signed_bin_size(fp_int *a); +void fp_read_signed_bin(fp_int *a, const unsigned char *b, int c); +void fp_to_signed_bin(fp_int *a, unsigned char *b); + +int fp_read_radix(fp_int *a, const char *str, int radix); + +int fp_radix_size(fp_int *a, int radix, int *size); +int fp_toradix(fp_int *a, char *str, int radix); +int fp_toradix_n(fp_int * a, char *str, int radix, int maxlen); + +#endif + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */ diff --git a/src/crypto/tfm/tfm_private.h b/src/crypto/tfm/tfm_private.h new file mode 100644 index 0000000..7f67225 --- /dev/null +++ b/src/crypto/tfm/tfm_private.h @@ -0,0 +1,125 @@ +/* TomsFastMath, a fast ISO C bignum library. + * + * This project is meant to fill in where LibTomMath + * falls short. That is speed ;-) + * + * This project is public domain and free for all purposes. + * + * Tom St Denis, tomstdenis@gmail.com + */ +#ifndef TFM_PRIVATE_H_ +#define TFM_PRIVATE_H_ + +#include "tfm.h" + +/* VARIOUS LOW LEVEL STUFFS */ +void s_fp_add(fp_int *a, fp_int *b, fp_int *c); +void s_fp_sub(fp_int *a, fp_int *b, fp_int *c); +void fp_reverse(unsigned char *s, int len); + +void fp_mul_comba(fp_int *A, fp_int *B, fp_int *C); + +#ifdef TFM_SMALL_SET +void fp_mul_comba_small(fp_int *A, fp_int *B, fp_int *C); +#endif + +#ifdef TFM_MUL3 +void fp_mul_comba3(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL4 +void fp_mul_comba4(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL6 +void fp_mul_comba6(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL7 +void fp_mul_comba7(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL8 +void fp_mul_comba8(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL9 +void fp_mul_comba9(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL12 +void fp_mul_comba12(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL17 +void fp_mul_comba17(fp_int *A, fp_int *B, fp_int *C); +#endif + +#ifdef TFM_MUL20 +void fp_mul_comba20(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL24 +void fp_mul_comba24(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL28 +void fp_mul_comba28(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL32 +void fp_mul_comba32(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL48 +void fp_mul_comba48(fp_int *A, fp_int *B, fp_int *C); +#endif +#ifdef TFM_MUL64 +void fp_mul_comba64(fp_int *A, fp_int *B, fp_int *C); +#endif + +void fp_sqr_comba(fp_int *A, fp_int *B); + +#ifdef TFM_SMALL_SET +void fp_sqr_comba_small(fp_int *A, fp_int *B); +#endif + +#ifdef TFM_SQR3 +void fp_sqr_comba3(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR4 +void fp_sqr_comba4(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR6 +void fp_sqr_comba6(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR7 +void fp_sqr_comba7(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR8 +void fp_sqr_comba8(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR9 +void fp_sqr_comba9(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR12 +void fp_sqr_comba12(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR17 +void fp_sqr_comba17(fp_int *A, fp_int *B); +#endif + +#ifdef TFM_SQR20 +void fp_sqr_comba20(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR24 +void fp_sqr_comba24(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR28 +void fp_sqr_comba28(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR32 +void fp_sqr_comba32(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR48 +void fp_sqr_comba48(fp_int *A, fp_int *B); +#endif +#ifdef TFM_SQR64 +void fp_sqr_comba64(fp_int *A, fp_int *B); +#endif +extern const char *fp_s_rmap; + +#endif + +/* $Source$ */ +/* $Revision$ */ +/* $Date$ */