Nixyri/godot
1
0
Fork 0
mirror of https://github.com/godotengine/godot.git synced 2024-10-23 13:41:43 +00:00
Code Issues Actions 1 Packages Projects Releases Wiki Activity

Update OpenSSL to version 1.0.2g

Browse Source
This commit is contained in:
mrezai 2016-04-15 19:03:35 +04:30
parent 880f4abda4
commit e97922f220
447 changed files with 32806 additions and 113880 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

11
drivers/builtin_openssl2/SCsub
View File

@ -3,6 +3,7 @@ Import('env')
openssl_sources = [
"builtin_openssl2/nocpuid.c",
"builtin_openssl2/ssl/t1_lib.c",
"builtin_openssl2/ssl/t1_ext.c",
"builtin_openssl2/ssl/s3_srvr.c",
"builtin_openssl2/ssl/t1_enc.c",
"builtin_openssl2/ssl/t1_meth.c",
@ -11,7 +12,6 @@ openssl_sources = [
"builtin_openssl2/ssl/tls_srp.c",
"builtin_openssl2/ssl/kssl.c",
"builtin_openssl2/ssl/d1_both.c",
"builtin_openssl2/ssl/d1_enc.c",
"builtin_openssl2/ssl/t1_clnt.c",
"builtin_openssl2/ssl/bio_ssl.c",
"builtin_openssl2/ssl/d1_srtp.c",
@ -209,12 +209,12 @@ openssl_sources = [
"builtin_openssl2/crypto/evp/c_all.c",
"builtin_openssl2/crypto/evp/m_md2.c",
"builtin_openssl2/crypto/evp/e_xcbc_d.c",
"builtin_openssl2/crypto/evp/evp_fips.c",
"builtin_openssl2/crypto/evp/pmeth_fn.c",
"builtin_openssl2/crypto/evp/p_lib.c",
"builtin_openssl2/crypto/evp/evp_key.c",
"builtin_openssl2/crypto/evp/encode.c",
"builtin_openssl2/crypto/evp/e_aes_cbc_hmac_sha1.c",
"builtin_openssl2/crypto/evp/e_aes_cbc_hmac_sha256.c",
"builtin_openssl2/crypto/evp/m_mdc2.c",
"builtin_openssl2/crypto/evp/e_null.c",
"builtin_openssl2/crypto/evp/p_sign.c",
@ -242,6 +242,7 @@ openssl_sources = [
"builtin_openssl2/crypto/ecdh/ech_ossl.c",
"builtin_openssl2/crypto/ecdh/ech_lib.c",
"builtin_openssl2/crypto/ecdh/ech_err.c",
"builtin_openssl2/crypto/ecdh/ech_kdf.c",
"builtin_openssl2/crypto/o_str.c",
"builtin_openssl2/crypto/conf/conf_api.c",
"builtin_openssl2/crypto/conf/conf_err.c",
@ -296,6 +297,7 @@ openssl_sources = [
"builtin_openssl2/crypto/cms/cms_env.c",
"builtin_openssl2/crypto/cms/cms_enc.c",
"builtin_openssl2/crypto/cms/cms_ess.c",
"builtin_openssl2/crypto/cms/cms_kari.c",
"builtin_openssl2/crypto/mem_dbg.c",
"builtin_openssl2/crypto/uid.c",
"builtin_openssl2/crypto/stack/stack.c",
@ -362,6 +364,7 @@ openssl_sources = [
"builtin_openssl2/crypto/x509v3/v3_genn.c",
"builtin_openssl2/crypto/x509v3/pcy_cache.c",
"builtin_openssl2/crypto/x509v3/v3_sxnet.c",
"builtin_openssl2/crypto/x509v3/v3_scts.c",
"builtin_openssl2/crypto/x509v3/v3err.c",
"builtin_openssl2/crypto/x509v3/v3_conf.c",
"builtin_openssl2/crypto/x509v3/v3_utl.c",
@ -420,7 +423,6 @@ openssl_sources = [
"builtin_openssl2/crypto/o_fips.c",
"builtin_openssl2/crypto/engine/eng_rdrand.c",
"builtin_openssl2/crypto/engine/eng_err.c",
"builtin_openssl2/crypto/engine/eng_rsax.c",
"builtin_openssl2/crypto/engine/tb_ecdsa.c",
"builtin_openssl2/crypto/engine/tb_rsa.c",
"builtin_openssl2/crypto/engine/tb_cipher.c",
@ -487,6 +489,8 @@ openssl_sources = [
"builtin_openssl2/crypto/dh/dh_ameth.c",
"builtin_openssl2/crypto/dh/dh_check.c",
"builtin_openssl2/crypto/dh/dh_err.c",
"builtin_openssl2/crypto/dh/dh_kdf.c",
"builtin_openssl2/crypto/dh/dh_rfc5114.c",
"builtin_openssl2/crypto/modes/ccm128.c",
"builtin_openssl2/crypto/modes/ofb128.c",
"builtin_openssl2/crypto/modes/cts128.c",
@ -495,6 +499,7 @@ openssl_sources = [
"builtin_openssl2/crypto/modes/cbc128.c",
"builtin_openssl2/crypto/modes/cfb128.c",
"builtin_openssl2/crypto/modes/xts128.c",
"builtin_openssl2/crypto/modes/wrap128.c",
"builtin_openssl2/crypto/camellia/cmll_cfb.c",
"builtin_openssl2/crypto/camellia/cmll_ecb.c",
"builtin_openssl2/crypto/camellia/cmll_utl.c",

186
drivers/builtin_openssl2/crypto/aes/aes_wrap.c
View File

@ -54,197 +54,19 @@
#include "cryptlib.h"
#include <openssl/aes.h>
#include <openssl/bio.h>
static const unsigned char default_iv[] = {
0xA6, 0xA6, 0xA6, 0xA6, 0xA6, 0xA6, 0xA6, 0xA6,
};
#include <openssl/modes.h>
int AES_wrap_key(AES_KEY *key, const unsigned char *iv,
unsigned char *out,
const unsigned char *in, unsigned int inlen)
{
unsigned char *A, B[16], *R;
unsigned int i, j, t;
if ((inlen & 0x7) || (inlen < 8))
return -1;
A = B;
t = 1;
memcpy(out + 8, in, inlen);
if (!iv)
iv = default_iv;
memcpy(A, iv, 8);
for (j = 0; j < 6; j++) {
R = out + 8;
for (i = 0; i < inlen; i += 8, t++, R += 8) {
memcpy(B + 8, R, 8);
AES_encrypt(B, B, key);
A[7] ^= (unsigned char)(t & 0xff);
if (t > 0xff) {
A[6] ^= (unsigned char)((t >> 8) & 0xff);
A[5] ^= (unsigned char)((t >> 16) & 0xff);
A[4] ^= (unsigned char)((t >> 24) & 0xff);
}
memcpy(R, B + 8, 8);
}
}
memcpy(out, A, 8);
return inlen + 8;
return CRYPTO_128_wrap(key, iv, out, in, inlen, (block128_f) AES_encrypt);
}
int AES_unwrap_key(AES_KEY *key, const unsigned char *iv,
unsigned char *out,
const unsigned char *in, unsigned int inlen)
{
unsigned char *A, B[16], *R;
unsigned int i, j, t;
inlen -= 8;
if (inlen & 0x7)
return -1;
if (inlen < 8)
return -1;
A = B;
t = 6 * (inlen >> 3);
memcpy(A, in, 8);
memcpy(out, in + 8, inlen);
for (j = 0; j < 6; j++) {
R = out + inlen - 8;
for (i = 0; i < inlen; i += 8, t--, R -= 8) {
A[7] ^= (unsigned char)(t & 0xff);
if (t > 0xff) {
A[6] ^= (unsigned char)((t >> 8) & 0xff);
A[5] ^= (unsigned char)((t >> 16) & 0xff);
A[4] ^= (unsigned char)((t >> 24) & 0xff);
}
memcpy(B + 8, R, 8);
AES_decrypt(B, B, key);
memcpy(R, B + 8, 8);
}
}
if (!iv)
iv = default_iv;
if (memcmp(A, iv, 8)) {
OPENSSL_cleanse(out, inlen);
return 0;
}
return inlen;
return CRYPTO_128_unwrap(key, iv, out, in, inlen,
(block128_f) AES_decrypt);
}
#ifdef AES_WRAP_TEST
int AES_wrap_unwrap_test(const unsigned char *kek, int keybits,
const unsigned char *iv,
const unsigned char *eout,
const unsigned char *key, int keylen)
{
unsigned char *otmp = NULL, *ptmp = NULL;
int r, ret = 0;
AES_KEY wctx;
otmp = OPENSSL_malloc(keylen + 8);
ptmp = OPENSSL_malloc(keylen);
if (!otmp || !ptmp)
return 0;
if (AES_set_encrypt_key(kek, keybits, &wctx))
goto err;
r = AES_wrap_key(&wctx, iv, otmp, key, keylen);
if (r <= 0)
goto err;
if (eout && memcmp(eout, otmp, keylen))
goto err;
if (AES_set_decrypt_key(kek, keybits, &wctx))
goto err;
r = AES_unwrap_key(&wctx, iv, ptmp, otmp, r);
if (memcmp(key, ptmp, keylen))
goto err;
ret = 1;
err:
if (otmp)
OPENSSL_free(otmp);
if (ptmp)
OPENSSL_free(ptmp);
return ret;
}
int main(int argc, char **argv)
{
static const unsigned char kek[] = {
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f,
0x10, 0x11, 0x12, 0x13, 0x14, 0x15, 0x16, 0x17,
0x18, 0x19, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f
};
static const unsigned char key[] = {
0x00, 0x11, 0x22, 0x33, 0x44, 0x55, 0x66, 0x77,
0x88, 0x99, 0xaa, 0xbb, 0xcc, 0xdd, 0xee, 0xff,
0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x07,
0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f
};
static const unsigned char e1[] = {
0x1f, 0xa6, 0x8b, 0x0a, 0x81, 0x12, 0xb4, 0x47,
0xae, 0xf3, 0x4b, 0xd8, 0xfb, 0x5a, 0x7b, 0x82,
0x9d, 0x3e, 0x86, 0x23, 0x71, 0xd2, 0xcf, 0xe5
};
static const unsigned char e2[] = {
0x96, 0x77, 0x8b, 0x25, 0xae, 0x6c, 0xa4, 0x35,
0xf9, 0x2b, 0x5b, 0x97, 0xc0, 0x50, 0xae, 0xd2,
0x46, 0x8a, 0xb8, 0xa1, 0x7a, 0xd8, 0x4e, 0x5d
};
static const unsigned char e3[] = {
0x64, 0xe8, 0xc3, 0xf9, 0xce, 0x0f, 0x5b, 0xa2,
0x63, 0xe9, 0x77, 0x79, 0x05, 0x81, 0x8a, 0x2a,
0x93, 0xc8, 0x19, 0x1e, 0x7d, 0x6e, 0x8a, 0xe7
};
static const unsigned char e4[] = {
0x03, 0x1d, 0x33, 0x26, 0x4e, 0x15, 0xd3, 0x32,
0x68, 0xf2, 0x4e, 0xc2, 0x60, 0x74, 0x3e, 0xdc,
0xe1, 0xc6, 0xc7, 0xdd, 0xee, 0x72, 0x5a, 0x93,
0x6b, 0xa8, 0x14, 0x91, 0x5c, 0x67, 0x62, 0xd2
};
static const unsigned char e5[] = {
0xa8, 0xf9, 0xbc, 0x16, 0x12, 0xc6, 0x8b, 0x3f,
0xf6, 0xe6, 0xf4, 0xfb, 0xe3, 0x0e, 0x71, 0xe4,
0x76, 0x9c, 0x8b, 0x80, 0xa3, 0x2c, 0xb8, 0x95,
0x8c, 0xd5, 0xd1, 0x7d, 0x6b, 0x25, 0x4d, 0xa1
};
static const unsigned char e6[] = {
0x28, 0xc9, 0xf4, 0x04, 0xc4, 0xb8, 0x10, 0xf4,
0xcb, 0xcc, 0xb3, 0x5c, 0xfb, 0x87, 0xf8, 0x26,
0x3f, 0x57, 0x86, 0xe2, 0xd8, 0x0e, 0xd3, 0x26,
0xcb, 0xc7, 0xf0, 0xe7, 0x1a, 0x99, 0xf4, 0x3b,
0xfb, 0x98, 0x8b, 0x9b, 0x7a, 0x02, 0xdd, 0x21
};
AES_KEY wctx, xctx;
int ret;
ret = AES_wrap_unwrap_test(kek, 128, NULL, e1, key, 16);
fprintf(stderr, "Key test result %d\n", ret);
ret = AES_wrap_unwrap_test(kek, 192, NULL, e2, key, 16);
fprintf(stderr, "Key test result %d\n", ret);
ret = AES_wrap_unwrap_test(kek, 256, NULL, e3, key, 16);
fprintf(stderr, "Key test result %d\n", ret);
ret = AES_wrap_unwrap_test(kek, 192, NULL, e4, key, 24);
fprintf(stderr, "Key test result %d\n", ret);
ret = AES_wrap_unwrap_test(kek, 256, NULL, e5, key, 24);
fprintf(stderr, "Key test result %d\n", ret);
ret = AES_wrap_unwrap_test(kek, 256, NULL, e6, key, 32);
fprintf(stderr, "Key test result %d\n", ret);
}
#endif

6
drivers/builtin_openssl2/crypto/aes/aes_x86core.c
View File

@ -89,8 +89,10 @@ typedef unsigned long long u64;
#endif
#undef ROTATE
#if defined(_MSC_VER) || defined(__ICC)
# define ROTATE(a,n) _lrotl(a,n)
#if defined(_MSC_VER)
# define ROTATE(a,n) _lrotl(a,n)
#elif defined(__ICC)
# define ROTATE(a,n) _rotl(a,n)
#elif defined(__GNUC__) && __GNUC__>=2
# if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
# define ROTATE(a,n) ({ register unsigned int ret; \

2980
drivers/builtin_openssl2/crypto/aes/asm/aes-586.pl
View File

File diff suppressed because it is too large Load Diff

1134
drivers/builtin_openssl2/crypto/aes/asm/aes-armv4.pl
View File

File diff suppressed because it is too large Load Diff

1123
drivers/builtin_openssl2/crypto/aes/asm/aes-ia64.S
View File

File diff suppressed because it is too large Load Diff

1611
drivers/builtin_openssl2/crypto/aes/asm/aes-mips.pl
View File

File diff suppressed because it is too large Load Diff

1022
drivers/builtin_openssl2/crypto/aes/asm/aes-parisc.pl
View File

File diff suppressed because it is too large Load Diff

1365
drivers/builtin_openssl2/crypto/aes/asm/aes-ppc.pl
View File

File diff suppressed because it is too large Load Diff

2237
drivers/builtin_openssl2/crypto/aes/asm/aes-s390x.pl
View File

File diff suppressed because it is too large Load Diff

1182
drivers/builtin_openssl2/crypto/aes/asm/aes-sparcv9.pl
View File

File diff suppressed because it is too large Load Diff

2819
drivers/builtin_openssl2/crypto/aes/asm/aes-x86_64.pl
View File

File diff suppressed because it is too large Load Diff

1250
drivers/builtin_openssl2/crypto/aes/asm/aesni-sha1-x86_64.pl
View File

File diff suppressed because it is too large Load Diff

2189
drivers/builtin_openssl2/crypto/aes/asm/aesni-x86.pl
View File

File diff suppressed because it is too large Load Diff

3071
drivers/builtin_openssl2/crypto/aes/asm/aesni-x86_64.pl
View File

File diff suppressed because it is too large Load Diff

3108
drivers/builtin_openssl2/crypto/aes/asm/bsaes-x86_64.pl
View File

File diff suppressed because it is too large Load Diff

903
drivers/builtin_openssl2/crypto/aes/asm/vpaes-x86.pl
View File

@ -1,903 +0,0 @@
#!/usr/bin/env perl
######################################################################
## Constant-time SSSE3 AES core implementation.
## version 0.1
##
## By Mike Hamburg (Stanford University), 2009
## Public domain.
##
## For details see http://shiftleft.org/papers/vector_aes/ and
## http://crypto.stanford.edu/vpaes/.
######################################################################
# September 2011.
#
# Port vpaes-x86_64.pl as 32-bit "almost" drop-in replacement for
# aes-586.pl. "Almost" refers to the fact that AES_cbc_encrypt
# doesn't handle partial vectors (doesn't have to if called from
# EVP only). "Drop-in" implies that this module doesn't share key
# schedule structure with the original nor does it make assumption
# about its alignment...
#
# Performance summary. aes-586.pl column lists large-block CBC
# encrypt/decrypt/with-hyper-threading-off(*) results in cycles per
# byte processed with 128-bit key, and vpaes-x86.pl column - [also
# large-block CBC] encrypt/decrypt.
#
# aes-586.pl vpaes-x86.pl
#
# Core 2(**) 29.1/42.3/18.3 22.0/25.6(***)
# Nehalem 27.9/40.4/18.1 10.3/12.0
# Atom 102./119./60.1 64.5/85.3(***)
#
# (*) "Hyper-threading" in the context refers rather to cache shared
# among multiple cores, than to specifically Intel HTT. As vast
# majority of contemporary cores share cache, slower code path
# is common place. In other words "with-hyper-threading-off"
# results are presented mostly for reference purposes.
#
# (**) "Core 2" refers to initial 65nm design, a.k.a. Conroe.
#
# (***) Less impressive improvement on Core 2 and Atom is due to slow
# pshufb, yet it's respectable +32%/65% improvement on Core 2
# and +58%/40% on Atom (as implied, over "hyper-threading-safe"
# code path).
#
# <appro@openssl.org>
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
push(@INC,"${dir}","${dir}../../perlasm");
require "x86asm.pl";
&asm_init($ARGV[0],"vpaes-x86.pl",$x86only = $ARGV[$#ARGV] eq "386");
$PREFIX="vpaes";
my ($round, $base, $magic, $key, $const, $inp, $out)=
("eax", "ebx", "ecx", "edx","ebp", "esi","edi");
&static_label("_vpaes_consts");
&static_label("_vpaes_schedule_low_round");
&set_label("_vpaes_consts",64);
$k_inv=-0x30; # inv, inva
&data_word(0x0D080180,0x0E05060F,0x0A0B0C02,0x04070309);
&data_word(0x0F0B0780,0x01040A06,0x02050809,0x030D0E0C);
$k_s0F=-0x10; # s0F
&data_word(0x0F0F0F0F,0x0F0F0F0F,0x0F0F0F0F,0x0F0F0F0F);
$k_ipt=0x00; # input transform (lo, hi)
&data_word(0x5A2A7000,0xC2B2E898,0x52227808,0xCABAE090);
&data_word(0x317C4D00,0x4C01307D,0xB0FDCC81,0xCD80B1FC);
$k_sb1=0x20; # sb1u, sb1t
&data_word(0xCB503E00,0xB19BE18F,0x142AF544,0xA5DF7A6E);
&data_word(0xFAE22300,0x3618D415,0x0D2ED9EF,0x3BF7CCC1);
$k_sb2=0x40; # sb2u, sb2t
&data_word(0x0B712400,0xE27A93C6,0xBC982FCD,0x5EB7E955);
&data_word(0x0AE12900,0x69EB8840,0xAB82234A,0xC2A163C8);
$k_sbo=0x60; # sbou, sbot
&data_word(0x6FBDC700,0xD0D26D17,0xC502A878,0x15AABF7A);
&data_word(0x5FBB6A00,0xCFE474A5,0x412B35FA,0x8E1E90D1);
$k_mc_forward=0x80; # mc_forward
&data_word(0x00030201,0x04070605,0x080B0A09,0x0C0F0E0D);
&data_word(0x04070605,0x080B0A09,0x0C0F0E0D,0x00030201);
&data_word(0x080B0A09,0x0C0F0E0D,0x00030201,0x04070605);
&data_word(0x0C0F0E0D,0x00030201,0x04070605,0x080B0A09);
$k_mc_backward=0xc0; # mc_backward
&data_word(0x02010003,0x06050407,0x0A09080B,0x0E0D0C0F);
&data_word(0x0E0D0C0F,0x02010003,0x06050407,0x0A09080B);
&data_word(0x0A09080B,0x0E0D0C0F,0x02010003,0x06050407);
&data_word(0x06050407,0x0A09080B,0x0E0D0C0F,0x02010003);
$k_sr=0x100; # sr
&data_word(0x03020100,0x07060504,0x0B0A0908,0x0F0E0D0C);
&data_word(0x0F0A0500,0x030E0904,0x07020D08,0x0B06010C);
&data_word(0x0B020900,0x0F060D04,0x030A0108,0x070E050C);
&data_word(0x070A0D00,0x0B0E0104,0x0F020508,0x0306090C);
$k_rcon=0x140; # rcon
&data_word(0xAF9DEEB6,0x1F8391B9,0x4D7C7D81,0x702A9808);
$k_s63=0x150; # s63: all equal to 0x63 transformed
&data_word(0x5B5B5B5B,0x5B5B5B5B,0x5B5B5B5B,0x5B5B5B5B);
$k_opt=0x160; # output transform
&data_word(0xD6B66000,0xFF9F4929,0xDEBE6808,0xF7974121);
&data_word(0x50BCEC00,0x01EDBD51,0xB05C0CE0,0xE10D5DB1);
$k_deskew=0x180; # deskew tables: inverts the sbox's "skew"
&data_word(0x47A4E300,0x07E4A340,0x5DBEF91A,0x1DFEB95A);
&data_word(0x83EA6900,0x5F36B5DC,0xF49D1E77,0x2841C2AB);
##
## Decryption stuff
## Key schedule constants
##
$k_dksd=0x1a0; # decryption key schedule: invskew x*D
&data_word(0xA3E44700,0xFEB91A5D,0x5A1DBEF9,0x0740E3A4);
&data_word(0xB5368300,0x41C277F4,0xAB289D1E,0x5FDC69EA);
$k_dksb=0x1c0; # decryption key schedule: invskew x*B
&data_word(0x8550D500,0x9A4FCA1F,0x1CC94C99,0x03D65386);
&data_word(0xB6FC4A00,0x115BEDA7,0x7E3482C8,0xD993256F);
$k_dkse=0x1e0; # decryption key schedule: invskew x*E + 0x63
&data_word(0x1FC9D600,0xD5031CCA,0x994F5086,0x53859A4C);
&data_word(0x4FDC7BE8,0xA2319605,0x20B31487,0xCD5EF96A);
$k_dks9=0x200; # decryption key schedule: invskew x*9
&data_word(0x7ED9A700,0xB6116FC8,0x82255BFC,0x4AED9334);
&data_word(0x27143300,0x45765162,0xE9DAFDCE,0x8BB89FAC);
##
## Decryption stuff
## Round function constants
##
$k_dipt=0x220; # decryption input transform
&data_word(0x0B545F00,0x0F505B04,0x114E451A,0x154A411E);
&data_word(0x60056500,0x86E383E6,0xF491F194,0x12771772);
$k_dsb9=0x240; # decryption sbox output *9*u, *9*t
&data_word(0x9A86D600,0x851C0353,0x4F994CC9,0xCAD51F50);
&data_word(0xECD74900,0xC03B1789,0xB2FBA565,0x725E2C9E);
$k_dsbd=0x260; # decryption sbox output *D*u, *D*t
&data_word(0xE6B1A200,0x7D57CCDF,0x882A4439,0xF56E9B13);
&data_word(0x24C6CB00,0x3CE2FAF7,0x15DEEFD3,0x2931180D);
$k_dsbb=0x280; # decryption sbox output *B*u, *B*t
&data_word(0x96B44200,0xD0226492,0xB0F2D404,0x602646F6);
&data_word(0xCD596700,0xC19498A6,0x3255AA6B,0xF3FF0C3E);
$k_dsbe=0x2a0; # decryption sbox output *E*u, *E*t
&data_word(0x26D4D000,0x46F29296,0x64B4F6B0,0x22426004);
&data_word(0xFFAAC100,0x0C55A6CD,0x98593E32,0x9467F36B);
$k_dsbo=0x2c0; # decryption sbox final output
&data_word(0x7EF94000,0x1387EA53,0xD4943E2D,0xC7AA6DB9);
&data_word(0x93441D00,0x12D7560F,0xD8C58E9C,0xCA4B8159);
&asciz ("Vector Permutation AES for x86/SSSE3, Mike Hamburg (Stanford University)");
&align (64);
&function_begin_B("_vpaes_preheat");
&add ($const,&DWP(0,"esp"));
&movdqa ("xmm7",&QWP($k_inv,$const));
&movdqa ("xmm6",&QWP($k_s0F,$const));
&ret ();
&function_end_B("_vpaes_preheat");
##
## _aes_encrypt_core
##
## AES-encrypt %xmm0.
##
## Inputs:
## %xmm0 = input
## %xmm6-%xmm7 as in _vpaes_preheat
## (%edx) = scheduled keys
##
## Output in %xmm0
## Clobbers %xmm1-%xmm5, %eax, %ebx, %ecx, %edx
##
##
&function_begin_B("_vpaes_encrypt_core");
&mov ($magic,16);
&mov ($round,&DWP(240,$key));
&movdqa ("xmm1","xmm6")
&movdqa ("xmm2",&QWP($k_ipt,$const));
&pandn ("xmm1","xmm0");
&movdqu ("xmm5",&QWP(0,$key));
&psrld ("xmm1",4);
&pand ("xmm0","xmm6");
&pshufb ("xmm2","xmm0");
&movdqa ("xmm0",&QWP($k_ipt+16,$const));
&pshufb ("xmm0","xmm1");
&pxor ("xmm2","xmm5");
&pxor ("xmm0","xmm2");
&add ($key,16);
&lea ($base,&DWP($k_mc_backward,$const));
&jmp (&label("enc_entry"));
&set_label("enc_loop",16);
# middle of middle round
&movdqa ("xmm4",&QWP($k_sb1,$const)); # 4 : sb1u
&pshufb ("xmm4","xmm2"); # 4 = sb1u
&pxor ("xmm4","xmm5"); # 4 = sb1u + k
&movdqa ("xmm0",&QWP($k_sb1+16,$const));# 0 : sb1t
&pshufb ("xmm0","xmm3"); # 0 = sb1t
&pxor ("xmm0","xmm4"); # 0 = A
&movdqa ("xmm5",&QWP($k_sb2,$const)); # 4 : sb2u
&pshufb ("xmm5","xmm2"); # 4 = sb2u
&movdqa ("xmm1",&QWP(-0x40,$base,$magic));# .Lk_mc_forward[]
&movdqa ("xmm2",&QWP($k_sb2+16,$const));# 2 : sb2t
&pshufb ("xmm2","xmm3"); # 2 = sb2t
&pxor ("xmm2","xmm5"); # 2 = 2A
&movdqa ("xmm4",&QWP(0,$base,$magic)); # .Lk_mc_backward[]
&movdqa ("xmm3","xmm0"); # 3 = A
&pshufb ("xmm0","xmm1"); # 0 = B
&add ($key,16); # next key
&pxor ("xmm0","xmm2"); # 0 = 2A+B
&pshufb ("xmm3","xmm4"); # 3 = D
&add ($magic,16); # next mc
&pxor ("xmm3","xmm0"); # 3 = 2A+B+D
&pshufb ("xmm0","xmm1"); # 0 = 2B+C
&and ($magic,0x30); # ... mod 4
&pxor ("xmm0","xmm3"); # 0 = 2A+3B+C+D
&sub ($round,1); # nr--
&set_label("enc_entry");
# top of round
&movdqa ("xmm1","xmm6"); # 1 : i
&pandn ("xmm1","xmm0"); # 1 = i<<4
&psrld ("xmm1",4); # 1 = i
&pand ("xmm0","xmm6"); # 0 = k
&movdqa ("xmm5",&QWP($k_inv+16,$const));# 2 : a/k
&pshufb ("xmm5","xmm0"); # 2 = a/k
&pxor ("xmm0","xmm1"); # 0 = j
&movdqa ("xmm3","xmm7"); # 3 : 1/i
&pshufb ("xmm3","xmm1"); # 3 = 1/i
&pxor ("xmm3","xmm5"); # 3 = iak = 1/i + a/k
&movdqa ("xmm4","xmm7"); # 4 : 1/j
&pshufb ("xmm4","xmm0"); # 4 = 1/j
&pxor ("xmm4","xmm5"); # 4 = jak = 1/j + a/k
&movdqa ("xmm2","xmm7"); # 2 : 1/iak
&pshufb ("xmm2","xmm3"); # 2 = 1/iak
&pxor ("xmm2","xmm0"); # 2 = io
&movdqa ("xmm3","xmm7"); # 3 : 1/jak
&movdqu ("xmm5",&QWP(0,$key));
&pshufb ("xmm3","xmm4"); # 3 = 1/jak
&pxor ("xmm3","xmm1"); # 3 = jo
&jnz (&label("enc_loop"));
# middle of last round
&movdqa ("xmm4",&QWP($k_sbo,$const)); # 3 : sbou .Lk_sbo
&movdqa ("xmm0",&QWP($k_sbo+16,$const));# 3 : sbot .Lk_sbo+16
&pshufb ("xmm4","xmm2"); # 4 = sbou
&pxor ("xmm4","xmm5"); # 4 = sb1u + k
&pshufb ("xmm0","xmm3"); # 0 = sb1t
&movdqa ("xmm1",&QWP(0x40,$base,$magic));# .Lk_sr[]
&pxor ("xmm0","xmm4"); # 0 = A
&pshufb ("xmm0","xmm1");
&ret ();
&function_end_B("_vpaes_encrypt_core");
##
## Decryption core
##
## Same API as encryption core.
##
&function_begin_B("_vpaes_decrypt_core");
&mov ($round,&DWP(240,$key));
&lea ($base,&DWP($k_dsbd,$const));
&movdqa ("xmm1","xmm6");
&movdqa ("xmm2",&QWP($k_dipt-$k_dsbd,$base));
&pandn ("xmm1","xmm0");
&mov ($magic,$round);
&psrld ("xmm1",4)
&movdqu ("xmm5",&QWP(0,$key));
&shl ($magic,4);
&pand ("xmm0","xmm6");
&pshufb ("xmm2","xmm0");
&movdqa ("xmm0",&QWP($k_dipt-$k_dsbd+16,$base));
&xor ($magic,0x30);
&pshufb ("xmm0","xmm1");
&and ($magic,0x30);
&pxor ("xmm2","xmm5");
&movdqa ("xmm5",&QWP($k_mc_forward+48,$const));
&pxor ("xmm0","xmm2");
&add ($key,16);
&lea ($magic,&DWP($k_sr-$k_dsbd,$base,$magic));
&jmp (&label("dec_entry"));
&set_label("dec_loop",16);
##
## Inverse mix columns
##
&movdqa ("xmm4",&QWP(-0x20,$base)); # 4 : sb9u
&pshufb ("xmm4","xmm2"); # 4 = sb9u
&pxor ("xmm4","xmm0");
&movdqa ("xmm0",&QWP(-0x10,$base)); # 0 : sb9t
&pshufb ("xmm0","xmm3"); # 0 = sb9t
&pxor ("xmm0","xmm4"); # 0 = ch
&add ($key,16); # next round key
&pshufb ("xmm0","xmm5"); # MC ch
&movdqa ("xmm4",&QWP(0,$base)); # 4 : sbdu
&pshufb ("xmm4","xmm2"); # 4 = sbdu
&pxor ("xmm4","xmm0"); # 4 = ch
&movdqa ("xmm0",&QWP(0x10,$base)); # 0 : sbdt
&pshufb ("xmm0","xmm3"); # 0 = sbdt
&pxor ("xmm0","xmm4"); # 0 = ch
&sub ($round,1); # nr--
&pshufb ("xmm0","xmm5"); # MC ch
&movdqa ("xmm4",&QWP(0x20,$base)); # 4 : sbbu
&pshufb ("xmm4","xmm2"); # 4 = sbbu
&pxor ("xmm4","xmm0"); # 4 = ch
&movdqa ("xmm0",&QWP(0x30,$base)); # 0 : sbbt
&pshufb ("xmm0","xmm3"); # 0 = sbbt
&pxor ("xmm0","xmm4"); # 0 = ch
&pshufb ("xmm0","xmm5"); # MC ch
&movdqa ("xmm4",&QWP(0x40,$base)); # 4 : sbeu
&pshufb ("xmm4","xmm2"); # 4 = sbeu
&pxor ("xmm4","xmm0"); # 4 = ch
&movdqa ("xmm0",&QWP(0x50,$base)); # 0 : sbet
&pshufb ("xmm0","xmm3"); # 0 = sbet
&pxor ("xmm0","xmm4"); # 0 = ch
&palignr("xmm5","xmm5",12);
&set_label("dec_entry");
# top of round
&movdqa ("xmm1","xmm6"); # 1 : i
&pandn ("xmm1","xmm0"); # 1 = i<<4
&psrld ("xmm1",4); # 1 = i
&pand ("xmm0","xmm6"); # 0 = k
&movdqa ("xmm2",&QWP($k_inv+16,$const));# 2 : a/k
&pshufb ("xmm2","xmm0"); # 2 = a/k
&pxor ("xmm0","xmm1"); # 0 = j
&movdqa ("xmm3","xmm7"); # 3 : 1/i
&pshufb ("xmm3","xmm1"); # 3 = 1/i
&pxor ("xmm3","xmm2"); # 3 = iak = 1/i + a/k
&movdqa ("xmm4","xmm7"); # 4 : 1/j
&pshufb ("xmm4","xmm0"); # 4 = 1/j
&pxor ("xmm4","xmm2"); # 4 = jak = 1/j + a/k
&movdqa ("xmm2","xmm7"); # 2 : 1/iak
&pshufb ("xmm2","xmm3"); # 2 = 1/iak
&pxor ("xmm2","xmm0"); # 2 = io
&movdqa ("xmm3","xmm7"); # 3 : 1/jak
&pshufb ("xmm3","xmm4"); # 3 = 1/jak
&pxor ("xmm3","xmm1"); # 3 = jo
&movdqu ("xmm0",&QWP(0,$key));
&jnz (&label("dec_loop"));
# middle of last round
&movdqa ("xmm4",&QWP(0x60,$base)); # 3 : sbou
&pshufb ("xmm4","xmm2"); # 4 = sbou
&pxor ("xmm4","xmm0"); # 4 = sb1u + k
&movdqa ("xmm0",&QWP(0x70,$base)); # 0 : sbot
&movdqa ("xmm2",&QWP(0,$magic));
&pshufb ("xmm0","xmm3"); # 0 = sb1t
&pxor ("xmm0","xmm4"); # 0 = A
&pshufb ("xmm0","xmm2");
&ret ();
&function_end_B("_vpaes_decrypt_core");
########################################################
## ##
## AES key schedule ##
## ##
########################################################
&function_begin_B("_vpaes_schedule_core");
&add ($const,&DWP(0,"esp"));
&movdqu ("xmm0",&QWP(0,$inp)); # load key (unaligned)
&movdqa ("xmm2",&QWP($k_rcon,$const)); # load rcon
# input transform
&movdqa ("xmm3","xmm0");
&lea ($base,&DWP($k_ipt,$const));
&movdqa (&QWP(4,"esp"),"xmm2"); # xmm8
&call ("_vpaes_schedule_transform");
&movdqa ("xmm7","xmm0");
&test ($out,$out);
&jnz (&label("schedule_am_decrypting"));
# encrypting, output zeroth round key after transform
&movdqu (&QWP(0,$key),"xmm0");
&jmp (&label("schedule_go"));
&set_label("schedule_am_decrypting");
# decrypting, output zeroth round key after shiftrows
&movdqa ("xmm1",&QWP($k_sr,$const,$magic));
&pshufb ("xmm3","xmm1");
&movdqu (&QWP(0,$key),"xmm3");
&xor ($magic,0x30);
&set_label("schedule_go");
&cmp ($round,192);
&ja (&label("schedule_256"));
&je (&label("schedule_192"));
# 128: fall though
##
## .schedule_128
##
## 128-bit specific part of key schedule.
##
## This schedule is really simple, because all its parts
## are accomplished by the subroutines.
##
&set_label("schedule_128");
&mov ($round,10);
&set_label("loop_schedule_128");
&call ("_vpaes_schedule_round");
&dec ($round);
&jz (&label("schedule_mangle_last"));
&call ("_vpaes_schedule_mangle"); # write output
&jmp (&label("loop_schedule_128"));
##
## .aes_schedule_192
##
## 192-bit specific part of key schedule.
##
## The main body of this schedule is the same as the 128-bit
## schedule, but with more smearing. The long, high side is
## stored in %xmm7 as before, and the short, low side is in
## the high bits of %xmm6.
##
## This schedule is somewhat nastier, however, because each
## round produces 192 bits of key material, or 1.5 round keys.
## Therefore, on each cycle we do 2 rounds and produce 3 round
## keys.
##
&set_label("schedule_192",16);
&movdqu ("xmm0",&QWP(8,$inp)); # load key part 2 (very unaligned)
&call ("_vpaes_schedule_transform"); # input transform
&movdqa ("xmm6","xmm0"); # save short part
&pxor ("xmm4","xmm4"); # clear 4
&movhlps("xmm6","xmm4"); # clobber low side with zeros
&mov ($round,4);
&set_label("loop_schedule_192");
&call ("_vpaes_schedule_round");
&palignr("xmm0","xmm6",8);
&call ("_vpaes_schedule_mangle"); # save key n
&call ("_vpaes_schedule_192_smear");
&call ("_vpaes_schedule_mangle"); # save key n+1
&call ("_vpaes_schedule_round");
&dec ($round);
&jz (&label("schedule_mangle_last"));
&call ("_vpaes_schedule_mangle"); # save key n+2
&call ("_vpaes_schedule_192_smear");
&jmp (&label("loop_schedule_192"));
##
## .aes_schedule_256
##
## 256-bit specific part of key schedule.
##
## The structure here is very similar to the 128-bit
## schedule, but with an additional "low side" in
## %xmm6. The low side's rounds are the same as the
## high side's, except no rcon and no rotation.
##
&set_label("schedule_256",16);
&movdqu ("xmm0",&QWP(16,$inp)); # load key part 2 (unaligned)
&call ("_vpaes_schedule_transform"); # input transform
&mov ($round,7);
&set_label("loop_schedule_256");
&call ("_vpaes_schedule_mangle"); # output low result
&movdqa ("xmm6","xmm0"); # save cur_lo in xmm6
# high round
&call ("_vpaes_schedule_round");
&dec ($round);
&jz (&label("schedule_mangle_last"));
&call ("_vpaes_schedule_mangle");
# low round. swap xmm7 and xmm6
&pshufd ("xmm0","xmm0",0xFF);
&movdqa (&QWP(20,"esp"),"xmm7");
&movdqa ("xmm7","xmm6");
&call ("_vpaes_schedule_low_round");
&movdqa ("xmm7",&QWP(20,"esp"));
&jmp (&label("loop_schedule_256"));
##
## .aes_schedule_mangle_last
##
## Mangler for last round of key schedule
## Mangles %xmm0
## when encrypting, outputs out(%xmm0) ^ 63
## when decrypting, outputs unskew(%xmm0)
##
## Always called right before return... jumps to cleanup and exits
##
&set_label("schedule_mangle_last",16);
# schedule last round key from xmm0
&lea ($base,&DWP($k_deskew,$const));
&test ($out,$out);
&jnz (&label("schedule_mangle_last_dec"));
# encrypting
&movdqa ("xmm1",&QWP($k_sr,$const,$magic));
&pshufb ("xmm0","xmm1"); # output permute
&lea ($base,&DWP($k_opt,$const)); # prepare to output transform
&add ($key,32);
&set_label("schedule_mangle_last_dec");
&add ($key,-16);
&pxor ("xmm0",&QWP($k_s63,$const));
&call ("_vpaes_schedule_transform"); # output transform
&movdqu (&QWP(0,$key),"xmm0"); # save last key
# cleanup
&pxor ("xmm0","xmm0");
&pxor ("xmm1","xmm1");
&pxor ("xmm2","xmm2");
&pxor ("xmm3","xmm3");
&pxor ("xmm4","xmm4");
&pxor ("xmm5","xmm5");
&pxor ("xmm6","xmm6");
&pxor ("xmm7","xmm7");
&ret ();
&function_end_B("_vpaes_schedule_core");
##
## .aes_schedule_192_smear
##
## Smear the short, low side in the 192-bit key schedule.
##
## Inputs:
## %xmm7: high side, b a x y
## %xmm6: low side, d c 0 0
## %xmm13: 0
##
## Outputs:
## %xmm6: b+c+d b+c 0 0
## %xmm0: b+c+d b+c b a
##
&function_begin_B("_vpaes_schedule_192_smear");
&pshufd ("xmm0","xmm6",0x80); # d c 0 0 -> c 0 0 0
&pxor ("xmm6","xmm0"); # -> c+d c 0 0
&pshufd ("xmm0","xmm7",0xFE); # b a _ _ -> b b b a
&pxor ("xmm6","xmm0"); # -> b+c+d b+c b a
&movdqa ("xmm0","xmm6");
&pxor ("xmm1","xmm1");
&movhlps("xmm6","xmm1"); # clobber low side with zeros
&ret ();
&function_end_B("_vpaes_schedule_192_smear");
##
## .aes_schedule_round
##
## Runs one main round of the key schedule on %xmm0, %xmm7
##
## Specifically, runs subbytes on the high dword of %xmm0
## then rotates it by one byte and xors into the low dword of
## %xmm7.
##
## Adds rcon from low byte of %xmm8, then rotates %xmm8 for
## next rcon.
##
## Smears the dwords of %xmm7 by xoring the low into the
## second low, result into third, result into highest.
##
## Returns results in %xmm7 = %xmm0.
## Clobbers %xmm1-%xmm5.
##
&function_begin_B("_vpaes_schedule_round");
# extract rcon from xmm8
&movdqa ("xmm2",&QWP(8,"esp")); # xmm8
&pxor ("xmm1","xmm1");
&palignr("xmm1","xmm2",15);
&palignr("xmm2","xmm2",15);
&pxor ("xmm7","xmm1");
# rotate
&pshufd ("xmm0","xmm0",0xFF);
&palignr("xmm0","xmm0",1);
# fall through...
&movdqa (&QWP(8,"esp"),"xmm2"); # xmm8
# low round: same as high round, but no rotation and no rcon.
&set_label("_vpaes_schedule_low_round");
# smear xmm7
&movdqa ("xmm1","xmm7");
&pslldq ("xmm7",4);
&pxor ("xmm7","xmm1");
&movdqa ("xmm1","xmm7");
&pslldq ("xmm7",8);
&pxor ("xmm7","xmm1");
&pxor ("xmm7",&QWP($k_s63,$const));
# subbyte
&movdqa ("xmm4",&QWP($k_s0F,$const));
&movdqa ("xmm5",&QWP($k_inv,$const)); # 4 : 1/j
&movdqa ("xmm1","xmm4");
&pandn ("xmm1","xmm0");
&psrld ("xmm1",4); # 1 = i
&pand ("xmm0","xmm4"); # 0 = k
&movdqa ("xmm2",&QWP($k_inv+16,$const));# 2 : a/k
&pshufb ("xmm2","xmm0"); # 2 = a/k
&pxor ("xmm0","xmm1"); # 0 = j
&movdqa ("xmm3","xmm5"); # 3 : 1/i
&pshufb ("xmm3","xmm1"); # 3 = 1/i
&pxor ("xmm3","xmm2"); # 3 = iak = 1/i + a/k
&movdqa ("xmm4","xmm5"); # 4 : 1/j
&pshufb ("xmm4","xmm0"); # 4 = 1/j
&pxor ("xmm4","xmm2"); # 4 = jak = 1/j + a/k
&movdqa ("xmm2","xmm5"); # 2 : 1/iak
&pshufb ("xmm2","xmm3"); # 2 = 1/iak
&pxor ("xmm2","xmm0"); # 2 = io
&movdqa ("xmm3","xmm5"); # 3 : 1/jak
&pshufb ("xmm3","xmm4"); # 3 = 1/jak
&pxor ("xmm3","xmm1"); # 3 = jo
&movdqa ("xmm4",&QWP($k_sb1,$const)); # 4 : sbou
&pshufb ("xmm4","xmm2"); # 4 = sbou
&movdqa ("xmm0",&QWP($k_sb1+16,$const));# 0 : sbot
&pshufb ("xmm0","xmm3"); # 0 = sb1t
&pxor ("xmm0","xmm4"); # 0 = sbox output
# add in smeared stuff
&pxor ("xmm0","xmm7");
&movdqa ("xmm7","xmm0");
&ret ();
&function_end_B("_vpaes_schedule_round");
##
## .aes_schedule_transform
##
## Linear-transform %xmm0 according to tables at (%ebx)
##
## Output in %xmm0
## Clobbers %xmm1, %xmm2
##
&function_begin_B("_vpaes_schedule_transform");
&movdqa ("xmm2",&QWP($k_s0F,$const));
&movdqa ("xmm1","xmm2");
&pandn ("xmm1","xmm0");
&psrld ("xmm1",4);
&pand ("xmm0","xmm2");
&movdqa ("xmm2",&QWP(0,$base));
&pshufb ("xmm2","xmm0");
&movdqa ("xmm0",&QWP(16,$base));
&pshufb ("xmm0","xmm1");
&pxor ("xmm0","xmm2");
&ret ();
&function_end_B("_vpaes_schedule_transform");
##
## .aes_schedule_mangle
##
## Mangle xmm0 from (basis-transformed) standard version
## to our version.
##
## On encrypt,
## xor with 0x63
## multiply by circulant 0,1,1,1
## apply shiftrows transform
##
## On decrypt,
## xor with 0x63
## multiply by "inverse mixcolumns" circulant E,B,D,9
## deskew
## apply shiftrows transform
##
##
## Writes out to (%edx), and increments or decrements it
## Keeps track of round number mod 4 in %ecx
## Preserves xmm0
## Clobbers xmm1-xmm5
##
&function_begin_B("_vpaes_schedule_mangle");
&movdqa ("xmm4","xmm0"); # save xmm0 for later
&movdqa ("xmm5",&QWP($k_mc_forward,$const));
&test ($out,$out);
&jnz (&label("schedule_mangle_dec"));
# encrypting
&add ($key,16);
&pxor ("xmm4",&QWP($k_s63,$const));
&pshufb ("xmm4","xmm5");
&movdqa ("xmm3","xmm4");
&pshufb ("xmm4","xmm5");
&pxor ("xmm3","xmm4");
&pshufb ("xmm4","xmm5");
&pxor ("xmm3","xmm4");
&jmp (&label("schedule_mangle_both"));
&set_label("schedule_mangle_dec",16);
# inverse mix columns
&movdqa ("xmm2",&QWP($k_s0F,$const));
&lea ($inp,&DWP($k_dksd,$const));
&movdqa ("xmm1","xmm2");
&pandn ("xmm1","xmm4");
&psrld ("xmm1",4); # 1 = hi
&pand ("xmm4","xmm2"); # 4 = lo
&movdqa ("xmm2",&QWP(0,$inp));
&pshufb ("xmm2","xmm4");
&movdqa ("xmm3",&QWP(0x10,$inp));
&pshufb ("xmm3","xmm1");
&pxor ("xmm3","xmm2");
&pshufb ("xmm3","xmm5");
&movdqa ("xmm2",&QWP(0x20,$inp));
&pshufb ("xmm2","xmm4");
&pxor ("xmm2","xmm3");
&movdqa ("xmm3",&QWP(0x30,$inp));
&pshufb ("xmm3","xmm1");
&pxor ("xmm3","xmm2");
&pshufb ("xmm3","xmm5");
&movdqa ("xmm2",&QWP(0x40,$inp));
&pshufb ("xmm2","xmm4");
&pxor ("xmm2","xmm3");
&movdqa ("xmm3",&QWP(0x50,$inp));
&pshufb ("xmm3","xmm1");
&pxor ("xmm3","xmm2");
&pshufb ("xmm3","xmm5");
&movdqa ("xmm2",&QWP(0x60,$inp));
&pshufb ("xmm2","xmm4");
&pxor ("xmm2","xmm3");
&movdqa ("xmm3",&QWP(0x70,$inp));
&pshufb ("xmm3","xmm1");
&pxor ("xmm3","xmm2");
&add ($key,-16);
&set_label("schedule_mangle_both");
&movdqa ("xmm1",&QWP($k_sr,$const,$magic));
&pshufb ("xmm3","xmm1");
&add ($magic,-16);
&and ($magic,0x30);
&movdqu (&QWP(0,$key),"xmm3");
&ret ();
&function_end_B("_vpaes_schedule_mangle");
#
# Interface to OpenSSL
#
&function_begin("${PREFIX}_set_encrypt_key");
&mov ($inp,&wparam(0)); # inp
&lea ($base,&DWP(-56,"esp"));
&mov ($round,&wparam(1)); # bits
&and ($base,-16);
&mov ($key,&wparam(2)); # key
&xchg ($base,"esp"); # alloca
&mov (&DWP(48,"esp"),$base);
&mov ($base,$round);
&shr ($base,5);
&add ($base,5);
&mov (&DWP(240,$key),$base); # AES_KEY->rounds = nbits/32+5;
&mov ($magic,0x30);
&mov ($out,0);
&lea ($const,&DWP(&label("_vpaes_consts")."+0x30-".&label("pic_point")));
&call ("_vpaes_schedule_core");
&set_label("pic_point");
&mov ("esp",&DWP(48,"esp"));
&xor ("eax","eax");
&function_end("${PREFIX}_set_encrypt_key");
&function_begin("${PREFIX}_set_decrypt_key");
&mov ($inp,&wparam(0)); # inp
&lea ($base,&DWP(-56,"esp"));
&mov ($round,&wparam(1)); # bits
&and ($base,-16);
&mov ($key,&wparam(2)); # key
&xchg ($base,"esp"); # alloca
&mov (&DWP(48,"esp"),$base);
&mov ($base,$round);
&shr ($base,5);
&add ($base,5);
&mov (&DWP(240,$key),$base); # AES_KEY->rounds = nbits/32+5;
&shl ($base,4);
&lea ($key,&DWP(16,$key,$base));
&mov ($out,1);
&mov ($magic,$round);
&shr ($magic,1);
&and ($magic,32);
&xor ($magic,32); # nbist==192?0:32;
&lea ($const,&DWP(&label("_vpaes_consts")."+0x30-".&label("pic_point")));
&call ("_vpaes_schedule_core");
&set_label("pic_point");
&mov ("esp",&DWP(48,"esp"));
&xor ("eax","eax");
&function_end("${PREFIX}_set_decrypt_key");
&function_begin("${PREFIX}_encrypt");
&lea ($const,&DWP(&label("_vpaes_consts")."+0x30-".&label("pic_point")));
&call ("_vpaes_preheat");
&set_label("pic_point");
&mov ($inp,&wparam(0)); # inp
&lea ($base,&DWP(-56,"esp"));
&mov ($out,&wparam(1)); # out
&and ($base,-16);
&mov ($key,&wparam(2)); # key
&xchg ($base,"esp"); # alloca
&mov (&DWP(48,"esp"),$base);
&movdqu ("xmm0",&QWP(0,$inp));
&call ("_vpaes_encrypt_core");
&movdqu (&QWP(0,$out),"xmm0");
&mov ("esp",&DWP(48,"esp"));
&function_end("${PREFIX}_encrypt");
&function_begin("${PREFIX}_decrypt");
&lea ($const,&DWP(&label("_vpaes_consts")."+0x30-".&label("pic_point")));
&call ("_vpaes_preheat");
&set_label("pic_point");
&mov ($inp,&wparam(0)); # inp
&lea ($base,&DWP(-56,"esp"));
&mov ($out,&wparam(1)); # out
&and ($base,-16);
&mov ($key,&wparam(2)); # key
&xchg ($base,"esp"); # alloca
&mov (&DWP(48,"esp"),$base);
&movdqu ("xmm0",&QWP(0,$inp));
&call ("_vpaes_decrypt_core");
&movdqu (&QWP(0,$out),"xmm0");
&mov ("esp",&DWP(48,"esp"));
&function_end("${PREFIX}_decrypt");
&function_begin("${PREFIX}_cbc_encrypt");
&mov ($inp,&wparam(0)); # inp
&mov ($out,&wparam(1)); # out
&mov ($round,&wparam(2)); # len
&mov ($key,&wparam(3)); # key
&sub ($round,16);
&jc (&label("cbc_abort"));
&lea ($base,&DWP(-56,"esp"));
&mov ($const,&wparam(4)); # ivp
&and ($base,-16);
&mov ($magic,&wparam(5)); # enc
&xchg ($base,"esp"); # alloca
&movdqu ("xmm1",&QWP(0,$const)); # load IV
&sub ($out,$inp);
&mov (&DWP(48,"esp"),$base);
&mov (&DWP(0,"esp"),$out); # save out
&mov (&DWP(4,"esp"),$key) # save key
&mov (&DWP(8,"esp"),$const); # save ivp
&mov ($out,$round); # $out works as $len
&lea ($const,&DWP(&label("_vpaes_consts")."+0x30-".&label("pic_point")));
&call ("_vpaes_preheat");
&set_label("pic_point");
&cmp ($magic,0);
&je (&label("cbc_dec_loop"));
&jmp (&label("cbc_enc_loop"));
&set_label("cbc_enc_loop",16);
&movdqu ("xmm0",&QWP(0,$inp)); # load input
&pxor ("xmm0","xmm1"); # inp^=iv
&call ("_vpaes_encrypt_core");
&mov ($base,&DWP(0,"esp")); # restore out
&mov ($key,&DWP(4,"esp")); # restore key
&movdqa ("xmm1","xmm0");
&movdqu (&QWP(0,$base,$inp),"xmm0"); # write output
&lea ($inp,&DWP(16,$inp));
&sub ($out,16);
&jnc (&label("cbc_enc_loop"));
&jmp (&label("cbc_done"));
&set_label("cbc_dec_loop",16);
&movdqu ("xmm0",&QWP(0,$inp)); # load input
&movdqa (&QWP(16,"esp"),"xmm1"); # save IV
&movdqa (&QWP(32,"esp"),"xmm0"); # save future IV
&call ("_vpaes_decrypt_core");
&mov ($base,&DWP(0,"esp")); # restore out
&mov ($key,&DWP(4,"esp")); # restore key
&pxor ("xmm0",&QWP(16,"esp")); # out^=iv
&movdqa ("xmm1",&QWP(32,"esp")); # load next IV
&movdqu (&QWP(0,$base,$inp),"xmm0"); # write output
&lea ($inp,&DWP(16,$inp));
&sub ($out,16);
&jnc (&label("cbc_dec_loop"));
&set_label("cbc_done");
&mov ($base,&DWP(8,"esp")); # restore ivp
&mov ("esp",&DWP(48,"esp"));
&movdqu (&QWP(0,$base),"xmm1"); # write IV
&set_label("cbc_abort");
&function_end("${PREFIX}_cbc_encrypt");
&asm_finish();

1207
drivers/builtin_openssl2/crypto/aes/asm/vpaes-x86_64.pl
View File

File diff suppressed because it is too large Load Diff

126
drivers/builtin_openssl2/crypto/alphacpuid.pl
View File

@ -1,126 +0,0 @@
#!/usr/bin/env perl
print <<'___';
.text
.set noat
.globl OPENSSL_cpuid_setup
.ent OPENSSL_cpuid_setup
OPENSSL_cpuid_setup:
.frame $30,0,$26
.prologue 0
ret ($26)
.end OPENSSL_cpuid_setup
.globl OPENSSL_wipe_cpu
.ent OPENSSL_wipe_cpu
OPENSSL_wipe_cpu:
.frame $30,0,$26
.prologue 0
clr $1
clr $2
clr $3
clr $4
clr $5
clr $6
clr $7
clr $8
clr $16
clr $17
clr $18
clr $19
clr $20
clr $21
clr $22
clr $23
clr $24
clr $25
clr $27
clr $at
clr $29
fclr $f0
fclr $f1
fclr $f10
fclr $f11
fclr $f12
fclr $f13
fclr $f14
fclr $f15
fclr $f16
fclr $f17
fclr $f18
fclr $f19
fclr $f20
fclr $f21
fclr $f22
fclr $f23
fclr $f24
fclr $f25
fclr $f26
fclr $f27
fclr $f28
fclr $f29
fclr $f30
mov $sp,$0
ret ($26)
.end OPENSSL_wipe_cpu
.globl OPENSSL_atomic_add
.ent OPENSSL_atomic_add
OPENSSL_atomic_add:
.frame $30,0,$26
.prologue 0
1: ldl_l $0,0($16)
addl $0,$17,$1
stl_c $1,0($16)
beq $1,1b
addl $0,$17,$0
ret ($26)
.end OPENSSL_atomic_add
.globl OPENSSL_rdtsc
.ent OPENSSL_rdtsc
OPENSSL_rdtsc:
.frame $30,0,$26
.prologue 0
rpcc $0
ret ($26)
.end OPENSSL_rdtsc
.globl OPENSSL_cleanse
.ent OPENSSL_cleanse
OPENSSL_cleanse:
.frame $30,0,$26
.prologue 0
beq $17,.Ldone
and $16,7,$0
bic $17,7,$at
beq $at,.Little
beq $0,.Laligned
.Little:
subq $0,8,$0
ldq_u $1,0($16)
mov $16,$2
.Lalign:
mskbl $1,$16,$1
lda $16,1($16)
subq $17,1,$17
addq $0,1,$0
beq $17,.Lout
bne $0,.Lalign
.Lout: stq_u $1,0($2)
beq $17,.Ldone
bic $17,7,$at
beq $at,.Little
.Laligned:
stq $31,0($16)
subq $17,8,$17
lda $16,8($16)
bic $17,7,$at
bne $at,.Laligned
bne $17,.Little
.Ldone: ret ($26)
.end OPENSSL_cleanse
___

39
drivers/builtin_openssl2/crypto/arm_arch.h
View File

@ -10,13 +10,24 @@
# define __ARMEL__
# endif
# elif defined(__GNUC__)
# if defined(__aarch64__)
# define __ARM_ARCH__ 8
# if __BYTE_ORDER__==__ORDER_BIG_ENDIAN__
# define __ARMEB__
# else
# define __ARMEL__
# endif
/*
* Why doesn't gcc define __ARM_ARCH__? Instead it defines
* bunch of below macros. See all_architectires[] table in
* gcc/config/arm/arm.c. On a side note it defines
* __ARMEL__/__ARMEB__ for little-/big-endian.
*/
# if defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || \
# elif defined(__ARM_ARCH)
# define __ARM_ARCH__ __ARM_ARCH
# elif defined(__ARM_ARCH_8A__)
# define __ARM_ARCH__ 8
# elif defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || \
defined(__ARM_ARCH_7R__)|| defined(__ARM_ARCH_7M__) || \
defined(__ARM_ARCH_7EM__)
# define __ARM_ARCH__ 7
@ -41,11 +52,27 @@
# include <openssl/fipssyms.h>
# endif
# if !__ASSEMBLER__
extern unsigned int OPENSSL_armcap_P;
# define ARMV7_NEON (1<<0)
# define ARMV7_TICK (1<<1)
# if !defined(__ARM_MAX_ARCH__)
# define __ARM_MAX_ARCH__ __ARM_ARCH__
# endif
# if __ARM_MAX_ARCH__<__ARM_ARCH__
# error "__ARM_MAX_ARCH__ can't be less than __ARM_ARCH__"
# elif __ARM_MAX_ARCH__!=__ARM_ARCH__
# if __ARM_ARCH__<7 && __ARM_MAX_ARCH__>=7 && defined(__ARMEB__)
# error "can't build universal big-endian binary"
# endif
# endif
# if !__ASSEMBLER__
extern unsigned int OPENSSL_armcap_P;
# endif
# define ARMV7_NEON (1<<0)
# define ARMV7_TICK (1<<1)
# define ARMV8_AES (1<<2)
# define ARMV8_SHA1 (1<<3)
# define ARMV8_SHA256 (1<<4)
# define ARMV8_PMULL (1<<5)
#endif

97
drivers/builtin_openssl2/crypto/armcap.c
View File

@ -7,8 +7,18 @@
#include "arm_arch.h"
unsigned int OPENSSL_armcap_P;
unsigned int OPENSSL_armcap_P = 0;
#if __ARM_MAX_ARCH__<7
void OPENSSL_cpuid_setup(void)
{
}
unsigned long OPENSSL_rdtsc(void)
{
return 0;
}
#else
static sigset_t all_masked;
static sigjmp_buf ill_jmp;
@ -22,9 +32,13 @@ static void ill_handler(int sig)
* ARM compilers support inline assembler...
*/
void _armv7_neon_probe(void);
unsigned int _armv7_tick(void);
void _armv8_aes_probe(void);
void _armv8_sha1_probe(void);
void _armv8_sha256_probe(void);
void _armv8_pmull_probe(void);
unsigned long _armv7_tick(void);
unsigned int OPENSSL_rdtsc(void)
unsigned long OPENSSL_rdtsc(void)
{
if (OPENSSL_armcap_P & ARMV7_TICK)
return _armv7_tick();
@ -32,9 +46,44 @@ unsigned int OPENSSL_rdtsc(void)
return 0;
}
#if defined(__GNUC__) && __GNUC__>=2
/*
* Use a weak reference to getauxval() so we can use it if it is available but
* don't break the build if it is not.
*/
# if defined(__GNUC__) && __GNUC__>=2
void OPENSSL_cpuid_setup(void) __attribute__ ((constructor));
#endif
extern unsigned long getauxval(unsigned long type) __attribute__ ((weak));
# else
static unsigned long (*getauxval) (unsigned long) = NULL;
# endif
/*
* ARM puts the the feature bits for Crypto Extensions in AT_HWCAP2, whereas
* AArch64 used AT_HWCAP.
*/
# if defined(__arm__) || defined (__arm)
# define HWCAP 16
/* AT_HWCAP */
# define HWCAP_NEON (1 << 12)
# define HWCAP_CE 26
/* AT_HWCAP2 */
# define HWCAP_CE_AES (1 << 0)
# define HWCAP_CE_PMULL (1 << 1)
# define HWCAP_CE_SHA1 (1 << 2)
# define HWCAP_CE_SHA256 (1 << 3)
# elif defined(__aarch64__)
# define HWCAP 16
/* AT_HWCAP */
# define HWCAP_NEON (1 << 1)
# define HWCAP_CE HWCAP
# define HWCAP_CE_AES (1 << 3)
# define HWCAP_CE_PMULL (1 << 4)
# define HWCAP_CE_SHA1 (1 << 5)
# define HWCAP_CE_SHA256 (1 << 6)
# endif
void OPENSSL_cpuid_setup(void)
{
char *e;
@ -47,7 +96,7 @@ void OPENSSL_cpuid_setup(void)
trigger = 1;
if ((e = getenv("OPENSSL_armcap"))) {
OPENSSL_armcap_P = strtoul(e, NULL, 0);
OPENSSL_armcap_P = (unsigned int)strtoul(e, NULL, 0);
return;
}
@ -67,9 +116,42 @@ void OPENSSL_cpuid_setup(void)
sigprocmask(SIG_SETMASK, &ill_act.sa_mask, &oset);
sigaction(SIGILL, &ill_act, &ill_oact);
if (sigsetjmp(ill_jmp, 1) == 0) {
if (getauxval != NULL) {
if (getauxval(HWCAP) & HWCAP_NEON) {
unsigned long hwcap = getauxval(HWCAP_CE);
OPENSSL_armcap_P |= ARMV7_NEON;
if (hwcap & HWCAP_CE_AES)
OPENSSL_armcap_P |= ARMV8_AES;
if (hwcap & HWCAP_CE_PMULL)
OPENSSL_armcap_P |= ARMV8_PMULL;
if (hwcap & HWCAP_CE_SHA1)
OPENSSL_armcap_P |= ARMV8_SHA1;
if (hwcap & HWCAP_CE_SHA256)
OPENSSL_armcap_P |= ARMV8_SHA256;
}
} else if (sigsetjmp(ill_jmp, 1) == 0) {
_armv7_neon_probe();
OPENSSL_armcap_P |= ARMV7_NEON;
if (sigsetjmp(ill_jmp, 1) == 0) {
_armv8_pmull_probe();
OPENSSL_armcap_P |= ARMV8_PMULL | ARMV8_AES;
} else if (sigsetjmp(ill_jmp, 1) == 0) {
_armv8_aes_probe();
OPENSSL_armcap_P |= ARMV8_AES;
}
if (sigsetjmp(ill_jmp, 1) == 0) {
_armv8_sha1_probe();
OPENSSL_armcap_P |= ARMV8_SHA1;
}
if (sigsetjmp(ill_jmp, 1) == 0) {
_armv8_sha256_probe();
OPENSSL_armcap_P |= ARMV8_SHA256;
}
}
if (sigsetjmp(ill_jmp, 1) == 0) {
_armv7_tick();
@ -79,3 +161,4 @@ void OPENSSL_cpuid_setup(void)
sigaction(SIGILL, &ill_oact, NULL);
sigprocmask(SIG_SETMASK, &oset, NULL);
}
#endif

154
drivers/builtin_openssl2/crypto/armv4cpuid.S
View File

@ -1,154 +0,0 @@
#include "arm_arch.h"
.text
.code 32
.align 5
.global _armv7_neon_probe
.type _armv7_neon_probe,%function
_armv7_neon_probe:
.word 0xf26ee1fe @ vorr q15,q15,q15
.word 0xe12fff1e @ bx lr
.size _armv7_neon_probe,.-_armv7_neon_probe
.global _armv7_tick
.type _armv7_tick,%function
_armv7_tick:
mrc p15,0,r0,c9,c13,0
.word 0xe12fff1e @ bx lr
.size _armv7_tick,.-_armv7_tick
.global OPENSSL_atomic_add
.type OPENSSL_atomic_add,%function
OPENSSL_atomic_add:
#if __ARM_ARCH__>=6
.Ladd: ldrex r2,[r0]
add r3,r2,r1
strex r2,r3,[r0]
cmp r2,#0
bne .Ladd
mov r0,r3
.word 0xe12fff1e @ bx lr
#else
stmdb sp!,{r4-r6,lr}
ldr r2,.Lspinlock
adr r3,.Lspinlock
mov r4,r0
mov r5,r1
add r6,r3,r2 @ &spinlock
b .+8
.Lspin: bl sched_yield
mov r0,#-1
swp r0,r0,[r6]
cmp r0,#0
bne .Lspin
ldr r2,[r4]
add r2,r2,r5
str r2,[r4]
str r0,[r6] @ release spinlock
ldmia sp!,{r4-r6,lr}
tst lr,#1
moveq pc,lr
.word 0xe12fff1e @ bx lr
#endif
.size OPENSSL_atomic_add,.-OPENSSL_atomic_add
.global OPENSSL_cleanse
.type OPENSSL_cleanse,%function
OPENSSL_cleanse:
eor ip,ip,ip
cmp r1,#7
subhs r1,r1,#4
bhs .Lot
cmp r1,#0
beq .Lcleanse_done
.Little:
strb ip,[r0],#1
subs r1,r1,#1
bhi .Little
b .Lcleanse_done
.Lot: tst r0,#3
beq .Laligned
strb ip,[r0],#1
sub r1,r1,#1
b .Lot
.Laligned:
str ip,[r0],#4
subs r1,r1,#4
bhs .Laligned
adds r1,r1,#4
bne .Little
.Lcleanse_done:
tst lr,#1
moveq pc,lr
.word 0xe12fff1e @ bx lr
.size OPENSSL_cleanse,.-OPENSSL_cleanse
.global OPENSSL_wipe_cpu
.type OPENSSL_wipe_cpu,%function
OPENSSL_wipe_cpu:
ldr r0,.LOPENSSL_armcap
adr r1,.LOPENSSL_armcap
ldr r0,[r1,r0]
eor r2,r2,r2
eor r3,r3,r3
eor ip,ip,ip
tst r0,#1
beq .Lwipe_done
.word 0xf3000150 @ veor q0, q0, q0
.word 0xf3022152 @ veor q1, q1, q1
.word 0xf3044154 @ veor q2, q2, q2
.word 0xf3066156 @ veor q3, q3, q3
.word 0xf34001f0 @ veor q8, q8, q8
.word 0xf34221f2 @ veor q9, q9, q9
.word 0xf34441f4 @ veor q10, q10, q10
.word 0xf34661f6 @ veor q11, q11, q11
.word 0xf34881f8 @ veor q12, q12, q12
.word 0xf34aa1fa @ veor q13, q13, q13
.word 0xf34cc1fc @ veor q14, q14, q14
.word 0xf34ee1fe @ veor q15, q15, q15
.Lwipe_done:
mov r0,sp
tst lr,#1
moveq pc,lr
.word 0xe12fff1e @ bx lr
.size OPENSSL_wipe_cpu,.-OPENSSL_wipe_cpu
.global OPENSSL_instrument_bus
.type OPENSSL_instrument_bus,%function
OPENSSL_instrument_bus:
eor r0,r0,r0
tst lr,#1
moveq pc,lr
.word 0xe12fff1e @ bx lr
.size OPENSSL_instrument_bus,.-OPENSSL_instrument_bus
.global OPENSSL_instrument_bus2
.type OPENSSL_instrument_bus2,%function
OPENSSL_instrument_bus2:
eor r0,r0,r0
tst lr,#1
moveq pc,lr
.word 0xe12fff1e @ bx lr
.size OPENSSL_instrument_bus2,.-OPENSSL_instrument_bus2
.align 5
.LOPENSSL_armcap:
.word OPENSSL_armcap_P-.LOPENSSL_armcap
#if __ARM_ARCH__>=6
.align 5
#else
.Lspinlock:
.word atomic_add_spinlock-.Lspinlock
.align 5
.data
.align 2
atomic_add_spinlock:
.word 0
#endif
.comm OPENSSL_armcap_P,4,4
.hidden OPENSSL_armcap_P

46
drivers/builtin_openssl2/crypto/asn1/a_gentm.c
View File

@ -65,6 +65,7 @@
#include "cryptlib.h"
#include "o_time.h"
#include <openssl/asn1.h>
#include "asn1_locl.h"
#if 0
@ -117,7 +118,7 @@ ASN1_GENERALIZEDTIME *d2i_ASN1_GENERALIZEDTIME(ASN1_GENERALIZEDTIME **a,
#endif
int ASN1_GENERALIZEDTIME_check(ASN1_GENERALIZEDTIME *d)
int asn1_generalizedtime_to_tm(struct tm *tm, const ASN1_GENERALIZEDTIME *d)
{
static const int min[9] = { 0, 0, 1, 1, 0, 0, 0, 0, 0 };
static const int max[9] = { 99, 99, 12, 31, 23, 59, 59, 12, 59 };
@ -139,6 +140,8 @@ int ASN1_GENERALIZEDTIME_check(ASN1_GENERALIZEDTIME *d)
for (i = 0; i < 7; i++) {
if ((i == 6) && ((a[o] == 'Z') || (a[o] == '+') || (a[o] == '-'))) {
i++;
if (tm)
tm->tm_sec = 0;
break;
}
if ((a[o] < '0') || (a[o] > '9'))
@ -155,6 +158,31 @@ int ASN1_GENERALIZEDTIME_check(ASN1_GENERALIZEDTIME *d)
if ((n < min[i]) || (n > max[i]))
goto err;
if (tm) {
switch (i) {
case 0:
tm->tm_year = n * 100 - 1900;
break;
case 1:
tm->tm_year += n;
break;
case 2:
tm->tm_mon = n - 1;
break;
case 3:
tm->tm_mday = n;
break;
case 4:
tm->tm_hour = n;
break;
case 5:
tm->tm_min = n;
break;
case 6:
tm->tm_sec = n;
break;
}
}
}
/*
* Optional fractional seconds: decimal point followed by one or more
@ -174,6 +202,7 @@ int ASN1_GENERALIZEDTIME_check(ASN1_GENERALIZEDTIME *d)
if (a[o] == 'Z')
o++;
else if ((a[o] == '+') || (a[o] == '-')) {
int offsign = a[o] == '-' ? -1 : 1, offset = 0;
o++;
if (o + 4 > l)
goto err;
@ -187,9 +216,17 @@ int ASN1_GENERALIZEDTIME_check(ASN1_GENERALIZEDTIME *d)
n = (n * 10) + a[o] - '0';
if ((n < min[i]) || (n > max[i]))
goto err;
if (tm) {
if (i == 7)
offset = n * 3600;
else if (i == 8)
offset += n * 60;
}
o++;
}
} else {
if (offset && !OPENSSL_gmtime_adj(tm, 0, offset * offsign))
return 0;
} else if (a[o]) {
/* Missing time zone information. */
goto err;
}
@ -198,6 +235,11 @@ int ASN1_GENERALIZEDTIME_check(ASN1_GENERALIZEDTIME *d)
return (0);
}
int ASN1_GENERALIZEDTIME_check(const ASN1_GENERALIZEDTIME *d)
{
return asn1_generalizedtime_to_tm(NULL, d);
}
int ASN1_GENERALIZEDTIME_set_string(ASN1_GENERALIZEDTIME *s, const char *str)
{
ASN1_GENERALIZEDTIME t;

30
drivers/builtin_openssl2/crypto/asn1/a_time.c
View File

@ -66,6 +66,7 @@
#include "cryptlib.h"
#include "o_time.h"
#include <openssl/asn1t.h>
#include "asn1_locl.h"
IMPLEMENT_ASN1_MSTRING(ASN1_TIME, B_ASN1_TIME)
@ -196,3 +197,32 @@ int ASN1_TIME_set_string(ASN1_TIME *s, const char *str)
return 1;
}
static int asn1_time_to_tm(struct tm *tm, const ASN1_TIME *t)
{
if (t == NULL) {
time_t now_t;
time(&now_t);
if (OPENSSL_gmtime(&now_t, tm))
return 1;
return 0;
}
if (t->type == V_ASN1_UTCTIME)
return asn1_utctime_to_tm(tm, t);
else if (t->type == V_ASN1_GENERALIZEDTIME)
return asn1_generalizedtime_to_tm(tm, t);
return 0;
}
int ASN1_TIME_diff(int *pday, int *psec,
const ASN1_TIME *from, const ASN1_TIME *to)
{
struct tm tm_from, tm_to;
if (!asn1_time_to_tm(&tm_from, from))
return 0;
if (!asn1_time_to_tm(&tm_to, to))
return 0;
return OPENSSL_gmtime_diff(pday, psec, &tm_from, &tm_to);
}

94
drivers/builtin_openssl2/crypto/asn1/a_utctm.c
View File

@ -61,6 +61,7 @@
#include "cryptlib.h"
#include "o_time.h"
#include <openssl/asn1.h>
#include "asn1_locl.h"
#if 0
int i2d_ASN1_UTCTIME(ASN1_UTCTIME *a, unsigned char **pp)
@ -109,7 +110,7 @@ ASN1_UTCTIME *d2i_ASN1_UTCTIME(ASN1_UTCTIME **a, unsigned char **pp,
#endif
int ASN1_UTCTIME_check(ASN1_UTCTIME *d)
int asn1_utctime_to_tm(struct tm *tm, const ASN1_UTCTIME *d)
{
static const int min[8] = { 0, 1, 1, 0, 0, 0, 0, 0 };
static const int max[8] = { 99, 12, 31, 23, 59, 59, 12, 59 };
@ -127,6 +128,8 @@ int ASN1_UTCTIME_check(ASN1_UTCTIME *d)
for (i = 0; i < 6; i++) {
if ((i == 5) && ((a[o] == 'Z') || (a[o] == '+') || (a[o] == '-'))) {
i++;
if (tm)
tm->tm_sec = 0;
break;
}
if ((a[o] < '0') || (a[o] > '9'))
@ -143,10 +146,33 @@ int ASN1_UTCTIME_check(ASN1_UTCTIME *d)
if ((n < min[i]) || (n > max[i]))
goto err;
if (tm) {
switch (i) {
case 0:
tm->tm_year = n < 50 ? n + 100 : n;
break;
case 1:
tm->tm_mon = n - 1;
break;
case 2:
tm->tm_mday = n;
break;
case 3:
tm->tm_hour = n;
break;
case 4:
tm->tm_min = n;
break;
case 5:
tm->tm_sec = n;
break;
}
}
}
if (a[o] == 'Z')
o++;
else if ((a[o] == '+') || (a[o] == '-')) {
int offsign = a[o] == '-' ? -1 : 1, offset = 0;
o++;
if (o + 4 > l)
goto err;
@ -160,12 +186,25 @@ int ASN1_UTCTIME_check(ASN1_UTCTIME *d)
n = (n * 10) + a[o] - '0';
if ((n < min[i]) || (n > max[i]))
goto err;
if (tm) {
if (i == 6)
offset = n * 3600;
else if (i == 7)
offset += n * 60;
}
o++;
}
if (offset && !OPENSSL_gmtime_adj(tm, 0, offset * offsign))
return 0;
}
return (o == l);
return o == l;
err:
return (0);
return 0;
}
int ASN1_UTCTIME_check(const ASN1_UTCTIME *d)
{
return asn1_utctime_to_tm(NULL, d);
}
int ASN1_UTCTIME_set_string(ASN1_UTCTIME *s, const char *str)
@ -249,43 +288,26 @@ ASN1_UTCTIME *ASN1_UTCTIME_adj(ASN1_UTCTIME *s, time_t t,
int ASN1_UTCTIME_cmp_time_t(const ASN1_UTCTIME *s, time_t t)
{
struct tm *tm;
struct tm data;
int offset;
int year;
struct tm stm, ttm;
int day, sec;
#define g2(p) (((p)[0]-'0')*10+(p)[1]-'0')
if (s->data[12] == 'Z')
offset = 0;
else {
offset = g2(s->data + 13) * 60 + g2(s->data + 15);
if (s->data[12] == '-')
offset = -offset;
}
t -= offset * 60; /* FIXME: may overflow in extreme cases */
tm = OPENSSL_gmtime(&t, &data);
/*
* NB: -1, 0, 1 already valid return values so use -2 to indicate error.
*/
if (tm == NULL)
if (!asn1_utctime_to_tm(&stm, s))
return -2;
#define return_cmp(a,b) if ((a)<(b)) return -1; else if ((a)>(b)) return 1
year = g2(s->data);
if (year < 50)
year += 100;
return_cmp(year, tm->tm_year);
return_cmp(g2(s->data + 2) - 1, tm->tm_mon);
return_cmp(g2(s->data + 4), tm->tm_mday);
return_cmp(g2(s->data + 6), tm->tm_hour);
return_cmp(g2(s->data + 8), tm->tm_min);
return_cmp(g2(s->data + 10), tm->tm_sec);
#undef g2
#undef return_cmp
if (!OPENSSL_gmtime(&t, &ttm))
return -2;
if (!OPENSSL_gmtime_diff(&day, &sec, &ttm, &stm))
return -2;
if (day > 0)
return 1;
if (day < 0)
return -1;
if (sec > 0)
return 1;
if (sec < 0)
return -1;
return 0;
}

24
drivers/builtin_openssl2/crypto/asn1/ameth_lib.c
View File

@ -68,6 +68,7 @@
extern const EVP_PKEY_ASN1_METHOD rsa_asn1_meths[];
extern const EVP_PKEY_ASN1_METHOD dsa_asn1_meths[];
extern const EVP_PKEY_ASN1_METHOD dh_asn1_meth;
extern const EVP_PKEY_ASN1_METHOD dhx_asn1_meth;
extern const EVP_PKEY_ASN1_METHOD eckey_asn1_meth;
extern const EVP_PKEY_ASN1_METHOD hmac_asn1_meth;
extern const EVP_PKEY_ASN1_METHOD cmac_asn1_meth;
@ -92,7 +93,10 @@ static const EVP_PKEY_ASN1_METHOD *standard_methods[] = {
&eckey_asn1_meth,
#endif
&hmac_asn1_meth,
&cmac_asn1_meth
&cmac_asn1_meth,
#ifndef OPENSSL_NO_DH
&dhx_asn1_meth
#endif
};
typedef int sk_cmp_fn_type(const char *const *a, const char *const *b);
@ -460,3 +464,21 @@ void EVP_PKEY_asn1_set_ctrl(EVP_PKEY_ASN1_METHOD *ameth,
{
ameth->pkey_ctrl = pkey_ctrl;
}
void EVP_PKEY_asn1_set_item(EVP_PKEY_ASN1_METHOD *ameth,
int (*item_verify) (EVP_MD_CTX *ctx,
const ASN1_ITEM *it,
void *asn,
X509_ALGOR *a,
ASN1_BIT_STRING *sig,
EVP_PKEY *pkey),
int (*item_sign) (EVP_MD_CTX *ctx,
const ASN1_ITEM *it,
void *asn,
X509_ALGOR *alg1,
X509_ALGOR *alg2,
ASN1_BIT_STRING *sig))
{
ameth->item_sign = item_sign;
ameth->item_verify = item_verify;
}

3
drivers/builtin_openssl2/crypto/asn1/asn1_locl.h
View File

@ -59,6 +59,9 @@
/* Internal ASN1 structures and functions: not for application use */
int asn1_utctime_to_tm(struct tm *tm, const ASN1_UTCTIME *d);
int asn1_generalizedtime_to_tm(struct tm *tm, const ASN1_GENERALIZEDTIME *d);
/* ASN1 print context structure */
struct asn1_pctx_st {

83
drivers/builtin_openssl2/crypto/asn1/charmap.pl
View File

@ -1,83 +0,0 @@
#!/usr/local/bin/perl -w
# Written by Dr Stephen N Henson (steve@openssl.org).
# Licensed under the terms of the OpenSSL license.
use strict;
my ($i, @arr);
# Set up an array with the type of ASCII characters
# Each set bit represents a character property.
# RFC2253 character properties
my $RFC2253_ESC = 1; # Character escaped with \
my $ESC_CTRL = 2; # Escaped control character
# These are used with RFC1779 quoting using "
my $NOESC_QUOTE = 8; # Not escaped if quoted
my $PSTRING_CHAR = 0x10; # Valid PrintableString character
my $RFC2253_FIRST_ESC = 0x20; # Escaped with \ if first character
my $RFC2253_LAST_ESC = 0x40; # Escaped with \ if last character
for($i = 0; $i < 128; $i++) {
# Set the RFC2253 escape characters (control)
$arr[$i] = 0;
if(($i < 32) || ($i > 126)) {
$arr[$i] |= $ESC_CTRL;
}
# Some PrintableString characters
if( ( ( $i >= ord("a")) && ( $i <= ord("z")) )
|| ( ( $i >= ord("A")) && ( $i <= ord("Z")) )
|| ( ( $i >= ord("0")) && ( $i <= ord("9")) ) ) {
$arr[$i] |= $PSTRING_CHAR;
}
}
# Now setup the rest
# Remaining RFC2253 escaped characters
$arr[ord(" ")] |= $NOESC_QUOTE | $RFC2253_FIRST_ESC | $RFC2253_LAST_ESC;
$arr[ord("#")] |= $NOESC_QUOTE | $RFC2253_FIRST_ESC;
$arr[ord(",")] |= $NOESC_QUOTE | $RFC2253_ESC;
$arr[ord("+")] |= $NOESC_QUOTE | $RFC2253_ESC;
$arr[ord("\"")] |= $RFC2253_ESC;
$arr[ord("\\")] |= $RFC2253_ESC;
$arr[ord("<")] |= $NOESC_QUOTE | $RFC2253_ESC;
$arr[ord(">")] |= $NOESC_QUOTE | $RFC2253_ESC;
$arr[ord(";")] |= $NOESC_QUOTE | $RFC2253_ESC;
# Remaining PrintableString characters
$arr[ord(" ")] |= $PSTRING_CHAR;
$arr[ord("'")] |= $PSTRING_CHAR;
$arr[ord("(")] |= $PSTRING_CHAR;
$arr[ord(")")] |= $PSTRING_CHAR;
$arr[ord("+")] |= $PSTRING_CHAR;
$arr[ord(",")] |= $PSTRING_CHAR;
$arr[ord("-")] |= $PSTRING_CHAR;
$arr[ord(".")] |= $PSTRING_CHAR;
$arr[ord("/")] |= $PSTRING_CHAR;
$arr[ord(":")] |= $PSTRING_CHAR;
$arr[ord("=")] |= $PSTRING_CHAR;
$arr[ord("?")] |= $PSTRING_CHAR;
# Now generate the C code
print <<EOF;
/* Auto generated with chartype.pl script.
* Mask of various character properties
*/
static unsigned char char_type[] = {
EOF
for($i = 0; $i < 128; $i++) {
print("\n") if($i && (($i % 16) == 0));
printf("%2d", $arr[$i]);
print(",") if ($i != 127);
}
print("\n};\n\n");

15
drivers/builtin_openssl2/crypto/asn1/t_x509.c
View File

@ -228,6 +228,21 @@ int X509_print_ex(BIO *bp, X509 *x, unsigned long nmflags,
}
}
if (!(cflag & X509_FLAG_NO_IDS)) {
if (ci->issuerUID) {
if (BIO_printf(bp, "%8sIssuer Unique ID: ", "") <= 0)
goto err;
if (!X509_signature_dump(bp, ci->issuerUID, 12))
goto err;
}
if (ci->subjectUID) {
if (BIO_printf(bp, "%8sSubject Unique ID: ", "") <= 0)
goto err;
if (!X509_signature_dump(bp, ci->subjectUID, 12))
goto err;
}
}
if (!(cflag & X509_FLAG_NO_EXTENSIONS))
X509V3_extensions_print(bp, "X509v3 extensions",
ci->extensions, cflag, 8);

14
drivers/builtin_openssl2/crypto/asn1/tasn_dec.c
View File

@ -717,7 +717,7 @@ static int asn1_d2i_ex_primitive(ASN1_VALUE **pval,
long plen;
char cst, inf, free_cont = 0;
const unsigned char *p;
BUF_MEM buf;
BUF_MEM buf = { 0, NULL, 0 };
const unsigned char *cont = NULL;
long len;
if (!pval) {
@ -793,7 +793,6 @@ static int asn1_d2i_ex_primitive(ASN1_VALUE **pval,
} else {
len = p - cont + plen;
p += plen;
buf.data = NULL;
}
} else if (cst) {
if (utype == V_ASN1_NULL || utype == V_ASN1_BOOLEAN
@ -802,9 +801,9 @@ static int asn1_d2i_ex_primitive(ASN1_VALUE **pval,
ASN1err(ASN1_F_ASN1_D2I_EX_PRIMITIVE, ASN1_R_TYPE_NOT_PRIMITIVE);
return 0;
}
buf.length = 0;
buf.max = 0;
buf.data = NULL;
/* Free any returned 'buf' content */
free_cont = 1;
/*
* Should really check the internal tags are correct but some things
* may get this wrong. The relevant specs say that constructed string
@ -812,18 +811,16 @@ static int asn1_d2i_ex_primitive(ASN1_VALUE **pval,
* So instead just check for UNIVERSAL class and ignore the tag.
*/
if (!asn1_collect(&buf, &p, plen, inf, -1, V_ASN1_UNIVERSAL, 0)) {
free_cont = 1;
goto err;
}
len = buf.length;
/* Append a final null to string */
if (!BUF_MEM_grow_clean(&buf, len + 1)) {
ASN1err(ASN1_F_ASN1_D2I_EX_PRIMITIVE, ERR_R_MALLOC_FAILURE);
return 0;
goto err;
}
buf.data[len] = 0;
cont = (const unsigned char *)buf.data;
free_cont = 1;
} else {
cont = p;
len = plen;
@ -831,6 +828,7 @@ static int asn1_d2i_ex_primitive(ASN1_VALUE **pval,
}
/* We now have content length and type: translate into a structure */
/* asn1_ex_c2i may reuse allocated buffer, and so sets free_cont to 0 */
if (!asn1_ex_c2i(pval, cont, len, utype, &free_cont, it))
goto err;

4
drivers/builtin_openssl2/crypto/asn1/x_crl.c
View File

@ -58,8 +58,8 @@
#include <stdio.h>
#include "cryptlib.h"
#include "asn1_locl.h"
#include <openssl/asn1t.h>
#include "asn1_locl.h"
#include <openssl/x509.h>
#include <openssl/x509v3.h>
@ -341,6 +341,8 @@ ASN1_SEQUENCE_ref(X509_CRL, crl_cb, CRYPTO_LOCK_X509_CRL) = {
IMPLEMENT_ASN1_FUNCTIONS(X509_REVOKED)
IMPLEMENT_ASN1_DUP_FUNCTION(X509_REVOKED)
IMPLEMENT_ASN1_FUNCTIONS(X509_CRL_INFO)
IMPLEMENT_ASN1_FUNCTIONS(X509_CRL)

20
drivers/builtin_openssl2/crypto/asn1/x_x509.c
View File

@ -207,3 +207,23 @@ int i2d_X509_AUX(X509 *a, unsigned char **pp)
length += i2d_X509_CERT_AUX(a->aux, pp);
return length;
}
int i2d_re_X509_tbs(X509 *x, unsigned char **pp)
{
x->cert_info->enc.modified = 1;
return i2d_X509_CINF(x->cert_info, pp);
}
void X509_get0_signature(ASN1_BIT_STRING **psig, X509_ALGOR **palg,
const X509 *x)
{
if (psig)
*psig = x->signature;
if (palg)
*palg = x->sig_alg;
}
int X509_get_signature_nid(const X509 *x)
{
return OBJ_obj2nid(x->sig_alg->algorithm);
}

7
drivers/builtin_openssl2/crypto/asn1/x_x509a.c
View File

@ -163,10 +163,13 @@ int X509_add1_reject_object(X509 *x, ASN1_OBJECT *obj)
if (!(objtmp = OBJ_dup(obj)))
return 0;
if (!(aux = aux_get(x)))
return 0;
goto err;
if (!aux->reject && !(aux->reject = sk_ASN1_OBJECT_new_null()))
return 0;
goto err;
return sk_ASN1_OBJECT_push(aux->reject, objtmp);
err:
ASN1_OBJECT_free(objtmp);
return 0;
}
void X509_trust_clear(X509 *x)

137
drivers/builtin_openssl2/crypto/bf/asm/bf-586.pl
View File

@ -1,137 +0,0 @@
#!/usr/local/bin/perl
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
push(@INC,"${dir}","${dir}../../perlasm");
require "x86asm.pl";
require "cbc.pl";
&asm_init($ARGV[0],"bf-586.pl",$ARGV[$#ARGV] eq "386");
$BF_ROUNDS=16;
$BF_OFF=($BF_ROUNDS+2)*4;
$L="edi";
$R="esi";
$P="ebp";
$tmp1="eax";
$tmp2="ebx";
$tmp3="ecx";
$tmp4="edx";
&BF_encrypt("BF_encrypt",1);
&BF_encrypt("BF_decrypt",0);
&cbc("BF_cbc_encrypt","BF_encrypt","BF_decrypt",1,4,5,3,-1,-1);
&asm_finish();
sub BF_encrypt
{
local($name,$enc)=@_;
&function_begin_B($name,"");
&comment("");
&push("ebp");
&push("ebx");
&mov($tmp2,&wparam(0));
&mov($P,&wparam(1));
&push("esi");
&push("edi");
&comment("Load the 2 words");
&mov($L,&DWP(0,$tmp2,"",0));
&mov($R,&DWP(4,$tmp2,"",0));
&xor( $tmp1, $tmp1);
# encrypting part
if ($enc)
{
&mov($tmp2,&DWP(0,$P,"",0));
&xor( $tmp3, $tmp3);
&xor($L,$tmp2);
for ($i=0; $i<$BF_ROUNDS; $i+=2)
{
&comment("");
&comment("Round $i");
&BF_ENCRYPT($i+1,$R,$L,$P,$tmp1,$tmp2,$tmp3,$tmp4,1);
&comment("");
&comment("Round ".sprintf("%d",$i+1));
&BF_ENCRYPT($i+2,$L,$R,$P,$tmp1,$tmp2,$tmp3,$tmp4,1);
}
# &mov($tmp1,&wparam(0)); In last loop
&mov($tmp4,&DWP(($BF_ROUNDS+1)*4,$P,"",0));
}
else
{
&mov($tmp2,&DWP(($BF_ROUNDS+1)*4,$P,"",0));
&xor( $tmp3, $tmp3);
&xor($L,$tmp2);
for ($i=$BF_ROUNDS; $i>0; $i-=2)
{
&comment("");
&comment("Round $i");
&BF_ENCRYPT($i,$R,$L,$P,$tmp1,$tmp2,$tmp3,$tmp4,0);
&comment("");
&comment("Round ".sprintf("%d",$i-1));
&BF_ENCRYPT($i-1,$L,$R,$P,$tmp1,$tmp2,$tmp3,$tmp4,0);
}
# &mov($tmp1,&wparam(0)); In last loop
&mov($tmp4,&DWP(0,$P,"",0));
}
&xor($R,$tmp4);
&mov(&DWP(4,$tmp1,"",0),$L);
&mov(&DWP(0,$tmp1,"",0),$R);
&function_end($name);
}
sub BF_ENCRYPT
{
local($i,$L,$R,$P,$tmp1,$tmp2,$tmp3,$tmp4,$enc)=@_;
&mov( $tmp4, &DWP(&n2a($i*4),$P,"",0)); # for next round
&mov( $tmp2, $R);
&xor( $L, $tmp4);
&shr( $tmp2, 16);
&mov( $tmp4, $R);
&movb( &LB($tmp1), &HB($tmp2)); # A
&and( $tmp2, 0xff); # B
&movb( &LB($tmp3), &HB($tmp4)); # C
&and( $tmp4, 0xff); # D
&mov( $tmp1, &DWP(&n2a($BF_OFF+0x0000),$P,$tmp1,4));
&mov( $tmp2, &DWP(&n2a($BF_OFF+0x0400),$P,$tmp2,4));
&add( $tmp2, $tmp1);
&mov( $tmp1, &DWP(&n2a($BF_OFF+0x0800),$P,$tmp3,4));
&xor( $tmp2, $tmp1);
&mov( $tmp4, &DWP(&n2a($BF_OFF+0x0C00),$P,$tmp4,4));
&add( $tmp2, $tmp4);
if (($enc && ($i != 16)) || ((!$enc) && ($i != 1)))
{ &xor( $tmp1, $tmp1); }
else
{
&comment("Load parameter 0 ($i) enc=$enc");
&mov($tmp1,&wparam(0));
} # In last loop
&xor( $L, $tmp2);
# delay
}
sub n2a
{
sprintf("%d",$_[0]);
}

127
drivers/builtin_openssl2/crypto/bf/asm/bf-686.pl
View File

@ -1,127 +0,0 @@
#!/usr/local/bin/perl
push(@INC,"perlasm","../../perlasm");
require "x86asm.pl";
require "cbc.pl";
&asm_init($ARGV[0],"bf-686.pl");
$BF_ROUNDS=16;
$BF_OFF=($BF_ROUNDS+2)*4;
$L="ecx";
$R="edx";
$P="edi";
$tot="esi";
$tmp1="eax";
$tmp2="ebx";
$tmp3="ebp";
&des_encrypt("BF_encrypt",1);
&des_encrypt("BF_decrypt",0);
&cbc("BF_cbc_encrypt","BF_encrypt","BF_decrypt",1,4,5,3,-1,-1);
&asm_finish();
&file_end();
sub des_encrypt
{
local($name,$enc)=@_;
&function_begin($name,"");
&comment("");
&comment("Load the 2 words");
&mov("eax",&wparam(0));
&mov($L,&DWP(0,"eax","",0));
&mov($R,&DWP(4,"eax","",0));
&comment("");
&comment("P pointer, s and enc flag");
&mov($P,&wparam(1));
&xor( $tmp1, $tmp1);
&xor( $tmp2, $tmp2);
# encrypting part
if ($enc)
{
&xor($L,&DWP(0,$P,"",0));
for ($i=0; $i<$BF_ROUNDS; $i+=2)
{
&comment("");
&comment("Round $i");
&BF_ENCRYPT($i+1,$R,$L,$P,$tot,$tmp1,$tmp2,$tmp3);
&comment("");
&comment("Round ".sprintf("%d",$i+1));
&BF_ENCRYPT($i+2,$L,$R,$P,$tot,$tmp1,$tmp2,$tmp3);
}
&xor($R,&DWP(($BF_ROUNDS+1)*4,$P,"",0));
&mov("eax",&wparam(0));
&mov(&DWP(0,"eax","",0),$R);
&mov(&DWP(4,"eax","",0),$L);
&function_end_A($name);
}
else
{
&xor($L,&DWP(($BF_ROUNDS+1)*4,$P,"",0));
for ($i=$BF_ROUNDS; $i>0; $i-=2)
{
&comment("");
&comment("Round $i");
&BF_ENCRYPT($i,$R,$L,$P,$tot,$tmp1,$tmp2,$tmp3);
&comment("");
&comment("Round ".sprintf("%d",$i-1));
&BF_ENCRYPT($i-1,$L,$R,$P,$tot,$tmp1,$tmp2,$tmp3);
}
&xor($R,&DWP(0,$P,"",0));
&mov("eax",&wparam(0));
&mov(&DWP(0,"eax","",0),$R);
&mov(&DWP(4,"eax","",0),$L);
&function_end_A($name);
}
&function_end_B($name);
}
sub BF_ENCRYPT
{
local($i,$L,$R,$P,$tot,$tmp1,$tmp2,$tmp3)=@_;
&rotr( $R, 16);
&mov( $tot, &DWP(&n2a($i*4),$P,"",0));
&movb( &LB($tmp1), &HB($R));
&movb( &LB($tmp2), &LB($R));
&rotr( $R, 16);
&xor( $L, $tot);
&mov( $tot, &DWP(&n2a($BF_OFF+0x0000),$P,$tmp1,4));
&mov( $tmp3, &DWP(&n2a($BF_OFF+0x0400),$P,$tmp2,4));
&movb( &LB($tmp1), &HB($R));
&movb( &LB($tmp2), &LB($R));
&add( $tot, $tmp3);
&mov( $tmp1, &DWP(&n2a($BF_OFF+0x0800),$P,$tmp1,4)); # delay
&xor( $tot, $tmp1);
&mov( $tmp3, &DWP(&n2a($BF_OFF+0x0C00),$P,$tmp2,4));
&add( $tot, $tmp3);
&xor( $tmp1, $tmp1);
&xor( $L, $tot);
# delay
}
sub n2a
{
sprintf("%d",$_[0]);
}

538
drivers/builtin_openssl2/crypto/bf/bftest.c
View File

@ -1,538 +0,0 @@
/* crypto/bf/bftest.c */
/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
* All rights reserved.
*
* This package is an SSL implementation written
* by Eric Young (eay@cryptsoft.com).
* The implementation was written so as to conform with Netscapes SSL.
*
* This library is free for commercial and non-commercial use as long as
* the following conditions are aheared to. The following conditions
* apply to all code found in this distribution, be it the RC4, RSA,
* lhash, DES, etc., code; not just the SSL code. The SSL documentation
* included with this distribution is covered by the same copyright terms
* except that the holder is Tim Hudson (tjh@cryptsoft.com).
*
* Copyright remains Eric Young's, and as such any Copyright notices in
* the code are not to be removed.
* If this package is used in a product, Eric Young should be given attribution
* as the author of the parts of the library used.
* This can be in the form of a textual message at program startup or
* in documentation (online or textual) provided with the package.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* "This product includes cryptographic software written by
* Eric Young (eay@cryptsoft.com)"
* The word 'cryptographic' can be left out if the rouines from the library
* being used are not cryptographic related :-).
* 4. If you include any Windows specific code (or a derivative thereof) from
* the apps directory (application code) you must include an acknowledgement:
* "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
*
* THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* The licence and distribution terms for any publically available version or
* derivative of this code cannot be changed. i.e. this code cannot simply be
* copied and put under another distribution licence
* [including the GNU Public Licence.]
*/
/*
* This has been a quickly hacked 'ideatest.c'. When I add tests for other
* RC2 modes, more of the code will be uncommented.
*/
#include <stdio.h>
#include <string.h>
#include <stdlib.h>
#include <openssl/opensslconf.h> /* To see if OPENSSL_NO_BF is defined */
#include "../e_os.h"
#ifdef OPENSSL_NO_BF
int main(int argc, char *argv[])
{
printf("No BF support\n");
return (0);
}
#else
# include <openssl/blowfish.h>
# ifdef CHARSET_EBCDIC
# include <openssl/ebcdic.h>
# endif
static char *bf_key[2] = {
"abcdefghijklmnopqrstuvwxyz",
"Who is John Galt?"
};
/* big endian */
static BF_LONG bf_plain[2][2] = {
{0x424c4f57L, 0x46495348L},
{0xfedcba98L, 0x76543210L}
};
static BF_LONG bf_cipher[2][2] = {
{0x324ed0feL, 0xf413a203L},
{0xcc91732bL, 0x8022f684L}
};
/************/
/* Lets use the DES test vectors :-) */
# define NUM_TESTS 34
static unsigned char ecb_data[NUM_TESTS][8] = {
{0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
{0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF},
{0x30, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
{0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11},
{0x01, 0x23, 0x45, 0x67, 0x89, 0xAB, 0xCD, 0xEF},
{0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11},
{0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
{0xFE, 0xDC, 0xBA, 0x98, 0x76, 0x54, 0x32, 0x10},
{0x7C, 0xA1, 0x10, 0x45, 0x4A, 0x1A, 0x6E, 0x57},
{0x01, 0x31, 0xD9, 0x61, 0x9D, 0xC1, 0x37, 0x6E},
{0x07, 0xA1, 0x13, 0x3E, 0x4A, 0x0B, 0x26, 0x86},
{0x38, 0x49, 0x67, 0x4C, 0x26, 0x02, 0x31, 0x9E},
{0x04, 0xB9, 0x15, 0xBA, 0x43, 0xFE, 0xB5, 0xB6},
{0x01, 0x13, 0xB9, 0x70, 0xFD, 0x34, 0xF2, 0xCE},
{0x01, 0x70, 0xF1, 0x75, 0x46, 0x8F, 0xB5, 0xE6},
{0x43, 0x29, 0x7F, 0xAD, 0x38, 0xE3, 0x73, 0xFE},
{0x07, 0xA7, 0x13, 0x70, 0x45, 0xDA, 0x2A, 0x16},
{0x04, 0x68, 0x91, 0x04, 0xC2, 0xFD, 0x3B, 0x2F},
{0x37, 0xD0, 0x6B, 0xB5, 0x16, 0xCB, 0x75, 0x46},
{0x1F, 0x08, 0x26, 0x0D, 0x1A, 0xC2, 0x46, 0x5E},
{0x58, 0x40, 0x23, 0x64, 0x1A, 0xBA, 0x61, 0x76},
{0x02, 0x58, 0x16, 0x16, 0x46, 0x29, 0xB0, 0x07},
{0x49, 0x79, 0x3E, 0xBC, 0x79, 0xB3, 0x25, 0x8F},
{0x4F, 0xB0, 0x5E, 0x15, 0x15, 0xAB, 0x73, 0xA7},
{0x49, 0xE9, 0x5D, 0x6D, 0x4C, 0xA2, 0x29, 0xBF},
{0x01, 0x83, 0x10, 0xDC, 0x40, 0x9B, 0x26, 0xD6},
{0x1C, 0x58, 0x7F, 0x1C, 0x13, 0x92, 0x4F, 0xEF},
{0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01, 0x01},
{0x1F, 0x1F, 0x1F, 0x1F, 0x0E, 0x0E, 0x0E, 0x0E},
{0xE0, 0xFE, 0xE0, 0xFE, 0xF1, 0xFE, 0xF1, 0xFE},
{0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
{0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF},
{0x01, 0x23, 0x45, 0x67, 0x89, 0xAB, 0xCD, 0xEF},
{0xFE, 0xDC, 0xBA, 0x98, 0x76, 0x54, 0x32, 0x10}
};
static unsigned char plain_data[NUM_TESTS][8] = {
{0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
{0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF},
{0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01},
{0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11},
{0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11, 0x11},
{0x01, 0x23, 0x45, 0x67, 0x89, 0xAB, 0xCD, 0xEF},
{0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
{0x01, 0x23, 0x45, 0x67, 0x89, 0xAB, 0xCD, 0xEF},
{0x01, 0xA1, 0xD6, 0xD0, 0x39, 0x77, 0x67, 0x42},
{0x5C, 0xD5, 0x4C, 0xA8, 0x3D, 0xEF, 0x57, 0xDA},
{0x02, 0x48, 0xD4, 0x38, 0x06, 0xF6, 0x71, 0x72},
{0x51, 0x45, 0x4B, 0x58, 0x2D, 0xDF, 0x44, 0x0A},
{0x42, 0xFD, 0x44, 0x30, 0x59, 0x57, 0x7F, 0xA2},
{0x05, 0x9B, 0x5E, 0x08, 0x51, 0xCF, 0x14, 0x3A},
{0x07, 0x56, 0xD8, 0xE0, 0x77, 0x47, 0x61, 0xD2},
{0x76, 0x25, 0x14, 0xB8, 0x29, 0xBF, 0x48, 0x6A},
{0x3B, 0xDD, 0x11, 0x90, 0x49, 0x37, 0x28, 0x02},
{0x26, 0x95, 0x5F, 0x68, 0x35, 0xAF, 0x60, 0x9A},
{0x16, 0x4D, 0x5E, 0x40, 0x4F, 0x27, 0x52, 0x32},
{0x6B, 0x05, 0x6E, 0x18, 0x75, 0x9F, 0x5C, 0xCA},
{0x00, 0x4B, 0xD6, 0xEF, 0x09, 0x17, 0x60, 0x62},
{0x48, 0x0D, 0x39, 0x00, 0x6E, 0xE7, 0x62, 0xF2},
{0x43, 0x75, 0x40, 0xC8, 0x69, 0x8F, 0x3C, 0xFA},
{0x07, 0x2D, 0x43, 0xA0, 0x77, 0x07, 0x52, 0x92},
{0x02, 0xFE, 0x55, 0x77, 0x81, 0x17, 0xF1, 0x2A},
{0x1D, 0x9D, 0x5C, 0x50, 0x18, 0xF7, 0x28, 0xC2},
{0x30, 0x55, 0x32, 0x28, 0x6D, 0x6F, 0x29, 0x5A},
{0x01, 0x23, 0x45, 0x67, 0x89, 0xAB, 0xCD, 0xEF},
{0x01, 0x23, 0x45, 0x67, 0x89, 0xAB, 0xCD, 0xEF},
{0x01, 0x23, 0x45, 0x67, 0x89, 0xAB, 0xCD, 0xEF},
{0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF},
{0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
{0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00},
{0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF, 0xFF}
};
static unsigned char cipher_data[NUM_TESTS][8] = {
{0x4E, 0xF9, 0x97, 0x45, 0x61, 0x98, 0xDD, 0x78},
{0x51, 0x86, 0x6F, 0xD5, 0xB8, 0x5E, 0xCB, 0x8A},
{0x7D, 0x85, 0x6F, 0x9A, 0x61, 0x30, 0x63, 0xF2},
{0x24, 0x66, 0xDD, 0x87, 0x8B, 0x96, 0x3C, 0x9D},
{0x61, 0xF9, 0xC3, 0x80, 0x22, 0x81, 0xB0, 0x96},
{0x7D, 0x0C, 0xC6, 0x30, 0xAF, 0xDA, 0x1E, 0xC7},
{0x4E, 0xF9, 0x97, 0x45, 0x61, 0x98, 0xDD, 0x78},
{0x0A, 0xCE, 0xAB, 0x0F, 0xC6, 0xA0, 0xA2, 0x8D},
{0x59, 0xC6, 0x82, 0x45, 0xEB, 0x05, 0x28, 0x2B},
{0xB1, 0xB8, 0xCC, 0x0B, 0x25, 0x0F, 0x09, 0xA0},
{0x17, 0x30, 0xE5, 0x77, 0x8B, 0xEA, 0x1D, 0xA4},
{0xA2, 0x5E, 0x78, 0x56, 0xCF, 0x26, 0x51, 0xEB},
{0x35, 0x38, 0x82, 0xB1, 0x09, 0xCE, 0x8F, 0x1A},
{0x48, 0xF4, 0xD0, 0x88, 0x4C, 0x37, 0x99, 0x18},
{0x43, 0x21, 0x93, 0xB7, 0x89, 0x51, 0xFC, 0x98},
{0x13, 0xF0, 0x41, 0x54, 0xD6, 0x9D, 0x1A, 0xE5},
{0x2E, 0xED, 0xDA, 0x93, 0xFF, 0xD3, 0x9C, 0x79},
{0xD8, 0x87, 0xE0, 0x39, 0x3C, 0x2D, 0xA6, 0xE3},
{0x5F, 0x99, 0xD0, 0x4F, 0x5B, 0x16, 0x39, 0x69},
{0x4A, 0x05, 0x7A, 0x3B, 0x24, 0xD3, 0x97, 0x7B},
{0x45, 0x20, 0x31, 0xC1, 0xE4, 0xFA, 0xDA, 0x8E},
{0x75, 0x55, 0xAE, 0x39, 0xF5, 0x9B, 0x87, 0xBD},
{0x53, 0xC5, 0x5F, 0x9C, 0xB4, 0x9F, 0xC0, 0x19},
{0x7A, 0x8E, 0x7B, 0xFA, 0x93, 0x7E, 0x89, 0xA3},
{0xCF, 0x9C, 0x5D, 0x7A, 0x49, 0x86, 0xAD, 0xB5},
{0xD1, 0xAB, 0xB2, 0x90, 0x65, 0x8B, 0xC7, 0x78},
{0x55, 0xCB, 0x37, 0x74, 0xD1, 0x3E, 0xF2, 0x01},
{0xFA, 0x34, 0xEC, 0x48, 0x47, 0xB2, 0x68, 0xB2},
{0xA7, 0x90, 0x79, 0x51, 0x08, 0xEA, 0x3C, 0xAE},
{0xC3, 0x9E, 0x07, 0x2D, 0x9F, 0xAC, 0x63, 0x1D},
{0x01, 0x49, 0x33, 0xE0, 0xCD, 0xAF, 0xF6, 0xE4},
{0xF2, 0x1E, 0x9A, 0x77, 0xB7, 0x1C, 0x49, 0xBC},
{0x24, 0x59, 0x46, 0x88, 0x57, 0x54, 0x36, 0x9A},
{0x6B, 0x5C, 0x5A, 0x9C, 0x5D, 0x9E, 0x0A, 0x5A},
};
static unsigned char cbc_key[16] = {
0x01, 0x23, 0x45, 0x67, 0x89, 0xab, 0xcd, 0xef,
0xf0, 0xe1, 0xd2, 0xc3, 0xb4, 0xa5, 0x96, 0x87
};
static unsigned char cbc_iv[8] =
{ 0xfe, 0xdc, 0xba, 0x98, 0x76, 0x54, 0x32, 0x10 };
static char cbc_data[40] = "7654321 Now is the time for ";
static unsigned char cbc_ok[32] = {
0x6B, 0x77, 0xB4, 0xD6, 0x30, 0x06, 0xDE, 0xE6,
0x05, 0xB1, 0x56, 0xE2, 0x74, 0x03, 0x97, 0x93,
0x58, 0xDE, 0xB9, 0xE7, 0x15, 0x46, 0x16, 0xD9,
0x59, 0xF1, 0x65, 0x2B, 0xD5, 0xFF, 0x92, 0xCC
};
static unsigned char cfb64_ok[] = {
0xE7, 0x32, 0x14, 0xA2, 0x82, 0x21, 0x39, 0xCA,
0xF2, 0x6E, 0xCF, 0x6D, 0x2E, 0xB9, 0xE7, 0x6E,
0x3D, 0xA3, 0xDE, 0x04, 0xD1, 0x51, 0x72, 0x00,
0x51, 0x9D, 0x57, 0xA6, 0xC3
};
static unsigned char ofb64_ok[] = {
0xE7, 0x32, 0x14, 0xA2, 0x82, 0x21, 0x39, 0xCA,
0x62, 0xB3, 0x43, 0xCC, 0x5B, 0x65, 0x58, 0x73,
0x10, 0xDD, 0x90, 0x8D, 0x0C, 0x24, 0x1B, 0x22,
0x63, 0xC2, 0xCF, 0x80, 0xDA
};
# define KEY_TEST_NUM 25
static unsigned char key_test[KEY_TEST_NUM] = {
0xf0, 0xe1, 0xd2, 0xc3, 0xb4, 0xa5, 0x96, 0x87,
0x78, 0x69, 0x5a, 0x4b, 0x3c, 0x2d, 0x1e, 0x0f,
0x00, 0x11, 0x22, 0x33, 0x44, 0x55, 0x66, 0x77,
0x88
};
static unsigned char key_data[8] =
{ 0xFE, 0xDC, 0xBA, 0x98, 0x76, 0x54, 0x32, 0x10 };
static unsigned char key_out[KEY_TEST_NUM][8] = {
{0xF9, 0xAD, 0x59, 0x7C, 0x49, 0xDB, 0x00, 0x5E},
{0xE9, 0x1D, 0x21, 0xC1, 0xD9, 0x61, 0xA6, 0xD6},
{0xE9, 0xC2, 0xB7, 0x0A, 0x1B, 0xC6, 0x5C, 0xF3},
{0xBE, 0x1E, 0x63, 0x94, 0x08, 0x64, 0x0F, 0x05},
{0xB3, 0x9E, 0x44, 0x48, 0x1B, 0xDB, 0x1E, 0x6E},
{0x94, 0x57, 0xAA, 0x83, 0xB1, 0x92, 0x8C, 0x0D},
{0x8B, 0xB7, 0x70, 0x32, 0xF9, 0x60, 0x62, 0x9D},
{0xE8, 0x7A, 0x24, 0x4E, 0x2C, 0xC8, 0x5E, 0x82},
{0x15, 0x75, 0x0E, 0x7A, 0x4F, 0x4E, 0xC5, 0x77},
{0x12, 0x2B, 0xA7, 0x0B, 0x3A, 0xB6, 0x4A, 0xE0},
{0x3A, 0x83, 0x3C, 0x9A, 0xFF, 0xC5, 0x37, 0xF6},
{0x94, 0x09, 0xDA, 0x87, 0xA9, 0x0F, 0x6B, 0xF2},
{0x88, 0x4F, 0x80, 0x62, 0x50, 0x60, 0xB8, 0xB4},
{0x1F, 0x85, 0x03, 0x1C, 0x19, 0xE1, 0x19, 0x68},
{0x79, 0xD9, 0x37, 0x3A, 0x71, 0x4C, 0xA3, 0x4F},
{0x93, 0x14, 0x28, 0x87, 0xEE, 0x3B, 0xE1, 0x5C},
{0x03, 0x42, 0x9E, 0x83, 0x8C, 0xE2, 0xD1, 0x4B},
{0xA4, 0x29, 0x9E, 0x27, 0x46, 0x9F, 0xF6, 0x7B},
{0xAF, 0xD5, 0xAE, 0xD1, 0xC1, 0xBC, 0x96, 0xA8},
{0x10, 0x85, 0x1C, 0x0E, 0x38, 0x58, 0xDA, 0x9F},
{0xE6, 0xF5, 0x1E, 0xD7, 0x9B, 0x9D, 0xB2, 0x1F},
{0x64, 0xA6, 0xE1, 0x4A, 0xFD, 0x36, 0xB4, 0x6F},
{0x80, 0xC7, 0xD7, 0xD4, 0x5A, 0x54, 0x79, 0xAD},
{0x05, 0x04, 0x4B, 0x62, 0xFA, 0x52, 0xD0, 0x80},
};
static int test(void);
static int print_test_data(void);
int main(int argc, char *argv[])
{
int ret;
if (argc > 1)
ret = print_test_data();
else
ret = test();
# ifdef OPENSSL_SYS_NETWARE
if (ret)
printf("ERROR: %d\n", ret);
# endif
EXIT(ret);
return (0);
}
static int print_test_data(void)
{
unsigned int i, j;
printf("ecb test data\n");
printf("key bytes\t\tclear bytes\t\tcipher bytes\n");
for (i = 0; i < NUM_TESTS; i++) {
for (j = 0; j < 8; j++)
printf("%02X", ecb_data[i][j]);
printf("\t");
for (j = 0; j < 8; j++)
printf("%02X", plain_data[i][j]);
printf("\t");
for (j = 0; j < 8; j++)
printf("%02X", cipher_data[i][j]);
printf("\n");
}
printf("set_key test data\n");
printf("data[8]= ");
for (j = 0; j < 8; j++)
printf("%02X", key_data[j]);
printf("\n");
for (i = 0; i < KEY_TEST_NUM - 1; i++) {
printf("c=");
for (j = 0; j < 8; j++)
printf("%02X", key_out[i][j]);
printf(" k[%2u]=", i + 1);
for (j = 0; j < i + 1; j++)
printf("%02X", key_test[j]);
printf("\n");
}
printf("\nchaining mode test data\n");
printf("key[16] = ");
for (j = 0; j < 16; j++)
printf("%02X", cbc_key[j]);
printf("\niv[8] = ");
for (j = 0; j < 8; j++)
printf("%02X", cbc_iv[j]);
printf("\ndata[%d] = '%s'", (int)strlen(cbc_data) + 1, cbc_data);
printf("\ndata[%d] = ", (int)strlen(cbc_data) + 1);
for (j = 0; j < strlen(cbc_data) + 1; j++)
printf("%02X", cbc_data[j]);
printf("\n");
printf("cbc cipher text\n");
printf("cipher[%d]= ", 32);
for (j = 0; j < 32; j++)
printf("%02X", cbc_ok[j]);
printf("\n");
printf("cfb64 cipher text\n");
printf("cipher[%d]= ", (int)strlen(cbc_data) + 1);
for (j = 0; j < strlen(cbc_data) + 1; j++)
printf("%02X", cfb64_ok[j]);
printf("\n");
printf("ofb64 cipher text\n");
printf("cipher[%d]= ", (int)strlen(cbc_data) + 1);
for (j = 0; j < strlen(cbc_data) + 1; j++)
printf("%02X", ofb64_ok[j]);
printf("\n");
return (0);
}
static int test(void)
{
unsigned char cbc_in[40], cbc_out[40], iv[8];
int i, n, err = 0;
BF_KEY key;
BF_LONG data[2];
unsigned char out[8];
BF_LONG len;
# ifdef CHARSET_EBCDIC
ebcdic2ascii(cbc_data, cbc_data, strlen(cbc_data));
# endif
printf("testing blowfish in raw ecb mode\n");
for (n = 0; n < 2; n++) {
# ifdef CHARSET_EBCDIC
ebcdic2ascii(bf_key[n], bf_key[n], strlen(bf_key[n]));
# endif
BF_set_key(&key, strlen(bf_key[n]), (unsigned char *)bf_key[n]);
data[0] = bf_plain[n][0];
data[1] = bf_plain[n][1];
BF_encrypt(data, &key);
if (memcmp(&(bf_cipher[n][0]), &(data[0]), 8) != 0) {
printf("BF_encrypt error encrypting\n");
printf("got :");
for (i = 0; i < 2; i++)
printf("%08lX ", (unsigned long)data[i]);
printf("\n");
printf("expected:");
for (i = 0; i < 2; i++)
printf("%08lX ", (unsigned long)bf_cipher[n][i]);
err = 1;
printf("\n");
}
BF_decrypt(&(data[0]), &key);
if (memcmp(&(bf_plain[n][0]), &(data[0]), 8) != 0) {
printf("BF_encrypt error decrypting\n");
printf("got :");
for (i = 0; i < 2; i++)
printf("%08lX ", (unsigned long)data[i]);
printf("\n");
printf("expected:");
for (i = 0; i < 2; i++)
printf("%08lX ", (unsigned long)bf_plain[n][i]);
printf("\n");
err = 1;
}
}
printf("testing blowfish in ecb mode\n");
for (n = 0; n < NUM_TESTS; n++) {
BF_set_key(&key, 8, ecb_data[n]);
BF_ecb_encrypt(&(plain_data[n][0]), out, &key, BF_ENCRYPT);
if (memcmp(&(cipher_data[n][0]), out, 8) != 0) {
printf("BF_ecb_encrypt blowfish error encrypting\n");
printf("got :");
for (i = 0; i < 8; i++)
printf("%02X ", out[i]);
printf("\n");
printf("expected:");
for (i = 0; i < 8; i++)
printf("%02X ", cipher_data[n][i]);
err = 1;
printf("\n");
}
BF_ecb_encrypt(out, out, &key, BF_DECRYPT);
if (memcmp(&(plain_data[n][0]), out, 8) != 0) {
printf("BF_ecb_encrypt error decrypting\n");
printf("got :");
for (i = 0; i < 8; i++)
printf("%02X ", out[i]);
printf("\n");
printf("expected:");
for (i = 0; i < 8; i++)
printf("%02X ", plain_data[n][i]);
printf("\n");
err = 1;
}
}
printf("testing blowfish set_key\n");
for (n = 1; n < KEY_TEST_NUM; n++) {
BF_set_key(&key, n, key_test);
BF_ecb_encrypt(key_data, out, &key, BF_ENCRYPT);
/* mips-sgi-irix6.5-gcc vv -mabi=64 bug workaround */
if (memcmp(out, &(key_out[i = n - 1][0]), 8) != 0) {
printf("blowfish setkey error\n");
err = 1;
}
}
printf("testing blowfish in cbc mode\n");
len = strlen(cbc_data) + 1;
BF_set_key(&key, 16, cbc_key);
memset(cbc_in, 0, sizeof cbc_in);
memset(cbc_out, 0, sizeof cbc_out);
memcpy(iv, cbc_iv, sizeof iv);
BF_cbc_encrypt((unsigned char *)cbc_data, cbc_out, len,
&key, iv, BF_ENCRYPT);
if (memcmp(cbc_out, cbc_ok, 32) != 0) {
err = 1;
printf("BF_cbc_encrypt encrypt error\n");
for (i = 0; i < 32; i++)
printf("0x%02X,", cbc_out[i]);
}
memcpy(iv, cbc_iv, 8);
BF_cbc_encrypt(cbc_out, cbc_in, len, &key, iv, BF_DECRYPT);
if (memcmp(cbc_in, cbc_data, strlen(cbc_data) + 1) != 0) {
printf("BF_cbc_encrypt decrypt error\n");
err = 1;
}
printf("testing blowfish in cfb64 mode\n");
BF_set_key(&key, 16, cbc_key);
memset(cbc_in, 0, 40);
memset(cbc_out, 0, 40);
memcpy(iv, cbc_iv, 8);
n = 0;
BF_cfb64_encrypt((unsigned char *)cbc_data, cbc_out, (long)13,
&key, iv, &n, BF_ENCRYPT);
BF_cfb64_encrypt((unsigned char *)&(cbc_data[13]), &(cbc_out[13]),
len - 13, &key, iv, &n, BF_ENCRYPT);
if (memcmp(cbc_out, cfb64_ok, (int)len) != 0) {
err = 1;
printf("BF_cfb64_encrypt encrypt error\n");
for (i = 0; i < (int)len; i++)
printf("0x%02X,", cbc_out[i]);
}
n = 0;
memcpy(iv, cbc_iv, 8);
BF_cfb64_encrypt(cbc_out, cbc_in, 17, &key, iv, &n, BF_DECRYPT);
BF_cfb64_encrypt(&(cbc_out[17]), &(cbc_in[17]), len - 17,
&key, iv, &n, BF_DECRYPT);
if (memcmp(cbc_in, cbc_data, (int)len) != 0) {
printf("BF_cfb64_encrypt decrypt error\n");
err = 1;
}
printf("testing blowfish in ofb64\n");
BF_set_key(&key, 16, cbc_key);
memset(cbc_in, 0, 40);
memset(cbc_out, 0, 40);
memcpy(iv, cbc_iv, 8);
n = 0;
BF_ofb64_encrypt((unsigned char *)cbc_data, cbc_out, (long)13, &key, iv,
&n);
BF_ofb64_encrypt((unsigned char *)&(cbc_data[13]), &(cbc_out[13]),
len - 13, &key, iv, &n);
if (memcmp(cbc_out, ofb64_ok, (int)len) != 0) {
err = 1;
printf("BF_ofb64_encrypt encrypt error\n");
for (i = 0; i < (int)len; i++)
printf("0x%02X,", cbc_out[i]);
}
n = 0;
memcpy(iv, cbc_iv, 8);
BF_ofb64_encrypt(cbc_out, cbc_in, 17, &key, iv, &n);
BF_ofb64_encrypt(&(cbc_out[17]), &(cbc_in[17]), len - 17, &key, iv, &n);
if (memcmp(cbc_in, cbc_data, (int)len) != 0) {
printf("BF_ofb64_encrypt decrypt error\n");
err = 1;
}
return (err);
}
#endif

25
drivers/builtin_openssl2/crypto/bio/b_dump.c
View File

@ -181,3 +181,28 @@ int BIO_dump_indent(BIO *bp, const char *s, int len, int indent)
{
return BIO_dump_indent_cb(write_bio, bp, s, len, indent);
}
int BIO_hex_string(BIO *out, int indent, int width, unsigned char *data,
int datalen)
{
int i, j = 0;
if (datalen < 1)
return 1;
for (i = 0; i < datalen - 1; i++) {
if (i && !j)
BIO_printf(out, "%*s", indent, "");
BIO_printf(out, "%02X:", data[i]);
j = (j + 1) % width;
if (!j)
BIO_printf(out, "\n");
}
if (i && !j)
BIO_printf(out, "%*s", indent, "");
BIO_printf(out, "%02X", data[datalen - 1]);
return 1;
}

8
drivers/builtin_openssl2/crypto/bio/b_sock.c
View File

@ -225,13 +225,17 @@ int BIO_get_port(const char *str, unsigned short *port_ptr)
int BIO_sock_error(int sock)
{
int j, i;
int size;
union {
size_t s;
int i;
} size;
# if defined(OPENSSL_SYS_BEOS_R5)
return 0;
# endif
size = sizeof(int);
/* heuristic way to adapt for platforms that expect 64-bit optlen */
size.s = 0, size.i = sizeof(j);
/*
* Note: under Windows the third parameter is of type (char *) whereas
* under other systems it is (void *) if you don't have a cast it will

2
drivers/builtin_openssl2/crypto/bio/bio_err.c
View File

@ -1,6 +1,6 @@
/* crypto/bio/bio_err.c */
/* ====================================================================
* Copyright (c) 1999-2011 The OpenSSL Project. All rights reserved.
* Copyright (c) 1999-2015 The OpenSSL Project. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions

2
drivers/builtin_openssl2/crypto/bio/bss_acpt.c
View File

@ -445,7 +445,7 @@ static int acpt_puts(BIO *bp, const char *str)
return (ret);
}
BIO *BIO_new_accept(char *str)
BIO *BIO_new_accept(const char *str)
{
BIO *ret;

2
drivers/builtin_openssl2/crypto/bio/bss_conn.c
View File

@ -594,7 +594,7 @@ static int conn_puts(BIO *bp, const char *str)
return (ret);
}
BIO *BIO_new_connect(char *str)
BIO *BIO_new_connect(const char *str)
{
BIO *ret;

84
drivers/builtin_openssl2/crypto/bio/bss_dgram.c
View File

@ -65,7 +65,7 @@
#include <openssl/bio.h>
#ifndef OPENSSL_NO_DGRAM
# if defined(OPENSSL_SYS_WIN32) || defined(OPENSSL_SYS_VMS)
# if defined(OPENSSL_SYS_VMS)
# include <sys/timeb.h>
# endif
@ -80,6 +80,10 @@
# define IP_MTU 14 /* linux is lame */
# endif
# if OPENSSL_USE_IPV6 && !defined(IPPROTO_IPV6)
# define IPPROTO_IPV6 41 /* windows is lame */
# endif
# if defined(__FreeBSD__) && defined(IN6_IS_ADDR_V4MAPPED)
/* Standard definition causes type-punning problems. */
# undef IN6_IS_ADDR_V4MAPPED
@ -496,8 +500,8 @@ static long dgram_ctrl(BIO *b, int cmd, long num, void *ptr)
int *ip;
struct sockaddr *to = NULL;
bio_dgram_data *data = NULL;
# if defined(OPENSSL_SYS_LINUX) && (defined(IP_MTU_DISCOVER) || defined(IP_MTU))
int sockopt_val = 0;
# if defined(OPENSSL_SYS_LINUX) && (defined(IP_MTU_DISCOVER) || defined(IP_MTU))
socklen_t sockopt_len; /* assume that system supporting IP_MTU is
* modern enough to define socklen_t */
socklen_t addr_len;
@ -880,6 +884,61 @@ static long dgram_ctrl(BIO *b, int cmd, long num, void *ptr)
ret = 0;
break;
# endif
case BIO_CTRL_DGRAM_SET_DONT_FRAG:
sockopt_val = num ? 1 : 0;
switch (data->peer.sa.sa_family) {
case AF_INET:
# if defined(IP_DONTFRAG)
if ((ret = setsockopt(b->num, IPPROTO_IP, IP_DONTFRAG,
&sockopt_val, sizeof(sockopt_val))) < 0) {
perror("setsockopt");
ret = -1;
}
# elif defined(OPENSSL_SYS_LINUX) && defined(IP_MTU_DISCOVER) && defined (IP_PMTUDISC_PROBE)
if ((sockopt_val = num ? IP_PMTUDISC_PROBE : IP_PMTUDISC_DONT),
(ret = setsockopt(b->num, IPPROTO_IP, IP_MTU_DISCOVER,
&sockopt_val, sizeof(sockopt_val))) < 0) {
perror("setsockopt");
ret = -1;
}
# elif defined(OPENSSL_SYS_WINDOWS) && defined(IP_DONTFRAGMENT)
if ((ret = setsockopt(b->num, IPPROTO_IP, IP_DONTFRAGMENT,
(const char *)&sockopt_val,
sizeof(sockopt_val))) < 0) {
perror("setsockopt");
ret = -1;
}
# else
ret = -1;
# endif
break;
# if OPENSSL_USE_IPV6
case AF_INET6:
# if defined(IPV6_DONTFRAG)
if ((ret = setsockopt(b->num, IPPROTO_IPV6, IPV6_DONTFRAG,
(const void *)&sockopt_val,
sizeof(sockopt_val))) < 0) {
perror("setsockopt");
ret = -1;
}
# elif defined(OPENSSL_SYS_LINUX) && defined(IPV6_MTUDISCOVER)
if ((sockopt_val = num ? IP_PMTUDISC_PROBE : IP_PMTUDISC_DONT),
(ret = setsockopt(b->num, IPPROTO_IPV6, IPV6_MTU_DISCOVER,
&sockopt_val, sizeof(sockopt_val))) < 0) {
perror("setsockopt");
ret = -1;
}
# else
ret = -1;
# endif
break;
# endif
default:
ret = -1;
break;
}
break;
case BIO_CTRL_DGRAM_GET_MTU_OVERHEAD:
ret = dgram_get_mtu_overhead(data);
break;
@ -1993,11 +2052,22 @@ int BIO_dgram_non_fatal_error(int err)
static void get_current_time(struct timeval *t)
{
# ifdef OPENSSL_SYS_WIN32
struct _timeb tb;
_ftime(&tb);
t->tv_sec = (long)tb.time;
t->tv_usec = (long)tb.millitm * 1000;
# if defined(_WIN32)
SYSTEMTIME st;
union {
unsigned __int64 ul;
FILETIME ft;
} now;
GetSystemTime(&st);
SystemTimeToFileTime(&st, &now.ft);
# ifdef __MINGW32__
now.ul -= 116444736000000000ULL;
# else
now.ul -= 116444736000000000UI64; /* re-bias to 1/1/1970 */
# endif
t->tv_sec = (long)(now.ul / 10000000);
t->tv_usec = ((int)(now.ul % 10000000)) / 10;
# elif defined(OPENSSL_SYS_VMS)
struct timeb tb;
ftime(&tb);

22
drivers/builtin_openssl2/crypto/bio/bss_fd.c
View File

@ -63,9 +63,27 @@
#if defined(OPENSSL_NO_POSIX_IO)
/*
* One can argue that one should implement dummy placeholder for
* BIO_s_fd here...
* Dummy placeholder for BIO_s_fd...
*/
BIO *BIO_new_fd(int fd, int close_flag)
{
return NULL;
}
int BIO_fd_non_fatal_error(int err)
{
return 0;
}
int BIO_fd_should_retry(int i)
{
return 0;
}
BIO_METHOD *BIO_s_fd(void)
{
return NULL;
}
#else
/*
* As for unconditional usage of "UPLINK" interface in this module.

6
drivers/builtin_openssl2/crypto/bio/bss_mem.c
View File

@ -91,7 +91,8 @@ BIO_METHOD *BIO_s_mem(void)
return (&mem_method);
}
BIO *BIO_new_mem_buf(void *buf, int len)
BIO *BIO_new_mem_buf(const void *buf, int len)
{
BIO *ret;
BUF_MEM *b;
@ -105,7 +106,8 @@ BIO *BIO_new_mem_buf(void *buf, int len)
if (!(ret = BIO_new(BIO_s_mem())))
return NULL;
b = (BUF_MEM *)ret->ptr;
b->data = buf;
/* Cast away const and trust in the MEM_RDONLY flag. */
b->data = (void *)buf;
b->length = sz;
b->max = sz;
ret->flags |= BIO_FLAGS_MEM_RDONLY;

321
drivers/builtin_openssl2/crypto/bn/asm/alpha-mont.pl
View File

@ -1,321 +0,0 @@
#!/usr/bin/env perl
#
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
#
# On 21264 RSA sign performance improves by 70/35/20/15 percent for
# 512/1024/2048/4096 bit key lengths. This is against vendor compiler
# instructed to '-tune host' code with in-line assembler. Other
# benchmarks improve by 15-20%. To anchor it to something else, the
# code provides approximately the same performance per GHz as AMD64.
# I.e. if you compare 1GHz 21264 and 2GHz Opteron, you'll observe ~2x
# difference.
# int bn_mul_mont(
$rp="a0"; # BN_ULONG *rp,
$ap="a1"; # const BN_ULONG *ap,
$bp="a2"; # const BN_ULONG *bp,
$np="a3"; # const BN_ULONG *np,
$n0="a4"; # const BN_ULONG *n0,
$num="a5"; # int num);
$lo0="t0";
$hi0="t1";
$lo1="t2";
$hi1="t3";
$aj="t4";
$bi="t5";
$nj="t6";
$tp="t7";
$alo="t8";
$ahi="t9";
$nlo="t10";
$nhi="t11";
$tj="t12";
$i="s3";
$j="s4";
$m1="s5";
$code=<<___;
#ifdef __linux__
#include <asm/regdef.h>
#else
#include <asm.h>
#include <regdef.h>
#endif
.text
.set noat
.set noreorder
.globl bn_mul_mont
.align 5
.ent bn_mul_mont
bn_mul_mont:
lda sp,-48(sp)
stq ra,0(sp)
stq s3,8(sp)
stq s4,16(sp)
stq s5,24(sp)
stq fp,32(sp)
mov sp,fp
.mask 0x0400f000,-48
.frame fp,48,ra
.prologue 0
.align 4
.set reorder
sextl $num,$num
mov 0,v0
cmplt $num,4,AT
bne AT,.Lexit
ldq $hi0,0($ap) # ap[0]
s8addq $num,16,AT
ldq $aj,8($ap)
subq sp,AT,sp
ldq $bi,0($bp) # bp[0]
lda AT,-4096(zero) # mov -4096,AT
ldq $n0,0($n0)
and sp,AT,sp
mulq $hi0,$bi,$lo0
ldq $hi1,0($np) # np[0]
umulh $hi0,$bi,$hi0
ldq $nj,8($np)
mulq $lo0,$n0,$m1
mulq $hi1,$m1,$lo1
umulh $hi1,$m1,$hi1
addq $lo1,$lo0,$lo1
cmpult $lo1,$lo0,AT
addq $hi1,AT,$hi1
mulq $aj,$bi,$alo
mov 2,$j
umulh $aj,$bi,$ahi
mov sp,$tp
mulq $nj,$m1,$nlo
s8addq $j,$ap,$aj
umulh $nj,$m1,$nhi
s8addq $j,$np,$nj
.align 4
.L1st:
.set noreorder
ldq $aj,0($aj)
addl $j,1,$j
ldq $nj,0($nj)
lda $tp,8($tp)
addq $alo,$hi0,$lo0
mulq $aj,$bi,$alo
cmpult $lo0,$hi0,AT
addq $nlo,$hi1,$lo1
mulq $nj,$m1,$nlo
addq $ahi,AT,$hi0
cmpult $lo1,$hi1,v0
cmplt $j,$num,$tj
umulh $aj,$bi,$ahi
addq $nhi,v0,$hi1
addq $lo1,$lo0,$lo1
s8addq $j,$ap,$aj
umulh $nj,$m1,$nhi
cmpult $lo1,$lo0,v0
addq $hi1,v0,$hi1
s8addq $j,$np,$nj
stq $lo1,-8($tp)
nop
unop
bne $tj,.L1st
.set reorder
addq $alo,$hi0,$lo0
addq $nlo,$hi1,$lo1
cmpult $lo0,$hi0,AT
cmpult $lo1,$hi1,v0
addq $ahi,AT,$hi0
addq $nhi,v0,$hi1
addq $lo1,$lo0,$lo1
cmpult $lo1,$lo0,v0
addq $hi1,v0,$hi1
stq $lo1,0($tp)
addq $hi1,$hi0,$hi1
cmpult $hi1,$hi0,AT
stq $hi1,8($tp)
stq AT,16($tp)
mov 1,$i
.align 4
.Louter:
s8addq $i,$bp,$bi
ldq $hi0,0($ap)
ldq $aj,8($ap)
ldq $bi,0($bi)
ldq $hi1,0($np)
ldq $nj,8($np)
ldq $tj,0(sp)
mulq $hi0,$bi,$lo0
umulh $hi0,$bi,$hi0
addq $lo0,$tj,$lo0
cmpult $lo0,$tj,AT
addq $hi0,AT,$hi0
mulq $lo0,$n0,$m1
mulq $hi1,$m1,$lo1
umulh $hi1,$m1,$hi1
addq $lo1,$lo0,$lo1
cmpult $lo1,$lo0,AT
mov 2,$j
addq $hi1,AT,$hi1
mulq $aj,$bi,$alo
mov sp,$tp
umulh $aj,$bi,$ahi
mulq $nj,$m1,$nlo
s8addq $j,$ap,$aj
umulh $nj,$m1,$nhi
.align 4
.Linner:
.set noreorder
ldq $tj,8($tp) #L0
nop #U1
ldq $aj,0($aj) #L1
s8addq $j,$np,$nj #U0
ldq $nj,0($nj) #L0
nop #U1
addq $alo,$hi0,$lo0 #L1
lda $tp,8($tp)
mulq $aj,$bi,$alo #U1
cmpult $lo0,$hi0,AT #L0
addq $nlo,$hi1,$lo1 #L1
addl $j,1,$j
mulq $nj,$m1,$nlo #U1
addq $ahi,AT,$hi0 #L0
addq $lo0,$tj,$lo0 #L1
cmpult $lo1,$hi1,v0 #U0
umulh $aj,$bi,$ahi #U1
cmpult $lo0,$tj,AT #L0
addq $lo1,$lo0,$lo1 #L1
addq $nhi,v0,$hi1 #U0
umulh $nj,$m1,$nhi #U1
s8addq $j,$ap,$aj #L0
cmpult $lo1,$lo0,v0 #L1
cmplt $j,$num,$tj #U0 # borrow $tj
addq $hi0,AT,$hi0 #L0
addq $hi1,v0,$hi1 #U1
stq $lo1,-8($tp) #L1
bne $tj,.Linner #U0
.set reorder
ldq $tj,8($tp)
addq $alo,$hi0,$lo0
addq $nlo,$hi1,$lo1
cmpult $lo0,$hi0,AT
cmpult $lo1,$hi1,v0
addq $ahi,AT,$hi0
addq $nhi,v0,$hi1
addq $lo0,$tj,$lo0
cmpult $lo0,$tj,AT
addq $hi0,AT,$hi0
ldq $tj,16($tp)
addq $lo1,$lo0,$j
cmpult $j,$lo0,v0
addq $hi1,v0,$hi1
addq $hi1,$hi0,$lo1
stq $j,0($tp)
cmpult $lo1,$hi0,$hi1
addq $lo1,$tj,$lo1
cmpult $lo1,$tj,AT
addl $i,1,$i
addq $hi1,AT,$hi1
stq $lo1,8($tp)
cmplt $i,$num,$tj # borrow $tj
stq $hi1,16($tp)
bne $tj,.Louter
s8addq $num,sp,$tj # &tp[num]
mov $rp,$bp # put rp aside
mov sp,$tp
mov sp,$ap
mov 0,$hi0 # clear borrow bit
.align 4
.Lsub: ldq $lo0,0($tp)
ldq $lo1,0($np)
lda $tp,8($tp)
lda $np,8($np)
subq $lo0,$lo1,$lo1 # tp[i]-np[i]
cmpult $lo0,$lo1,AT
subq $lo1,$hi0,$lo0
cmpult $lo1,$lo0,$hi0
or $hi0,AT,$hi0
stq $lo0,0($rp)
cmpult $tp,$tj,v0
lda $rp,8($rp)
bne v0,.Lsub
subq $hi1,$hi0,$hi0 # handle upmost overflow bit
mov sp,$tp
mov $bp,$rp # restore rp
and sp,$hi0,$ap
bic $bp,$hi0,$bp
bis $bp,$ap,$ap # ap=borrow?tp:rp
.align 4
.Lcopy: ldq $aj,0($ap) # copy or in-place refresh
lda $tp,8($tp)
lda $rp,8($rp)
lda $ap,8($ap)
stq zero,-8($tp) # zap tp
cmpult $tp,$tj,AT
stq $aj,-8($rp)
bne AT,.Lcopy
mov 1,v0
.Lexit:
.set noreorder
mov fp,sp
/*ldq ra,0(sp)*/
ldq s3,8(sp)
ldq s4,16(sp)
ldq s5,24(sp)
ldq fp,32(sp)
lda sp,48(sp)
ret (ra)
.end bn_mul_mont
.ascii "Montgomery Multiplication for Alpha, CRYPTOGAMS by <appro\@openssl.org>"
.align 2
___
print $code;
close STDOUT;

278
drivers/builtin_openssl2/crypto/bn/asm/armv4-gf2m.pl
View File

@ -1,278 +0,0 @@
#!/usr/bin/env perl
#
# ====================================================================
# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
#
# May 2011
#
# The module implements bn_GF2m_mul_2x2 polynomial multiplication
# used in bn_gf2m.c. It's kind of low-hanging mechanical port from
# C for the time being... Except that it has two code paths: pure
# integer code suitable for any ARMv4 and later CPU and NEON code
# suitable for ARMv7. Pure integer 1x1 multiplication subroutine runs
# in ~45 cycles on dual-issue core such as Cortex A8, which is ~50%
# faster than compiler-generated code. For ECDH and ECDSA verify (but
# not for ECDSA sign) it means 25%-45% improvement depending on key
# length, more for longer keys. Even though NEON 1x1 multiplication
# runs in even less cycles, ~30, improvement is measurable only on
# longer keys. One has to optimize code elsewhere to get NEON glow...
while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
open STDOUT,">$output";
sub Dlo() { shift=~m|q([1]?[0-9])|?"d".($1*2):""; }
sub Dhi() { shift=~m|q([1]?[0-9])|?"d".($1*2+1):""; }
sub Q() { shift=~m|d([1-3]?[02468])|?"q".($1/2):""; }
$code=<<___;
#include "arm_arch.h"
.text
.code 32
#if __ARM_ARCH__>=7
.fpu neon
.type mul_1x1_neon,%function
.align 5
mul_1x1_neon:
vshl.u64 `&Dlo("q1")`,d16,#8 @ q1-q3 are slided $a
vmull.p8 `&Q("d0")`,d16,d17 @ a·bb
vshl.u64 `&Dlo("q2")`,d16,#16
vmull.p8 q1,`&Dlo("q1")`,d17 @ a<<8·bb
vshl.u64 `&Dlo("q3")`,d16,#24
vmull.p8 q2,`&Dlo("q2")`,d17 @ a<<16·bb
vshr.u64 `&Dlo("q1")`,#8
vmull.p8 q3,`&Dlo("q3")`,d17 @ a<<24·bb
vshl.u64 `&Dhi("q1")`,#24
veor d0,`&Dlo("q1")`
vshr.u64 `&Dlo("q2")`,#16
veor d0,`&Dhi("q1")`
vshl.u64 `&Dhi("q2")`,#16
veor d0,`&Dlo("q2")`
vshr.u64 `&Dlo("q3")`,#24
veor d0,`&Dhi("q2")`
vshl.u64 `&Dhi("q3")`,#8
veor d0,`&Dlo("q3")`
veor d0,`&Dhi("q3")`
bx lr
.size mul_1x1_neon,.-mul_1x1_neon
#endif
___
################
# private interface to mul_1x1_ialu
#
$a="r1";
$b="r0";
($a0,$a1,$a2,$a12,$a4,$a14)=
($hi,$lo,$t0,$t1, $i0,$i1 )=map("r$_",(4..9),12);
$mask="r12";
$code.=<<___;
.type mul_1x1_ialu,%function
.align 5
mul_1x1_ialu:
mov $a0,#0
bic $a1,$a,#3<<30 @ a1=a&0x3fffffff
str $a0,[sp,#0] @ tab[0]=0
add $a2,$a1,$a1 @ a2=a1<<1
str $a1,[sp,#4] @ tab[1]=a1
eor $a12,$a1,$a2 @ a1^a2
str $a2,[sp,#8] @ tab[2]=a2
mov $a4,$a1,lsl#2 @ a4=a1<<2
str $a12,[sp,#12] @ tab[3]=a1^a2
eor $a14,$a1,$a4 @ a1^a4
str $a4,[sp,#16] @ tab[4]=a4
eor $a0,$a2,$a4 @ a2^a4
str $a14,[sp,#20] @ tab[5]=a1^a4
eor $a12,$a12,$a4 @ a1^a2^a4
str $a0,[sp,#24] @ tab[6]=a2^a4
and $i0,$mask,$b,lsl#2
str $a12,[sp,#28] @ tab[7]=a1^a2^a4
and $i1,$mask,$b,lsr#1
ldr $lo,[sp,$i0] @ tab[b & 0x7]
and $i0,$mask,$b,lsr#4
ldr $t1,[sp,$i1] @ tab[b >> 3 & 0x7]
and $i1,$mask,$b,lsr#7
ldr $t0,[sp,$i0] @ tab[b >> 6 & 0x7]
eor $lo,$lo,$t1,lsl#3 @ stall
mov $hi,$t1,lsr#29
ldr $t1,[sp,$i1] @ tab[b >> 9 & 0x7]
and $i0,$mask,$b,lsr#10
eor $lo,$lo,$t0,lsl#6
eor $hi,$hi,$t0,lsr#26
ldr $t0,[sp,$i0] @ tab[b >> 12 & 0x7]
and $i1,$mask,$b,lsr#13
eor $lo,$lo,$t1,lsl#9
eor $hi,$hi,$t1,lsr#23
ldr $t1,[sp,$i1] @ tab[b >> 15 & 0x7]
and $i0,$mask,$b,lsr#16
eor $lo,$lo,$t0,lsl#12
eor $hi,$hi,$t0,lsr#20
ldr $t0,[sp,$i0] @ tab[b >> 18 & 0x7]
and $i1,$mask,$b,lsr#19
eor $lo,$lo,$t1,lsl#15
eor $hi,$hi,$t1,lsr#17
ldr $t1,[sp,$i1] @ tab[b >> 21 & 0x7]
and $i0,$mask,$b,lsr#22
eor $lo,$lo,$t0,lsl#18
eor $hi,$hi,$t0,lsr#14
ldr $t0,[sp,$i0] @ tab[b >> 24 & 0x7]
and $i1,$mask,$b,lsr#25
eor $lo,$lo,$t1,lsl#21
eor $hi,$hi,$t1,lsr#11
ldr $t1,[sp,$i1] @ tab[b >> 27 & 0x7]
tst $a,#1<<30
and $i0,$mask,$b,lsr#28
eor $lo,$lo,$t0,lsl#24
eor $hi,$hi,$t0,lsr#8
ldr $t0,[sp,$i0] @ tab[b >> 30 ]
eorne $lo,$lo,$b,lsl#30
eorne $hi,$hi,$b,lsr#2
tst $a,#1<<31
eor $lo,$lo,$t1,lsl#27
eor $hi,$hi,$t1,lsr#5
eorne $lo,$lo,$b,lsl#31
eorne $hi,$hi,$b,lsr#1
eor $lo,$lo,$t0,lsl#30
eor $hi,$hi,$t0,lsr#2
mov pc,lr
.size mul_1x1_ialu,.-mul_1x1_ialu
___
################
# void bn_GF2m_mul_2x2(BN_ULONG *r,
# BN_ULONG a1,BN_ULONG a0,
# BN_ULONG b1,BN_ULONG b0); # r[3..0]=a1a0·b1b0
($A1,$B1,$A0,$B0,$A1B1,$A0B0)=map("d$_",(18..23));
$code.=<<___;
.global bn_GF2m_mul_2x2
.type bn_GF2m_mul_2x2,%function
.align 5
bn_GF2m_mul_2x2:
#if __ARM_ARCH__>=7
ldr r12,.LOPENSSL_armcap
.Lpic: ldr r12,[pc,r12]
tst r12,#1
beq .Lialu
veor $A1,$A1
vmov.32 $B1,r3,r3 @ two copies of b1
vmov.32 ${A1}[0],r1 @ a1
veor $A0,$A0
vld1.32 ${B0}[],[sp,:32] @ two copies of b0
vmov.32 ${A0}[0],r2 @ a0
mov r12,lr
vmov d16,$A1
vmov d17,$B1
bl mul_1x1_neon @ a1·b1
vmov $A1B1,d0
vmov d16,$A0
vmov d17,$B0
bl mul_1x1_neon @ a0·b0
vmov $A0B0,d0
veor d16,$A0,$A1
veor d17,$B0,$B1
veor $A0,$A0B0,$A1B1
bl mul_1x1_neon @ (a0+a1)·(b0+b1)
veor d0,$A0 @ (a0+a1)·(b0+b1)-a0·b0-a1·b1
vshl.u64 d1,d0,#32
vshr.u64 d0,d0,#32
veor $A0B0,d1
veor $A1B1,d0
vst1.32 {${A0B0}[0]},[r0,:32]!
vst1.32 {${A0B0}[1]},[r0,:32]!
vst1.32 {${A1B1}[0]},[r0,:32]!
vst1.32 {${A1B1}[1]},[r0,:32]
bx r12
.align 4
.Lialu:
#endif
___
$ret="r10"; # reassigned 1st argument
$code.=<<___;
stmdb sp!,{r4-r10,lr}
mov $ret,r0 @ reassign 1st argument
mov $b,r3 @ $b=b1
ldr r3,[sp,#32] @ load b0
mov $mask,#7<<2
sub sp,sp,#32 @ allocate tab[8]
bl mul_1x1_ialu @ a1·b1
str $lo,[$ret,#8]
str $hi,[$ret,#12]
eor $b,$b,r3 @ flip b0 and b1
eor $a,$a,r2 @ flip a0 and a1
eor r3,r3,$b
eor r2,r2,$a
eor $b,$b,r3
eor $a,$a,r2
bl mul_1x1_ialu @ a0·b0
str $lo,[$ret]
str $hi,[$ret,#4]
eor $a,$a,r2
eor $b,$b,r3
bl mul_1x1_ialu @ (a1+a0)·(b1+b0)
___
@r=map("r$_",(6..9));
$code.=<<___;
ldmia $ret,{@r[0]-@r[3]}
eor $lo,$lo,$hi
eor $hi,$hi,@r[1]
eor $lo,$lo,@r[0]
eor $hi,$hi,@r[2]
eor $lo,$lo,@r[3]
eor $hi,$hi,@r[3]
str $hi,[$ret,#8]
eor $lo,$lo,$hi
add sp,sp,#32 @ destroy tab[8]
str $lo,[$ret,#4]
#if __ARM_ARCH__>=5
ldmia sp!,{r4-r10,pc}
#else
ldmia sp!,{r4-r10,lr}
tst lr,#1
moveq pc,lr @ be binary compatible with V4, yet
bx lr @ interoperable with Thumb ISA:-)
#endif
.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2
#if __ARM_ARCH__>=7
.align 5
.LOPENSSL_armcap:
.word OPENSSL_armcap_P-(.Lpic+8)
#endif
.asciz "GF(2^m) Multiplication for ARMv4/NEON, CRYPTOGAMS by <appro\@openssl.org>"
.align 5
.comm OPENSSL_armcap_P,4,4
___
$code =~ s/\`([^\`]*)\`/eval $1/gem;
$code =~ s/\bbx\s+lr\b/.word\t0xe12fff1e/gm; # make it possible to compile with -march=armv4
print $code;
close STDOUT; # enforce flush

204
drivers/builtin_openssl2/crypto/bn/asm/armv4-mont.pl
View File

@ -1,204 +0,0 @@
#!/usr/bin/env perl
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# January 2007.
# Montgomery multiplication for ARMv4.
#
# Performance improvement naturally varies among CPU implementations
# and compilers. The code was observed to provide +65-35% improvement
# [depending on key length, less for longer keys] on ARM920T, and
# +115-80% on Intel IXP425. This is compared to pre-bn_mul_mont code
# base and compiler generated code with in-lined umull and even umlal
# instructions. The latter means that this code didn't really have an
# "advantage" of utilizing some "secret" instruction.
#
# The code is interoperable with Thumb ISA and is rather compact, less
# than 1/2KB. Windows CE port would be trivial, as it's exclusively
# about decorations, ABI and instruction syntax are identical.
while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
open STDOUT,">$output";
$num="r0"; # starts as num argument, but holds &tp[num-1]
$ap="r1";
$bp="r2"; $bi="r2"; $rp="r2";
$np="r3";
$tp="r4";
$aj="r5";
$nj="r6";
$tj="r7";
$n0="r8";
########### # r9 is reserved by ELF as platform specific, e.g. TLS pointer
$alo="r10"; # sl, gcc uses it to keep @GOT
$ahi="r11"; # fp
$nlo="r12"; # ip
########### # r13 is stack pointer
$nhi="r14"; # lr
########### # r15 is program counter
#### argument block layout relative to &tp[num-1], a.k.a. $num
$_rp="$num,#12*4";
# ap permanently resides in r1
$_bp="$num,#13*4";
# np permanently resides in r3
$_n0="$num,#14*4";
$_num="$num,#15*4"; $_bpend=$_num;
$code=<<___;
.text
.global bn_mul_mont
.type bn_mul_mont,%function
.align 2
bn_mul_mont:
stmdb sp!,{r0,r2} @ sp points at argument block
ldr $num,[sp,#3*4] @ load num
cmp $num,#2
movlt r0,#0
addlt sp,sp,#2*4
blt .Labrt
stmdb sp!,{r4-r12,lr} @ save 10 registers
mov $num,$num,lsl#2 @ rescale $num for byte count
sub sp,sp,$num @ alloca(4*num)
sub sp,sp,#4 @ +extra dword
sub $num,$num,#4 @ "num=num-1"
add $tp,$bp,$num @ &bp[num-1]
add $num,sp,$num @ $num to point at &tp[num-1]
ldr $n0,[$_n0] @ &n0
ldr $bi,[$bp] @ bp[0]
ldr $aj,[$ap],#4 @ ap[0],ap++
ldr $nj,[$np],#4 @ np[0],np++
ldr $n0,[$n0] @ *n0
str $tp,[$_bpend] @ save &bp[num]
umull $alo,$ahi,$aj,$bi @ ap[0]*bp[0]
str $n0,[$_n0] @ save n0 value
mul $n0,$alo,$n0 @ "tp[0]"*n0
mov $nlo,#0
umlal $alo,$nlo,$nj,$n0 @ np[0]*n0+"t[0]"
mov $tp,sp
.L1st:
ldr $aj,[$ap],#4 @ ap[j],ap++
mov $alo,$ahi
ldr $nj,[$np],#4 @ np[j],np++
mov $ahi,#0
umlal $alo,$ahi,$aj,$bi @ ap[j]*bp[0]
mov $nhi,#0
umlal $nlo,$nhi,$nj,$n0 @ np[j]*n0
adds $nlo,$nlo,$alo
str $nlo,[$tp],#4 @ tp[j-1]=,tp++
adc $nlo,$nhi,#0
cmp $tp,$num
bne .L1st
adds $nlo,$nlo,$ahi
ldr $tp,[$_bp] @ restore bp
mov $nhi,#0
ldr $n0,[$_n0] @ restore n0
adc $nhi,$nhi,#0
str $nlo,[$num] @ tp[num-1]=
str $nhi,[$num,#4] @ tp[num]=
.Louter:
sub $tj,$num,sp @ "original" $num-1 value
sub $ap,$ap,$tj @ "rewind" ap to &ap[1]
ldr $bi,[$tp,#4]! @ *(++bp)
sub $np,$np,$tj @ "rewind" np to &np[1]
ldr $aj,[$ap,#-4] @ ap[0]
ldr $alo,[sp] @ tp[0]
ldr $nj,[$np,#-4] @ np[0]
ldr $tj,[sp,#4] @ tp[1]
mov $ahi,#0
umlal $alo,$ahi,$aj,$bi @ ap[0]*bp[i]+tp[0]
str $tp,[$_bp] @ save bp
mul $n0,$alo,$n0
mov $nlo,#0
umlal $alo,$nlo,$nj,$n0 @ np[0]*n0+"tp[0]"
mov $tp,sp
.Linner:
ldr $aj,[$ap],#4 @ ap[j],ap++
adds $alo,$ahi,$tj @ +=tp[j]
ldr $nj,[$np],#4 @ np[j],np++
mov $ahi,#0
umlal $alo,$ahi,$aj,$bi @ ap[j]*bp[i]
mov $nhi,#0
umlal $nlo,$nhi,$nj,$n0 @ np[j]*n0
adc $ahi,$ahi,#0
ldr $tj,[$tp,#8] @ tp[j+1]
adds $nlo,$nlo,$alo
str $nlo,[$tp],#4 @ tp[j-1]=,tp++
adc $nlo,$nhi,#0
cmp $tp,$num
bne .Linner
adds $nlo,$nlo,$ahi
mov $nhi,#0
ldr $tp,[$_bp] @ restore bp
adc $nhi,$nhi,#0
ldr $n0,[$_n0] @ restore n0
adds $nlo,$nlo,$tj
ldr $tj,[$_bpend] @ restore &bp[num]
adc $nhi,$nhi,#0
str $nlo,[$num] @ tp[num-1]=
str $nhi,[$num,#4] @ tp[num]=
cmp $tp,$tj
bne .Louter
ldr $rp,[$_rp] @ pull rp
add $num,$num,#4 @ $num to point at &tp[num]
sub $aj,$num,sp @ "original" num value
mov $tp,sp @ "rewind" $tp
mov $ap,$tp @ "borrow" $ap
sub $np,$np,$aj @ "rewind" $np to &np[0]
subs $tj,$tj,$tj @ "clear" carry flag
.Lsub: ldr $tj,[$tp],#4
ldr $nj,[$np],#4
sbcs $tj,$tj,$nj @ tp[j]-np[j]
str $tj,[$rp],#4 @ rp[j]=
teq $tp,$num @ preserve carry
bne .Lsub
sbcs $nhi,$nhi,#0 @ upmost carry
mov $tp,sp @ "rewind" $tp
sub $rp,$rp,$aj @ "rewind" $rp
and $ap,$tp,$nhi
bic $np,$rp,$nhi
orr $ap,$ap,$np @ ap=borrow?tp:rp
.Lcopy: ldr $tj,[$ap],#4 @ copy or in-place refresh
str sp,[$tp],#4 @ zap tp
str $tj,[$rp],#4
cmp $tp,$num
bne .Lcopy
add sp,$num,#4 @ skip over tp[num+1]
ldmia sp!,{r4-r12,lr} @ restore registers
add sp,sp,#2*4 @ skip over {r0,r2}
mov r0,#1
.Labrt: tst lr,#1
moveq pc,lr @ be binary compatible with V4, yet
bx lr @ interoperable with Thumb ISA:-)
.size bn_mul_mont,.-bn_mul_mont
.asciz "Montgomery multiplication for ARMv4, CRYPTOGAMS by <appro\@openssl.org>"
.align 2
___
$code =~ s/\bbx\s+lr\b/.word\t0xe12fff1e/gm; # make it possible to compile with -march=armv4
print $code;
close STDOUT;

774
drivers/builtin_openssl2/crypto/bn/asm/bn-586.pl
View File

@ -1,774 +0,0 @@
#!/usr/local/bin/perl
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
push(@INC,"${dir}","${dir}../../perlasm");
require "x86asm.pl";
&asm_init($ARGV[0],$0);
$sse2=0;
for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
&external_label("OPENSSL_ia32cap_P") if ($sse2);
&bn_mul_add_words("bn_mul_add_words");
&bn_mul_words("bn_mul_words");
&bn_sqr_words("bn_sqr_words");
&bn_div_words("bn_div_words");
&bn_add_words("bn_add_words");
&bn_sub_words("bn_sub_words");
&bn_sub_part_words("bn_sub_part_words");
&asm_finish();
sub bn_mul_add_words
{
local($name)=@_;
&function_begin_B($name,$sse2?"EXTRN\t_OPENSSL_ia32cap_P:DWORD":"");
$r="eax";
$a="edx";
$c="ecx";
if ($sse2) {
&picmeup("eax","OPENSSL_ia32cap_P");
&bt(&DWP(0,"eax"),26);
&jnc(&label("maw_non_sse2"));
&mov($r,&wparam(0));
&mov($a,&wparam(1));
&mov($c,&wparam(2));
&movd("mm0",&wparam(3)); # mm0 = w
&pxor("mm1","mm1"); # mm1 = carry_in
&jmp(&label("maw_sse2_entry"));
&set_label("maw_sse2_unrolled",16);
&movd("mm3",&DWP(0,$r,"",0)); # mm3 = r[0]
&paddq("mm1","mm3"); # mm1 = carry_in + r[0]
&movd("mm2",&DWP(0,$a,"",0)); # mm2 = a[0]
&pmuludq("mm2","mm0"); # mm2 = w*a[0]
&movd("mm4",&DWP(4,$a,"",0)); # mm4 = a[1]
&pmuludq("mm4","mm0"); # mm4 = w*a[1]
&movd("mm6",&DWP(8,$a,"",0)); # mm6 = a[2]
&pmuludq("mm6","mm0"); # mm6 = w*a[2]
&movd("mm7",&DWP(12,$a,"",0)); # mm7 = a[3]
&pmuludq("mm7","mm0"); # mm7 = w*a[3]
&paddq("mm1","mm2"); # mm1 = carry_in + r[0] + w*a[0]
&movd("mm3",&DWP(4,$r,"",0)); # mm3 = r[1]
&paddq("mm3","mm4"); # mm3 = r[1] + w*a[1]
&movd("mm5",&DWP(8,$r,"",0)); # mm5 = r[2]
&paddq("mm5","mm6"); # mm5 = r[2] + w*a[2]
&movd("mm4",&DWP(12,$r,"",0)); # mm4 = r[3]
&paddq("mm7","mm4"); # mm7 = r[3] + w*a[3]
&movd(&DWP(0,$r,"",0),"mm1");
&movd("mm2",&DWP(16,$a,"",0)); # mm2 = a[4]
&pmuludq("mm2","mm0"); # mm2 = w*a[4]
&psrlq("mm1",32); # mm1 = carry0
&movd("mm4",&DWP(20,$a,"",0)); # mm4 = a[5]
&pmuludq("mm4","mm0"); # mm4 = w*a[5]
&paddq("mm1","mm3"); # mm1 = carry0 + r[1] + w*a[1]
&movd("mm6",&DWP(24,$a,"",0)); # mm6 = a[6]
&pmuludq("mm6","mm0"); # mm6 = w*a[6]
&movd(&DWP(4,$r,"",0),"mm1");
&psrlq("mm1",32); # mm1 = carry1
&movd("mm3",&DWP(28,$a,"",0)); # mm3 = a[7]
&add($a,32);
&pmuludq("mm3","mm0"); # mm3 = w*a[7]
&paddq("mm1","mm5"); # mm1 = carry1 + r[2] + w*a[2]
&movd("mm5",&DWP(16,$r,"",0)); # mm5 = r[4]
&paddq("mm2","mm5"); # mm2 = r[4] + w*a[4]
&movd(&DWP(8,$r,"",0),"mm1");
&psrlq("mm1",32); # mm1 = carry2
&paddq("mm1","mm7"); # mm1 = carry2 + r[3] + w*a[3]
&movd("mm5",&DWP(20,$r,"",0)); # mm5 = r[5]
&paddq("mm4","mm5"); # mm4 = r[5] + w*a[5]
&movd(&DWP(12,$r,"",0),"mm1");
&psrlq("mm1",32); # mm1 = carry3
&paddq("mm1","mm2"); # mm1 = carry3 + r[4] + w*a[4]
&movd("mm5",&DWP(24,$r,"",0)); # mm5 = r[6]
&paddq("mm6","mm5"); # mm6 = r[6] + w*a[6]
&movd(&DWP(16,$r,"",0),"mm1");
&psrlq("mm1",32); # mm1 = carry4
&paddq("mm1","mm4"); # mm1 = carry4 + r[5] + w*a[5]
&movd("mm5",&DWP(28,$r,"",0)); # mm5 = r[7]
&paddq("mm3","mm5"); # mm3 = r[7] + w*a[7]
&movd(&DWP(20,$r,"",0),"mm1");
&psrlq("mm1",32); # mm1 = carry5
&paddq("mm1","mm6"); # mm1 = carry5 + r[6] + w*a[6]
&movd(&DWP(24,$r,"",0),"mm1");
&psrlq("mm1",32); # mm1 = carry6
&paddq("mm1","mm3"); # mm1 = carry6 + r[7] + w*a[7]
&movd(&DWP(28,$r,"",0),"mm1");
&lea($r,&DWP(32,$r));
&psrlq("mm1",32); # mm1 = carry_out
&sub($c,8);
&jz(&label("maw_sse2_exit"));
&set_label("maw_sse2_entry");
&test($c,0xfffffff8);
&jnz(&label("maw_sse2_unrolled"));
&set_label("maw_sse2_loop",4);
&movd("mm2",&DWP(0,$a)); # mm2 = a[i]
&movd("mm3",&DWP(0,$r)); # mm3 = r[i]
&pmuludq("mm2","mm0"); # a[i] *= w
&lea($a,&DWP(4,$a));
&paddq("mm1","mm3"); # carry += r[i]
&paddq("mm1","mm2"); # carry += a[i]*w
&movd(&DWP(0,$r),"mm1"); # r[i] = carry_low
&sub($c,1);
&psrlq("mm1",32); # carry = carry_high
&lea($r,&DWP(4,$r));
&jnz(&label("maw_sse2_loop"));
&set_label("maw_sse2_exit");
&movd("eax","mm1"); # c = carry_out
&emms();
&ret();
&set_label("maw_non_sse2",16);
}
# function_begin prologue
&push("ebp");
&push("ebx");
&push("esi");
&push("edi");
&comment("");
$Low="eax";
$High="edx";
$a="ebx";
$w="ebp";
$r="edi";
$c="esi";
&xor($c,$c); # clear carry
&mov($r,&wparam(0)); #
&mov("ecx",&wparam(2)); #
&mov($a,&wparam(1)); #
&and("ecx",0xfffffff8); # num / 8
&mov($w,&wparam(3)); #
&push("ecx"); # Up the stack for a tmp variable
&jz(&label("maw_finish"));
&set_label("maw_loop",16);
for ($i=0; $i<32; $i+=4)
{
&comment("Round $i");
&mov("eax",&DWP($i,$a)); # *a
&mul($w); # *a * w
&add("eax",$c); # L(t)+= c
&adc("edx",0); # H(t)+=carry
&add("eax",&DWP($i,$r)); # L(t)+= *r
&adc("edx",0); # H(t)+=carry
&mov(&DWP($i,$r),"eax"); # *r= L(t);
&mov($c,"edx"); # c= H(t);
}
&comment("");
&sub("ecx",8);
&lea($a,&DWP(32,$a));
&lea($r,&DWP(32,$r));
&jnz(&label("maw_loop"));
&set_label("maw_finish",0);
&mov("ecx",&wparam(2)); # get num
&and("ecx",7);
&jnz(&label("maw_finish2")); # helps branch prediction
&jmp(&label("maw_end"));
&set_label("maw_finish2",1);
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov("eax",&DWP($i*4,$a)); # *a
&mul($w); # *a * w
&add("eax",$c); # L(t)+=c
&adc("edx",0); # H(t)+=carry
&add("eax",&DWP($i*4,$r)); # L(t)+= *r
&adc("edx",0); # H(t)+=carry
&dec("ecx") if ($i != 7-1);
&mov(&DWP($i*4,$r),"eax"); # *r= L(t);
&mov($c,"edx"); # c= H(t);
&jz(&label("maw_end")) if ($i != 7-1);
}
&set_label("maw_end",0);
&mov("eax",$c);
&pop("ecx"); # clear variable from
&function_end($name);
}
sub bn_mul_words
{
local($name)=@_;
&function_begin_B($name,$sse2?"EXTRN\t_OPENSSL_ia32cap_P:DWORD":"");
$r="eax";
$a="edx";
$c="ecx";
if ($sse2) {
&picmeup("eax","OPENSSL_ia32cap_P");
&bt(&DWP(0,"eax"),26);
&jnc(&label("mw_non_sse2"));
&mov($r,&wparam(0));
&mov($a,&wparam(1));
&mov($c,&wparam(2));
&movd("mm0",&wparam(3)); # mm0 = w
&pxor("mm1","mm1"); # mm1 = carry = 0
&set_label("mw_sse2_loop",16);
&movd("mm2",&DWP(0,$a)); # mm2 = a[i]
&pmuludq("mm2","mm0"); # a[i] *= w
&lea($a,&DWP(4,$a));
&paddq("mm1","mm2"); # carry += a[i]*w
&movd(&DWP(0,$r),"mm1"); # r[i] = carry_low
&sub($c,1);
&psrlq("mm1",32); # carry = carry_high
&lea($r,&DWP(4,$r));
&jnz(&label("mw_sse2_loop"));
&movd("eax","mm1"); # return carry
&emms();
&ret();
&set_label("mw_non_sse2",16);
}
# function_begin prologue
&push("ebp");
&push("ebx");
&push("esi");
&push("edi");
&comment("");
$Low="eax";
$High="edx";
$a="ebx";
$w="ecx";
$r="edi";
$c="esi";
$num="ebp";
&xor($c,$c); # clear carry
&mov($r,&wparam(0)); #
&mov($a,&wparam(1)); #
&mov($num,&wparam(2)); #
&mov($w,&wparam(3)); #
&and($num,0xfffffff8); # num / 8
&jz(&label("mw_finish"));
&set_label("mw_loop",0);
for ($i=0; $i<32; $i+=4)
{
&comment("Round $i");
&mov("eax",&DWP($i,$a,"",0)); # *a
&mul($w); # *a * w
&add("eax",$c); # L(t)+=c
# XXX
&adc("edx",0); # H(t)+=carry
&mov(&DWP($i,$r,"",0),"eax"); # *r= L(t);
&mov($c,"edx"); # c= H(t);
}
&comment("");
&add($a,32);
&add($r,32);
&sub($num,8);
&jz(&label("mw_finish"));
&jmp(&label("mw_loop"));
&set_label("mw_finish",0);
&mov($num,&wparam(2)); # get num
&and($num,7);
&jnz(&label("mw_finish2"));
&jmp(&label("mw_end"));
&set_label("mw_finish2",1);
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov("eax",&DWP($i*4,$a,"",0));# *a
&mul($w); # *a * w
&add("eax",$c); # L(t)+=c
# XXX
&adc("edx",0); # H(t)+=carry
&mov(&DWP($i*4,$r,"",0),"eax");# *r= L(t);
&mov($c,"edx"); # c= H(t);
&dec($num) if ($i != 7-1);
&jz(&label("mw_end")) if ($i != 7-1);
}
&set_label("mw_end",0);
&mov("eax",$c);
&function_end($name);
}
sub bn_sqr_words
{
local($name)=@_;
&function_begin_B($name,$sse2?"EXTRN\t_OPENSSL_ia32cap_P:DWORD":"");
$r="eax";
$a="edx";
$c="ecx";
if ($sse2) {
&picmeup("eax","OPENSSL_ia32cap_P");
&bt(&DWP(0,"eax"),26);
&jnc(&label("sqr_non_sse2"));
&mov($r,&wparam(0));
&mov($a,&wparam(1));
&mov($c,&wparam(2));
&set_label("sqr_sse2_loop",16);
&movd("mm0",&DWP(0,$a)); # mm0 = a[i]
&pmuludq("mm0","mm0"); # a[i] *= a[i]
&lea($a,&DWP(4,$a)); # a++
&movq(&QWP(0,$r),"mm0"); # r[i] = a[i]*a[i]
&sub($c,1);
&lea($r,&DWP(8,$r)); # r += 2
&jnz(&label("sqr_sse2_loop"));
&emms();
&ret();
&set_label("sqr_non_sse2",16);
}
# function_begin prologue
&push("ebp");
&push("ebx");
&push("esi");
&push("edi");
&comment("");
$r="esi";
$a="edi";
$num="ebx";
&mov($r,&wparam(0)); #
&mov($a,&wparam(1)); #
&mov($num,&wparam(2)); #
&and($num,0xfffffff8); # num / 8
&jz(&label("sw_finish"));
&set_label("sw_loop",0);
for ($i=0; $i<32; $i+=4)
{
&comment("Round $i");
&mov("eax",&DWP($i,$a,"",0)); # *a
# XXX
&mul("eax"); # *a * *a
&mov(&DWP($i*2,$r,"",0),"eax"); #
&mov(&DWP($i*2+4,$r,"",0),"edx");#
}
&comment("");
&add($a,32);
&add($r,64);
&sub($num,8);
&jnz(&label("sw_loop"));
&set_label("sw_finish",0);
&mov($num,&wparam(2)); # get num
&and($num,7);
&jz(&label("sw_end"));
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov("eax",&DWP($i*4,$a,"",0)); # *a
# XXX
&mul("eax"); # *a * *a
&mov(&DWP($i*8,$r,"",0),"eax"); #
&dec($num) if ($i != 7-1);
&mov(&DWP($i*8+4,$r,"",0),"edx");
&jz(&label("sw_end")) if ($i != 7-1);
}
&set_label("sw_end",0);
&function_end($name);
}
sub bn_div_words
{
local($name)=@_;
&function_begin_B($name,"");
&mov("edx",&wparam(0)); #
&mov("eax",&wparam(1)); #
&mov("ecx",&wparam(2)); #
&div("ecx");
&ret();
&function_end_B($name);
}
sub bn_add_words
{
local($name)=@_;
&function_begin($name,"");
&comment("");
$a="esi";
$b="edi";
$c="eax";
$r="ebx";
$tmp1="ecx";
$tmp2="edx";
$num="ebp";
&mov($r,&wparam(0)); # get r
&mov($a,&wparam(1)); # get a
&mov($b,&wparam(2)); # get b
&mov($num,&wparam(3)); # get num
&xor($c,$c); # clear carry
&and($num,0xfffffff8); # num / 8
&jz(&label("aw_finish"));
&set_label("aw_loop",0);
for ($i=0; $i<8; $i++)
{
&comment("Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov($tmp2,&DWP($i*4,$b,"",0)); # *b
&add($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&add($tmp1,$tmp2);
&adc($c,0);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
}
&comment("");
&add($a,32);
&add($b,32);
&add($r,32);
&sub($num,8);
&jnz(&label("aw_loop"));
&set_label("aw_finish",0);
&mov($num,&wparam(3)); # get num
&and($num,7);
&jz(&label("aw_end"));
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov($tmp2,&DWP($i*4,$b,"",0));# *b
&add($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&add($tmp1,$tmp2);
&adc($c,0);
&dec($num) if ($i != 6);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
&jz(&label("aw_end")) if ($i != 6);
}
&set_label("aw_end",0);
# &mov("eax",$c); # $c is "eax"
&function_end($name);
}
sub bn_sub_words
{
local($name)=@_;
&function_begin($name,"");
&comment("");
$a="esi";
$b="edi";
$c="eax";
$r="ebx";
$tmp1="ecx";
$tmp2="edx";
$num="ebp";
&mov($r,&wparam(0)); # get r
&mov($a,&wparam(1)); # get a
&mov($b,&wparam(2)); # get b
&mov($num,&wparam(3)); # get num
&xor($c,$c); # clear carry
&and($num,0xfffffff8); # num / 8
&jz(&label("aw_finish"));
&set_label("aw_loop",0);
for ($i=0; $i<8; $i++)
{
&comment("Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov($tmp2,&DWP($i*4,$b,"",0)); # *b
&sub($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&sub($tmp1,$tmp2);
&adc($c,0);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
}
&comment("");
&add($a,32);
&add($b,32);
&add($r,32);
&sub($num,8);
&jnz(&label("aw_loop"));
&set_label("aw_finish",0);
&mov($num,&wparam(3)); # get num
&and($num,7);
&jz(&label("aw_end"));
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov($tmp2,&DWP($i*4,$b,"",0));# *b
&sub($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&sub($tmp1,$tmp2);
&adc($c,0);
&dec($num) if ($i != 6);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
&jz(&label("aw_end")) if ($i != 6);
}
&set_label("aw_end",0);
# &mov("eax",$c); # $c is "eax"
&function_end($name);
}
sub bn_sub_part_words
{
local($name)=@_;
&function_begin($name,"");
&comment("");
$a="esi";
$b="edi";
$c="eax";
$r="ebx";
$tmp1="ecx";
$tmp2="edx";
$num="ebp";
&mov($r,&wparam(0)); # get r
&mov($a,&wparam(1)); # get a
&mov($b,&wparam(2)); # get b
&mov($num,&wparam(3)); # get num
&xor($c,$c); # clear carry
&and($num,0xfffffff8); # num / 8
&jz(&label("aw_finish"));
&set_label("aw_loop",0);
for ($i=0; $i<8; $i++)
{
&comment("Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov($tmp2,&DWP($i*4,$b,"",0)); # *b
&sub($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&sub($tmp1,$tmp2);
&adc($c,0);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
}
&comment("");
&add($a,32);
&add($b,32);
&add($r,32);
&sub($num,8);
&jnz(&label("aw_loop"));
&set_label("aw_finish",0);
&mov($num,&wparam(3)); # get num
&and($num,7);
&jz(&label("aw_end"));
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov($tmp1,&DWP(0,$a,"",0)); # *a
&mov($tmp2,&DWP(0,$b,"",0));# *b
&sub($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&sub($tmp1,$tmp2);
&adc($c,0);
&mov(&DWP(0,$r,"",0),$tmp1); # *r
&add($a, 4);
&add($b, 4);
&add($r, 4);
&dec($num) if ($i != 6);
&jz(&label("aw_end")) if ($i != 6);
}
&set_label("aw_end",0);
&cmp(&wparam(4),0);
&je(&label("pw_end"));
&mov($num,&wparam(4)); # get dl
&cmp($num,0);
&je(&label("pw_end"));
&jge(&label("pw_pos"));
&comment("pw_neg");
&mov($tmp2,0);
&sub($tmp2,$num);
&mov($num,$tmp2);
&and($num,0xfffffff8); # num / 8
&jz(&label("pw_neg_finish"));
&set_label("pw_neg_loop",0);
for ($i=0; $i<8; $i++)
{
&comment("dl<0 Round $i");
&mov($tmp1,0);
&mov($tmp2,&DWP($i*4,$b,"",0)); # *b
&sub($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&sub($tmp1,$tmp2);
&adc($c,0);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
}
&comment("");
&add($b,32);
&add($r,32);
&sub($num,8);
&jnz(&label("pw_neg_loop"));
&set_label("pw_neg_finish",0);
&mov($tmp2,&wparam(4)); # get dl
&mov($num,0);
&sub($num,$tmp2);
&and($num,7);
&jz(&label("pw_end"));
for ($i=0; $i<7; $i++)
{
&comment("dl<0 Tail Round $i");
&mov($tmp1,0);
&mov($tmp2,&DWP($i*4,$b,"",0));# *b
&sub($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&sub($tmp1,$tmp2);
&adc($c,0);
&dec($num) if ($i != 6);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
&jz(&label("pw_end")) if ($i != 6);
}
&jmp(&label("pw_end"));
&set_label("pw_pos",0);
&and($num,0xfffffff8); # num / 8
&jz(&label("pw_pos_finish"));
&set_label("pw_pos_loop",0);
for ($i=0; $i<8; $i++)
{
&comment("dl>0 Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&sub($tmp1,$c);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
&jnc(&label("pw_nc".$i));
}
&comment("");
&add($a,32);
&add($r,32);
&sub($num,8);
&jnz(&label("pw_pos_loop"));
&set_label("pw_pos_finish",0);
&mov($num,&wparam(4)); # get dl
&and($num,7);
&jz(&label("pw_end"));
for ($i=0; $i<7; $i++)
{
&comment("dl>0 Tail Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&sub($tmp1,$c);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
&jnc(&label("pw_tail_nc".$i));
&dec($num) if ($i != 6);
&jz(&label("pw_end")) if ($i != 6);
}
&mov($c,1);
&jmp(&label("pw_end"));
&set_label("pw_nc_loop",0);
for ($i=0; $i<8; $i++)
{
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
&set_label("pw_nc".$i,0);
}
&comment("");
&add($a,32);
&add($r,32);
&sub($num,8);
&jnz(&label("pw_nc_loop"));
&mov($num,&wparam(4)); # get dl
&and($num,7);
&jz(&label("pw_nc_end"));
for ($i=0; $i<7; $i++)
{
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
&set_label("pw_tail_nc".$i,0);
&dec($num) if ($i != 6);
&jz(&label("pw_nc_end")) if ($i != 6);
}
&set_label("pw_nc_end",0);
&mov($c,0);
&set_label("pw_end",0);
# &mov("eax",$c); # $c is "eax"
&function_end($name);
}

287
drivers/builtin_openssl2/crypto/bn/asm/co-586.pl
View File

@ -1,287 +0,0 @@
#!/usr/local/bin/perl
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
push(@INC,"${dir}","${dir}../../perlasm");
require "x86asm.pl";
&asm_init($ARGV[0],$0);
&bn_mul_comba("bn_mul_comba8",8);
&bn_mul_comba("bn_mul_comba4",4);
&bn_sqr_comba("bn_sqr_comba8",8);
&bn_sqr_comba("bn_sqr_comba4",4);
&asm_finish();
sub mul_add_c
{
local($a,$ai,$b,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
# pos == -1 if eax and edx are pre-loaded, 0 to load from next
# words, and 1 if load return value
&comment("mul a[$ai]*b[$bi]");
# "eax" and "edx" will always be pre-loaded.
# &mov("eax",&DWP($ai*4,$a,"",0)) ;
# &mov("edx",&DWP($bi*4,$b,"",0));
&mul("edx");
&add($c0,"eax");
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # laod next a
&mov("eax",&wparam(0)) if $pos > 0; # load r[]
###
&adc($c1,"edx");
&mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 0; # laod next b
&mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 1; # laod next b
###
&adc($c2,0);
# is pos > 1, it means it is the last loop
&mov(&DWP($i*4,"eax","",0),$c0) if $pos > 0; # save r[];
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # laod next a
}
sub sqr_add_c
{
local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
# pos == -1 if eax and edx are pre-loaded, 0 to load from next
# words, and 1 if load return value
&comment("sqr a[$ai]*a[$bi]");
# "eax" and "edx" will always be pre-loaded.
# &mov("eax",&DWP($ai*4,$a,"",0)) ;
# &mov("edx",&DWP($bi*4,$b,"",0));
if ($ai == $bi)
{ &mul("eax");}
else
{ &mul("edx");}
&add($c0,"eax");
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a
###
&adc($c1,"edx");
&mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos == 1) && ($na != $nb);
###
&adc($c2,0);
# is pos > 1, it means it is the last loop
&mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[];
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b
}
sub sqr_add_c2
{
local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
# pos == -1 if eax and edx are pre-loaded, 0 to load from next
# words, and 1 if load return value
&comment("sqr a[$ai]*a[$bi]");
# "eax" and "edx" will always be pre-loaded.
# &mov("eax",&DWP($ai*4,$a,"",0)) ;
# &mov("edx",&DWP($bi*4,$a,"",0));
if ($ai == $bi)
{ &mul("eax");}
else
{ &mul("edx");}
&add("eax","eax");
###
&adc("edx","edx");
###
&adc($c2,0);
&add($c0,"eax");
&adc($c1,"edx");
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b
&adc($c2,0);
&mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[];
&mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos <= 1) && ($na != $nb);
###
}
sub bn_mul_comba
{
local($name,$num)=@_;
local($a,$b,$c0,$c1,$c2);
local($i,$as,$ae,$bs,$be,$ai,$bi);
local($tot,$end);
&function_begin_B($name,"");
$c0="ebx";
$c1="ecx";
$c2="ebp";
$a="esi";
$b="edi";
$as=0;
$ae=0;
$bs=0;
$be=0;
$tot=$num+$num-1;
&push("esi");
&mov($a,&wparam(1));
&push("edi");
&mov($b,&wparam(2));
&push("ebp");
&push("ebx");
&xor($c0,$c0);
&mov("eax",&DWP(0,$a,"",0)); # load the first word
&xor($c1,$c1);
&mov("edx",&DWP(0,$b,"",0)); # load the first second
for ($i=0; $i<$tot; $i++)
{
$ai=$as;
$bi=$bs;
$end=$be+1;
&comment("################## Calculate word $i");
for ($j=$bs; $j<$end; $j++)
{
&xor($c2,$c2) if ($j == $bs);
if (($j+1) == $end)
{
$v=1;
$v=2 if (($i+1) == $tot);
}
else
{ $v=0; }
if (($j+1) != $end)
{
$na=($ai-1);
$nb=($bi+1);
}
else
{
$na=$as+($i < ($num-1));
$nb=$bs+($i >= ($num-1));
}
#printf STDERR "[$ai,$bi] -> [$na,$nb]\n";
&mul_add_c($a,$ai,$b,$bi,$c0,$c1,$c2,$v,$i,$na,$nb);
if ($v)
{
&comment("saved r[$i]");
# &mov("eax",&wparam(0));
# &mov(&DWP($i*4,"eax","",0),$c0);
($c0,$c1,$c2)=($c1,$c2,$c0);
}
$ai--;
$bi++;
}
$as++ if ($i < ($num-1));
$ae++ if ($i >= ($num-1));
$bs++ if ($i >= ($num-1));
$be++ if ($i < ($num-1));
}
&comment("save r[$i]");
# &mov("eax",&wparam(0));
&mov(&DWP($i*4,"eax","",0),$c0);
&pop("ebx");
&pop("ebp");
&pop("edi");
&pop("esi");
&ret();
&function_end_B($name);
}
sub bn_sqr_comba
{
local($name,$num)=@_;
local($r,$a,$c0,$c1,$c2)=@_;
local($i,$as,$ae,$bs,$be,$ai,$bi);
local($b,$tot,$end,$half);
&function_begin_B($name,"");
$c0="ebx";
$c1="ecx";
$c2="ebp";
$a="esi";
$r="edi";
&push("esi");
&push("edi");
&push("ebp");
&push("ebx");
&mov($r,&wparam(0));
&mov($a,&wparam(1));
&xor($c0,$c0);
&xor($c1,$c1);
&mov("eax",&DWP(0,$a,"",0)); # load the first word
$as=0;
$ae=0;
$bs=0;
$be=0;
$tot=$num+$num-1;
for ($i=0; $i<$tot; $i++)
{
$ai=$as;
$bi=$bs;
$end=$be+1;
&comment("############### Calculate word $i");
for ($j=$bs; $j<$end; $j++)
{
&xor($c2,$c2) if ($j == $bs);
if (($ai-1) < ($bi+1))
{
$v=1;
$v=2 if ($i+1) == $tot;
}
else
{ $v=0; }
if (!$v)
{
$na=$ai-1;
$nb=$bi+1;
}
else
{
$na=$as+($i < ($num-1));
$nb=$bs+($i >= ($num-1));
}
if ($ai == $bi)
{
&sqr_add_c($r,$a,$ai,$bi,
$c0,$c1,$c2,$v,$i,$na,$nb);
}
else
{
&sqr_add_c2($r,$a,$ai,$bi,
$c0,$c1,$c2,$v,$i,$na,$nb);
}
if ($v)
{
&comment("saved r[$i]");
#&mov(&DWP($i*4,$r,"",0),$c0);
($c0,$c1,$c2)=($c1,$c2,$c0);
last;
}
$ai--;
$bi++;
}
$as++ if ($i < ($num-1));
$ae++ if ($i >= ($num-1));
$bs++ if ($i >= ($num-1));
$be++ if ($i < ($num-1));
}
&mov(&DWP($i*4,$r,"",0),$c0);
&pop("ebx");
&pop("ebp");
&pop("edi");
&pop("esi");
&ret();
&function_end_B($name);
}

851
drivers/builtin_openssl2/crypto/bn/asm/ia64-mont.pl
View File

@ -1,851 +0,0 @@
#!/usr/bin/env perl
#
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# January 2010
#
# "Teaser" Montgomery multiplication module for IA-64. There are
# several possibilities for improvement:
#
# - modulo-scheduling outer loop would eliminate quite a number of
# stalls after ldf8, xma and getf.sig outside inner loop and
# improve shorter key performance;
# - shorter vector support [with input vectors being fetched only
# once] should be added;
# - 2x unroll with help of n0[1] would make the code scalable on
# "wider" IA-64, "wider" than Itanium 2 that is, which is not of
# acute interest, because upcoming Tukwila's individual cores are
# reportedly based on Itanium 2 design;
# - dedicated squaring procedure(?);
#
# January 2010
#
# Shorter vector support is implemented by zero-padding ap and np
# vectors up to 8 elements, or 512 bits. This means that 256-bit
# inputs will be processed only 2 times faster than 512-bit inputs,
# not 4 [as one would expect, because algorithm complexity is n^2].
# The reason for padding is that inputs shorter than 512 bits won't
# be processed faster anyway, because minimal critical path of the
# core loop happens to match 512-bit timing. Either way, it resulted
# in >100% improvement of 512-bit RSA sign benchmark and 50% - of
# 1024-bit one [in comparison to original version of *this* module].
#
# So far 'openssl speed rsa dsa' output on 900MHz Itanium 2 *with*
# this module is:
# sign verify sign/s verify/s
# rsa 512 bits 0.000290s 0.000024s 3452.8 42031.4
# rsa 1024 bits 0.000793s 0.000058s 1261.7 17172.0
# rsa 2048 bits 0.005908s 0.000148s 169.3 6754.0
# rsa 4096 bits 0.033456s 0.000469s 29.9 2133.6
# dsa 512 bits 0.000253s 0.000198s 3949.9 5057.0
# dsa 1024 bits 0.000585s 0.000607s 1708.4 1647.4
# dsa 2048 bits 0.001453s 0.001703s 688.1 587.4
#
# ... and *without* (but still with ia64.S):
#
# rsa 512 bits 0.000670s 0.000041s 1491.8 24145.5
# rsa 1024 bits 0.001988s 0.000080s 502.9 12499.3
# rsa 2048 bits 0.008702s 0.000189s 114.9 5293.9
# rsa 4096 bits 0.043860s 0.000533s 22.8 1875.9
# dsa 512 bits 0.000441s 0.000427s 2265.3 2340.6
# dsa 1024 bits 0.000823s 0.000867s 1215.6 1153.2
# dsa 2048 bits 0.001894s 0.002179s 528.1 458.9
#
# As it can be seen, RSA sign performance improves by 130-30%,
# hereafter less for longer keys, while verify - by 74-13%.
# DSA performance improves by 115-30%.
if ($^O eq "hpux") {
$ADDP="addp4";
for (@ARGV) { $ADDP="add" if (/[\+DD|\-mlp]64/); }
} else { $ADDP="add"; }
$code=<<___;
.explicit
.text
// int bn_mul_mont (BN_ULONG *rp,const BN_ULONG *ap,
// const BN_ULONG *bp,const BN_ULONG *np,
// const BN_ULONG *n0p,int num);
.align 64
.global bn_mul_mont#
.proc bn_mul_mont#
bn_mul_mont:
.prologue
.body
{ .mmi; cmp4.le p6,p7=2,r37;;
(p6) cmp4.lt.unc p8,p9=8,r37
mov ret0=r0 };;
{ .bbb;
(p9) br.cond.dptk.many bn_mul_mont_8
(p8) br.cond.dpnt.many bn_mul_mont_general
(p7) br.ret.spnt.many b0 };;
.endp bn_mul_mont#
prevfs=r2; prevpr=r3; prevlc=r10; prevsp=r11;
rptr=r8; aptr=r9; bptr=r14; nptr=r15;
tptr=r16; // &tp[0]
tp_1=r17; // &tp[-1]
num=r18; len=r19; lc=r20;
topbit=r21; // carry bit from tmp[num]
n0=f6;
m0=f7;
bi=f8;
.align 64
.local bn_mul_mont_general#
.proc bn_mul_mont_general#
bn_mul_mont_general:
.prologue
{ .mmi; .save ar.pfs,prevfs
alloc prevfs=ar.pfs,6,2,0,8
$ADDP aptr=0,in1
.save ar.lc,prevlc
mov prevlc=ar.lc }
{ .mmi; .vframe prevsp
mov prevsp=sp
$ADDP bptr=0,in2
.save pr,prevpr
mov prevpr=pr };;
.body
.rotf alo[6],nlo[4],ahi[8],nhi[6]
.rotr a[3],n[3],t[2]
{ .mmi; ldf8 bi=[bptr],8 // (*bp++)
ldf8 alo[4]=[aptr],16 // ap[0]
$ADDP r30=8,in1 };;
{ .mmi; ldf8 alo[3]=[r30],16 // ap[1]
ldf8 alo[2]=[aptr],16 // ap[2]
$ADDP in4=0,in4 };;
{ .mmi; ldf8 alo[1]=[r30] // ap[3]
ldf8 n0=[in4] // n0
$ADDP rptr=0,in0 }
{ .mmi; $ADDP nptr=0,in3
mov r31=16
zxt4 num=in5 };;
{ .mmi; ldf8 nlo[2]=[nptr],8 // np[0]
shladd len=num,3,r0
shladd r31=num,3,r31 };;
{ .mmi; ldf8 nlo[1]=[nptr],8 // np[1]
add lc=-5,num
sub r31=sp,r31 };;
{ .mfb; and sp=-16,r31 // alloca
xmpy.hu ahi[2]=alo[4],bi // ap[0]*bp[0]
nop.b 0 }
{ .mfb; nop.m 0
xmpy.lu alo[4]=alo[4],bi
brp.loop.imp .L1st_ctop,.L1st_cend-16
};;
{ .mfi; nop.m 0
xma.hu ahi[1]=alo[3],bi,ahi[2] // ap[1]*bp[0]
add tp_1=8,sp }
{ .mfi; nop.m 0
xma.lu alo[3]=alo[3],bi,ahi[2]
mov pr.rot=0x20001f<<16
// ------^----- (p40) at first (p23)
// ----------^^ p[16:20]=1
};;
{ .mfi; nop.m 0
xmpy.lu m0=alo[4],n0 // (ap[0]*bp[0])*n0
mov ar.lc=lc }
{ .mfi; nop.m 0
fcvt.fxu.s1 nhi[1]=f0
mov ar.ec=8 };;
.align 32
.L1st_ctop:
.pred.rel "mutex",p40,p42
{ .mfi; (p16) ldf8 alo[0]=[aptr],8 // *(aptr++)
(p18) xma.hu ahi[0]=alo[2],bi,ahi[1]
(p40) add n[2]=n[2],a[2] } // (p23) }
{ .mfi; (p18) ldf8 nlo[0]=[nptr],8 // *(nptr++)(p16)
(p18) xma.lu alo[2]=alo[2],bi,ahi[1]
(p42) add n[2]=n[2],a[2],1 };; // (p23)
{ .mfi; (p21) getf.sig a[0]=alo[5]
(p20) xma.hu nhi[0]=nlo[2],m0,nhi[1]
(p42) cmp.leu p41,p39=n[2],a[2] } // (p23)
{ .mfi; (p23) st8 [tp_1]=n[2],8
(p20) xma.lu nlo[2]=nlo[2],m0,nhi[1]
(p40) cmp.ltu p41,p39=n[2],a[2] } // (p23)
{ .mmb; (p21) getf.sig n[0]=nlo[3]
(p16) nop.m 0
br.ctop.sptk .L1st_ctop };;
.L1st_cend:
{ .mmi; getf.sig a[0]=ahi[6] // (p24)
getf.sig n[0]=nhi[4]
add num=-1,num };; // num--
{ .mmi; .pred.rel "mutex",p40,p42
(p40) add n[0]=n[0],a[0]
(p42) add n[0]=n[0],a[0],1
sub aptr=aptr,len };; // rewind
{ .mmi; .pred.rel "mutex",p40,p42
(p40) cmp.ltu p41,p39=n[0],a[0]
(p42) cmp.leu p41,p39=n[0],a[0]
sub nptr=nptr,len };;
{ .mmi; .pred.rel "mutex",p39,p41
(p39) add topbit=r0,r0
(p41) add topbit=r0,r0,1
nop.i 0 }
{ .mmi; st8 [tp_1]=n[0]
add tptr=16,sp
add tp_1=8,sp };;
.Louter:
{ .mmi; ldf8 bi=[bptr],8 // (*bp++)
ldf8 ahi[3]=[tptr] // tp[0]
add r30=8,aptr };;
{ .mmi; ldf8 alo[4]=[aptr],16 // ap[0]
ldf8 alo[3]=[r30],16 // ap[1]
add r31=8,nptr };;
{ .mfb; ldf8 alo[2]=[aptr],16 // ap[2]
xma.hu ahi[2]=alo[4],bi,ahi[3] // ap[0]*bp[i]+tp[0]
brp.loop.imp .Linner_ctop,.Linner_cend-16
}
{ .mfb; ldf8 alo[1]=[r30] // ap[3]
xma.lu alo[4]=alo[4],bi,ahi[3]
clrrrb.pr };;
{ .mfi; ldf8 nlo[2]=[nptr],16 // np[0]
xma.hu ahi[1]=alo[3],bi,ahi[2] // ap[1]*bp[i]
nop.i 0 }
{ .mfi; ldf8 nlo[1]=[r31] // np[1]
xma.lu alo[3]=alo[3],bi,ahi[2]
mov pr.rot=0x20101f<<16
// ------^----- (p40) at first (p23)
// --------^--- (p30) at first (p22)
// ----------^^ p[16:20]=1
};;
{ .mfi; st8 [tptr]=r0 // tp[0] is already accounted
xmpy.lu m0=alo[4],n0 // (ap[0]*bp[i]+tp[0])*n0
mov ar.lc=lc }
{ .mfi;
fcvt.fxu.s1 nhi[1]=f0
mov ar.ec=8 };;
// This loop spins in 4*(n+7) ticks on Itanium 2 and should spin in
// 7*(n+7) ticks on Itanium (the one codenamed Merced). Factor of 7
// in latter case accounts for two-tick pipeline stall, which means
// that its performance would be ~20% lower than optimal one. No
// attempt was made to address this, because original Itanium is
// hardly represented out in the wild...
.align 32
.Linner_ctop:
.pred.rel "mutex",p40,p42
.pred.rel "mutex",p30,p32
{ .mfi; (p16) ldf8 alo[0]=[aptr],8 // *(aptr++)
(p18) xma.hu ahi[0]=alo[2],bi,ahi[1]
(p40) add n[2]=n[2],a[2] } // (p23)
{ .mfi; (p16) nop.m 0
(p18) xma.lu alo[2]=alo[2],bi,ahi[1]
(p42) add n[2]=n[2],a[2],1 };; // (p23)
{ .mfi; (p21) getf.sig a[0]=alo[5]
(p16) nop.f 0
(p40) cmp.ltu p41,p39=n[2],a[2] } // (p23)
{ .mfi; (p21) ld8 t[0]=[tptr],8
(p16) nop.f 0
(p42) cmp.leu p41,p39=n[2],a[2] };; // (p23)
{ .mfi; (p18) ldf8 nlo[0]=[nptr],8 // *(nptr++)
(p20) xma.hu nhi[0]=nlo[2],m0,nhi[1]
(p30) add a[1]=a[1],t[1] } // (p22)
{ .mfi; (p16) nop.m 0
(p20) xma.lu nlo[2]=nlo[2],m0,nhi[1]
(p32) add a[1]=a[1],t[1],1 };; // (p22)
{ .mmi; (p21) getf.sig n[0]=nlo[3]
(p16) nop.m 0
(p30) cmp.ltu p31,p29=a[1],t[1] } // (p22)
{ .mmb; (p23) st8 [tp_1]=n[2],8
(p32) cmp.leu p31,p29=a[1],t[1] // (p22)
br.ctop.sptk .Linner_ctop };;
.Linner_cend:
{ .mmi; getf.sig a[0]=ahi[6] // (p24)
getf.sig n[0]=nhi[4]
nop.i 0 };;
{ .mmi; .pred.rel "mutex",p31,p33
(p31) add a[0]=a[0],topbit
(p33) add a[0]=a[0],topbit,1
mov topbit=r0 };;
{ .mfi; .pred.rel "mutex",p31,p33
(p31) cmp.ltu p32,p30=a[0],topbit
(p33) cmp.leu p32,p30=a[0],topbit
}
{ .mfi; .pred.rel "mutex",p40,p42
(p40) add n[0]=n[0],a[0]
(p42) add n[0]=n[0],a[0],1
};;
{ .mmi; .pred.rel "mutex",p44,p46
(p40) cmp.ltu p41,p39=n[0],a[0]
(p42) cmp.leu p41,p39=n[0],a[0]
(p32) add topbit=r0,r0,1 }
{ .mmi; st8 [tp_1]=n[0],8
cmp4.ne p6,p0=1,num
sub aptr=aptr,len };; // rewind
{ .mmi; sub nptr=nptr,len
(p41) add topbit=r0,r0,1
add tptr=16,sp }
{ .mmb; add tp_1=8,sp
add num=-1,num // num--
(p6) br.cond.sptk.many .Louter };;
{ .mbb; add lc=4,lc
brp.loop.imp .Lsub_ctop,.Lsub_cend-16
clrrrb.pr };;
{ .mii; nop.m 0
mov pr.rot=0x10001<<16
// ------^---- (p33) at first (p17)
mov ar.lc=lc }
{ .mii; nop.m 0
mov ar.ec=3
nop.i 0 };;
.Lsub_ctop:
.pred.rel "mutex",p33,p35
{ .mfi; (p16) ld8 t[0]=[tptr],8 // t=*(tp++)
(p16) nop.f 0
(p33) sub n[1]=t[1],n[1] } // (p17)
{ .mfi; (p16) ld8 n[0]=[nptr],8 // n=*(np++)
(p16) nop.f 0
(p35) sub n[1]=t[1],n[1],1 };; // (p17)
{ .mib; (p18) st8 [rptr]=n[2],8 // *(rp++)=r
(p33) cmp.gtu p34,p32=n[1],t[1] // (p17)
(p18) nop.b 0 }
{ .mib; (p18) nop.m 0
(p35) cmp.geu p34,p32=n[1],t[1] // (p17)
br.ctop.sptk .Lsub_ctop };;
.Lsub_cend:
{ .mmb; .pred.rel "mutex",p34,p36
(p34) sub topbit=topbit,r0 // (p19)
(p36) sub topbit=topbit,r0,1
brp.loop.imp .Lcopy_ctop,.Lcopy_cend-16
}
{ .mmb; sub rptr=rptr,len // rewind
sub tptr=tptr,len
clrrrb.pr };;
{ .mmi; and aptr=tptr,topbit
andcm bptr=rptr,topbit
mov pr.rot=1<<16 };;
{ .mii; or nptr=aptr,bptr
mov ar.lc=lc
mov ar.ec=3 };;
.Lcopy_ctop:
{ .mmb; (p16) ld8 n[0]=[nptr],8
(p18) st8 [tptr]=r0,8
(p16) nop.b 0 }
{ .mmb; (p16) nop.m 0
(p18) st8 [rptr]=n[2],8
br.ctop.sptk .Lcopy_ctop };;
.Lcopy_cend:
{ .mmi; mov ret0=1 // signal "handled"
rum 1<<5 // clear um.mfh
mov ar.lc=prevlc }
{ .mib; .restore sp
mov sp=prevsp
mov pr=prevpr,0x1ffff
br.ret.sptk.many b0 };;
.endp bn_mul_mont_general#
a1=r16; a2=r17; a3=r18; a4=r19; a5=r20; a6=r21; a7=r22; a8=r23;
n1=r24; n2=r25; n3=r26; n4=r27; n5=r28; n6=r29; n7=r30; n8=r31;
t0=r15;
ai0=f8; ai1=f9; ai2=f10; ai3=f11; ai4=f12; ai5=f13; ai6=f14; ai7=f15;
ni0=f16; ni1=f17; ni2=f18; ni3=f19; ni4=f20; ni5=f21; ni6=f22; ni7=f23;
.align 64
.skip 48 // aligns loop body
.local bn_mul_mont_8#
.proc bn_mul_mont_8#
bn_mul_mont_8:
.prologue
{ .mmi; .save ar.pfs,prevfs
alloc prevfs=ar.pfs,6,2,0,8
.vframe prevsp
mov prevsp=sp
.save ar.lc,prevlc
mov prevlc=ar.lc }
{ .mmi; add r17=-6*16,sp
add sp=-7*16,sp
.save pr,prevpr
mov prevpr=pr };;
{ .mmi; .save.gf 0,0x10
stf.spill [sp]=f16,-16
.save.gf 0,0x20
stf.spill [r17]=f17,32
add r16=-5*16,prevsp};;
{ .mmi; .save.gf 0,0x40
stf.spill [r16]=f18,32
.save.gf 0,0x80
stf.spill [r17]=f19,32
$ADDP aptr=0,in1 };;
{ .mmi; .save.gf 0,0x100
stf.spill [r16]=f20,32
.save.gf 0,0x200
stf.spill [r17]=f21,32
$ADDP r29=8,in1 };;
{ .mmi; .save.gf 0,0x400
stf.spill [r16]=f22
.save.gf 0,0x800
stf.spill [r17]=f23
$ADDP rptr=0,in0 };;
.body
.rotf bj[8],mj[2],tf[2],alo[10],ahi[10],nlo[10],nhi[10]
.rotr t[8]
// load input vectors padding them to 8 elements
{ .mmi; ldf8 ai0=[aptr],16 // ap[0]
ldf8 ai1=[r29],16 // ap[1]
$ADDP bptr=0,in2 }
{ .mmi; $ADDP r30=8,in2
$ADDP nptr=0,in3
$ADDP r31=8,in3 };;
{ .mmi; ldf8 bj[7]=[bptr],16 // bp[0]
ldf8 bj[6]=[r30],16 // bp[1]
cmp4.le p4,p5=3,in5 }
{ .mmi; ldf8 ni0=[nptr],16 // np[0]
ldf8 ni1=[r31],16 // np[1]
cmp4.le p6,p7=4,in5 };;
{ .mfi; (p4)ldf8 ai2=[aptr],16 // ap[2]
(p5)fcvt.fxu ai2=f0
cmp4.le p8,p9=5,in5 }
{ .mfi; (p6)ldf8 ai3=[r29],16 // ap[3]
(p7)fcvt.fxu ai3=f0
cmp4.le p10,p11=6,in5 }
{ .mfi; (p4)ldf8 bj[5]=[bptr],16 // bp[2]
(p5)fcvt.fxu bj[5]=f0
cmp4.le p12,p13=7,in5 }
{ .mfi; (p6)ldf8 bj[4]=[r30],16 // bp[3]
(p7)fcvt.fxu bj[4]=f0
cmp4.le p14,p15=8,in5 }
{ .mfi; (p4)ldf8 ni2=[nptr],16 // np[2]
(p5)fcvt.fxu ni2=f0
addp4 r28=-1,in5 }
{ .mfi; (p6)ldf8 ni3=[r31],16 // np[3]
(p7)fcvt.fxu ni3=f0
$ADDP in4=0,in4 };;
{ .mfi; ldf8 n0=[in4]
fcvt.fxu tf[1]=f0
nop.i 0 }
{ .mfi; (p8)ldf8 ai4=[aptr],16 // ap[4]
(p9)fcvt.fxu ai4=f0
mov t[0]=r0 }
{ .mfi; (p10)ldf8 ai5=[r29],16 // ap[5]
(p11)fcvt.fxu ai5=f0
mov t[1]=r0 }
{ .mfi; (p8)ldf8 bj[3]=[bptr],16 // bp[4]
(p9)fcvt.fxu bj[3]=f0
mov t[2]=r0 }
{ .mfi; (p10)ldf8 bj[2]=[r30],16 // bp[5]
(p11)fcvt.fxu bj[2]=f0
mov t[3]=r0 }
{ .mfi; (p8)ldf8 ni4=[nptr],16 // np[4]
(p9)fcvt.fxu ni4=f0
mov t[4]=r0 }
{ .mfi; (p10)ldf8 ni5=[r31],16 // np[5]
(p11)fcvt.fxu ni5=f0
mov t[5]=r0 };;
{ .mfi; (p12)ldf8 ai6=[aptr],16 // ap[6]
(p13)fcvt.fxu ai6=f0
mov t[6]=r0 }
{ .mfi; (p14)ldf8 ai7=[r29],16 // ap[7]
(p15)fcvt.fxu ai7=f0
mov t[7]=r0 }
{ .mfi; (p12)ldf8 bj[1]=[bptr],16 // bp[6]
(p13)fcvt.fxu bj[1]=f0
mov ar.lc=r28 }
{ .mfi; (p14)ldf8 bj[0]=[r30],16 // bp[7]
(p15)fcvt.fxu bj[0]=f0
mov ar.ec=1 }
{ .mfi; (p12)ldf8 ni6=[nptr],16 // np[6]
(p13)fcvt.fxu ni6=f0
mov pr.rot=1<<16 }
{ .mfb; (p14)ldf8 ni7=[r31],16 // np[7]
(p15)fcvt.fxu ni7=f0
brp.loop.imp .Louter_8_ctop,.Louter_8_cend-16
};;
// The loop is scheduled for 32*n ticks on Itanium 2. Actual attempt
// to measure with help of Interval Time Counter indicated that the
// factor is a tad higher: 33 or 34, if not 35. Exact measurement and
// addressing the issue is problematic, because I don't have access
// to platform-specific instruction-level profiler. On Itanium it
// should run in 56*n ticks, because of higher xma latency...
.Louter_8_ctop:
.pred.rel "mutex",p40,p42
.pred.rel "mutex",p48,p50
{ .mfi; (p16) nop.m 0 // 0:
(p16) xma.hu ahi[0]=ai0,bj[7],tf[1] // ap[0]*b[i]+t[0]
(p40) add a3=a3,n3 } // (p17) a3+=n3
{ .mfi; (p42) add a3=a3,n3,1
(p16) xma.lu alo[0]=ai0,bj[7],tf[1]
(p16) nop.i 0 };;
{ .mii; (p17) getf.sig a7=alo[8] // 1:
(p48) add t[6]=t[6],a3 // (p17) t[6]+=a3
(p50) add t[6]=t[6],a3,1 };;
{ .mfi; (p17) getf.sig a8=ahi[8] // 2:
(p17) xma.hu nhi[7]=ni6,mj[1],nhi[6] // np[6]*m0
(p40) cmp.ltu p43,p41=a3,n3 }
{ .mfi; (p42) cmp.leu p43,p41=a3,n3
(p17) xma.lu nlo[7]=ni6,mj[1],nhi[6]
(p16) nop.i 0 };;
{ .mii; (p17) getf.sig n5=nlo[6] // 3:
(p48) cmp.ltu p51,p49=t[6],a3
(p50) cmp.leu p51,p49=t[6],a3 };;
.pred.rel "mutex",p41,p43
.pred.rel "mutex",p49,p51
{ .mfi; (p16) nop.m 0 // 4:
(p16) xma.hu ahi[1]=ai1,bj[7],ahi[0] // ap[1]*b[i]
(p41) add a4=a4,n4 } // (p17) a4+=n4
{ .mfi; (p43) add a4=a4,n4,1
(p16) xma.lu alo[1]=ai1,bj[7],ahi[0]
(p16) nop.i 0 };;
{ .mfi; (p49) add t[5]=t[5],a4 // 5: (p17) t[5]+=a4
(p16) xmpy.lu mj[0]=alo[0],n0 // (ap[0]*b[i]+t[0])*n0
(p51) add t[5]=t[5],a4,1 };;
{ .mfi; (p16) nop.m 0 // 6:
(p17) xma.hu nhi[8]=ni7,mj[1],nhi[7] // np[7]*m0
(p41) cmp.ltu p42,p40=a4,n4 }
{ .mfi; (p43) cmp.leu p42,p40=a4,n4
(p17) xma.lu nlo[8]=ni7,mj[1],nhi[7]
(p16) nop.i 0 };;
{ .mii; (p17) getf.sig n6=nlo[7] // 7:
(p49) cmp.ltu p50,p48=t[5],a4
(p51) cmp.leu p50,p48=t[5],a4 };;
.pred.rel "mutex",p40,p42
.pred.rel "mutex",p48,p50
{ .mfi; (p16) nop.m 0 // 8:
(p16) xma.hu ahi[2]=ai2,bj[7],ahi[1] // ap[2]*b[i]
(p40) add a5=a5,n5 } // (p17) a5+=n5
{ .mfi; (p42) add a5=a5,n5,1
(p16) xma.lu alo[2]=ai2,bj[7],ahi[1]
(p16) nop.i 0 };;
{ .mii; (p16) getf.sig a1=alo[1] // 9:
(p48) add t[4]=t[4],a5 // p(17) t[4]+=a5
(p50) add t[4]=t[4],a5,1 };;
{ .mfi; (p16) nop.m 0 // 10:
(p16) xma.hu nhi[0]=ni0,mj[0],alo[0] // np[0]*m0
(p40) cmp.ltu p43,p41=a5,n5 }
{ .mfi; (p42) cmp.leu p43,p41=a5,n5
(p16) xma.lu nlo[0]=ni0,mj[0],alo[0]
(p16) nop.i 0 };;
{ .mii; (p17) getf.sig n7=nlo[8] // 11:
(p48) cmp.ltu p51,p49=t[4],a5
(p50) cmp.leu p51,p49=t[4],a5 };;
.pred.rel "mutex",p41,p43
.pred.rel "mutex",p49,p51
{ .mfi; (p17) getf.sig n8=nhi[8] // 12:
(p16) xma.hu ahi[3]=ai3,bj[7],ahi[2] // ap[3]*b[i]
(p41) add a6=a6,n6 } // (p17) a6+=n6
{ .mfi; (p43) add a6=a6,n6,1
(p16) xma.lu alo[3]=ai3,bj[7],ahi[2]
(p16) nop.i 0 };;
{ .mii; (p16) getf.sig a2=alo[2] // 13:
(p49) add t[3]=t[3],a6 // (p17) t[3]+=a6
(p51) add t[3]=t[3],a6,1 };;
{ .mfi; (p16) nop.m 0 // 14:
(p16) xma.hu nhi[1]=ni1,mj[0],nhi[0] // np[1]*m0
(p41) cmp.ltu p42,p40=a6,n6 }
{ .mfi; (p43) cmp.leu p42,p40=a6,n6
(p16) xma.lu nlo[1]=ni1,mj[0],nhi[0]
(p16) nop.i 0 };;
{ .mii; (p16) nop.m 0 // 15:
(p49) cmp.ltu p50,p48=t[3],a6
(p51) cmp.leu p50,p48=t[3],a6 };;
.pred.rel "mutex",p40,p42
.pred.rel "mutex",p48,p50
{ .mfi; (p16) nop.m 0 // 16:
(p16) xma.hu ahi[4]=ai4,bj[7],ahi[3] // ap[4]*b[i]
(p40) add a7=a7,n7 } // (p17) a7+=n7
{ .mfi; (p42) add a7=a7,n7,1
(p16) xma.lu alo[4]=ai4,bj[7],ahi[3]
(p16) nop.i 0 };;
{ .mii; (p16) getf.sig a3=alo[3] // 17:
(p48) add t[2]=t[2],a7 // (p17) t[2]+=a7
(p50) add t[2]=t[2],a7,1 };;
{ .mfi; (p16) nop.m 0 // 18:
(p16) xma.hu nhi[2]=ni2,mj[0],nhi[1] // np[2]*m0
(p40) cmp.ltu p43,p41=a7,n7 }
{ .mfi; (p42) cmp.leu p43,p41=a7,n7
(p16) xma.lu nlo[2]=ni2,mj[0],nhi[1]
(p16) nop.i 0 };;
{ .mii; (p16) getf.sig n1=nlo[1] // 19:
(p48) cmp.ltu p51,p49=t[2],a7
(p50) cmp.leu p51,p49=t[2],a7 };;
.pred.rel "mutex",p41,p43
.pred.rel "mutex",p49,p51
{ .mfi; (p16) nop.m 0 // 20:
(p16) xma.hu ahi[5]=ai5,bj[7],ahi[4] // ap[5]*b[i]
(p41) add a8=a8,n8 } // (p17) a8+=n8
{ .mfi; (p43) add a8=a8,n8,1
(p16) xma.lu alo[5]=ai5,bj[7],ahi[4]
(p16) nop.i 0 };;
{ .mii; (p16) getf.sig a4=alo[4] // 21:
(p49) add t[1]=t[1],a8 // (p17) t[1]+=a8
(p51) add t[1]=t[1],a8,1 };;
{ .mfi; (p16) nop.m 0 // 22:
(p16) xma.hu nhi[3]=ni3,mj[0],nhi[2] // np[3]*m0
(p41) cmp.ltu p42,p40=a8,n8 }
{ .mfi; (p43) cmp.leu p42,p40=a8,n8
(p16) xma.lu nlo[3]=ni3,mj[0],nhi[2]
(p16) nop.i 0 };;
{ .mii; (p16) getf.sig n2=nlo[2] // 23:
(p49) cmp.ltu p50,p48=t[1],a8
(p51) cmp.leu p50,p48=t[1],a8 };;
{ .mfi; (p16) nop.m 0 // 24:
(p16) xma.hu ahi[6]=ai6,bj[7],ahi[5] // ap[6]*b[i]
(p16) add a1=a1,n1 } // (p16) a1+=n1
{ .mfi; (p16) nop.m 0
(p16) xma.lu alo[6]=ai6,bj[7],ahi[5]
(p17) mov t[0]=r0 };;
{ .mii; (p16) getf.sig a5=alo[5] // 25:
(p16) add t0=t[7],a1 // (p16) t[7]+=a1
(p42) add t[0]=t[0],r0,1 };;
{ .mfi; (p16) setf.sig tf[0]=t0 // 26:
(p16) xma.hu nhi[4]=ni4,mj[0],nhi[3] // np[4]*m0
(p50) add t[0]=t[0],r0,1 }
{ .mfi; (p16) cmp.ltu.unc p42,p40=a1,n1
(p16) xma.lu nlo[4]=ni4,mj[0],nhi[3]
(p16) nop.i 0 };;
{ .mii; (p16) getf.sig n3=nlo[3] // 27:
(p16) cmp.ltu.unc p50,p48=t0,a1
(p16) nop.i 0 };;
.pred.rel "mutex",p40,p42
.pred.rel "mutex",p48,p50
{ .mfi; (p16) nop.m 0 // 28:
(p16) xma.hu ahi[7]=ai7,bj[7],ahi[6] // ap[7]*b[i]
(p40) add a2=a2,n2 } // (p16) a2+=n2
{ .mfi; (p42) add a2=a2,n2,1
(p16) xma.lu alo[7]=ai7,bj[7],ahi[6]
(p16) nop.i 0 };;
{ .mii; (p16) getf.sig a6=alo[6] // 29:
(p48) add t[6]=t[6],a2 // (p16) t[6]+=a2
(p50) add t[6]=t[6],a2,1 };;
{ .mfi; (p16) nop.m 0 // 30:
(p16) xma.hu nhi[5]=ni5,mj[0],nhi[4] // np[5]*m0
(p40) cmp.ltu p41,p39=a2,n2 }
{ .mfi; (p42) cmp.leu p41,p39=a2,n2
(p16) xma.lu nlo[5]=ni5,mj[0],nhi[4]
(p16) nop.i 0 };;
{ .mfi; (p16) getf.sig n4=nlo[4] // 31:
(p16) nop.f 0
(p48) cmp.ltu p49,p47=t[6],a2 }
{ .mfb; (p50) cmp.leu p49,p47=t[6],a2
(p16) nop.f 0
br.ctop.sptk.many .Louter_8_ctop };;
.Louter_8_cend:
// above loop has to execute one more time, without (p16), which is
// replaced with merged move of np[8] to GPR bank
.pred.rel "mutex",p40,p42
.pred.rel "mutex",p48,p50
{ .mmi; (p0) getf.sig n1=ni0 // 0:
(p40) add a3=a3,n3 // (p17) a3+=n3
(p42) add a3=a3,n3,1 };;
{ .mii; (p17) getf.sig a7=alo[8] // 1:
(p48) add t[6]=t[6],a3 // (p17) t[6]+=a3
(p50) add t[6]=t[6],a3,1 };;
{ .mfi; (p17) getf.sig a8=ahi[8] // 2:
(p17) xma.hu nhi[7]=ni6,mj[1],nhi[6] // np[6]*m0
(p40) cmp.ltu p43,p41=a3,n3 }
{ .mfi; (p42) cmp.leu p43,p41=a3,n3
(p17) xma.lu nlo[7]=ni6,mj[1],nhi[6]
(p0) nop.i 0 };;
{ .mii; (p17) getf.sig n5=nlo[6] // 3:
(p48) cmp.ltu p51,p49=t[6],a3
(p50) cmp.leu p51,p49=t[6],a3 };;
.pred.rel "mutex",p41,p43
.pred.rel "mutex",p49,p51
{ .mmi; (p0) getf.sig n2=ni1 // 4:
(p41) add a4=a4,n4 // (p17) a4+=n4
(p43) add a4=a4,n4,1 };;
{ .mfi; (p49) add t[5]=t[5],a4 // 5: (p17) t[5]+=a4
(p0) nop.f 0
(p51) add t[5]=t[5],a4,1 };;
{ .mfi; (p0) getf.sig n3=ni2 // 6:
(p17) xma.hu nhi[8]=ni7,mj[1],nhi[7] // np[7]*m0
(p41) cmp.ltu p42,p40=a4,n4 }
{ .mfi; (p43) cmp.leu p42,p40=a4,n4
(p17) xma.lu nlo[8]=ni7,mj[1],nhi[7]
(p0) nop.i 0 };;
{ .mii; (p17) getf.sig n6=nlo[7] // 7:
(p49) cmp.ltu p50,p48=t[5],a4
(p51) cmp.leu p50,p48=t[5],a4 };;
.pred.rel "mutex",p40,p42
.pred.rel "mutex",p48,p50
{ .mii; (p0) getf.sig n4=ni3 // 8:
(p40) add a5=a5,n5 // (p17) a5+=n5
(p42) add a5=a5,n5,1 };;
{ .mii; (p0) nop.m 0 // 9:
(p48) add t[4]=t[4],a5 // p(17) t[4]+=a5
(p50) add t[4]=t[4],a5,1 };;
{ .mii; (p0) nop.m 0 // 10:
(p40) cmp.ltu p43,p41=a5,n5
(p42) cmp.leu p43,p41=a5,n5 };;
{ .mii; (p17) getf.sig n7=nlo[8] // 11:
(p48) cmp.ltu p51,p49=t[4],a5
(p50) cmp.leu p51,p49=t[4],a5 };;
.pred.rel "mutex",p41,p43
.pred.rel "mutex",p49,p51
{ .mii; (p17) getf.sig n8=nhi[8] // 12:
(p41) add a6=a6,n6 // (p17) a6+=n6
(p43) add a6=a6,n6,1 };;
{ .mii; (p0) getf.sig n5=ni4 // 13:
(p49) add t[3]=t[3],a6 // (p17) t[3]+=a6
(p51) add t[3]=t[3],a6,1 };;
{ .mii; (p0) nop.m 0 // 14:
(p41) cmp.ltu p42,p40=a6,n6
(p43) cmp.leu p42,p40=a6,n6 };;
{ .mii; (p0) getf.sig n6=ni5 // 15:
(p49) cmp.ltu p50,p48=t[3],a6
(p51) cmp.leu p50,p48=t[3],a6 };;
.pred.rel "mutex",p40,p42
.pred.rel "mutex",p48,p50
{ .mii; (p0) nop.m 0 // 16:
(p40) add a7=a7,n7 // (p17) a7+=n7
(p42) add a7=a7,n7,1 };;
{ .mii; (p0) nop.m 0 // 17:
(p48) add t[2]=t[2],a7 // (p17) t[2]+=a7
(p50) add t[2]=t[2],a7,1 };;
{ .mii; (p0) nop.m 0 // 18:
(p40) cmp.ltu p43,p41=a7,n7
(p42) cmp.leu p43,p41=a7,n7 };;
{ .mii; (p0) getf.sig n7=ni6 // 19:
(p48) cmp.ltu p51,p49=t[2],a7
(p50) cmp.leu p51,p49=t[2],a7 };;
.pred.rel "mutex",p41,p43
.pred.rel "mutex",p49,p51
{ .mii; (p0) nop.m 0 // 20:
(p41) add a8=a8,n8 // (p17) a8+=n8
(p43) add a8=a8,n8,1 };;
{ .mmi; (p0) nop.m 0 // 21:
(p49) add t[1]=t[1],a8 // (p17) t[1]+=a8
(p51) add t[1]=t[1],a8,1 }
{ .mmi; (p17) mov t[0]=r0
(p41) cmp.ltu p42,p40=a8,n8
(p43) cmp.leu p42,p40=a8,n8 };;
{ .mmi; (p0) getf.sig n8=ni7 // 22:
(p49) cmp.ltu p50,p48=t[1],a8
(p51) cmp.leu p50,p48=t[1],a8 }
{ .mmi; (p42) add t[0]=t[0],r0,1
(p0) add r16=-7*16,prevsp
(p0) add r17=-6*16,prevsp };;
// subtract np[8] from carrybit|tmp[8]
// carrybit|tmp[8] layout upon exit from above loop is:
// t[0]|t[1]|t[2]|t[3]|t[4]|t[5]|t[6]|t[7]|t0 (least significant)
{ .mmi; (p50)add t[0]=t[0],r0,1
add r18=-5*16,prevsp
sub n1=t0,n1 };;
{ .mmi; cmp.gtu p34,p32=n1,t0;;
.pred.rel "mutex",p32,p34
(p32)sub n2=t[7],n2
(p34)sub n2=t[7],n2,1 };;
{ .mii; (p32)cmp.gtu p35,p33=n2,t[7]
(p34)cmp.geu p35,p33=n2,t[7];;
.pred.rel "mutex",p33,p35
(p33)sub n3=t[6],n3 }
{ .mmi; (p35)sub n3=t[6],n3,1;;
(p33)cmp.gtu p34,p32=n3,t[6]
(p35)cmp.geu p34,p32=n3,t[6] };;
.pred.rel "mutex",p32,p34
{ .mii; (p32)sub n4=t[5],n4
(p34)sub n4=t[5],n4,1;;
(p32)cmp.gtu p35,p33=n4,t[5] }
{ .mmi; (p34)cmp.geu p35,p33=n4,t[5];;
.pred.rel "mutex",p33,p35
(p33)sub n5=t[4],n5
(p35)sub n5=t[4],n5,1 };;
{ .mii; (p33)cmp.gtu p34,p32=n5,t[4]
(p35)cmp.geu p34,p32=n5,t[4];;
.pred.rel "mutex",p32,p34
(p32)sub n6=t[3],n6 }
{ .mmi; (p34)sub n6=t[3],n6,1;;
(p32)cmp.gtu p35,p33=n6,t[3]
(p34)cmp.geu p35,p33=n6,t[3] };;
.pred.rel "mutex",p33,p35
{ .mii; (p33)sub n7=t[2],n7
(p35)sub n7=t[2],n7,1;;
(p33)cmp.gtu p34,p32=n7,t[2] }
{ .mmi; (p35)cmp.geu p34,p32=n7,t[2];;
.pred.rel "mutex",p32,p34
(p32)sub n8=t[1],n8
(p34)sub n8=t[1],n8,1 };;
{ .mii; (p32)cmp.gtu p35,p33=n8,t[1]
(p34)cmp.geu p35,p33=n8,t[1];;
.pred.rel "mutex",p33,p35
(p33)sub a8=t[0],r0 }
{ .mmi; (p35)sub a8=t[0],r0,1;;
(p33)cmp.gtu p34,p32=a8,t[0]
(p35)cmp.geu p34,p32=a8,t[0] };;
// save the result, either tmp[num] or tmp[num]-np[num]
.pred.rel "mutex",p32,p34
{ .mmi; (p32)st8 [rptr]=n1,8
(p34)st8 [rptr]=t0,8
add r19=-4*16,prevsp};;
{ .mmb; (p32)st8 [rptr]=n2,8
(p34)st8 [rptr]=t[7],8
(p5)br.cond.dpnt.few .Ldone };;
{ .mmb; (p32)st8 [rptr]=n3,8
(p34)st8 [rptr]=t[6],8
(p7)br.cond.dpnt.few .Ldone };;
{ .mmb; (p32)st8 [rptr]=n4,8
(p34)st8 [rptr]=t[5],8
(p9)br.cond.dpnt.few .Ldone };;
{ .mmb; (p32)st8 [rptr]=n5,8
(p34)st8 [rptr]=t[4],8
(p11)br.cond.dpnt.few .Ldone };;
{ .mmb; (p32)st8 [rptr]=n6,8
(p34)st8 [rptr]=t[3],8
(p13)br.cond.dpnt.few .Ldone };;
{ .mmb; (p32)st8 [rptr]=n7,8
(p34)st8 [rptr]=t[2],8
(p15)br.cond.dpnt.few .Ldone };;
{ .mmb; (p32)st8 [rptr]=n8,8
(p34)st8 [rptr]=t[1],8
nop.b 0 };;
.Ldone: // epilogue
{ .mmi; ldf.fill f16=[r16],64
ldf.fill f17=[r17],64
nop.i 0 }
{ .mmi; ldf.fill f18=[r18],64
ldf.fill f19=[r19],64
mov pr=prevpr,0x1ffff };;
{ .mmi; ldf.fill f20=[r16]
ldf.fill f21=[r17]
mov ar.lc=prevlc }
{ .mmi; ldf.fill f22=[r18]
ldf.fill f23=[r19]
mov ret0=1 } // signal "handled"
{ .mib; rum 1<<5
.restore sp
mov sp=prevsp
br.ret.sptk.many b0 };;
.endp bn_mul_mont_8#
.type copyright#,\@object
copyright:
stringz "Montgomery multiplication for IA-64, CRYPTOGAMS by <appro\@openssl.org>"
___
$output=shift and open STDOUT,">$output";
print $code;
close STDOUT;

1555
drivers/builtin_openssl2/crypto/bn/asm/ia64.S
View File

File diff suppressed because it is too large Load Diff

426
drivers/builtin_openssl2/crypto/bn/asm/mips-mont.pl
View File

@ -1,426 +0,0 @@
#!/usr/bin/env perl
#
# ====================================================================
# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# This module doesn't present direct interest for OpenSSL, because it
# doesn't provide better performance for longer keys, at least not on
# in-order-execution cores. While 512-bit RSA sign operations can be
# 65% faster in 64-bit mode, 1024-bit ones are only 15% faster, and
# 4096-bit ones are up to 15% slower. In 32-bit mode it varies from
# 16% improvement for 512-bit RSA sign to -33% for 4096-bit RSA
# verify:-( All comparisons are against bn_mul_mont-free assembler.
# The module might be of interest to embedded system developers, as
# the code is smaller than 1KB, yet offers >3x improvement on MIPS64
# and 75-30% [less for longer keys] on MIPS32 over compiler-generated
# code.
######################################################################
# There is a number of MIPS ABI in use, O32 and N32/64 are most
# widely used. Then there is a new contender: NUBI. It appears that if
# one picks the latter, it's possible to arrange code in ABI neutral
# manner. Therefore let's stick to NUBI register layout:
#
($zero,$at,$t0,$t1,$t2)=map("\$$_",(0..2,24,25));
($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7,$s8,$s9,$s10,$s11)=map("\$$_",(12..23));
($gp,$tp,$sp,$fp,$ra)=map("\$$_",(3,28..31));
#
# The return value is placed in $a0. Following coding rules facilitate
# interoperability:
#
# - never ever touch $tp, "thread pointer", former $gp;
# - copy return value to $t0, former $v0 [or to $a0 if you're adapting
# old code];
# - on O32 populate $a4-$a7 with 'lw $aN,4*N($sp)' if necessary;
#
# For reference here is register layout for N32/64 MIPS ABIs:
#
# ($zero,$at,$v0,$v1)=map("\$$_",(0..3));
# ($a0,$a1,$a2,$a3,$a4,$a5,$a6,$a7)=map("\$$_",(4..11));
# ($t0,$t1,$t2,$t3,$t8,$t9)=map("\$$_",(12..15,24,25));
# ($s0,$s1,$s2,$s3,$s4,$s5,$s6,$s7)=map("\$$_",(16..23));
# ($gp,$sp,$fp,$ra)=map("\$$_",(28..31));
#
$flavour = shift; # supported flavours are o32,n32,64,nubi32,nubi64
if ($flavour =~ /64|n32/i) {
$PTR_ADD="dadd"; # incidentally works even on n32
$PTR_SUB="dsub"; # incidentally works even on n32
$REG_S="sd";
$REG_L="ld";
$SZREG=8;
} else {
$PTR_ADD="add";
$PTR_SUB="sub";
$REG_S="sw";
$REG_L="lw";
$SZREG=4;
}
$SAVED_REGS_MASK = ($flavour =~ /nubi/i) ? 0x00fff000 : 0x00ff0000;
#
# <appro@openssl.org>
#
######################################################################
while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
open STDOUT,">$output";
if ($flavour =~ /64|n32/i) {
$LD="ld";
$ST="sd";
$MULTU="dmultu";
$ADDU="daddu";
$SUBU="dsubu";
$BNSZ=8;
} else {
$LD="lw";
$ST="sw";
$MULTU="multu";
$ADDU="addu";
$SUBU="subu";
$BNSZ=4;
}
# int bn_mul_mont(
$rp=$a0; # BN_ULONG *rp,
$ap=$a1; # const BN_ULONG *ap,
$bp=$a2; # const BN_ULONG *bp,
$np=$a3; # const BN_ULONG *np,
$n0=$a4; # const BN_ULONG *n0,
$num=$a5; # int num);
$lo0=$a6;
$hi0=$a7;
$lo1=$t1;
$hi1=$t2;
$aj=$s0;
$bi=$s1;
$nj=$s2;
$tp=$s3;
$alo=$s4;
$ahi=$s5;
$nlo=$s6;
$nhi=$s7;
$tj=$s8;
$i=$s9;
$j=$s10;
$m1=$s11;
$FRAMESIZE=14;
$code=<<___;
.text
.set noat
.set noreorder
.align 5
.globl bn_mul_mont
.ent bn_mul_mont
bn_mul_mont:
___
$code.=<<___ if ($flavour =~ /o32/i);
lw $n0,16($sp)
lw $num,20($sp)
___
$code.=<<___;
slt $at,$num,4
bnez $at,1f
li $t0,0
slt $at,$num,17 # on in-order CPU
bnez $at,bn_mul_mont_internal
nop
1: jr $ra
li $a0,0
.end bn_mul_mont
.align 5
.ent bn_mul_mont_internal
bn_mul_mont_internal:
.frame $fp,$FRAMESIZE*$SZREG,$ra
.mask 0x40000000|$SAVED_REGS_MASK,-$SZREG
$PTR_SUB $sp,$FRAMESIZE*$SZREG
$REG_S $fp,($FRAMESIZE-1)*$SZREG($sp)
$REG_S $s11,($FRAMESIZE-2)*$SZREG($sp)
$REG_S $s10,($FRAMESIZE-3)*$SZREG($sp)
$REG_S $s9,($FRAMESIZE-4)*$SZREG($sp)
$REG_S $s8,($FRAMESIZE-5)*$SZREG($sp)
$REG_S $s7,($FRAMESIZE-6)*$SZREG($sp)
$REG_S $s6,($FRAMESIZE-7)*$SZREG($sp)
$REG_S $s5,($FRAMESIZE-8)*$SZREG($sp)
$REG_S $s4,($FRAMESIZE-9)*$SZREG($sp)
___
$code.=<<___ if ($flavour =~ /nubi/i);
$REG_S $s3,($FRAMESIZE-10)*$SZREG($sp)
$REG_S $s2,($FRAMESIZE-11)*$SZREG($sp)
$REG_S $s1,($FRAMESIZE-12)*$SZREG($sp)
$REG_S $s0,($FRAMESIZE-13)*$SZREG($sp)
___
$code.=<<___;
move $fp,$sp
.set reorder
$LD $n0,0($n0)
$LD $bi,0($bp) # bp[0]
$LD $aj,0($ap) # ap[0]
$LD $nj,0($np) # np[0]
$PTR_SUB $sp,2*$BNSZ # place for two extra words
sll $num,`log($BNSZ)/log(2)`
li $at,-4096
$PTR_SUB $sp,$num
and $sp,$at
$MULTU $aj,$bi
$LD $alo,$BNSZ($ap)
$LD $nlo,$BNSZ($np)
mflo $lo0
mfhi $hi0
$MULTU $lo0,$n0
mflo $m1
$MULTU $alo,$bi
mflo $alo
mfhi $ahi
$MULTU $nj,$m1
mflo $lo1
mfhi $hi1
$MULTU $nlo,$m1
$ADDU $lo1,$lo0
sltu $at,$lo1,$lo0
$ADDU $hi1,$at
mflo $nlo
mfhi $nhi
move $tp,$sp
li $j,2*$BNSZ
.align 4
.L1st:
.set noreorder
$PTR_ADD $aj,$ap,$j
$PTR_ADD $nj,$np,$j
$LD $aj,($aj)
$LD $nj,($nj)
$MULTU $aj,$bi
$ADDU $lo0,$alo,$hi0
$ADDU $lo1,$nlo,$hi1
sltu $at,$lo0,$hi0
sltu $t0,$lo1,$hi1
$ADDU $hi0,$ahi,$at
$ADDU $hi1,$nhi,$t0
mflo $alo
mfhi $ahi
$ADDU $lo1,$lo0
sltu $at,$lo1,$lo0
$MULTU $nj,$m1
$ADDU $hi1,$at
addu $j,$BNSZ
$ST $lo1,($tp)
sltu $t0,$j,$num
mflo $nlo
mfhi $nhi
bnez $t0,.L1st
$PTR_ADD $tp,$BNSZ
.set reorder
$ADDU $lo0,$alo,$hi0
sltu $at,$lo0,$hi0
$ADDU $hi0,$ahi,$at
$ADDU $lo1,$nlo,$hi1
sltu $t0,$lo1,$hi1
$ADDU $hi1,$nhi,$t0
$ADDU $lo1,$lo0
sltu $at,$lo1,$lo0
$ADDU $hi1,$at
$ST $lo1,($tp)
$ADDU $hi1,$hi0
sltu $at,$hi1,$hi0
$ST $hi1,$BNSZ($tp)
$ST $at,2*$BNSZ($tp)
li $i,$BNSZ
.align 4
.Louter:
$PTR_ADD $bi,$bp,$i
$LD $bi,($bi)
$LD $aj,($ap)
$LD $alo,$BNSZ($ap)
$LD $tj,($sp)
$MULTU $aj,$bi
$LD $nj,($np)
$LD $nlo,$BNSZ($np)
mflo $lo0
mfhi $hi0
$ADDU $lo0,$tj
$MULTU $lo0,$n0
sltu $at,$lo0,$tj
$ADDU $hi0,$at
mflo $m1
$MULTU $alo,$bi
mflo $alo
mfhi $ahi
$MULTU $nj,$m1
mflo $lo1
mfhi $hi1
$MULTU $nlo,$m1
$ADDU $lo1,$lo0
sltu $at,$lo1,$lo0
$ADDU $hi1,$at
mflo $nlo
mfhi $nhi
move $tp,$sp
li $j,2*$BNSZ
$LD $tj,$BNSZ($tp)
.align 4
.Linner:
.set noreorder
$PTR_ADD $aj,$ap,$j
$PTR_ADD $nj,$np,$j
$LD $aj,($aj)
$LD $nj,($nj)
$MULTU $aj,$bi
$ADDU $lo0,$alo,$hi0
$ADDU $lo1,$nlo,$hi1
sltu $at,$lo0,$hi0
sltu $t0,$lo1,$hi1
$ADDU $hi0,$ahi,$at
$ADDU $hi1,$nhi,$t0
mflo $alo
mfhi $ahi
$ADDU $lo0,$tj
addu $j,$BNSZ
$MULTU $nj,$m1
sltu $at,$lo0,$tj
$ADDU $lo1,$lo0
$ADDU $hi0,$at
sltu $t0,$lo1,$lo0
$LD $tj,2*$BNSZ($tp)
$ADDU $hi1,$t0
sltu $at,$j,$num
mflo $nlo
mfhi $nhi
$ST $lo1,($tp)
bnez $at,.Linner
$PTR_ADD $tp,$BNSZ
.set reorder
$ADDU $lo0,$alo,$hi0
sltu $at,$lo0,$hi0
$ADDU $hi0,$ahi,$at
$ADDU $lo0,$tj
sltu $t0,$lo0,$tj
$ADDU $hi0,$t0
$LD $tj,2*$BNSZ($tp)
$ADDU $lo1,$nlo,$hi1
sltu $at,$lo1,$hi1
$ADDU $hi1,$nhi,$at
$ADDU $lo1,$lo0
sltu $t0,$lo1,$lo0
$ADDU $hi1,$t0
$ST $lo1,($tp)
$ADDU $lo1,$hi1,$hi0
sltu $hi1,$lo1,$hi0
$ADDU $lo1,$tj
sltu $at,$lo1,$tj
$ADDU $hi1,$at
$ST $lo1,$BNSZ($tp)
$ST $hi1,2*$BNSZ($tp)
addu $i,$BNSZ
sltu $t0,$i,$num
bnez $t0,.Louter
.set noreorder
$PTR_ADD $tj,$sp,$num # &tp[num]
move $tp,$sp
move $ap,$sp
li $hi0,0 # clear borrow bit
.align 4
.Lsub: $LD $lo0,($tp)
$LD $lo1,($np)
$PTR_ADD $tp,$BNSZ
$PTR_ADD $np,$BNSZ
$SUBU $lo1,$lo0,$lo1 # tp[i]-np[i]
sgtu $at,$lo1,$lo0
$SUBU $lo0,$lo1,$hi0
sgtu $hi0,$lo0,$lo1
$ST $lo0,($rp)
or $hi0,$at
sltu $at,$tp,$tj
bnez $at,.Lsub
$PTR_ADD $rp,$BNSZ
$SUBU $hi0,$hi1,$hi0 # handle upmost overflow bit
move $tp,$sp
$PTR_SUB $rp,$num # restore rp
not $hi1,$hi0
and $ap,$hi0,$sp
and $bp,$hi1,$rp
or $ap,$ap,$bp # ap=borrow?tp:rp
.align 4
.Lcopy: $LD $aj,($ap)
$PTR_ADD $ap,$BNSZ
$ST $zero,($tp)
$PTR_ADD $tp,$BNSZ
sltu $at,$tp,$tj
$ST $aj,($rp)
bnez $at,.Lcopy
$PTR_ADD $rp,$BNSZ
li $a0,1
li $t0,1
.set noreorder
move $sp,$fp
$REG_L $fp,($FRAMESIZE-1)*$SZREG($sp)
$REG_L $s11,($FRAMESIZE-2)*$SZREG($sp)
$REG_L $s10,($FRAMESIZE-3)*$SZREG($sp)
$REG_L $s9,($FRAMESIZE-4)*$SZREG($sp)
$REG_L $s8,($FRAMESIZE-5)*$SZREG($sp)
$REG_L $s7,($FRAMESIZE-6)*$SZREG($sp)
$REG_L $s6,($FRAMESIZE-7)*$SZREG($sp)
$REG_L $s5,($FRAMESIZE-8)*$SZREG($sp)
$REG_L $s4,($FRAMESIZE-9)*$SZREG($sp)
___
$code.=<<___ if ($flavour =~ /nubi/i);
$REG_L $s3,($FRAMESIZE-10)*$SZREG($sp)
$REG_L $s2,($FRAMESIZE-11)*$SZREG($sp)
$REG_L $s1,($FRAMESIZE-12)*$SZREG($sp)
$REG_L $s0,($FRAMESIZE-13)*$SZREG($sp)
___
$code.=<<___;
jr $ra
$PTR_ADD $sp,$FRAMESIZE*$SZREG
.end bn_mul_mont_internal
.rdata
.asciiz "Montgomery Multiplication for MIPS, CRYPTOGAMS by <appro\@openssl.org>"
___
$code =~ s/\`([^\`]*)\`/eval $1/gem;
print $code;
close STDOUT;

2234
drivers/builtin_openssl2/crypto/bn/asm/mips.pl
View File

File diff suppressed because it is too large Load Diff

327
drivers/builtin_openssl2/crypto/bn/asm/mips3-mont.pl
View File

@ -1,327 +0,0 @@
#!/usr/bin/env perl
#
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# This module doesn't present direct interest for OpenSSL, because it
# doesn't provide better performance for longer keys. While 512-bit
# RSA private key operations are 40% faster, 1024-bit ones are hardly
# faster at all, while longer key operations are slower by up to 20%.
# It might be of interest to embedded system developers though, as
# it's smaller than 1KB, yet offers ~3x improvement over compiler
# generated code.
#
# The module targets N32 and N64 MIPS ABIs and currently is a bit
# IRIX-centric, i.e. is likely to require adaptation for other OSes.
# int bn_mul_mont(
$rp="a0"; # BN_ULONG *rp,
$ap="a1"; # const BN_ULONG *ap,
$bp="a2"; # const BN_ULONG *bp,
$np="a3"; # const BN_ULONG *np,
$n0="a4"; # const BN_ULONG *n0,
$num="a5"; # int num);
$lo0="a6";
$hi0="a7";
$lo1="v0";
$hi1="v1";
$aj="t0";
$bi="t1";
$nj="t2";
$tp="t3";
$alo="s0";
$ahi="s1";
$nlo="s2";
$nhi="s3";
$tj="s4";
$i="s5";
$j="s6";
$fp="t8";
$m1="t9";
$FRAME=8*(2+8);
$code=<<___;
#include <asm.h>
#include <regdef.h>
.text
.set noat
.set reorder
.align 5
.globl bn_mul_mont
.ent bn_mul_mont
bn_mul_mont:
.set noreorder
PTR_SUB sp,64
move $fp,sp
.frame $fp,64,ra
slt AT,$num,4
li v0,0
beqzl AT,.Lproceed
nop
jr ra
PTR_ADD sp,$fp,64
.set reorder
.align 5
.Lproceed:
ld $n0,0($n0)
ld $bi,0($bp) # bp[0]
ld $aj,0($ap) # ap[0]
ld $nj,0($np) # np[0]
PTR_SUB sp,16 # place for two extra words
sll $num,3
li AT,-4096
PTR_SUB sp,$num
and sp,AT
sd s0,0($fp)
sd s1,8($fp)
sd s2,16($fp)
sd s3,24($fp)
sd s4,32($fp)
sd s5,40($fp)
sd s6,48($fp)
sd s7,56($fp)
dmultu $aj,$bi
ld $alo,8($ap)
ld $nlo,8($np)
mflo $lo0
mfhi $hi0
dmultu $lo0,$n0
mflo $m1
dmultu $alo,$bi
mflo $alo
mfhi $ahi
dmultu $nj,$m1
mflo $lo1
mfhi $hi1
dmultu $nlo,$m1
daddu $lo1,$lo0
sltu AT,$lo1,$lo0
daddu $hi1,AT
mflo $nlo
mfhi $nhi
move $tp,sp
li $j,16
.align 4
.L1st:
.set noreorder
PTR_ADD $aj,$ap,$j
ld $aj,($aj)
PTR_ADD $nj,$np,$j
ld $nj,($nj)
dmultu $aj,$bi
daddu $lo0,$alo,$hi0
daddu $lo1,$nlo,$hi1
sltu AT,$lo0,$hi0
sltu s7,$lo1,$hi1
daddu $hi0,$ahi,AT
daddu $hi1,$nhi,s7
mflo $alo
mfhi $ahi
daddu $lo1,$lo0
sltu AT,$lo1,$lo0
dmultu $nj,$m1
daddu $hi1,AT
addu $j,8
sd $lo1,($tp)
sltu s7,$j,$num
mflo $nlo
mfhi $nhi
bnez s7,.L1st
PTR_ADD $tp,8
.set reorder
daddu $lo0,$alo,$hi0
sltu AT,$lo0,$hi0
daddu $hi0,$ahi,AT
daddu $lo1,$nlo,$hi1
sltu s7,$lo1,$hi1
daddu $hi1,$nhi,s7
daddu $lo1,$lo0
sltu AT,$lo1,$lo0
daddu $hi1,AT
sd $lo1,($tp)
daddu $hi1,$hi0
sltu AT,$hi1,$hi0
sd $hi1,8($tp)
sd AT,16($tp)
li $i,8
.align 4
.Louter:
PTR_ADD $bi,$bp,$i
ld $bi,($bi)
ld $aj,($ap)
ld $alo,8($ap)
ld $tj,(sp)
dmultu $aj,$bi
ld $nj,($np)
ld $nlo,8($np)
mflo $lo0
mfhi $hi0
daddu $lo0,$tj
dmultu $lo0,$n0
sltu AT,$lo0,$tj
daddu $hi0,AT
mflo $m1
dmultu $alo,$bi
mflo $alo
mfhi $ahi
dmultu $nj,$m1
mflo $lo1
mfhi $hi1
dmultu $nlo,$m1
daddu $lo1,$lo0
sltu AT,$lo1,$lo0
daddu $hi1,AT
mflo $nlo
mfhi $nhi
move $tp,sp
li $j,16
ld $tj,8($tp)
.align 4
.Linner:
.set noreorder
PTR_ADD $aj,$ap,$j
ld $aj,($aj)
PTR_ADD $nj,$np,$j
ld $nj,($nj)
dmultu $aj,$bi
daddu $lo0,$alo,$hi0
daddu $lo1,$nlo,$hi1
sltu AT,$lo0,$hi0
sltu s7,$lo1,$hi1
daddu $hi0,$ahi,AT
daddu $hi1,$nhi,s7
mflo $alo
mfhi $ahi
daddu $lo0,$tj
addu $j,8
dmultu $nj,$m1
sltu AT,$lo0,$tj
daddu $lo1,$lo0
daddu $hi0,AT
sltu s7,$lo1,$lo0
ld $tj,16($tp)
daddu $hi1,s7
sltu AT,$j,$num
mflo $nlo
mfhi $nhi
sd $lo1,($tp)
bnez AT,.Linner
PTR_ADD $tp,8
.set reorder
daddu $lo0,$alo,$hi0
sltu AT,$lo0,$hi0
daddu $hi0,$ahi,AT
daddu $lo0,$tj
sltu s7,$lo0,$tj
daddu $hi0,s7
ld $tj,16($tp)
daddu $lo1,$nlo,$hi1
sltu AT,$lo1,$hi1
daddu $hi1,$nhi,AT
daddu $lo1,$lo0
sltu s7,$lo1,$lo0
daddu $hi1,s7
sd $lo1,($tp)
daddu $lo1,$hi1,$hi0
sltu $hi1,$lo1,$hi0
daddu $lo1,$tj
sltu AT,$lo1,$tj
daddu $hi1,AT
sd $lo1,8($tp)
sd $hi1,16($tp)
addu $i,8
sltu s7,$i,$num
bnez s7,.Louter
.set noreorder
PTR_ADD $tj,sp,$num # &tp[num]
move $tp,sp
move $ap,sp
li $hi0,0 # clear borrow bit
.align 4
.Lsub: ld $lo0,($tp)
ld $lo1,($np)
PTR_ADD $tp,8
PTR_ADD $np,8
dsubu $lo1,$lo0,$lo1 # tp[i]-np[i]
sgtu AT,$lo1,$lo0
dsubu $lo0,$lo1,$hi0
sgtu $hi0,$lo0,$lo1
sd $lo0,($rp)
or $hi0,AT
sltu AT,$tp,$tj
bnez AT,.Lsub
PTR_ADD $rp,8
dsubu $hi0,$hi1,$hi0 # handle upmost overflow bit
move $tp,sp
PTR_SUB $rp,$num # restore rp
not $hi1,$hi0
and $ap,$hi0,sp
and $bp,$hi1,$rp
or $ap,$ap,$bp # ap=borrow?tp:rp
.align 4
.Lcopy: ld $aj,($ap)
PTR_ADD $ap,8
PTR_ADD $tp,8
sd zero,-8($tp)
sltu AT,$tp,$tj
sd $aj,($rp)
bnez AT,.Lcopy
PTR_ADD $rp,8
ld s0,0($fp)
ld s1,8($fp)
ld s2,16($fp)
ld s3,24($fp)
ld s4,32($fp)
ld s5,40($fp)
ld s6,48($fp)
ld s7,56($fp)
li v0,1
jr ra
PTR_ADD sp,$fp,64
.set reorder
END(bn_mul_mont)
.rdata
.asciiz "Montgomery Multiplication for MIPS III/IV, CRYPTOGAMS by <appro\@openssl.org>"
___
print $code;
close STDOUT;

2201
drivers/builtin_openssl2/crypto/bn/asm/mips3.s
View File

File diff suppressed because it is too large Load Diff

1497
drivers/builtin_openssl2/crypto/bn/asm/modexp512-x86_64.pl
View File

File diff suppressed because it is too large Load Diff

1618
drivers/builtin_openssl2/crypto/bn/asm/pa-risc2.s
View File

File diff suppressed because it is too large Load Diff

1605
drivers/builtin_openssl2/crypto/bn/asm/pa-risc2W.s
View File

File diff suppressed because it is too large Load Diff

995
drivers/builtin_openssl2/crypto/bn/asm/parisc-mont.pl
View File

@ -1,995 +0,0 @@
#!/usr/bin/env perl
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# On PA-7100LC this module performs ~90-50% better, less for longer
# keys, than code generated by gcc 3.2 for PA-RISC 1.1. Latter means
# that compiler utilized xmpyu instruction to perform 32x32=64-bit
# multiplication, which in turn means that "baseline" performance was
# optimal in respect to instruction set capabilities. Fair comparison
# with vendor compiler is problematic, because OpenSSL doesn't define
# BN_LLONG [presumably] for historical reasons, which drives compiler
# toward 4 times 16x16=32-bit multiplicatons [plus complementary
# shifts and additions] instead. This means that you should observe
# several times improvement over code generated by vendor compiler
# for PA-RISC 1.1, but the "baseline" is far from optimal. The actual
# improvement coefficient was never collected on PA-7100LC, or any
# other 1.1 CPU, because I don't have access to such machine with
# vendor compiler. But to give you a taste, PA-RISC 1.1 code path
# reportedly outperformed code generated by cc +DA1.1 +O3 by factor
# of ~5x on PA-8600.
#
# On PA-RISC 2.0 it has to compete with pa-risc2[W].s, which is
# reportedly ~2x faster than vendor compiler generated code [according
# to comment in pa-risc2[W].s]. Here comes a catch. Execution core of
# this implementation is actually 32-bit one, in the sense that it
# operates on 32-bit values. But pa-risc2[W].s operates on arrays of
# 64-bit BN_LONGs... How do they interoperate then? No problem. This
# module picks halves of 64-bit values in reverse order and pretends
# they were 32-bit BN_LONGs. But can 32-bit core compete with "pure"
# 64-bit code such as pa-risc2[W].s then? Well, the thing is that
# 32x32=64-bit multiplication is the best even PA-RISC 2.0 can do,
# i.e. there is no "wider" multiplication like on most other 64-bit
# platforms. This means that even being effectively 32-bit, this
# implementation performs "64-bit" computational task in same amount
# of arithmetic operations, most notably multiplications. It requires
# more memory references, most notably to tp[num], but this doesn't
# seem to exhaust memory port capacity. And indeed, dedicated PA-RISC
# 2.0 code path provides virtually same performance as pa-risc2[W].s:
# it's ~10% better for shortest key length and ~10% worse for longest
# one.
#
# In case it wasn't clear. The module has two distinct code paths:
# PA-RISC 1.1 and PA-RISC 2.0 ones. Latter features carry-free 64-bit
# additions and 64-bit integer loads, not to mention specific
# instruction scheduling. In 64-bit build naturally only 2.0 code path
# is assembled. In 32-bit application context both code paths are
# assembled, PA-RISC 2.0 CPU is detected at run-time and proper path
# is taken automatically. Also, in 32-bit build the module imposes
# couple of limitations: vector lengths has to be even and vector
# addresses has to be 64-bit aligned. Normally neither is a problem:
# most common key lengths are even and vectors are commonly malloc-ed,
# which ensures alignment.
#
# Special thanks to polarhome.com for providing HP-UX account on
# PA-RISC 1.1 machine, and to correspondent who chose to remain
# anonymous for testing the code on PA-RISC 2.0 machine.
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
$flavour = shift;
$output = shift;
open STDOUT,">$output";
if ($flavour =~ /64/) {
$LEVEL ="2.0W";
$SIZE_T =8;
$FRAME_MARKER =80;
$SAVED_RP =16;
$PUSH ="std";
$PUSHMA ="std,ma";
$POP ="ldd";
$POPMB ="ldd,mb";
$BN_SZ =$SIZE_T;
} else {
$LEVEL ="1.1"; #$LEVEL.="\n\t.ALLOW\t2.0";
$SIZE_T =4;
$FRAME_MARKER =48;
$SAVED_RP =20;
$PUSH ="stw";
$PUSHMA ="stwm";
$POP ="ldw";
$POPMB ="ldwm";
$BN_SZ =$SIZE_T;
if (open CONF,"<${dir}../../opensslconf.h") {
while(<CONF>) {
if (m/#\s*define\s+SIXTY_FOUR_BIT/) {
$BN_SZ=8;
$LEVEL="2.0";
last;
}
}
close CONF;
}
}
$FRAME=8*$SIZE_T+$FRAME_MARKER; # 8 saved regs + frame marker
# [+ argument transfer]
$LOCALS=$FRAME-$FRAME_MARKER;
$FRAME+=32; # local variables
$tp="%r31";
$ti1="%r29";
$ti0="%r28";
$rp="%r26";
$ap="%r25";
$bp="%r24";
$np="%r23";
$n0="%r22"; # passed through stack in 32-bit
$num="%r21"; # passed through stack in 32-bit
$idx="%r20";
$arrsz="%r19";
$nm1="%r7";
$nm0="%r6";
$ab1="%r5";
$ab0="%r4";
$fp="%r3";
$hi1="%r2";
$hi0="%r1";
$xfer=$n0; # accomodates [-16..15] offset in fld[dw]s
$fm0="%fr4"; $fti=$fm0;
$fbi="%fr5L";
$fn0="%fr5R";
$fai="%fr6"; $fab0="%fr7"; $fab1="%fr8";
$fni="%fr9"; $fnm0="%fr10"; $fnm1="%fr11";
$code=<<___;
.LEVEL $LEVEL
.SPACE \$TEXT\$
.SUBSPA \$CODE\$,QUAD=0,ALIGN=8,ACCESS=0x2C,CODE_ONLY
.EXPORT bn_mul_mont,ENTRY,ARGW0=GR,ARGW1=GR,ARGW2=GR,ARGW3=GR
.ALIGN 64
bn_mul_mont
.PROC
.CALLINFO FRAME=`$FRAME-8*$SIZE_T`,NO_CALLS,SAVE_RP,SAVE_SP,ENTRY_GR=6
.ENTRY
$PUSH %r2,-$SAVED_RP(%sp) ; standard prologue
$PUSHMA %r3,$FRAME(%sp)
$PUSH %r4,`-$FRAME+1*$SIZE_T`(%sp)
$PUSH %r5,`-$FRAME+2*$SIZE_T`(%sp)
$PUSH %r6,`-$FRAME+3*$SIZE_T`(%sp)
$PUSH %r7,`-$FRAME+4*$SIZE_T`(%sp)
$PUSH %r8,`-$FRAME+5*$SIZE_T`(%sp)
$PUSH %r9,`-$FRAME+6*$SIZE_T`(%sp)
$PUSH %r10,`-$FRAME+7*$SIZE_T`(%sp)
ldo -$FRAME(%sp),$fp
___
$code.=<<___ if ($SIZE_T==4);
ldw `-$FRAME_MARKER-4`($fp),$n0
ldw `-$FRAME_MARKER-8`($fp),$num
nop
nop ; alignment
___
$code.=<<___ if ($BN_SZ==4);
comiclr,<= 6,$num,%r0 ; are vectors long enough?
b L\$abort
ldi 0,%r28 ; signal "unhandled"
add,ev %r0,$num,$num ; is $num even?
b L\$abort
nop
or $ap,$np,$ti1
extru,= $ti1,31,3,%r0 ; are ap and np 64-bit aligned?
b L\$abort
nop
nop ; alignment
nop
fldws 0($n0),${fn0}
fldws,ma 4($bp),${fbi} ; bp[0]
___
$code.=<<___ if ($BN_SZ==8);
comib,> 3,$num,L\$abort ; are vectors long enough?
ldi 0,%r28 ; signal "unhandled"
addl $num,$num,$num ; I operate on 32-bit values
fldws 4($n0),${fn0} ; only low part of n0
fldws 4($bp),${fbi} ; bp[0] in flipped word order
___
$code.=<<___;
fldds 0($ap),${fai} ; ap[0,1]
fldds 0($np),${fni} ; np[0,1]
sh2addl $num,%r0,$arrsz
ldi 31,$hi0
ldo 36($arrsz),$hi1 ; space for tp[num+1]
andcm $hi1,$hi0,$hi1 ; align
addl $hi1,%sp,%sp
$PUSH $fp,-$SIZE_T(%sp)
ldo `$LOCALS+16`($fp),$xfer
ldo `$LOCALS+32+4`($fp),$tp
xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[0]
xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[0]
xmpyu ${fn0},${fab0}R,${fm0}
addl $arrsz,$ap,$ap ; point at the end
addl $arrsz,$np,$np
subi 0,$arrsz,$idx ; j=0
ldo 8($idx),$idx ; j++++
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[0]*m
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[1]*m
fstds ${fab0},-16($xfer)
fstds ${fnm0},-8($xfer)
fstds ${fab1},0($xfer)
fstds ${fnm1},8($xfer)
flddx $idx($ap),${fai} ; ap[2,3]
flddx $idx($np),${fni} ; np[2,3]
___
$code.=<<___ if ($BN_SZ==4);
mtctl $hi0,%cr11 ; $hi0 still holds 31
extrd,u,*= $hi0,%sar,1,$hi0 ; executes on PA-RISC 1.0
b L\$parisc11
nop
___
$code.=<<___; # PA-RISC 2.0 code-path
xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0]
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
ldd -16($xfer),$ab0
fstds ${fab0},-16($xfer)
extrd,u $ab0,31,32,$hi0
extrd,u $ab0,63,32,$ab0
ldd -8($xfer),$nm0
fstds ${fnm0},-8($xfer)
ldo 8($idx),$idx ; j++++
addl $ab0,$nm0,$nm0 ; low part is discarded
extrd,u $nm0,31,32,$hi1
L\$1st
xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[0]
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m
ldd 0($xfer),$ab1
fstds ${fab1},0($xfer)
addl $hi0,$ab1,$ab1
extrd,u $ab1,31,32,$hi0
ldd 8($xfer),$nm1
fstds ${fnm1},8($xfer)
extrd,u $ab1,63,32,$ab1
addl $hi1,$nm1,$nm1
flddx $idx($ap),${fai} ; ap[j,j+1]
flddx $idx($np),${fni} ; np[j,j+1]
addl $ab1,$nm1,$nm1
extrd,u $nm1,31,32,$hi1
xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0]
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
ldd -16($xfer),$ab0
fstds ${fab0},-16($xfer)
addl $hi0,$ab0,$ab0
extrd,u $ab0,31,32,$hi0
ldd -8($xfer),$nm0
fstds ${fnm0},-8($xfer)
extrd,u $ab0,63,32,$ab0
addl $hi1,$nm0,$nm0
stw $nm1,-4($tp) ; tp[j-1]
addl $ab0,$nm0,$nm0
stw,ma $nm0,8($tp) ; tp[j-1]
addib,<> 8,$idx,L\$1st ; j++++
extrd,u $nm0,31,32,$hi1
xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[0]
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m
ldd 0($xfer),$ab1
fstds ${fab1},0($xfer)
addl $hi0,$ab1,$ab1
extrd,u $ab1,31,32,$hi0
ldd 8($xfer),$nm1
fstds ${fnm1},8($xfer)
extrd,u $ab1,63,32,$ab1
addl $hi1,$nm1,$nm1
ldd -16($xfer),$ab0
addl $ab1,$nm1,$nm1
ldd -8($xfer),$nm0
extrd,u $nm1,31,32,$hi1
addl $hi0,$ab0,$ab0
extrd,u $ab0,31,32,$hi0
stw $nm1,-4($tp) ; tp[j-1]
extrd,u $ab0,63,32,$ab0
addl $hi1,$nm0,$nm0
ldd 0($xfer),$ab1
addl $ab0,$nm0,$nm0
ldd,mb 8($xfer),$nm1
extrd,u $nm0,31,32,$hi1
stw,ma $nm0,8($tp) ; tp[j-1]
ldo -1($num),$num ; i--
subi 0,$arrsz,$idx ; j=0
___
$code.=<<___ if ($BN_SZ==4);
fldws,ma 4($bp),${fbi} ; bp[1]
___
$code.=<<___ if ($BN_SZ==8);
fldws 0($bp),${fbi} ; bp[1] in flipped word order
___
$code.=<<___;
flddx $idx($ap),${fai} ; ap[0,1]
flddx $idx($np),${fni} ; np[0,1]
fldws 8($xfer),${fti}R ; tp[0]
addl $hi0,$ab1,$ab1
extrd,u $ab1,31,32,$hi0
extrd,u $ab1,63,32,$ab1
ldo 8($idx),$idx ; j++++
xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[1]
xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[1]
addl $hi1,$nm1,$nm1
addl $ab1,$nm1,$nm1
extrd,u $nm1,31,32,$hi1
fstws,mb ${fab0}L,-8($xfer) ; save high part
stw $nm1,-4($tp) ; tp[j-1]
fcpy,sgl %fr0,${fti}L ; zero high part
fcpy,sgl %fr0,${fab0}L
addl $hi1,$hi0,$hi0
extrd,u $hi0,31,32,$hi1
fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double
fcnvxf,dbl,dbl ${fab0},${fab0}
stw $hi0,0($tp)
stw $hi1,4($tp)
fadd,dbl ${fti},${fab0},${fab0} ; add tp[0]
fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int
xmpyu ${fn0},${fab0}R,${fm0}
ldo `$LOCALS+32+4`($fp),$tp
L\$outer
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[0]*m
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[1]*m
fstds ${fab0},-16($xfer) ; 33-bit value
fstds ${fnm0},-8($xfer)
flddx $idx($ap),${fai} ; ap[2]
flddx $idx($np),${fni} ; np[2]
ldo 8($idx),$idx ; j++++
ldd -16($xfer),$ab0 ; 33-bit value
ldd -8($xfer),$nm0
ldw 0($xfer),$hi0 ; high part
xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i]
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
extrd,u $ab0,31,32,$ti0 ; carry bit
extrd,u $ab0,63,32,$ab0
fstds ${fab1},0($xfer)
addl $ti0,$hi0,$hi0 ; account carry bit
fstds ${fnm1},8($xfer)
addl $ab0,$nm0,$nm0 ; low part is discarded
ldw 0($tp),$ti1 ; tp[1]
extrd,u $nm0,31,32,$hi1
fstds ${fab0},-16($xfer)
fstds ${fnm0},-8($xfer)
L\$inner
xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[i]
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m
ldd 0($xfer),$ab1
fstds ${fab1},0($xfer)
addl $hi0,$ti1,$ti1
addl $ti1,$ab1,$ab1
ldd 8($xfer),$nm1
fstds ${fnm1},8($xfer)
extrd,u $ab1,31,32,$hi0
extrd,u $ab1,63,32,$ab1
flddx $idx($ap),${fai} ; ap[j,j+1]
flddx $idx($np),${fni} ; np[j,j+1]
addl $hi1,$nm1,$nm1
addl $ab1,$nm1,$nm1
ldw 4($tp),$ti0 ; tp[j]
stw $nm1,-4($tp) ; tp[j-1]
xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i]
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
ldd -16($xfer),$ab0
fstds ${fab0},-16($xfer)
addl $hi0,$ti0,$ti0
addl $ti0,$ab0,$ab0
ldd -8($xfer),$nm0
fstds ${fnm0},-8($xfer)
extrd,u $ab0,31,32,$hi0
extrd,u $nm1,31,32,$hi1
ldw 8($tp),$ti1 ; tp[j]
extrd,u $ab0,63,32,$ab0
addl $hi1,$nm0,$nm0
addl $ab0,$nm0,$nm0
stw,ma $nm0,8($tp) ; tp[j-1]
addib,<> 8,$idx,L\$inner ; j++++
extrd,u $nm0,31,32,$hi1
xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[i]
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m
ldd 0($xfer),$ab1
fstds ${fab1},0($xfer)
addl $hi0,$ti1,$ti1
addl $ti1,$ab1,$ab1
ldd 8($xfer),$nm1
fstds ${fnm1},8($xfer)
extrd,u $ab1,31,32,$hi0
extrd,u $ab1,63,32,$ab1
ldw 4($tp),$ti0 ; tp[j]
addl $hi1,$nm1,$nm1
addl $ab1,$nm1,$nm1
ldd -16($xfer),$ab0
ldd -8($xfer),$nm0
extrd,u $nm1,31,32,$hi1
addl $hi0,$ab0,$ab0
addl $ti0,$ab0,$ab0
stw $nm1,-4($tp) ; tp[j-1]
extrd,u $ab0,31,32,$hi0
ldw 8($tp),$ti1 ; tp[j]
extrd,u $ab0,63,32,$ab0
addl $hi1,$nm0,$nm0
ldd 0($xfer),$ab1
addl $ab0,$nm0,$nm0
ldd,mb 8($xfer),$nm1
extrd,u $nm0,31,32,$hi1
stw,ma $nm0,8($tp) ; tp[j-1]
addib,= -1,$num,L\$outerdone ; i--
subi 0,$arrsz,$idx ; j=0
___
$code.=<<___ if ($BN_SZ==4);
fldws,ma 4($bp),${fbi} ; bp[i]
___
$code.=<<___ if ($BN_SZ==8);
ldi 12,$ti0 ; bp[i] in flipped word order
addl,ev %r0,$num,$num
ldi -4,$ti0
addl $ti0,$bp,$bp
fldws 0($bp),${fbi}
___
$code.=<<___;
flddx $idx($ap),${fai} ; ap[0]
addl $hi0,$ab1,$ab1
flddx $idx($np),${fni} ; np[0]
fldws 8($xfer),${fti}R ; tp[0]
addl $ti1,$ab1,$ab1
extrd,u $ab1,31,32,$hi0
extrd,u $ab1,63,32,$ab1
ldo 8($idx),$idx ; j++++
xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[i]
xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[i]
ldw 4($tp),$ti0 ; tp[j]
addl $hi1,$nm1,$nm1
fstws,mb ${fab0}L,-8($xfer) ; save high part
addl $ab1,$nm1,$nm1
extrd,u $nm1,31,32,$hi1
fcpy,sgl %fr0,${fti}L ; zero high part
fcpy,sgl %fr0,${fab0}L
stw $nm1,-4($tp) ; tp[j-1]
fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double
fcnvxf,dbl,dbl ${fab0},${fab0}
addl $hi1,$hi0,$hi0
fadd,dbl ${fti},${fab0},${fab0} ; add tp[0]
addl $ti0,$hi0,$hi0
extrd,u $hi0,31,32,$hi1
fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int
stw $hi0,0($tp)
stw $hi1,4($tp)
xmpyu ${fn0},${fab0}R,${fm0}
b L\$outer
ldo `$LOCALS+32+4`($fp),$tp
L\$outerdone
addl $hi0,$ab1,$ab1
addl $ti1,$ab1,$ab1
extrd,u $ab1,31,32,$hi0
extrd,u $ab1,63,32,$ab1
ldw 4($tp),$ti0 ; tp[j]
addl $hi1,$nm1,$nm1
addl $ab1,$nm1,$nm1
extrd,u $nm1,31,32,$hi1
stw $nm1,-4($tp) ; tp[j-1]
addl $hi1,$hi0,$hi0
addl $ti0,$hi0,$hi0
extrd,u $hi0,31,32,$hi1
stw $hi0,0($tp)
stw $hi1,4($tp)
ldo `$LOCALS+32`($fp),$tp
sub %r0,%r0,%r0 ; clear borrow
___
$code.=<<___ if ($BN_SZ==4);
ldws,ma 4($tp),$ti0
extru,= $rp,31,3,%r0 ; is rp 64-bit aligned?
b L\$sub_pa11
addl $tp,$arrsz,$tp
L\$sub
ldwx $idx($np),$hi0
subb $ti0,$hi0,$hi1
ldwx $idx($tp),$ti0
addib,<> 4,$idx,L\$sub
stws,ma $hi1,4($rp)
subb $ti0,%r0,$hi1
ldo -4($tp),$tp
___
$code.=<<___ if ($BN_SZ==8);
ldd,ma 8($tp),$ti0
L\$sub
ldd $idx($np),$hi0
shrpd $ti0,$ti0,32,$ti0 ; flip word order
std $ti0,-8($tp) ; save flipped value
sub,db $ti0,$hi0,$hi1
ldd,ma 8($tp),$ti0
addib,<> 8,$idx,L\$sub
std,ma $hi1,8($rp)
extrd,u $ti0,31,32,$ti0 ; carry in flipped word order
sub,db $ti0,%r0,$hi1
ldo -8($tp),$tp
___
$code.=<<___;
and $tp,$hi1,$ap
andcm $rp,$hi1,$bp
or $ap,$bp,$np
sub $rp,$arrsz,$rp ; rewind rp
subi 0,$arrsz,$idx
ldo `$LOCALS+32`($fp),$tp
L\$copy
ldd $idx($np),$hi0
std,ma %r0,8($tp)
addib,<> 8,$idx,.-8 ; L\$copy
std,ma $hi0,8($rp)
___
if ($BN_SZ==4) { # PA-RISC 1.1 code-path
$ablo=$ab0;
$abhi=$ab1;
$nmlo0=$nm0;
$nmhi0=$nm1;
$nmlo1="%r9";
$nmhi1="%r8";
$code.=<<___;
b L\$done
nop
.ALIGN 8
L\$parisc11
xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0]
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
ldw -12($xfer),$ablo
ldw -16($xfer),$hi0
ldw -4($xfer),$nmlo0
ldw -8($xfer),$nmhi0
fstds ${fab0},-16($xfer)
fstds ${fnm0},-8($xfer)
ldo 8($idx),$idx ; j++++
add $ablo,$nmlo0,$nmlo0 ; discarded
addc %r0,$nmhi0,$hi1
ldw 4($xfer),$ablo
ldw 0($xfer),$abhi
nop
L\$1st_pa11
xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[0]
flddx $idx($ap),${fai} ; ap[j,j+1]
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m
flddx $idx($np),${fni} ; np[j,j+1]
add $hi0,$ablo,$ablo
ldw 12($xfer),$nmlo1
addc %r0,$abhi,$hi0
ldw 8($xfer),$nmhi1
add $ablo,$nmlo1,$nmlo1
fstds ${fab1},0($xfer)
addc %r0,$nmhi1,$nmhi1
fstds ${fnm1},8($xfer)
add $hi1,$nmlo1,$nmlo1
ldw -12($xfer),$ablo
addc %r0,$nmhi1,$hi1
ldw -16($xfer),$abhi
xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[0]
ldw -4($xfer),$nmlo0
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
ldw -8($xfer),$nmhi0
add $hi0,$ablo,$ablo
stw $nmlo1,-4($tp) ; tp[j-1]
addc %r0,$abhi,$hi0
fstds ${fab0},-16($xfer)
add $ablo,$nmlo0,$nmlo0
fstds ${fnm0},-8($xfer)
addc %r0,$nmhi0,$nmhi0
ldw 0($xfer),$abhi
add $hi1,$nmlo0,$nmlo0
ldw 4($xfer),$ablo
stws,ma $nmlo0,8($tp) ; tp[j-1]
addib,<> 8,$idx,L\$1st_pa11 ; j++++
addc %r0,$nmhi0,$hi1
ldw 8($xfer),$nmhi1
ldw 12($xfer),$nmlo1
xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[0]
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m
add $hi0,$ablo,$ablo
fstds ${fab1},0($xfer)
addc %r0,$abhi,$hi0
fstds ${fnm1},8($xfer)
add $ablo,$nmlo1,$nmlo1
ldw -16($xfer),$abhi
addc %r0,$nmhi1,$nmhi1
ldw -12($xfer),$ablo
add $hi1,$nmlo1,$nmlo1
ldw -8($xfer),$nmhi0
addc %r0,$nmhi1,$hi1
ldw -4($xfer),$nmlo0
add $hi0,$ablo,$ablo
stw $nmlo1,-4($tp) ; tp[j-1]
addc %r0,$abhi,$hi0
ldw 0($xfer),$abhi
add $ablo,$nmlo0,$nmlo0
ldw 4($xfer),$ablo
addc %r0,$nmhi0,$nmhi0
ldws,mb 8($xfer),$nmhi1
add $hi1,$nmlo0,$nmlo0
ldw 4($xfer),$nmlo1
addc %r0,$nmhi0,$hi1
stws,ma $nmlo0,8($tp) ; tp[j-1]
ldo -1($num),$num ; i--
subi 0,$arrsz,$idx ; j=0
fldws,ma 4($bp),${fbi} ; bp[1]
flddx $idx($ap),${fai} ; ap[0,1]
flddx $idx($np),${fni} ; np[0,1]
fldws 8($xfer),${fti}R ; tp[0]
add $hi0,$ablo,$ablo
addc %r0,$abhi,$hi0
ldo 8($idx),$idx ; j++++
xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[1]
xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[1]
add $hi1,$nmlo1,$nmlo1
addc %r0,$nmhi1,$nmhi1
add $ablo,$nmlo1,$nmlo1
addc %r0,$nmhi1,$hi1
fstws,mb ${fab0}L,-8($xfer) ; save high part
stw $nmlo1,-4($tp) ; tp[j-1]
fcpy,sgl %fr0,${fti}L ; zero high part
fcpy,sgl %fr0,${fab0}L
add $hi1,$hi0,$hi0
addc %r0,%r0,$hi1
fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double
fcnvxf,dbl,dbl ${fab0},${fab0}
stw $hi0,0($tp)
stw $hi1,4($tp)
fadd,dbl ${fti},${fab0},${fab0} ; add tp[0]
fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int
xmpyu ${fn0},${fab0}R,${fm0}
ldo `$LOCALS+32+4`($fp),$tp
L\$outer_pa11
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[0]*m
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[1]*m
fstds ${fab0},-16($xfer) ; 33-bit value
fstds ${fnm0},-8($xfer)
flddx $idx($ap),${fai} ; ap[2,3]
flddx $idx($np),${fni} ; np[2,3]
ldw -16($xfer),$abhi ; carry bit actually
ldo 8($idx),$idx ; j++++
ldw -12($xfer),$ablo
ldw -8($xfer),$nmhi0
ldw -4($xfer),$nmlo0
ldw 0($xfer),$hi0 ; high part
xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i]
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
fstds ${fab1},0($xfer)
addl $abhi,$hi0,$hi0 ; account carry bit
fstds ${fnm1},8($xfer)
add $ablo,$nmlo0,$nmlo0 ; discarded
ldw 0($tp),$ti1 ; tp[1]
addc %r0,$nmhi0,$hi1
fstds ${fab0},-16($xfer)
fstds ${fnm0},-8($xfer)
ldw 4($xfer),$ablo
ldw 0($xfer),$abhi
L\$inner_pa11
xmpyu ${fai}R,${fbi},${fab1} ; ap[j+1]*bp[i]
flddx $idx($ap),${fai} ; ap[j,j+1]
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j+1]*m
flddx $idx($np),${fni} ; np[j,j+1]
add $hi0,$ablo,$ablo
ldw 4($tp),$ti0 ; tp[j]
addc %r0,$abhi,$abhi
ldw 12($xfer),$nmlo1
add $ti1,$ablo,$ablo
ldw 8($xfer),$nmhi1
addc %r0,$abhi,$hi0
fstds ${fab1},0($xfer)
add $ablo,$nmlo1,$nmlo1
fstds ${fnm1},8($xfer)
addc %r0,$nmhi1,$nmhi1
ldw -12($xfer),$ablo
add $hi1,$nmlo1,$nmlo1
ldw -16($xfer),$abhi
addc %r0,$nmhi1,$hi1
xmpyu ${fai}L,${fbi},${fab0} ; ap[j]*bp[i]
ldw 8($tp),$ti1 ; tp[j]
xmpyu ${fni}L,${fm0}R,${fnm0} ; np[j]*m
ldw -4($xfer),$nmlo0
add $hi0,$ablo,$ablo
ldw -8($xfer),$nmhi0
addc %r0,$abhi,$abhi
stw $nmlo1,-4($tp) ; tp[j-1]
add $ti0,$ablo,$ablo
fstds ${fab0},-16($xfer)
addc %r0,$abhi,$hi0
fstds ${fnm0},-8($xfer)
add $ablo,$nmlo0,$nmlo0
ldw 4($xfer),$ablo
addc %r0,$nmhi0,$nmhi0
ldw 0($xfer),$abhi
add $hi1,$nmlo0,$nmlo0
stws,ma $nmlo0,8($tp) ; tp[j-1]
addib,<> 8,$idx,L\$inner_pa11 ; j++++
addc %r0,$nmhi0,$hi1
xmpyu ${fai}R,${fbi},${fab1} ; ap[j]*bp[i]
ldw 12($xfer),$nmlo1
xmpyu ${fni}R,${fm0}R,${fnm1} ; np[j]*m
ldw 8($xfer),$nmhi1
add $hi0,$ablo,$ablo
ldw 4($tp),$ti0 ; tp[j]
addc %r0,$abhi,$abhi
fstds ${fab1},0($xfer)
add $ti1,$ablo,$ablo
fstds ${fnm1},8($xfer)
addc %r0,$abhi,$hi0
ldw -16($xfer),$abhi
add $ablo,$nmlo1,$nmlo1
ldw -12($xfer),$ablo
addc %r0,$nmhi1,$nmhi1
ldw -8($xfer),$nmhi0
add $hi1,$nmlo1,$nmlo1
ldw -4($xfer),$nmlo0
addc %r0,$nmhi1,$hi1
add $hi0,$ablo,$ablo
stw $nmlo1,-4($tp) ; tp[j-1]
addc %r0,$abhi,$abhi
add $ti0,$ablo,$ablo
ldw 8($tp),$ti1 ; tp[j]
addc %r0,$abhi,$hi0
ldw 0($xfer),$abhi
add $ablo,$nmlo0,$nmlo0
ldw 4($xfer),$ablo
addc %r0,$nmhi0,$nmhi0
ldws,mb 8($xfer),$nmhi1
add $hi1,$nmlo0,$nmlo0
ldw 4($xfer),$nmlo1
addc %r0,$nmhi0,$hi1
stws,ma $nmlo0,8($tp) ; tp[j-1]
addib,= -1,$num,L\$outerdone_pa11; i--
subi 0,$arrsz,$idx ; j=0
fldws,ma 4($bp),${fbi} ; bp[i]
flddx $idx($ap),${fai} ; ap[0]
add $hi0,$ablo,$ablo
addc %r0,$abhi,$abhi
flddx $idx($np),${fni} ; np[0]
fldws 8($xfer),${fti}R ; tp[0]
add $ti1,$ablo,$ablo
addc %r0,$abhi,$hi0
ldo 8($idx),$idx ; j++++
xmpyu ${fai}L,${fbi},${fab0} ; ap[0]*bp[i]
xmpyu ${fai}R,${fbi},${fab1} ; ap[1]*bp[i]
ldw 4($tp),$ti0 ; tp[j]
add $hi1,$nmlo1,$nmlo1
addc %r0,$nmhi1,$nmhi1
fstws,mb ${fab0}L,-8($xfer) ; save high part
add $ablo,$nmlo1,$nmlo1
addc %r0,$nmhi1,$hi1
fcpy,sgl %fr0,${fti}L ; zero high part
fcpy,sgl %fr0,${fab0}L
stw $nmlo1,-4($tp) ; tp[j-1]
fcnvxf,dbl,dbl ${fti},${fti} ; 32-bit unsigned int -> double
fcnvxf,dbl,dbl ${fab0},${fab0}
add $hi1,$hi0,$hi0
addc %r0,%r0,$hi1
fadd,dbl ${fti},${fab0},${fab0} ; add tp[0]
add $ti0,$hi0,$hi0
addc %r0,$hi1,$hi1
fcnvfx,dbl,dbl ${fab0},${fab0} ; double -> 33-bit unsigned int
stw $hi0,0($tp)
stw $hi1,4($tp)
xmpyu ${fn0},${fab0}R,${fm0}
b L\$outer_pa11
ldo `$LOCALS+32+4`($fp),$tp
L\$outerdone_pa11
add $hi0,$ablo,$ablo
addc %r0,$abhi,$abhi
add $ti1,$ablo,$ablo
addc %r0,$abhi,$hi0
ldw 4($tp),$ti0 ; tp[j]
add $hi1,$nmlo1,$nmlo1
addc %r0,$nmhi1,$nmhi1
add $ablo,$nmlo1,$nmlo1
addc %r0,$nmhi1,$hi1
stw $nmlo1,-4($tp) ; tp[j-1]
add $hi1,$hi0,$hi0
addc %r0,%r0,$hi1
add $ti0,$hi0,$hi0
addc %r0,$hi1,$hi1
stw $hi0,0($tp)
stw $hi1,4($tp)
ldo `$LOCALS+32+4`($fp),$tp
sub %r0,%r0,%r0 ; clear borrow
ldw -4($tp),$ti0
addl $tp,$arrsz,$tp
L\$sub_pa11
ldwx $idx($np),$hi0
subb $ti0,$hi0,$hi1
ldwx $idx($tp),$ti0
addib,<> 4,$idx,L\$sub_pa11
stws,ma $hi1,4($rp)
subb $ti0,%r0,$hi1
ldo -4($tp),$tp
and $tp,$hi1,$ap
andcm $rp,$hi1,$bp
or $ap,$bp,$np
sub $rp,$arrsz,$rp ; rewind rp
subi 0,$arrsz,$idx
ldo `$LOCALS+32`($fp),$tp
L\$copy_pa11
ldwx $idx($np),$hi0
stws,ma %r0,4($tp)
addib,<> 4,$idx,L\$copy_pa11
stws,ma $hi0,4($rp)
nop ; alignment
L\$done
___
}
$code.=<<___;
ldi 1,%r28 ; signal "handled"
ldo $FRAME($fp),%sp ; destroy tp[num+1]
$POP `-$FRAME-$SAVED_RP`(%sp),%r2 ; standard epilogue
$POP `-$FRAME+1*$SIZE_T`(%sp),%r4
$POP `-$FRAME+2*$SIZE_T`(%sp),%r5
$POP `-$FRAME+3*$SIZE_T`(%sp),%r6
$POP `-$FRAME+4*$SIZE_T`(%sp),%r7
$POP `-$FRAME+5*$SIZE_T`(%sp),%r8
$POP `-$FRAME+6*$SIZE_T`(%sp),%r9
$POP `-$FRAME+7*$SIZE_T`(%sp),%r10
L\$abort
bv (%r2)
.EXIT
$POPMB -$FRAME(%sp),%r3
.PROCEND
.STRINGZ "Montgomery Multiplication for PA-RISC, CRYPTOGAMS by <appro\@openssl.org>"
___
# Explicitly encode PA-RISC 2.0 instructions used in this module, so
# that it can be compiled with .LEVEL 1.0. It should be noted that I
# wouldn't have to do this, if GNU assembler understood .ALLOW 2.0
# directive...
my $ldd = sub {
my ($mod,$args) = @_;
my $orig = "ldd$mod\t$args";
if ($args =~ /%r([0-9]+)\(%r([0-9]+)\),%r([0-9]+)/) # format 4
{ my $opcode=(0x03<<26)|($2<<21)|($1<<16)|(3<<6)|$3;
sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
}
elsif ($args =~ /(\-?[0-9]+)\(%r([0-9]+)\),%r([0-9]+)/) # format 5
{ my $opcode=(0x03<<26)|($2<<21)|(1<<12)|(3<<6)|$3;
$opcode|=(($1&0xF)<<17)|(($1&0x10)<<12); # encode offset
$opcode|=(1<<5) if ($mod =~ /^,m/);
$opcode|=(1<<13) if ($mod =~ /^,mb/);
sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
}
else { "\t".$orig; }
};
my $std = sub {
my ($mod,$args) = @_;
my $orig = "std$mod\t$args";
if ($args =~ /%r([0-9]+),(\-?[0-9]+)\(%r([0-9]+)\)/) # format 6
{ my $opcode=(0x03<<26)|($3<<21)|($1<<16)|(1<<12)|(0xB<<6);
$opcode|=(($2&0xF)<<1)|(($2&0x10)>>4); # encode offset
$opcode|=(1<<5) if ($mod =~ /^,m/);
$opcode|=(1<<13) if ($mod =~ /^,mb/);
sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
}
else { "\t".$orig; }
};
my $extrd = sub {
my ($mod,$args) = @_;
my $orig = "extrd$mod\t$args";
# I only have ",u" completer, it's implicitly encoded...
if ($args =~ /%r([0-9]+),([0-9]+),([0-9]+),%r([0-9]+)/) # format 15
{ my $opcode=(0x36<<26)|($1<<21)|($4<<16);
my $len=32-$3;
$opcode |= (($2&0x20)<<6)|(($2&0x1f)<<5); # encode pos
$opcode |= (($len&0x20)<<7)|($len&0x1f); # encode len
sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
}
elsif ($args =~ /%r([0-9]+),%sar,([0-9]+),%r([0-9]+)/) # format 12
{ my $opcode=(0x34<<26)|($1<<21)|($3<<16)|(2<<11)|(1<<9);
my $len=32-$2;
$opcode |= (($len&0x20)<<3)|($len&0x1f); # encode len
$opcode |= (1<<13) if ($mod =~ /,\**=/);
sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
}
else { "\t".$orig; }
};
my $shrpd = sub {
my ($mod,$args) = @_;
my $orig = "shrpd$mod\t$args";
if ($args =~ /%r([0-9]+),%r([0-9]+),([0-9]+),%r([0-9]+)/) # format 14
{ my $opcode=(0x34<<26)|($2<<21)|($1<<16)|(1<<10)|$4;
my $cpos=63-$3;
$opcode |= (($cpos&0x20)<<6)|(($cpos&0x1f)<<5); # encode sa
sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig;
}
else { "\t".$orig; }
};
my $sub = sub {
my ($mod,$args) = @_;
my $orig = "sub$mod\t$args";
if ($mod eq ",db" && $args =~ /%r([0-9]+),%r([0-9]+),%r([0-9]+)/) {
my $opcode=(0x02<<26)|($2<<21)|($1<<16)|$3;
$opcode|=(1<<10); # e1
$opcode|=(1<<8); # e2
$opcode|=(1<<5); # d
sprintf "\t.WORD\t0x%08x\t; %s",$opcode,$orig
}
else { "\t".$orig; }
};
sub assemble {
my ($mnemonic,$mod,$args)=@_;
my $opcode = eval("\$$mnemonic");
ref($opcode) eq 'CODE' ? &$opcode($mod,$args) : "\t$mnemonic$mod\t$args";
}
foreach (split("\n",$code)) {
s/\`([^\`]*)\`/eval $1/ge;
# flip word order in 64-bit mode...
s/(xmpyu\s+)($fai|$fni)([LR])/$1.$2.($3 eq "L"?"R":"L")/e if ($BN_SZ==8);
# assemble 2.0 instructions in 32-bit mode...
s/^\s+([a-z]+)([\S]*)\s+([\S]*)/&assemble($1,$2,$3)/e if ($BN_SZ==4);
s/\bbv\b/bve/gm if ($SIZE_T==8);
print $_,"\n";
}
close STDOUT;

334
drivers/builtin_openssl2/crypto/bn/asm/ppc-mont.pl
View File

@ -1,334 +0,0 @@
#!/usr/bin/env perl
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# April 2006
# "Teaser" Montgomery multiplication module for PowerPC. It's possible
# to gain a bit more by modulo-scheduling outer loop, then dedicated
# squaring procedure should give further 20% and code can be adapted
# for 32-bit application running on 64-bit CPU. As for the latter.
# It won't be able to achieve "native" 64-bit performance, because in
# 32-bit application context every addc instruction will have to be
# expanded as addc, twice right shift by 32 and finally adde, etc.
# So far RSA *sign* performance improvement over pre-bn_mul_mont asm
# for 64-bit application running on PPC970/G5 is:
#
# 512-bit +65%
# 1024-bit +35%
# 2048-bit +18%
# 4096-bit +4%
$flavour = shift;
if ($flavour =~ /32/) {
$BITS= 32;
$BNSZ= $BITS/8;
$SIZE_T=4;
$RZONE= 224;
$LD= "lwz"; # load
$LDU= "lwzu"; # load and update
$LDX= "lwzx"; # load indexed
$ST= "stw"; # store
$STU= "stwu"; # store and update
$STX= "stwx"; # store indexed
$STUX= "stwux"; # store indexed and update
$UMULL= "mullw"; # unsigned multiply low
$UMULH= "mulhwu"; # unsigned multiply high
$UCMP= "cmplw"; # unsigned compare
$SHRI= "srwi"; # unsigned shift right by immediate
$PUSH= $ST;
$POP= $LD;
} elsif ($flavour =~ /64/) {
$BITS= 64;
$BNSZ= $BITS/8;
$SIZE_T=8;
$RZONE= 288;
# same as above, but 64-bit mnemonics...
$LD= "ld"; # load
$LDU= "ldu"; # load and update
$LDX= "ldx"; # load indexed
$ST= "std"; # store
$STU= "stdu"; # store and update
$STX= "stdx"; # store indexed
$STUX= "stdux"; # store indexed and update
$UMULL= "mulld"; # unsigned multiply low
$UMULH= "mulhdu"; # unsigned multiply high
$UCMP= "cmpld"; # unsigned compare
$SHRI= "srdi"; # unsigned shift right by immediate
$PUSH= $ST;
$POP= $LD;
} else { die "nonsense $flavour"; }
$FRAME=8*$SIZE_T+$RZONE;
$LOCALS=8*$SIZE_T;
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
( $xlate="${dir}ppc-xlate.pl" and -f $xlate ) or
( $xlate="${dir}../../perlasm/ppc-xlate.pl" and -f $xlate) or
die "can't locate ppc-xlate.pl";
open STDOUT,"| $^X $xlate $flavour ".shift || die "can't call $xlate: $!";
$sp="r1";
$toc="r2";
$rp="r3"; $ovf="r3";
$ap="r4";
$bp="r5";
$np="r6";
$n0="r7";
$num="r8";
$rp="r9"; # $rp is reassigned
$aj="r10";
$nj="r11";
$tj="r12";
# non-volatile registers
$i="r20";
$j="r21";
$tp="r22";
$m0="r23";
$m1="r24";
$lo0="r25";
$hi0="r26";
$lo1="r27";
$hi1="r28";
$alo="r29";
$ahi="r30";
$nlo="r31";
#
$nhi="r0";
$code=<<___;
.machine "any"
.text
.globl .bn_mul_mont_int
.align 4
.bn_mul_mont_int:
cmpwi $num,4
mr $rp,r3 ; $rp is reassigned
li r3,0
bltlr
___
$code.=<<___ if ($BNSZ==4);
cmpwi $num,32 ; longer key performance is not better
bgelr
___
$code.=<<___;
slwi $num,$num,`log($BNSZ)/log(2)`
li $tj,-4096
addi $ovf,$num,$FRAME
subf $ovf,$ovf,$sp ; $sp-$ovf
and $ovf,$ovf,$tj ; minimize TLB usage
subf $ovf,$sp,$ovf ; $ovf-$sp
mr $tj,$sp
srwi $num,$num,`log($BNSZ)/log(2)`
$STUX $sp,$sp,$ovf
$PUSH r20,`-12*$SIZE_T`($tj)
$PUSH r21,`-11*$SIZE_T`($tj)
$PUSH r22,`-10*$SIZE_T`($tj)
$PUSH r23,`-9*$SIZE_T`($tj)
$PUSH r24,`-8*$SIZE_T`($tj)
$PUSH r25,`-7*$SIZE_T`($tj)
$PUSH r26,`-6*$SIZE_T`($tj)
$PUSH r27,`-5*$SIZE_T`($tj)
$PUSH r28,`-4*$SIZE_T`($tj)
$PUSH r29,`-3*$SIZE_T`($tj)
$PUSH r30,`-2*$SIZE_T`($tj)
$PUSH r31,`-1*$SIZE_T`($tj)
$LD $n0,0($n0) ; pull n0[0] value
addi $num,$num,-2 ; adjust $num for counter register
$LD $m0,0($bp) ; m0=bp[0]
$LD $aj,0($ap) ; ap[0]
addi $tp,$sp,$LOCALS
$UMULL $lo0,$aj,$m0 ; ap[0]*bp[0]
$UMULH $hi0,$aj,$m0
$LD $aj,$BNSZ($ap) ; ap[1]
$LD $nj,0($np) ; np[0]
$UMULL $m1,$lo0,$n0 ; "tp[0]"*n0
$UMULL $alo,$aj,$m0 ; ap[1]*bp[0]
$UMULH $ahi,$aj,$m0
$UMULL $lo1,$nj,$m1 ; np[0]*m1
$UMULH $hi1,$nj,$m1
$LD $nj,$BNSZ($np) ; np[1]
addc $lo1,$lo1,$lo0
addze $hi1,$hi1
$UMULL $nlo,$nj,$m1 ; np[1]*m1
$UMULH $nhi,$nj,$m1
mtctr $num
li $j,`2*$BNSZ`
.align 4
L1st:
$LDX $aj,$ap,$j ; ap[j]
addc $lo0,$alo,$hi0
$LDX $nj,$np,$j ; np[j]
addze $hi0,$ahi
$UMULL $alo,$aj,$m0 ; ap[j]*bp[0]
addc $lo1,$nlo,$hi1
$UMULH $ahi,$aj,$m0
addze $hi1,$nhi
$UMULL $nlo,$nj,$m1 ; np[j]*m1
addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[0]
$UMULH $nhi,$nj,$m1
addze $hi1,$hi1
$ST $lo1,0($tp) ; tp[j-1]
addi $j,$j,$BNSZ ; j++
addi $tp,$tp,$BNSZ ; tp++
bdnz- L1st
;L1st
addc $lo0,$alo,$hi0
addze $hi0,$ahi
addc $lo1,$nlo,$hi1
addze $hi1,$nhi
addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[0]
addze $hi1,$hi1
$ST $lo1,0($tp) ; tp[j-1]
li $ovf,0
addc $hi1,$hi1,$hi0
addze $ovf,$ovf ; upmost overflow bit
$ST $hi1,$BNSZ($tp)
li $i,$BNSZ
.align 4
Louter:
$LDX $m0,$bp,$i ; m0=bp[i]
$LD $aj,0($ap) ; ap[0]
addi $tp,$sp,$LOCALS
$LD $tj,$LOCALS($sp); tp[0]
$UMULL $lo0,$aj,$m0 ; ap[0]*bp[i]
$UMULH $hi0,$aj,$m0
$LD $aj,$BNSZ($ap) ; ap[1]
$LD $nj,0($np) ; np[0]
addc $lo0,$lo0,$tj ; ap[0]*bp[i]+tp[0]
$UMULL $alo,$aj,$m0 ; ap[j]*bp[i]
addze $hi0,$hi0
$UMULL $m1,$lo0,$n0 ; tp[0]*n0
$UMULH $ahi,$aj,$m0
$UMULL $lo1,$nj,$m1 ; np[0]*m1
$UMULH $hi1,$nj,$m1
$LD $nj,$BNSZ($np) ; np[1]
addc $lo1,$lo1,$lo0
$UMULL $nlo,$nj,$m1 ; np[1]*m1
addze $hi1,$hi1
$UMULH $nhi,$nj,$m1
mtctr $num
li $j,`2*$BNSZ`
.align 4
Linner:
$LDX $aj,$ap,$j ; ap[j]
addc $lo0,$alo,$hi0
$LD $tj,$BNSZ($tp) ; tp[j]
addze $hi0,$ahi
$LDX $nj,$np,$j ; np[j]
addc $lo1,$nlo,$hi1
$UMULL $alo,$aj,$m0 ; ap[j]*bp[i]
addze $hi1,$nhi
$UMULH $ahi,$aj,$m0
addc $lo0,$lo0,$tj ; ap[j]*bp[i]+tp[j]
$UMULL $nlo,$nj,$m1 ; np[j]*m1
addze $hi0,$hi0
$UMULH $nhi,$nj,$m1
addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[i]+tp[j]
addi $j,$j,$BNSZ ; j++
addze $hi1,$hi1
$ST $lo1,0($tp) ; tp[j-1]
addi $tp,$tp,$BNSZ ; tp++
bdnz- Linner
;Linner
$LD $tj,$BNSZ($tp) ; tp[j]
addc $lo0,$alo,$hi0
addze $hi0,$ahi
addc $lo0,$lo0,$tj ; ap[j]*bp[i]+tp[j]
addze $hi0,$hi0
addc $lo1,$nlo,$hi1
addze $hi1,$nhi
addc $lo1,$lo1,$lo0 ; np[j]*m1+ap[j]*bp[i]+tp[j]
addze $hi1,$hi1
$ST $lo1,0($tp) ; tp[j-1]
addic $ovf,$ovf,-1 ; move upmost overflow to XER[CA]
li $ovf,0
adde $hi1,$hi1,$hi0
addze $ovf,$ovf
$ST $hi1,$BNSZ($tp)
;
slwi $tj,$num,`log($BNSZ)/log(2)`
$UCMP $i,$tj
addi $i,$i,$BNSZ
ble- Louter
addi $num,$num,2 ; restore $num
subfc $j,$j,$j ; j=0 and "clear" XER[CA]
addi $tp,$sp,$LOCALS
mtctr $num
.align 4
Lsub: $LDX $tj,$tp,$j
$LDX $nj,$np,$j
subfe $aj,$nj,$tj ; tp[j]-np[j]
$STX $aj,$rp,$j
addi $j,$j,$BNSZ
bdnz- Lsub
li $j,0
mtctr $num
subfe $ovf,$j,$ovf ; handle upmost overflow bit
and $ap,$tp,$ovf
andc $np,$rp,$ovf
or $ap,$ap,$np ; ap=borrow?tp:rp
.align 4
Lcopy: ; copy or in-place refresh
$LDX $tj,$ap,$j
$STX $tj,$rp,$j
$STX $j,$tp,$j ; zap at once
addi $j,$j,$BNSZ
bdnz- Lcopy
$POP $tj,0($sp)
li r3,1
$POP r20,`-12*$SIZE_T`($tj)
$POP r21,`-11*$SIZE_T`($tj)
$POP r22,`-10*$SIZE_T`($tj)
$POP r23,`-9*$SIZE_T`($tj)
$POP r24,`-8*$SIZE_T`($tj)
$POP r25,`-7*$SIZE_T`($tj)
$POP r26,`-6*$SIZE_T`($tj)
$POP r27,`-5*$SIZE_T`($tj)
$POP r28,`-4*$SIZE_T`($tj)
$POP r29,`-3*$SIZE_T`($tj)
$POP r30,`-2*$SIZE_T`($tj)
$POP r31,`-1*$SIZE_T`($tj)
mr $sp,$tj
blr
.long 0
.byte 0,12,4,0,0x80,12,6,0
.long 0
.asciz "Montgomery Multiplication for PPC, CRYPTOGAMS by <appro\@openssl.org>"
___
$code =~ s/\`([^\`]*)\`/eval $1/gem;
print $code;
close STDOUT;

1998
drivers/builtin_openssl2/crypto/bn/asm/ppc.pl
View File

File diff suppressed because it is too large Load Diff

1088
drivers/builtin_openssl2/crypto/bn/asm/ppc64-mont.pl
View File

File diff suppressed because it is too large Load Diff

221
drivers/builtin_openssl2/crypto/bn/asm/s390x-gf2m.pl
View File

@ -1,221 +0,0 @@
#!/usr/bin/env perl
#
# ====================================================================
# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
#
# May 2011
#
# The module implements bn_GF2m_mul_2x2 polynomial multiplication used
# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for
# the time being... gcc 4.3 appeared to generate poor code, therefore
# the effort. And indeed, the module delivers 55%-90%(*) improvement
# on haviest ECDSA verify and ECDH benchmarks for 163- and 571-bit
# key lengths on z990, 30%-55%(*) - on z10, and 70%-110%(*) - on z196.
# This is for 64-bit build. In 32-bit "highgprs" case improvement is
# even higher, for example on z990 it was measured 80%-150%. ECDSA
# sign is modest 9%-12% faster. Keep in mind that these coefficients
# are not ones for bn_GF2m_mul_2x2 itself, as not all CPU time is
# burnt in it...
#
# (*) gcc 4.1 was observed to deliver better results than gcc 4.3,
# so that improvement coefficients can vary from one specific
# setup to another.
$flavour = shift;
if ($flavour =~ /3[12]/) {
$SIZE_T=4;
$g="";
} else {
$SIZE_T=8;
$g="g";
}
while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
open STDOUT,">$output";
$stdframe=16*$SIZE_T+4*8;
$rp="%r2";
$a1="%r3";
$a0="%r4";
$b1="%r5";
$b0="%r6";
$ra="%r14";
$sp="%r15";
@T=("%r0","%r1");
@i=("%r12","%r13");
($a1,$a2,$a4,$a8,$a12,$a48)=map("%r$_",(6..11));
($lo,$hi,$b)=map("%r$_",(3..5)); $a=$lo; $mask=$a8;
$code.=<<___;
.text
.type _mul_1x1,\@function
.align 16
_mul_1x1:
lgr $a1,$a
sllg $a2,$a,1
sllg $a4,$a,2
sllg $a8,$a,3
srag $lo,$a1,63 # broadcast 63rd bit
nihh $a1,0x1fff
srag @i[0],$a2,63 # broadcast 62nd bit
nihh $a2,0x3fff
srag @i[1],$a4,63 # broadcast 61st bit
nihh $a4,0x7fff
ngr $lo,$b
ngr @i[0],$b
ngr @i[1],$b
lghi @T[0],0
lgr $a12,$a1
stg @T[0],`$stdframe+0*8`($sp) # tab[0]=0
xgr $a12,$a2
stg $a1,`$stdframe+1*8`($sp) # tab[1]=a1
lgr $a48,$a4
stg $a2,`$stdframe+2*8`($sp) # tab[2]=a2
xgr $a48,$a8
stg $a12,`$stdframe+3*8`($sp) # tab[3]=a1^a2
xgr $a1,$a4
stg $a4,`$stdframe+4*8`($sp) # tab[4]=a4
xgr $a2,$a4
stg $a1,`$stdframe+5*8`($sp) # tab[5]=a1^a4
xgr $a12,$a4
stg $a2,`$stdframe+6*8`($sp) # tab[6]=a2^a4
xgr $a1,$a48
stg $a12,`$stdframe+7*8`($sp) # tab[7]=a1^a2^a4
xgr $a2,$a48
stg $a8,`$stdframe+8*8`($sp) # tab[8]=a8
xgr $a12,$a48
stg $a1,`$stdframe+9*8`($sp) # tab[9]=a1^a8
xgr $a1,$a4
stg $a2,`$stdframe+10*8`($sp) # tab[10]=a2^a8
xgr $a2,$a4
stg $a12,`$stdframe+11*8`($sp) # tab[11]=a1^a2^a8
xgr $a12,$a4
stg $a48,`$stdframe+12*8`($sp) # tab[12]=a4^a8
srlg $hi,$lo,1
stg $a1,`$stdframe+13*8`($sp) # tab[13]=a1^a4^a8
sllg $lo,$lo,63
stg $a2,`$stdframe+14*8`($sp) # tab[14]=a2^a4^a8
srlg @T[0],@i[0],2
stg $a12,`$stdframe+15*8`($sp) # tab[15]=a1^a2^a4^a8
lghi $mask,`0xf<<3`
sllg $a1,@i[0],62
sllg @i[0],$b,3
srlg @T[1],@i[1],3
ngr @i[0],$mask
sllg $a2,@i[1],61
srlg @i[1],$b,4-3
xgr $hi,@T[0]
ngr @i[1],$mask
xgr $lo,$a1
xgr $hi,@T[1]
xgr $lo,$a2
xg $lo,$stdframe(@i[0],$sp)
srlg @i[0],$b,8-3
ngr @i[0],$mask
___
for($n=1;$n<14;$n++) {
$code.=<<___;
lg @T[1],$stdframe(@i[1],$sp)
srlg @i[1],$b,`($n+2)*4`-3
sllg @T[0],@T[1],`$n*4`
ngr @i[1],$mask
srlg @T[1],@T[1],`64-$n*4`
xgr $lo,@T[0]
xgr $hi,@T[1]
___
push(@i,shift(@i)); push(@T,shift(@T));
}
$code.=<<___;
lg @T[1],$stdframe(@i[1],$sp)
sllg @T[0],@T[1],`$n*4`
srlg @T[1],@T[1],`64-$n*4`
xgr $lo,@T[0]
xgr $hi,@T[1]
lg @T[0],$stdframe(@i[0],$sp)
sllg @T[1],@T[0],`($n+1)*4`
srlg @T[0],@T[0],`64-($n+1)*4`
xgr $lo,@T[1]
xgr $hi,@T[0]
br $ra
.size _mul_1x1,.-_mul_1x1
.globl bn_GF2m_mul_2x2
.type bn_GF2m_mul_2x2,\@function
.align 16
bn_GF2m_mul_2x2:
stm${g} %r3,%r15,3*$SIZE_T($sp)
lghi %r1,-$stdframe-128
la %r0,0($sp)
la $sp,0(%r1,$sp) # alloca
st${g} %r0,0($sp) # back chain
___
if ($SIZE_T==8) {
my @r=map("%r$_",(6..9));
$code.=<<___;
bras $ra,_mul_1x1 # a1·b1
stmg $lo,$hi,16($rp)
lg $a,`$stdframe+128+4*$SIZE_T`($sp)
lg $b,`$stdframe+128+6*$SIZE_T`($sp)
bras $ra,_mul_1x1 # a0·b0
stmg $lo,$hi,0($rp)
lg $a,`$stdframe+128+3*$SIZE_T`($sp)
lg $b,`$stdframe+128+5*$SIZE_T`($sp)
xg $a,`$stdframe+128+4*$SIZE_T`($sp)
xg $b,`$stdframe+128+6*$SIZE_T`($sp)
bras $ra,_mul_1x1 # (a0+a1)·(b0+b1)
lmg @r[0],@r[3],0($rp)
xgr $lo,$hi
xgr $hi,@r[1]
xgr $lo,@r[0]
xgr $hi,@r[2]
xgr $lo,@r[3]
xgr $hi,@r[3]
xgr $lo,$hi
stg $hi,16($rp)
stg $lo,8($rp)
___
} else {
$code.=<<___;
sllg %r3,%r3,32
sllg %r5,%r5,32
or %r3,%r4
or %r5,%r6
bras $ra,_mul_1x1
rllg $lo,$lo,32
rllg $hi,$hi,32
stmg $lo,$hi,0($rp)
___
}
$code.=<<___;
lm${g} %r6,%r15,`$stdframe+128+6*$SIZE_T`($sp)
br $ra
.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2
.string "GF(2^m) Multiplication for s390x, CRYPTOGAMS by <appro\@openssl.org>"
___
$code =~ s/\`([^\`]*)\`/eval($1)/gem;
print $code;
close STDOUT;

277
drivers/builtin_openssl2/crypto/bn/asm/s390x-mont.pl
View File

@ -1,277 +0,0 @@
#!/usr/bin/env perl
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# April 2007.
#
# Performance improvement over vanilla C code varies from 85% to 45%
# depending on key length and benchmark. Unfortunately in this context
# these are not very impressive results [for code that utilizes "wide"
# 64x64=128-bit multiplication, which is not commonly available to C
# programmers], at least hand-coded bn_asm.c replacement is known to
# provide 30-40% better results for longest keys. Well, on a second
# thought it's not very surprising, because z-CPUs are single-issue
# and _strictly_ in-order execution, while bn_mul_mont is more or less
# dependent on CPU ability to pipe-line instructions and have several
# of them "in-flight" at the same time. I mean while other methods,
# for example Karatsuba, aim to minimize amount of multiplications at
# the cost of other operations increase, bn_mul_mont aim to neatly
# "overlap" multiplications and the other operations [and on most
# platforms even minimize the amount of the other operations, in
# particular references to memory]. But it's possible to improve this
# module performance by implementing dedicated squaring code-path and
# possibly by unrolling loops...
# January 2009.
#
# Reschedule to minimize/avoid Address Generation Interlock hazard,
# make inner loops counter-based.
# November 2010.
#
# Adapt for -m31 build. If kernel supports what's called "highgprs"
# feature on Linux [see /proc/cpuinfo], it's possible to use 64-bit
# instructions and achieve "64-bit" performance even in 31-bit legacy
# application context. The feature is not specific to any particular
# processor, as long as it's "z-CPU". Latter implies that the code
# remains z/Architecture specific. Compatibility with 32-bit BN_ULONG
# is achieved by swapping words after 64-bit loads, follow _dswap-s.
# On z990 it was measured to perform 2.6-2.2 times better than
# compiler-generated code, less for longer keys...
$flavour = shift;
if ($flavour =~ /3[12]/) {
$SIZE_T=4;
$g="";
} else {
$SIZE_T=8;
$g="g";
}
while (($output=shift) && ($output!~/^\w[\w\-]*\.\w+$/)) {}
open STDOUT,">$output";
$stdframe=16*$SIZE_T+4*8;
$mn0="%r0";
$num="%r1";
# int bn_mul_mont(
$rp="%r2"; # BN_ULONG *rp,
$ap="%r3"; # const BN_ULONG *ap,
$bp="%r4"; # const BN_ULONG *bp,
$np="%r5"; # const BN_ULONG *np,
$n0="%r6"; # const BN_ULONG *n0,
#$num="160(%r15)" # int num);
$bi="%r2"; # zaps rp
$j="%r7";
$ahi="%r8";
$alo="%r9";
$nhi="%r10";
$nlo="%r11";
$AHI="%r12";
$NHI="%r13";
$count="%r14";
$sp="%r15";
$code.=<<___;
.text
.globl bn_mul_mont
.type bn_mul_mont,\@function
bn_mul_mont:
lgf $num,`$stdframe+$SIZE_T-4`($sp) # pull $num
sla $num,`log($SIZE_T)/log(2)` # $num to enumerate bytes
la $bp,0($num,$bp)
st${g} %r2,2*$SIZE_T($sp)
cghi $num,16 #
lghi %r2,0 #
blr %r14 # if($num<16) return 0;
___
$code.=<<___ if ($flavour =~ /3[12]/);
tmll $num,4
bnzr %r14 # if ($num&1) return 0;
___
$code.=<<___ if ($flavour !~ /3[12]/);
cghi $num,96 #
bhr %r14 # if($num>96) return 0;
___
$code.=<<___;
stm${g} %r3,%r15,3*$SIZE_T($sp)
lghi $rp,-$stdframe-8 # leave room for carry bit
lcgr $j,$num # -$num
lgr %r0,$sp
la $rp,0($rp,$sp)
la $sp,0($j,$rp) # alloca
st${g} %r0,0($sp) # back chain
sra $num,3 # restore $num
la $bp,0($j,$bp) # restore $bp
ahi $num,-1 # adjust $num for inner loop
lg $n0,0($n0) # pull n0
_dswap $n0
lg $bi,0($bp)
_dswap $bi
lg $alo,0($ap)
_dswap $alo
mlgr $ahi,$bi # ap[0]*bp[0]
lgr $AHI,$ahi
lgr $mn0,$alo # "tp[0]"*n0
msgr $mn0,$n0
lg $nlo,0($np) #
_dswap $nlo
mlgr $nhi,$mn0 # np[0]*m1
algr $nlo,$alo # +="tp[0]"
lghi $NHI,0
alcgr $NHI,$nhi
la $j,8(%r0) # j=1
lr $count,$num
.align 16
.L1st:
lg $alo,0($j,$ap)
_dswap $alo
mlgr $ahi,$bi # ap[j]*bp[0]
algr $alo,$AHI
lghi $AHI,0
alcgr $AHI,$ahi
lg $nlo,0($j,$np)
_dswap $nlo
mlgr $nhi,$mn0 # np[j]*m1
algr $nlo,$NHI
lghi $NHI,0
alcgr $nhi,$NHI # +="tp[j]"
algr $nlo,$alo
alcgr $NHI,$nhi
stg $nlo,$stdframe-8($j,$sp) # tp[j-1]=
la $j,8($j) # j++
brct $count,.L1st
algr $NHI,$AHI
lghi $AHI,0
alcgr $AHI,$AHI # upmost overflow bit
stg $NHI,$stdframe-8($j,$sp)
stg $AHI,$stdframe($j,$sp)
la $bp,8($bp) # bp++
.Louter:
lg $bi,0($bp) # bp[i]
_dswap $bi
lg $alo,0($ap)
_dswap $alo
mlgr $ahi,$bi # ap[0]*bp[i]
alg $alo,$stdframe($sp) # +=tp[0]
lghi $AHI,0
alcgr $AHI,$ahi
lgr $mn0,$alo
msgr $mn0,$n0 # tp[0]*n0
lg $nlo,0($np) # np[0]
_dswap $nlo
mlgr $nhi,$mn0 # np[0]*m1
algr $nlo,$alo # +="tp[0]"
lghi $NHI,0
alcgr $NHI,$nhi
la $j,8(%r0) # j=1
lr $count,$num
.align 16
.Linner:
lg $alo,0($j,$ap)
_dswap $alo
mlgr $ahi,$bi # ap[j]*bp[i]
algr $alo,$AHI
lghi $AHI,0
alcgr $ahi,$AHI
alg $alo,$stdframe($j,$sp)# +=tp[j]
alcgr $AHI,$ahi
lg $nlo,0($j,$np)
_dswap $nlo
mlgr $nhi,$mn0 # np[j]*m1
algr $nlo,$NHI
lghi $NHI,0
alcgr $nhi,$NHI
algr $nlo,$alo # +="tp[j]"
alcgr $NHI,$nhi
stg $nlo,$stdframe-8($j,$sp) # tp[j-1]=
la $j,8($j) # j++
brct $count,.Linner
algr $NHI,$AHI
lghi $AHI,0
alcgr $AHI,$AHI
alg $NHI,$stdframe($j,$sp)# accumulate previous upmost overflow bit
lghi $ahi,0
alcgr $AHI,$ahi # new upmost overflow bit
stg $NHI,$stdframe-8($j,$sp)
stg $AHI,$stdframe($j,$sp)
la $bp,8($bp) # bp++
cl${g} $bp,`$stdframe+8+4*$SIZE_T`($j,$sp) # compare to &bp[num]
jne .Louter
l${g} $rp,`$stdframe+8+2*$SIZE_T`($j,$sp) # reincarnate rp
la $ap,$stdframe($sp)
ahi $num,1 # restore $num, incidentally clears "borrow"
la $j,0(%r0)
lr $count,$num
.Lsub: lg $alo,0($j,$ap)
lg $nlo,0($j,$np)
_dswap $nlo
slbgr $alo,$nlo
stg $alo,0($j,$rp)
la $j,8($j)
brct $count,.Lsub
lghi $ahi,0
slbgr $AHI,$ahi # handle upmost carry
ngr $ap,$AHI
lghi $np,-1
xgr $np,$AHI
ngr $np,$rp
ogr $ap,$np # ap=borrow?tp:rp
la $j,0(%r0)
lgr $count,$num
.Lcopy: lg $alo,0($j,$ap) # copy or in-place refresh
_dswap $alo
stg $j,$stdframe($j,$sp) # zap tp
stg $alo,0($j,$rp)
la $j,8($j)
brct $count,.Lcopy
la %r1,`$stdframe+8+6*$SIZE_T`($j,$sp)
lm${g} %r6,%r15,0(%r1)
lghi %r2,1 # signal "processed"
br %r14
.size bn_mul_mont,.-bn_mul_mont
.string "Montgomery Multiplication for s390x, CRYPTOGAMS by <appro\@openssl.org>"
___
foreach (split("\n",$code)) {
s/\`([^\`]*)\`/eval $1/ge;
s/_dswap\s+(%r[0-9]+)/sprintf("rllg\t%s,%s,32",$1,$1) if($SIZE_T==4)/e;
print $_,"\n";
}
close STDOUT;

678
drivers/builtin_openssl2/crypto/bn/asm/s390x.S
View File

@ -1,678 +0,0 @@
.ident "s390x.S, version 1.1"
// ====================================================================
// Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
// project.
//
// Rights for redistribution and usage in source and binary forms are
// granted according to the OpenSSL license. Warranty of any kind is
// disclaimed.
// ====================================================================
.text
#define zero %r0
// BN_ULONG bn_mul_add_words(BN_ULONG *r2,BN_ULONG *r3,int r4,BN_ULONG r5);
.globl bn_mul_add_words
.type bn_mul_add_words,@function
.align 4
bn_mul_add_words:
lghi zero,0 // zero = 0
la %r1,0(%r2) // put rp aside
lghi %r2,0 // i=0;
ltgfr %r4,%r4
bler %r14 // if (len<=0) return 0;
stmg %r6,%r10,48(%r15)
lghi %r10,3
lghi %r8,0 // carry = 0
nr %r10,%r4 // len%4
sra %r4,2 // cnt=len/4
jz .Loop1_madd // carry is incidentally cleared if branch taken
algr zero,zero // clear carry
.Loop4_madd:
lg %r7,0(%r2,%r3) // ap[i]
mlgr %r6,%r5 // *=w
alcgr %r7,%r8 // +=carry
alcgr %r6,zero
alg %r7,0(%r2,%r1) // +=rp[i]
stg %r7,0(%r2,%r1) // rp[i]=
lg %r9,8(%r2,%r3)
mlgr %r8,%r5
alcgr %r9,%r6
alcgr %r8,zero
alg %r9,8(%r2,%r1)
stg %r9,8(%r2,%r1)
lg %r7,16(%r2,%r3)
mlgr %r6,%r5
alcgr %r7,%r8
alcgr %r6,zero
alg %r7,16(%r2,%r1)
stg %r7,16(%r2,%r1)
lg %r9,24(%r2,%r3)
mlgr %r8,%r5
alcgr %r9,%r6
alcgr %r8,zero
alg %r9,24(%r2,%r1)
stg %r9,24(%r2,%r1)
la %r2,32(%r2) // i+=4
brct %r4,.Loop4_madd
la %r10,1(%r10) // see if len%4 is zero ...
brct %r10,.Loop1_madd // without touching condition code:-)
.Lend_madd:
alcgr %r8,zero // collect carry bit
lgr %r2,%r8
lmg %r6,%r10,48(%r15)
br %r14
.Loop1_madd:
lg %r7,0(%r2,%r3) // ap[i]
mlgr %r6,%r5 // *=w
alcgr %r7,%r8 // +=carry
alcgr %r6,zero
alg %r7,0(%r2,%r1) // +=rp[i]
stg %r7,0(%r2,%r1) // rp[i]=
lgr %r8,%r6
la %r2,8(%r2) // i++
brct %r10,.Loop1_madd
j .Lend_madd
.size bn_mul_add_words,.-bn_mul_add_words
// BN_ULONG bn_mul_words(BN_ULONG *r2,BN_ULONG *r3,int r4,BN_ULONG r5);
.globl bn_mul_words
.type bn_mul_words,@function
.align 4
bn_mul_words:
lghi zero,0 // zero = 0
la %r1,0(%r2) // put rp aside
lghi %r2,0 // i=0;
ltgfr %r4,%r4
bler %r14 // if (len<=0) return 0;
stmg %r6,%r10,48(%r15)
lghi %r10,3
lghi %r8,0 // carry = 0
nr %r10,%r4 // len%4
sra %r4,2 // cnt=len/4
jz .Loop1_mul // carry is incidentally cleared if branch taken
algr zero,zero // clear carry
.Loop4_mul:
lg %r7,0(%r2,%r3) // ap[i]
mlgr %r6,%r5 // *=w
alcgr %r7,%r8 // +=carry
stg %r7,0(%r2,%r1) // rp[i]=
lg %r9,8(%r2,%r3)
mlgr %r8,%r5
alcgr %r9,%r6
stg %r9,8(%r2,%r1)
lg %r7,16(%r2,%r3)
mlgr %r6,%r5
alcgr %r7,%r8
stg %r7,16(%r2,%r1)
lg %r9,24(%r2,%r3)
mlgr %r8,%r5
alcgr %r9,%r6
stg %r9,24(%r2,%r1)
la %r2,32(%r2) // i+=4
brct %r4,.Loop4_mul
la %r10,1(%r10) // see if len%4 is zero ...
brct %r10,.Loop1_mul // without touching condition code:-)
.Lend_mul:
alcgr %r8,zero // collect carry bit
lgr %r2,%r8
lmg %r6,%r10,48(%r15)
br %r14
.Loop1_mul:
lg %r7,0(%r2,%r3) // ap[i]
mlgr %r6,%r5 // *=w
alcgr %r7,%r8 // +=carry
stg %r7,0(%r2,%r1) // rp[i]=
lgr %r8,%r6
la %r2,8(%r2) // i++
brct %r10,.Loop1_mul
j .Lend_mul
.size bn_mul_words,.-bn_mul_words
// void bn_sqr_words(BN_ULONG *r2,BN_ULONG *r2,int r4)
.globl bn_sqr_words
.type bn_sqr_words,@function
.align 4
bn_sqr_words:
ltgfr %r4,%r4
bler %r14
stmg %r6,%r7,48(%r15)
srag %r1,%r4,2 // cnt=len/4
jz .Loop1_sqr
.Loop4_sqr:
lg %r7,0(%r3)
mlgr %r6,%r7
stg %r7,0(%r2)
stg %r6,8(%r2)
lg %r7,8(%r3)
mlgr %r6,%r7
stg %r7,16(%r2)
stg %r6,24(%r2)
lg %r7,16(%r3)
mlgr %r6,%r7
stg %r7,32(%r2)
stg %r6,40(%r2)
lg %r7,24(%r3)
mlgr %r6,%r7
stg %r7,48(%r2)
stg %r6,56(%r2)
la %r3,32(%r3)
la %r2,64(%r2)
brct %r1,.Loop4_sqr
lghi %r1,3
nr %r4,%r1 // cnt=len%4
jz .Lend_sqr
.Loop1_sqr:
lg %r7,0(%r3)
mlgr %r6,%r7
stg %r7,0(%r2)
stg %r6,8(%r2)
la %r3,8(%r3)
la %r2,16(%r2)
brct %r4,.Loop1_sqr
.Lend_sqr:
lmg %r6,%r7,48(%r15)
br %r14
.size bn_sqr_words,.-bn_sqr_words
// BN_ULONG bn_div_words(BN_ULONG h,BN_ULONG l,BN_ULONG d);
.globl bn_div_words
.type bn_div_words,@function
.align 4
bn_div_words:
dlgr %r2,%r4
lgr %r2,%r3
br %r14
.size bn_div_words,.-bn_div_words
// BN_ULONG bn_add_words(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4,int r5);
.globl bn_add_words
.type bn_add_words,@function
.align 4
bn_add_words:
la %r1,0(%r2) // put rp aside
lghi %r2,0 // i=0
ltgfr %r5,%r5
bler %r14 // if (len<=0) return 0;
stg %r6,48(%r15)
lghi %r6,3
nr %r6,%r5 // len%4
sra %r5,2 // len/4, use sra because it sets condition code
jz .Loop1_add // carry is incidentally cleared if branch taken
algr %r2,%r2 // clear carry
.Loop4_add:
lg %r0,0(%r2,%r3)
alcg %r0,0(%r2,%r4)
stg %r0,0(%r2,%r1)
lg %r0,8(%r2,%r3)
alcg %r0,8(%r2,%r4)
stg %r0,8(%r2,%r1)
lg %r0,16(%r2,%r3)
alcg %r0,16(%r2,%r4)
stg %r0,16(%r2,%r1)
lg %r0,24(%r2,%r3)
alcg %r0,24(%r2,%r4)
stg %r0,24(%r2,%r1)
la %r2,32(%r2) // i+=4
brct %r5,.Loop4_add
la %r6,1(%r6) // see if len%4 is zero ...
brct %r6,.Loop1_add // without touching condition code:-)
.Lexit_add:
lghi %r2,0
alcgr %r2,%r2
lg %r6,48(%r15)
br %r14
.Loop1_add:
lg %r0,0(%r2,%r3)
alcg %r0,0(%r2,%r4)
stg %r0,0(%r2,%r1)
la %r2,8(%r2) // i++
brct %r6,.Loop1_add
j .Lexit_add
.size bn_add_words,.-bn_add_words
// BN_ULONG bn_sub_words(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4,int r5);
.globl bn_sub_words
.type bn_sub_words,@function
.align 4
bn_sub_words:
la %r1,0(%r2) // put rp aside
lghi %r2,0 // i=0
ltgfr %r5,%r5
bler %r14 // if (len<=0) return 0;
stg %r6,48(%r15)
lghi %r6,3
nr %r6,%r5 // len%4
sra %r5,2 // len/4, use sra because it sets condition code
jnz .Loop4_sub // borrow is incidentally cleared if branch taken
slgr %r2,%r2 // clear borrow
.Loop1_sub:
lg %r0,0(%r2,%r3)
slbg %r0,0(%r2,%r4)
stg %r0,0(%r2,%r1)
la %r2,8(%r2) // i++
brct %r6,.Loop1_sub
j .Lexit_sub
.Loop4_sub:
lg %r0,0(%r2,%r3)
slbg %r0,0(%r2,%r4)
stg %r0,0(%r2,%r1)
lg %r0,8(%r2,%r3)
slbg %r0,8(%r2,%r4)
stg %r0,8(%r2,%r1)
lg %r0,16(%r2,%r3)
slbg %r0,16(%r2,%r4)
stg %r0,16(%r2,%r1)
lg %r0,24(%r2,%r3)
slbg %r0,24(%r2,%r4)
stg %r0,24(%r2,%r1)
la %r2,32(%r2) // i+=4
brct %r5,.Loop4_sub
la %r6,1(%r6) // see if len%4 is zero ...
brct %r6,.Loop1_sub // without touching condition code:-)
.Lexit_sub:
lghi %r2,0
slbgr %r2,%r2
lcgr %r2,%r2
lg %r6,48(%r15)
br %r14
.size bn_sub_words,.-bn_sub_words
#define c1 %r1
#define c2 %r5
#define c3 %r8
#define mul_add_c(ai,bi,c1,c2,c3) \
lg %r7,ai*8(%r3); \
mlg %r6,bi*8(%r4); \
algr c1,%r7; \
alcgr c2,%r6; \
alcgr c3,zero
// void bn_mul_comba8(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4);
.globl bn_mul_comba8
.type bn_mul_comba8,@function
.align 4
bn_mul_comba8:
stmg %r6,%r8,48(%r15)
lghi c1,0
lghi c2,0
lghi c3,0
lghi zero,0
mul_add_c(0,0,c1,c2,c3);
stg c1,0*8(%r2)
lghi c1,0
mul_add_c(0,1,c2,c3,c1);
mul_add_c(1,0,c2,c3,c1);
stg c2,1*8(%r2)
lghi c2,0
mul_add_c(2,0,c3,c1,c2);
mul_add_c(1,1,c3,c1,c2);
mul_add_c(0,2,c3,c1,c2);
stg c3,2*8(%r2)
lghi c3,0
mul_add_c(0,3,c1,c2,c3);
mul_add_c(1,2,c1,c2,c3);
mul_add_c(2,1,c1,c2,c3);
mul_add_c(3,0,c1,c2,c3);
stg c1,3*8(%r2)
lghi c1,0
mul_add_c(4,0,c2,c3,c1);
mul_add_c(3,1,c2,c3,c1);
mul_add_c(2,2,c2,c3,c1);
mul_add_c(1,3,c2,c3,c1);
mul_add_c(0,4,c2,c3,c1);
stg c2,4*8(%r2)
lghi c2,0
mul_add_c(0,5,c3,c1,c2);
mul_add_c(1,4,c3,c1,c2);
mul_add_c(2,3,c3,c1,c2);
mul_add_c(3,2,c3,c1,c2);
mul_add_c(4,1,c3,c1,c2);
mul_add_c(5,0,c3,c1,c2);
stg c3,5*8(%r2)
lghi c3,0
mul_add_c(6,0,c1,c2,c3);
mul_add_c(5,1,c1,c2,c3);
mul_add_c(4,2,c1,c2,c3);
mul_add_c(3,3,c1,c2,c3);
mul_add_c(2,4,c1,c2,c3);
mul_add_c(1,5,c1,c2,c3);
mul_add_c(0,6,c1,c2,c3);
stg c1,6*8(%r2)
lghi c1,0
mul_add_c(0,7,c2,c3,c1);
mul_add_c(1,6,c2,c3,c1);
mul_add_c(2,5,c2,c3,c1);
mul_add_c(3,4,c2,c3,c1);
mul_add_c(4,3,c2,c3,c1);
mul_add_c(5,2,c2,c3,c1);
mul_add_c(6,1,c2,c3,c1);
mul_add_c(7,0,c2,c3,c1);
stg c2,7*8(%r2)
lghi c2,0
mul_add_c(7,1,c3,c1,c2);
mul_add_c(6,2,c3,c1,c2);
mul_add_c(5,3,c3,c1,c2);
mul_add_c(4,4,c3,c1,c2);
mul_add_c(3,5,c3,c1,c2);
mul_add_c(2,6,c3,c1,c2);
mul_add_c(1,7,c3,c1,c2);
stg c3,8*8(%r2)
lghi c3,0
mul_add_c(2,7,c1,c2,c3);
mul_add_c(3,6,c1,c2,c3);
mul_add_c(4,5,c1,c2,c3);
mul_add_c(5,4,c1,c2,c3);
mul_add_c(6,3,c1,c2,c3);
mul_add_c(7,2,c1,c2,c3);
stg c1,9*8(%r2)
lghi c1,0
mul_add_c(7,3,c2,c3,c1);
mul_add_c(6,4,c2,c3,c1);
mul_add_c(5,5,c2,c3,c1);
mul_add_c(4,6,c2,c3,c1);
mul_add_c(3,7,c2,c3,c1);
stg c2,10*8(%r2)
lghi c2,0
mul_add_c(4,7,c3,c1,c2);
mul_add_c(5,6,c3,c1,c2);
mul_add_c(6,5,c3,c1,c2);
mul_add_c(7,4,c3,c1,c2);
stg c3,11*8(%r2)
lghi c3,0
mul_add_c(7,5,c1,c2,c3);
mul_add_c(6,6,c1,c2,c3);
mul_add_c(5,7,c1,c2,c3);
stg c1,12*8(%r2)
lghi c1,0
mul_add_c(6,7,c2,c3,c1);
mul_add_c(7,6,c2,c3,c1);
stg c2,13*8(%r2)
lghi c2,0
mul_add_c(7,7,c3,c1,c2);
stg c3,14*8(%r2)
stg c1,15*8(%r2)
lmg %r6,%r8,48(%r15)
br %r14
.size bn_mul_comba8,.-bn_mul_comba8
// void bn_mul_comba4(BN_ULONG *r2,BN_ULONG *r3,BN_ULONG *r4);
.globl bn_mul_comba4
.type bn_mul_comba4,@function
.align 4
bn_mul_comba4:
stmg %r6,%r8,48(%r15)
lghi c1,0
lghi c2,0
lghi c3,0
lghi zero,0
mul_add_c(0,0,c1,c2,c3);
stg c1,0*8(%r3)
lghi c1,0
mul_add_c(0,1,c2,c3,c1);
mul_add_c(1,0,c2,c3,c1);
stg c2,1*8(%r2)
lghi c2,0
mul_add_c(2,0,c3,c1,c2);
mul_add_c(1,1,c3,c1,c2);
mul_add_c(0,2,c3,c1,c2);
stg c3,2*8(%r2)
lghi c3,0
mul_add_c(0,3,c1,c2,c3);
mul_add_c(1,2,c1,c2,c3);
mul_add_c(2,1,c1,c2,c3);
mul_add_c(3,0,c1,c2,c3);
stg c1,3*8(%r2)
lghi c1,0
mul_add_c(3,1,c2,c3,c1);
mul_add_c(2,2,c2,c3,c1);
mul_add_c(1,3,c2,c3,c1);
stg c2,4*8(%r2)
lghi c2,0
mul_add_c(2,3,c3,c1,c2);
mul_add_c(3,2,c3,c1,c2);
stg c3,5*8(%r2)
lghi c3,0
mul_add_c(3,3,c1,c2,c3);
stg c1,6*8(%r2)
stg c2,7*8(%r2)
stmg %r6,%r8,48(%r15)
br %r14
.size bn_mul_comba4,.-bn_mul_comba4
#define sqr_add_c(ai,c1,c2,c3) \
lg %r7,ai*8(%r3); \
mlgr %r6,%r7; \
algr c1,%r7; \
alcgr c2,%r6; \
alcgr c3,zero
#define sqr_add_c2(ai,aj,c1,c2,c3) \
lg %r7,ai*8(%r3); \
mlg %r6,aj*8(%r3); \
algr c1,%r7; \
alcgr c2,%r6; \
alcgr c3,zero; \
algr c1,%r7; \
alcgr c2,%r6; \
alcgr c3,zero
// void bn_sqr_comba8(BN_ULONG *r2,BN_ULONG *r3);
.globl bn_sqr_comba8
.type bn_sqr_comba8,@function
.align 4
bn_sqr_comba8:
stmg %r6,%r8,48(%r15)
lghi c1,0
lghi c2,0
lghi c3,0
lghi zero,0
sqr_add_c(0,c1,c2,c3);
stg c1,0*8(%r2)
lghi c1,0
sqr_add_c2(1,0,c2,c3,c1);
stg c2,1*8(%r2)
lghi c2,0
sqr_add_c(1,c3,c1,c2);
sqr_add_c2(2,0,c3,c1,c2);
stg c3,2*8(%r2)
lghi c3,0
sqr_add_c2(3,0,c1,c2,c3);
sqr_add_c2(2,1,c1,c2,c3);
stg c1,3*8(%r2)
lghi c1,0
sqr_add_c(2,c2,c3,c1);
sqr_add_c2(3,1,c2,c3,c1);
sqr_add_c2(4,0,c2,c3,c1);
stg c2,4*8(%r2)
lghi c2,0
sqr_add_c2(5,0,c3,c1,c2);
sqr_add_c2(4,1,c3,c1,c2);
sqr_add_c2(3,2,c3,c1,c2);
stg c3,5*8(%r2)
lghi c3,0
sqr_add_c(3,c1,c2,c3);
sqr_add_c2(4,2,c1,c2,c3);
sqr_add_c2(5,1,c1,c2,c3);
sqr_add_c2(6,0,c1,c2,c3);
stg c1,6*8(%r2)
lghi c1,0
sqr_add_c2(7,0,c2,c3,c1);
sqr_add_c2(6,1,c2,c3,c1);
sqr_add_c2(5,2,c2,c3,c1);
sqr_add_c2(4,3,c2,c3,c1);
stg c2,7*8(%r2)
lghi c2,0
sqr_add_c(4,c3,c1,c2);
sqr_add_c2(5,3,c3,c1,c2);
sqr_add_c2(6,2,c3,c1,c2);
sqr_add_c2(7,1,c3,c1,c2);
stg c3,8*8(%r2)
lghi c3,0
sqr_add_c2(7,2,c1,c2,c3);
sqr_add_c2(6,3,c1,c2,c3);
sqr_add_c2(5,4,c1,c2,c3);
stg c1,9*8(%r2)
lghi c1,0
sqr_add_c(5,c2,c3,c1);
sqr_add_c2(6,4,c2,c3,c1);
sqr_add_c2(7,3,c2,c3,c1);
stg c2,10*8(%r2)
lghi c2,0
sqr_add_c2(7,4,c3,c1,c2);
sqr_add_c2(6,5,c3,c1,c2);
stg c3,11*8(%r2)
lghi c3,0
sqr_add_c(6,c1,c2,c3);
sqr_add_c2(7,5,c1,c2,c3);
stg c1,12*8(%r2)
lghi c1,0
sqr_add_c2(7,6,c2,c3,c1);
stg c2,13*8(%r2)
lghi c2,0
sqr_add_c(7,c3,c1,c2);
stg c3,14*8(%r2)
stg c1,15*8(%r2)
lmg %r6,%r8,48(%r15)
br %r14
.size bn_sqr_comba8,.-bn_sqr_comba8
// void bn_sqr_comba4(BN_ULONG *r2,BN_ULONG *r3);
.globl bn_sqr_comba4
.type bn_sqr_comba4,@function
.align 4
bn_sqr_comba4:
stmg %r6,%r8,48(%r15)
lghi c1,0
lghi c2,0
lghi c3,0
lghi zero,0
sqr_add_c(0,c1,c2,c3);
stg c1,0*8(%r2)
lghi c1,0
sqr_add_c2(1,0,c2,c3,c1);
stg c2,1*8(%r2)
lghi c2,0
sqr_add_c(1,c3,c1,c2);
sqr_add_c2(2,0,c3,c1,c2);
stg c3,2*8(%r2)
lghi c3,0
sqr_add_c2(3,0,c1,c2,c3);
sqr_add_c2(2,1,c1,c2,c3);
stg c1,3*8(%r2)
lghi c1,0
sqr_add_c(2,c2,c3,c1);
sqr_add_c2(3,1,c2,c3,c1);
stg c2,4*8(%r2)
lghi c2,0
sqr_add_c2(3,2,c3,c1,c2);
stg c3,5*8(%r2)
lghi c3,0
sqr_add_c(3,c1,c2,c3);
stg c1,6*8(%r2)
stg c2,7*8(%r2)
lmg %r6,%r8,48(%r15)
br %r14
.size bn_sqr_comba4,.-bn_sqr_comba4

1458
drivers/builtin_openssl2/crypto/bn/asm/sparcv8.S
View File

File diff suppressed because it is too large Load Diff

1558
drivers/builtin_openssl2/crypto/bn/asm/sparcv8plus.S
View File

File diff suppressed because it is too large Load Diff

606
drivers/builtin_openssl2/crypto/bn/asm/sparcv9-mont.pl
View File

@ -1,606 +0,0 @@
#!/usr/bin/env perl
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# December 2005
#
# Pure SPARCv9/8+ and IALU-only bn_mul_mont implementation. The reasons
# for undertaken effort are multiple. First of all, UltraSPARC is not
# the whole SPARCv9 universe and other VIS-free implementations deserve
# optimized code as much. Secondly, newly introduced UltraSPARC T1,
# a.k.a. Niagara, has shared FPU and concurrent FPU-intensive pathes,
# such as sparcv9a-mont, will simply sink it. Yes, T1 is equipped with
# several integrated RSA/DSA accelerator circuits accessible through
# kernel driver [only(*)], but having decent user-land software
# implementation is important too. Finally, reasons like desire to
# experiment with dedicated squaring procedure. Yes, this module
# implements one, because it was easiest to draft it in SPARCv9
# instructions...
# (*) Engine accessing the driver in question is on my TODO list.
# For reference, acceleator is estimated to give 6 to 10 times
# improvement on single-threaded RSA sign. It should be noted
# that 6-10x improvement coefficient does not actually mean
# something extraordinary in terms of absolute [single-threaded]
# performance, as SPARCv9 instruction set is by all means least
# suitable for high performance crypto among other 64 bit
# platforms. 6-10x factor simply places T1 in same performance
# domain as say AMD64 and IA-64. Improvement of RSA verify don't
# appear impressive at all, but it's the sign operation which is
# far more critical/interesting.
# You might notice that inner loops are modulo-scheduled:-) This has
# essentially negligible impact on UltraSPARC performance, it's
# Fujitsu SPARC64 V users who should notice and hopefully appreciate
# the advantage... Currently this module surpasses sparcv9a-mont.pl
# by ~20% on UltraSPARC-III and later cores, but recall that sparcv9a
# module still have hidden potential [see TODO list there], which is
# estimated to be larger than 20%...
# int bn_mul_mont(
$rp="%i0"; # BN_ULONG *rp,
$ap="%i1"; # const BN_ULONG *ap,
$bp="%i2"; # const BN_ULONG *bp,
$np="%i3"; # const BN_ULONG *np,
$n0="%i4"; # const BN_ULONG *n0,
$num="%i5"; # int num);
$bits=32;
for (@ARGV) { $bits=64 if (/\-m64/ || /\-xarch\=v9/); }
if ($bits==64) { $bias=2047; $frame=192; }
else { $bias=0; $frame=128; }
$car0="%o0";
$car1="%o1";
$car2="%o2"; # 1 bit
$acc0="%o3";
$acc1="%o4";
$mask="%g1"; # 32 bits, what a waste...
$tmp0="%g4";
$tmp1="%g5";
$i="%l0";
$j="%l1";
$mul0="%l2";
$mul1="%l3";
$tp="%l4";
$apj="%l5";
$npj="%l6";
$tpj="%l7";
$fname="bn_mul_mont_int";
$code=<<___;
.section ".text",#alloc,#execinstr
.global $fname
.align 32
$fname:
cmp %o5,4 ! 128 bits minimum
bge,pt %icc,.Lenter
sethi %hi(0xffffffff),$mask
retl
clr %o0
.align 32
.Lenter:
save %sp,-$frame,%sp
sll $num,2,$num ! num*=4
or $mask,%lo(0xffffffff),$mask
ld [$n0],$n0
cmp $ap,$bp
and $num,$mask,$num
ld [$bp],$mul0 ! bp[0]
nop
add %sp,$bias,%o7 ! real top of stack
ld [$ap],$car0 ! ap[0] ! redundant in squaring context
sub %o7,$num,%o7
ld [$ap+4],$apj ! ap[1]
and %o7,-1024,%o7
ld [$np],$car1 ! np[0]
sub %o7,$bias,%sp ! alloca
ld [$np+4],$npj ! np[1]
be,pt `$bits==32?"%icc":"%xcc"`,.Lbn_sqr_mont
mov 12,$j
mulx $car0,$mul0,$car0 ! ap[0]*bp[0]
mulx $apj,$mul0,$tmp0 !prologue! ap[1]*bp[0]
and $car0,$mask,$acc0
add %sp,$bias+$frame,$tp
ld [$ap+8],$apj !prologue!
mulx $n0,$acc0,$mul1 ! "t[0]"*n0
and $mul1,$mask,$mul1
mulx $car1,$mul1,$car1 ! np[0]*"t[0]"*n0
mulx $npj,$mul1,$acc1 !prologue! np[1]*"t[0]"*n0
srlx $car0,32,$car0
add $acc0,$car1,$car1
ld [$np+8],$npj !prologue!
srlx $car1,32,$car1
mov $tmp0,$acc0 !prologue!
.L1st:
mulx $apj,$mul0,$tmp0
mulx $npj,$mul1,$tmp1
add $acc0,$car0,$car0
ld [$ap+$j],$apj ! ap[j]
and $car0,$mask,$acc0
add $acc1,$car1,$car1
ld [$np+$j],$npj ! np[j]
srlx $car0,32,$car0
add $acc0,$car1,$car1
add $j,4,$j ! j++
mov $tmp0,$acc0
st $car1,[$tp]
cmp $j,$num
mov $tmp1,$acc1
srlx $car1,32,$car1
bl %icc,.L1st
add $tp,4,$tp ! tp++
!.L1st
mulx $apj,$mul0,$tmp0 !epilogue!
mulx $npj,$mul1,$tmp1
add $acc0,$car0,$car0
and $car0,$mask,$acc0
add $acc1,$car1,$car1
srlx $car0,32,$car0
add $acc0,$car1,$car1
st $car1,[$tp]
srlx $car1,32,$car1
add $tmp0,$car0,$car0
and $car0,$mask,$acc0
add $tmp1,$car1,$car1
srlx $car0,32,$car0
add $acc0,$car1,$car1
st $car1,[$tp+4]
srlx $car1,32,$car1
add $car0,$car1,$car1
st $car1,[$tp+8]
srlx $car1,32,$car2
mov 4,$i ! i++
ld [$bp+4],$mul0 ! bp[1]
.Louter:
add %sp,$bias+$frame,$tp
ld [$ap],$car0 ! ap[0]
ld [$ap+4],$apj ! ap[1]
ld [$np],$car1 ! np[0]
ld [$np+4],$npj ! np[1]
ld [$tp],$tmp1 ! tp[0]
ld [$tp+4],$tpj ! tp[1]
mov 12,$j
mulx $car0,$mul0,$car0
mulx $apj,$mul0,$tmp0 !prologue!
add $tmp1,$car0,$car0
ld [$ap+8],$apj !prologue!
and $car0,$mask,$acc0
mulx $n0,$acc0,$mul1
and $mul1,$mask,$mul1
mulx $car1,$mul1,$car1
mulx $npj,$mul1,$acc1 !prologue!
srlx $car0,32,$car0
add $acc0,$car1,$car1
ld [$np+8],$npj !prologue!
srlx $car1,32,$car1
mov $tmp0,$acc0 !prologue!
.Linner:
mulx $apj,$mul0,$tmp0
mulx $npj,$mul1,$tmp1
add $tpj,$car0,$car0
ld [$ap+$j],$apj ! ap[j]
add $acc0,$car0,$car0
add $acc1,$car1,$car1
ld [$np+$j],$npj ! np[j]
and $car0,$mask,$acc0
ld [$tp+8],$tpj ! tp[j]
srlx $car0,32,$car0
add $acc0,$car1,$car1
add $j,4,$j ! j++
mov $tmp0,$acc0
st $car1,[$tp] ! tp[j-1]
srlx $car1,32,$car1
mov $tmp1,$acc1
cmp $j,$num
bl %icc,.Linner
add $tp,4,$tp ! tp++
!.Linner
mulx $apj,$mul0,$tmp0 !epilogue!
mulx $npj,$mul1,$tmp1
add $tpj,$car0,$car0
add $acc0,$car0,$car0
ld [$tp+8],$tpj ! tp[j]
and $car0,$mask,$acc0
add $acc1,$car1,$car1
srlx $car0,32,$car0
add $acc0,$car1,$car1
st $car1,[$tp] ! tp[j-1]
srlx $car1,32,$car1
add $tpj,$car0,$car0
add $tmp0,$car0,$car0
and $car0,$mask,$acc0
add $tmp1,$car1,$car1
add $acc0,$car1,$car1
st $car1,[$tp+4] ! tp[j-1]
srlx $car0,32,$car0
add $i,4,$i ! i++
srlx $car1,32,$car1
add $car0,$car1,$car1
cmp $i,$num
add $car2,$car1,$car1
st $car1,[$tp+8]
srlx $car1,32,$car2
bl,a %icc,.Louter
ld [$bp+$i],$mul0 ! bp[i]
!.Louter
add $tp,12,$tp
.Ltail:
add $np,$num,$np
add $rp,$num,$rp
mov $tp,$ap
sub %g0,$num,%o7 ! k=-num
ba .Lsub
subcc %g0,%g0,%g0 ! clear %icc.c
.align 16
.Lsub:
ld [$tp+%o7],%o0
ld [$np+%o7],%o1
subccc %o0,%o1,%o1 ! tp[j]-np[j]
add $rp,%o7,$i
add %o7,4,%o7
brnz %o7,.Lsub
st %o1,[$i]
subc $car2,0,$car2 ! handle upmost overflow bit
and $tp,$car2,$ap
andn $rp,$car2,$np
or $ap,$np,$ap
sub %g0,$num,%o7
.Lcopy:
ld [$ap+%o7],%o0 ! copy or in-place refresh
st %g0,[$tp+%o7] ! zap tp
st %o0,[$rp+%o7]
add %o7,4,%o7
brnz %o7,.Lcopy
nop
mov 1,%i0
ret
restore
___
########
######## .Lbn_sqr_mont gives up to 20% *overall* improvement over
######## code without following dedicated squaring procedure.
########
$sbit="%i2"; # re-use $bp!
$code.=<<___;
.align 32
.Lbn_sqr_mont:
mulx $mul0,$mul0,$car0 ! ap[0]*ap[0]
mulx $apj,$mul0,$tmp0 !prologue!
and $car0,$mask,$acc0
add %sp,$bias+$frame,$tp
ld [$ap+8],$apj !prologue!
mulx $n0,$acc0,$mul1 ! "t[0]"*n0
srlx $car0,32,$car0
and $mul1,$mask,$mul1
mulx $car1,$mul1,$car1 ! np[0]*"t[0]"*n0
mulx $npj,$mul1,$acc1 !prologue!
and $car0,1,$sbit
ld [$np+8],$npj !prologue!
srlx $car0,1,$car0
add $acc0,$car1,$car1
srlx $car1,32,$car1
mov $tmp0,$acc0 !prologue!
.Lsqr_1st:
mulx $apj,$mul0,$tmp0
mulx $npj,$mul1,$tmp1
add $acc0,$car0,$car0 ! ap[j]*a0+c0
add $acc1,$car1,$car1
ld [$ap+$j],$apj ! ap[j]
and $car0,$mask,$acc0
ld [$np+$j],$npj ! np[j]
srlx $car0,32,$car0
add $acc0,$acc0,$acc0
or $sbit,$acc0,$acc0
mov $tmp1,$acc1
srlx $acc0,32,$sbit
add $j,4,$j ! j++
and $acc0,$mask,$acc0
cmp $j,$num
add $acc0,$car1,$car1
st $car1,[$tp]
mov $tmp0,$acc0
srlx $car1,32,$car1
bl %icc,.Lsqr_1st
add $tp,4,$tp ! tp++
!.Lsqr_1st
mulx $apj,$mul0,$tmp0 ! epilogue
mulx $npj,$mul1,$tmp1
add $acc0,$car0,$car0 ! ap[j]*a0+c0
add $acc1,$car1,$car1
and $car0,$mask,$acc0
srlx $car0,32,$car0
add $acc0,$acc0,$acc0
or $sbit,$acc0,$acc0
srlx $acc0,32,$sbit
and $acc0,$mask,$acc0
add $acc0,$car1,$car1
st $car1,[$tp]
srlx $car1,32,$car1
add $tmp0,$car0,$car0 ! ap[j]*a0+c0
add $tmp1,$car1,$car1
and $car0,$mask,$acc0
srlx $car0,32,$car0
add $acc0,$acc0,$acc0
or $sbit,$acc0,$acc0
srlx $acc0,32,$sbit
and $acc0,$mask,$acc0
add $acc0,$car1,$car1
st $car1,[$tp+4]
srlx $car1,32,$car1
add $car0,$car0,$car0
or $sbit,$car0,$car0
add $car0,$car1,$car1
st $car1,[$tp+8]
srlx $car1,32,$car2
ld [%sp+$bias+$frame],$tmp0 ! tp[0]
ld [%sp+$bias+$frame+4],$tmp1 ! tp[1]
ld [%sp+$bias+$frame+8],$tpj ! tp[2]
ld [$ap+4],$mul0 ! ap[1]
ld [$ap+8],$apj ! ap[2]
ld [$np],$car1 ! np[0]
ld [$np+4],$npj ! np[1]
mulx $n0,$tmp0,$mul1
mulx $mul0,$mul0,$car0
and $mul1,$mask,$mul1
mulx $car1,$mul1,$car1
mulx $npj,$mul1,$acc1
add $tmp0,$car1,$car1
and $car0,$mask,$acc0
ld [$np+8],$npj ! np[2]
srlx $car1,32,$car1
add $tmp1,$car1,$car1
srlx $car0,32,$car0
add $acc0,$car1,$car1
and $car0,1,$sbit
add $acc1,$car1,$car1
srlx $car0,1,$car0
mov 12,$j
st $car1,[%sp+$bias+$frame] ! tp[0]=
srlx $car1,32,$car1
add %sp,$bias+$frame+4,$tp
.Lsqr_2nd:
mulx $apj,$mul0,$acc0
mulx $npj,$mul1,$acc1
add $acc0,$car0,$car0
add $tpj,$car1,$car1
ld [$ap+$j],$apj ! ap[j]
and $car0,$mask,$acc0
ld [$np+$j],$npj ! np[j]
srlx $car0,32,$car0
add $acc1,$car1,$car1
ld [$tp+8],$tpj ! tp[j]
add $acc0,$acc0,$acc0
add $j,4,$j ! j++
or $sbit,$acc0,$acc0
srlx $acc0,32,$sbit
and $acc0,$mask,$acc0
cmp $j,$num
add $acc0,$car1,$car1
st $car1,[$tp] ! tp[j-1]
srlx $car1,32,$car1
bl %icc,.Lsqr_2nd
add $tp,4,$tp ! tp++
!.Lsqr_2nd
mulx $apj,$mul0,$acc0
mulx $npj,$mul1,$acc1
add $acc0,$car0,$car0
add $tpj,$car1,$car1
and $car0,$mask,$acc0
srlx $car0,32,$car0
add $acc1,$car1,$car1
add $acc0,$acc0,$acc0
or $sbit,$acc0,$acc0
srlx $acc0,32,$sbit
and $acc0,$mask,$acc0
add $acc0,$car1,$car1
st $car1,[$tp] ! tp[j-1]
srlx $car1,32,$car1
add $car0,$car0,$car0
or $sbit,$car0,$car0
add $car0,$car1,$car1
add $car2,$car1,$car1
st $car1,[$tp+4]
srlx $car1,32,$car2
ld [%sp+$bias+$frame],$tmp1 ! tp[0]
ld [%sp+$bias+$frame+4],$tpj ! tp[1]
ld [$ap+8],$mul0 ! ap[2]
ld [$np],$car1 ! np[0]
ld [$np+4],$npj ! np[1]
mulx $n0,$tmp1,$mul1
and $mul1,$mask,$mul1
mov 8,$i
mulx $mul0,$mul0,$car0
mulx $car1,$mul1,$car1
and $car0,$mask,$acc0
add $tmp1,$car1,$car1
srlx $car0,32,$car0
add %sp,$bias+$frame,$tp
srlx $car1,32,$car1
and $car0,1,$sbit
srlx $car0,1,$car0
mov 4,$j
.Lsqr_outer:
.Lsqr_inner1:
mulx $npj,$mul1,$acc1
add $tpj,$car1,$car1
add $j,4,$j
ld [$tp+8],$tpj
cmp $j,$i
add $acc1,$car1,$car1
ld [$np+$j],$npj
st $car1,[$tp]
srlx $car1,32,$car1
bl %icc,.Lsqr_inner1
add $tp,4,$tp
!.Lsqr_inner1
add $j,4,$j
ld [$ap+$j],$apj ! ap[j]
mulx $npj,$mul1,$acc1
add $tpj,$car1,$car1
ld [$np+$j],$npj ! np[j]
add $acc0,$car1,$car1
ld [$tp+8],$tpj ! tp[j]
add $acc1,$car1,$car1
st $car1,[$tp]
srlx $car1,32,$car1
add $j,4,$j
cmp $j,$num
be,pn %icc,.Lsqr_no_inner2
add $tp,4,$tp
.Lsqr_inner2:
mulx $apj,$mul0,$acc0
mulx $npj,$mul1,$acc1
add $tpj,$car1,$car1
add $acc0,$car0,$car0
ld [$ap+$j],$apj ! ap[j]
and $car0,$mask,$acc0
ld [$np+$j],$npj ! np[j]
srlx $car0,32,$car0
add $acc0,$acc0,$acc0
ld [$tp+8],$tpj ! tp[j]
or $sbit,$acc0,$acc0
add $j,4,$j ! j++
srlx $acc0,32,$sbit
and $acc0,$mask,$acc0
cmp $j,$num
add $acc0,$car1,$car1
add $acc1,$car1,$car1
st $car1,[$tp] ! tp[j-1]
srlx $car1,32,$car1
bl %icc,.Lsqr_inner2
add $tp,4,$tp ! tp++
.Lsqr_no_inner2:
mulx $apj,$mul0,$acc0
mulx $npj,$mul1,$acc1
add $tpj,$car1,$car1
add $acc0,$car0,$car0
and $car0,$mask,$acc0
srlx $car0,32,$car0
add $acc0,$acc0,$acc0
or $sbit,$acc0,$acc0
srlx $acc0,32,$sbit
and $acc0,$mask,$acc0
add $acc0,$car1,$car1
add $acc1,$car1,$car1
st $car1,[$tp] ! tp[j-1]
srlx $car1,32,$car1
add $car0,$car0,$car0
or $sbit,$car0,$car0
add $car0,$car1,$car1
add $car2,$car1,$car1
st $car1,[$tp+4]
srlx $car1,32,$car2
add $i,4,$i ! i++
ld [%sp+$bias+$frame],$tmp1 ! tp[0]
ld [%sp+$bias+$frame+4],$tpj ! tp[1]
ld [$ap+$i],$mul0 ! ap[j]
ld [$np],$car1 ! np[0]
ld [$np+4],$npj ! np[1]
mulx $n0,$tmp1,$mul1
and $mul1,$mask,$mul1
add $i,4,$tmp0
mulx $mul0,$mul0,$car0
mulx $car1,$mul1,$car1
and $car0,$mask,$acc0
add $tmp1,$car1,$car1
srlx $car0,32,$car0
add %sp,$bias+$frame,$tp
srlx $car1,32,$car1
and $car0,1,$sbit
srlx $car0,1,$car0
cmp $tmp0,$num ! i<num-1
bl %icc,.Lsqr_outer
mov 4,$j
.Lsqr_last:
mulx $npj,$mul1,$acc1
add $tpj,$car1,$car1
add $j,4,$j
ld [$tp+8],$tpj
cmp $j,$i
add $acc1,$car1,$car1
ld [$np+$j],$npj
st $car1,[$tp]
srlx $car1,32,$car1
bl %icc,.Lsqr_last
add $tp,4,$tp
!.Lsqr_last
mulx $npj,$mul1,$acc1
add $tpj,$car1,$car1
add $acc0,$car1,$car1
add $acc1,$car1,$car1
st $car1,[$tp]
srlx $car1,32,$car1
add $car0,$car0,$car0 ! recover $car0
or $sbit,$car0,$car0
add $car0,$car1,$car1
add $car2,$car1,$car1
st $car1,[$tp+4]
srlx $car1,32,$car2
ba .Ltail
add $tp,8,$tp
.type $fname,#function
.size $fname,(.-$fname)
.asciz "Montgomery Multipltication for SPARCv9, CRYPTOGAMS by <appro\@openssl.org>"
.align 32
___
$code =~ s/\`([^\`]*)\`/eval($1)/gem;
print $code;
close STDOUT;

882
drivers/builtin_openssl2/crypto/bn/asm/sparcv9a-mont.pl
View File

@ -1,882 +0,0 @@
#!/usr/bin/env perl
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# October 2005
#
# "Teaser" Montgomery multiplication module for UltraSPARC. Why FPU?
# Because unlike integer multiplier, which simply stalls whole CPU,
# FPU is fully pipelined and can effectively emit 48 bit partial
# product every cycle. Why not blended SPARC v9? One can argue that
# making this module dependent on UltraSPARC VIS extension limits its
# binary compatibility. Well yes, it does exclude SPARC64 prior-V(!)
# implementations from compatibility matrix. But the rest, whole Sun
# UltraSPARC family and brand new Fujitsu's SPARC64 V, all support
# VIS extension instructions used in this module. This is considered
# good enough to not care about HAL SPARC64 users [if any] who have
# integer-only pure SPARCv9 module to "fall down" to.
# USI&II cores currently exhibit uniform 2x improvement [over pre-
# bn_mul_mont codebase] for all key lengths and benchmarks. On USIII
# performance improves few percents for shorter keys and worsens few
# percents for longer keys. This is because USIII integer multiplier
# is >3x faster than USI&II one, which is harder to match [but see
# TODO list below]. It should also be noted that SPARC64 V features
# out-of-order execution, which *might* mean that integer multiplier
# is pipelined, which in turn *might* be impossible to match... On
# additional note, SPARC64 V implements FP Multiply-Add instruction,
# which is perfectly usable in this context... In other words, as far
# as Fujitsu SPARC64 V goes, talk to the author:-)
# The implementation implies following "non-natural" limitations on
# input arguments:
# - num may not be less than 4;
# - num has to be even;
# Failure to meet either condition has no fatal effects, simply
# doesn't give any performance gain.
# TODO:
# - modulo-schedule inner loop for better performance (on in-order
# execution core such as UltraSPARC this shall result in further
# noticeable(!) improvement);
# - dedicated squaring procedure[?];
######################################################################
# November 2006
#
# Modulo-scheduled inner loops allow to interleave floating point and
# integer instructions and minimize Read-After-Write penalties. This
# results in *further* 20-50% perfromance improvement [depending on
# key length, more for longer keys] on USI&II cores and 30-80% - on
# USIII&IV.
$fname="bn_mul_mont_fpu";
$bits=32;
for (@ARGV) { $bits=64 if (/\-m64/ || /\-xarch\=v9/); }
if ($bits==64) {
$bias=2047;
$frame=192;
} else {
$bias=0;
$frame=128; # 96 rounded up to largest known cache-line
}
$locals=64;
# In order to provide for 32-/64-bit ABI duality, I keep integers wider
# than 32 bit in %g1-%g4 and %o0-%o5. %l0-%l7 and %i0-%i5 are used
# exclusively for pointers, indexes and other small values...
# int bn_mul_mont(
$rp="%i0"; # BN_ULONG *rp,
$ap="%i1"; # const BN_ULONG *ap,
$bp="%i2"; # const BN_ULONG *bp,
$np="%i3"; # const BN_ULONG *np,
$n0="%i4"; # const BN_ULONG *n0,
$num="%i5"; # int num);
$tp="%l0"; # t[num]
$ap_l="%l1"; # a[num],n[num] are smashed to 32-bit words and saved
$ap_h="%l2"; # to these four vectors as double-precision FP values.
$np_l="%l3"; # This way a bunch of fxtods are eliminated in second
$np_h="%l4"; # loop and L1-cache aliasing is minimized...
$i="%l5";
$j="%l6";
$mask="%l7"; # 16-bit mask, 0xffff
$n0="%g4"; # reassigned(!) to "64-bit" register
$carry="%i4"; # %i4 reused(!) for a carry bit
# FP register naming chart
#
# ..HILO
# dcba
# --------
# LOa
# LOb
# LOc
# LOd
# HIa
# HIb
# HIc
# HId
# ..a
# ..b
$ba="%f0"; $bb="%f2"; $bc="%f4"; $bd="%f6";
$na="%f8"; $nb="%f10"; $nc="%f12"; $nd="%f14";
$alo="%f16"; $alo_="%f17"; $ahi="%f18"; $ahi_="%f19";
$nlo="%f20"; $nlo_="%f21"; $nhi="%f22"; $nhi_="%f23";
$dota="%f24"; $dotb="%f26";
$aloa="%f32"; $alob="%f34"; $aloc="%f36"; $alod="%f38";
$ahia="%f40"; $ahib="%f42"; $ahic="%f44"; $ahid="%f46";
$nloa="%f48"; $nlob="%f50"; $nloc="%f52"; $nlod="%f54";
$nhia="%f56"; $nhib="%f58"; $nhic="%f60"; $nhid="%f62";
$ASI_FL16_P=0xD2; # magic ASI value to engage 16-bit FP load
$code=<<___;
.section ".text",#alloc,#execinstr
.global $fname
.align 32
$fname:
save %sp,-$frame-$locals,%sp
cmp $num,4
bl,a,pn %icc,.Lret
clr %i0
andcc $num,1,%g0 ! $num has to be even...
bnz,a,pn %icc,.Lret
clr %i0 ! signal "unsupported input value"
srl $num,1,$num
sethi %hi(0xffff),$mask
ld [%i4+0],$n0 ! $n0 reassigned, remember?
or $mask,%lo(0xffff),$mask
ld [%i4+4],%o0
sllx %o0,32,%o0
or %o0,$n0,$n0 ! $n0=n0[1].n0[0]
sll $num,3,$num ! num*=8
add %sp,$bias,%o0 ! real top of stack
sll $num,2,%o1
add %o1,$num,%o1 ! %o1=num*5
sub %o0,%o1,%o0
and %o0,-2048,%o0 ! optimize TLB utilization
sub %o0,$bias,%sp ! alloca(5*num*8)
rd %asi,%o7 ! save %asi
add %sp,$bias+$frame+$locals,$tp
add $tp,$num,$ap_l
add $ap_l,$num,$ap_l ! [an]p_[lh] point at the vectors' ends !
add $ap_l,$num,$ap_h
add $ap_h,$num,$np_l
add $np_l,$num,$np_h
wr %g0,$ASI_FL16_P,%asi ! setup %asi for 16-bit FP loads
add $rp,$num,$rp ! readjust input pointers to point
add $ap,$num,$ap ! at the ends too...
add $bp,$num,$bp
add $np,$num,$np
stx %o7,[%sp+$bias+$frame+48] ! save %asi
sub %g0,$num,$i ! i=-num
sub %g0,$num,$j ! j=-num
add $ap,$j,%o3
add $bp,$i,%o4
ld [%o3+4],%g1 ! bp[0]
ld [%o3+0],%o0
ld [%o4+4],%g5 ! ap[0]
sllx %g1,32,%g1
ld [%o4+0],%o1
sllx %g5,32,%g5
or %g1,%o0,%o0
or %g5,%o1,%o1
add $np,$j,%o5
mulx %o1,%o0,%o0 ! ap[0]*bp[0]
mulx $n0,%o0,%o0 ! ap[0]*bp[0]*n0
stx %o0,[%sp+$bias+$frame+0]
ld [%o3+0],$alo_ ! load a[j] as pair of 32-bit words
fzeros $alo
ld [%o3+4],$ahi_
fzeros $ahi
ld [%o5+0],$nlo_ ! load n[j] as pair of 32-bit words
fzeros $nlo
ld [%o5+4],$nhi_
fzeros $nhi
! transfer b[i] to FPU as 4x16-bit values
ldda [%o4+2]%asi,$ba
fxtod $alo,$alo
ldda [%o4+0]%asi,$bb
fxtod $ahi,$ahi
ldda [%o4+6]%asi,$bc
fxtod $nlo,$nlo
ldda [%o4+4]%asi,$bd
fxtod $nhi,$nhi
! transfer ap[0]*b[0]*n0 to FPU as 4x16-bit values
ldda [%sp+$bias+$frame+6]%asi,$na
fxtod $ba,$ba
ldda [%sp+$bias+$frame+4]%asi,$nb
fxtod $bb,$bb
ldda [%sp+$bias+$frame+2]%asi,$nc
fxtod $bc,$bc
ldda [%sp+$bias+$frame+0]%asi,$nd
fxtod $bd,$bd
std $alo,[$ap_l+$j] ! save smashed ap[j] in double format
fxtod $na,$na
std $ahi,[$ap_h+$j]
fxtod $nb,$nb
std $nlo,[$np_l+$j] ! save smashed np[j] in double format
fxtod $nc,$nc
std $nhi,[$np_h+$j]
fxtod $nd,$nd
fmuld $alo,$ba,$aloa
fmuld $nlo,$na,$nloa
fmuld $alo,$bb,$alob
fmuld $nlo,$nb,$nlob
fmuld $alo,$bc,$aloc
faddd $aloa,$nloa,$nloa
fmuld $nlo,$nc,$nloc
fmuld $alo,$bd,$alod
faddd $alob,$nlob,$nlob
fmuld $nlo,$nd,$nlod
fmuld $ahi,$ba,$ahia
faddd $aloc,$nloc,$nloc
fmuld $nhi,$na,$nhia
fmuld $ahi,$bb,$ahib
faddd $alod,$nlod,$nlod
fmuld $nhi,$nb,$nhib
fmuld $ahi,$bc,$ahic
faddd $ahia,$nhia,$nhia
fmuld $nhi,$nc,$nhic
fmuld $ahi,$bd,$ahid
faddd $ahib,$nhib,$nhib
fmuld $nhi,$nd,$nhid
faddd $ahic,$nhic,$dota ! $nhic
faddd $ahid,$nhid,$dotb ! $nhid
faddd $nloc,$nhia,$nloc
faddd $nlod,$nhib,$nlod
fdtox $nloa,$nloa
fdtox $nlob,$nlob
fdtox $nloc,$nloc
fdtox $nlod,$nlod
std $nloa,[%sp+$bias+$frame+0]
add $j,8,$j
std $nlob,[%sp+$bias+$frame+8]
add $ap,$j,%o4
std $nloc,[%sp+$bias+$frame+16]
add $np,$j,%o5
std $nlod,[%sp+$bias+$frame+24]
ld [%o4+0],$alo_ ! load a[j] as pair of 32-bit words
fzeros $alo
ld [%o4+4],$ahi_
fzeros $ahi
ld [%o5+0],$nlo_ ! load n[j] as pair of 32-bit words
fzeros $nlo
ld [%o5+4],$nhi_
fzeros $nhi
fxtod $alo,$alo
fxtod $ahi,$ahi
fxtod $nlo,$nlo
fxtod $nhi,$nhi
ldx [%sp+$bias+$frame+0],%o0
fmuld $alo,$ba,$aloa
ldx [%sp+$bias+$frame+8],%o1
fmuld $nlo,$na,$nloa
ldx [%sp+$bias+$frame+16],%o2
fmuld $alo,$bb,$alob
ldx [%sp+$bias+$frame+24],%o3
fmuld $nlo,$nb,$nlob
srlx %o0,16,%o7
std $alo,[$ap_l+$j] ! save smashed ap[j] in double format
fmuld $alo,$bc,$aloc
add %o7,%o1,%o1
std $ahi,[$ap_h+$j]
faddd $aloa,$nloa,$nloa
fmuld $nlo,$nc,$nloc
srlx %o1,16,%o7
std $nlo,[$np_l+$j] ! save smashed np[j] in double format
fmuld $alo,$bd,$alod
add %o7,%o2,%o2
std $nhi,[$np_h+$j]
faddd $alob,$nlob,$nlob
fmuld $nlo,$nd,$nlod
srlx %o2,16,%o7
fmuld $ahi,$ba,$ahia
add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
faddd $aloc,$nloc,$nloc
fmuld $nhi,$na,$nhia
!and %o0,$mask,%o0
!and %o1,$mask,%o1
!and %o2,$mask,%o2
!sllx %o1,16,%o1
!sllx %o2,32,%o2
!sllx %o3,48,%o7
!or %o1,%o0,%o0
!or %o2,%o0,%o0
!or %o7,%o0,%o0 ! 64-bit result
srlx %o3,16,%g1 ! 34-bit carry
fmuld $ahi,$bb,$ahib
faddd $alod,$nlod,$nlod
fmuld $nhi,$nb,$nhib
fmuld $ahi,$bc,$ahic
faddd $ahia,$nhia,$nhia
fmuld $nhi,$nc,$nhic
fmuld $ahi,$bd,$ahid
faddd $ahib,$nhib,$nhib
fmuld $nhi,$nd,$nhid
faddd $dota,$nloa,$nloa
faddd $dotb,$nlob,$nlob
faddd $ahic,$nhic,$dota ! $nhic
faddd $ahid,$nhid,$dotb ! $nhid
faddd $nloc,$nhia,$nloc
faddd $nlod,$nhib,$nlod
fdtox $nloa,$nloa
fdtox $nlob,$nlob
fdtox $nloc,$nloc
fdtox $nlod,$nlod
std $nloa,[%sp+$bias+$frame+0]
std $nlob,[%sp+$bias+$frame+8]
addcc $j,8,$j
std $nloc,[%sp+$bias+$frame+16]
bz,pn %icc,.L1stskip
std $nlod,[%sp+$bias+$frame+24]
.align 32 ! incidentally already aligned !
.L1st:
add $ap,$j,%o4
add $np,$j,%o5
ld [%o4+0],$alo_ ! load a[j] as pair of 32-bit words
fzeros $alo
ld [%o4+4],$ahi_
fzeros $ahi
ld [%o5+0],$nlo_ ! load n[j] as pair of 32-bit words
fzeros $nlo
ld [%o5+4],$nhi_
fzeros $nhi
fxtod $alo,$alo
fxtod $ahi,$ahi
fxtod $nlo,$nlo
fxtod $nhi,$nhi
ldx [%sp+$bias+$frame+0],%o0
fmuld $alo,$ba,$aloa
ldx [%sp+$bias+$frame+8],%o1
fmuld $nlo,$na,$nloa
ldx [%sp+$bias+$frame+16],%o2
fmuld $alo,$bb,$alob
ldx [%sp+$bias+$frame+24],%o3
fmuld $nlo,$nb,$nlob
srlx %o0,16,%o7
std $alo,[$ap_l+$j] ! save smashed ap[j] in double format
fmuld $alo,$bc,$aloc
add %o7,%o1,%o1
std $ahi,[$ap_h+$j]
faddd $aloa,$nloa,$nloa
fmuld $nlo,$nc,$nloc
srlx %o1,16,%o7
std $nlo,[$np_l+$j] ! save smashed np[j] in double format
fmuld $alo,$bd,$alod
add %o7,%o2,%o2
std $nhi,[$np_h+$j]
faddd $alob,$nlob,$nlob
fmuld $nlo,$nd,$nlod
srlx %o2,16,%o7
fmuld $ahi,$ba,$ahia
add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
and %o0,$mask,%o0
faddd $aloc,$nloc,$nloc
fmuld $nhi,$na,$nhia
and %o1,$mask,%o1
and %o2,$mask,%o2
fmuld $ahi,$bb,$ahib
sllx %o1,16,%o1
faddd $alod,$nlod,$nlod
fmuld $nhi,$nb,$nhib
sllx %o2,32,%o2
fmuld $ahi,$bc,$ahic
sllx %o3,48,%o7
or %o1,%o0,%o0
faddd $ahia,$nhia,$nhia
fmuld $nhi,$nc,$nhic
or %o2,%o0,%o0
fmuld $ahi,$bd,$ahid
or %o7,%o0,%o0 ! 64-bit result
faddd $ahib,$nhib,$nhib
fmuld $nhi,$nd,$nhid
addcc %g1,%o0,%o0
faddd $dota,$nloa,$nloa
srlx %o3,16,%g1 ! 34-bit carry
faddd $dotb,$nlob,$nlob
bcs,a %xcc,.+8
add %g1,1,%g1
stx %o0,[$tp] ! tp[j-1]=
faddd $ahic,$nhic,$dota ! $nhic
faddd $ahid,$nhid,$dotb ! $nhid
faddd $nloc,$nhia,$nloc
faddd $nlod,$nhib,$nlod
fdtox $nloa,$nloa
fdtox $nlob,$nlob
fdtox $nloc,$nloc
fdtox $nlod,$nlod
std $nloa,[%sp+$bias+$frame+0]
std $nlob,[%sp+$bias+$frame+8]
std $nloc,[%sp+$bias+$frame+16]
std $nlod,[%sp+$bias+$frame+24]
addcc $j,8,$j
bnz,pt %icc,.L1st
add $tp,8,$tp
.L1stskip:
fdtox $dota,$dota
fdtox $dotb,$dotb
ldx [%sp+$bias+$frame+0],%o0
ldx [%sp+$bias+$frame+8],%o1
ldx [%sp+$bias+$frame+16],%o2
ldx [%sp+$bias+$frame+24],%o3
srlx %o0,16,%o7
std $dota,[%sp+$bias+$frame+32]
add %o7,%o1,%o1
std $dotb,[%sp+$bias+$frame+40]
srlx %o1,16,%o7
add %o7,%o2,%o2
srlx %o2,16,%o7
add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
and %o0,$mask,%o0
and %o1,$mask,%o1
and %o2,$mask,%o2
sllx %o1,16,%o1
sllx %o2,32,%o2
sllx %o3,48,%o7
or %o1,%o0,%o0
or %o2,%o0,%o0
or %o7,%o0,%o0 ! 64-bit result
ldx [%sp+$bias+$frame+32],%o4
addcc %g1,%o0,%o0
ldx [%sp+$bias+$frame+40],%o5
srlx %o3,16,%g1 ! 34-bit carry
bcs,a %xcc,.+8
add %g1,1,%g1
stx %o0,[$tp] ! tp[j-1]=
add $tp,8,$tp
srlx %o4,16,%o7
add %o7,%o5,%o5
and %o4,$mask,%o4
sllx %o5,16,%o7
or %o7,%o4,%o4
addcc %g1,%o4,%o4
srlx %o5,48,%g1
bcs,a %xcc,.+8
add %g1,1,%g1
mov %g1,$carry
stx %o4,[$tp] ! tp[num-1]=
ba .Louter
add $i,8,$i
.align 32
.Louter:
sub %g0,$num,$j ! j=-num
add %sp,$bias+$frame+$locals,$tp
add $ap,$j,%o3
add $bp,$i,%o4
ld [%o3+4],%g1 ! bp[i]
ld [%o3+0],%o0
ld [%o4+4],%g5 ! ap[0]
sllx %g1,32,%g1
ld [%o4+0],%o1
sllx %g5,32,%g5
or %g1,%o0,%o0
or %g5,%o1,%o1
ldx [$tp],%o2 ! tp[0]
mulx %o1,%o0,%o0
addcc %o2,%o0,%o0
mulx $n0,%o0,%o0 ! (ap[0]*bp[i]+t[0])*n0
stx %o0,[%sp+$bias+$frame+0]
! transfer b[i] to FPU as 4x16-bit values
ldda [%o4+2]%asi,$ba
ldda [%o4+0]%asi,$bb
ldda [%o4+6]%asi,$bc
ldda [%o4+4]%asi,$bd
! transfer (ap[0]*b[i]+t[0])*n0 to FPU as 4x16-bit values
ldda [%sp+$bias+$frame+6]%asi,$na
fxtod $ba,$ba
ldda [%sp+$bias+$frame+4]%asi,$nb
fxtod $bb,$bb
ldda [%sp+$bias+$frame+2]%asi,$nc
fxtod $bc,$bc
ldda [%sp+$bias+$frame+0]%asi,$nd
fxtod $bd,$bd
ldd [$ap_l+$j],$alo ! load a[j] in double format
fxtod $na,$na
ldd [$ap_h+$j],$ahi
fxtod $nb,$nb
ldd [$np_l+$j],$nlo ! load n[j] in double format
fxtod $nc,$nc
ldd [$np_h+$j],$nhi
fxtod $nd,$nd
fmuld $alo,$ba,$aloa
fmuld $nlo,$na,$nloa
fmuld $alo,$bb,$alob
fmuld $nlo,$nb,$nlob
fmuld $alo,$bc,$aloc
faddd $aloa,$nloa,$nloa
fmuld $nlo,$nc,$nloc
fmuld $alo,$bd,$alod
faddd $alob,$nlob,$nlob
fmuld $nlo,$nd,$nlod
fmuld $ahi,$ba,$ahia
faddd $aloc,$nloc,$nloc
fmuld $nhi,$na,$nhia
fmuld $ahi,$bb,$ahib
faddd $alod,$nlod,$nlod
fmuld $nhi,$nb,$nhib
fmuld $ahi,$bc,$ahic
faddd $ahia,$nhia,$nhia
fmuld $nhi,$nc,$nhic
fmuld $ahi,$bd,$ahid
faddd $ahib,$nhib,$nhib
fmuld $nhi,$nd,$nhid
faddd $ahic,$nhic,$dota ! $nhic
faddd $ahid,$nhid,$dotb ! $nhid
faddd $nloc,$nhia,$nloc
faddd $nlod,$nhib,$nlod
fdtox $nloa,$nloa
fdtox $nlob,$nlob
fdtox $nloc,$nloc
fdtox $nlod,$nlod
std $nloa,[%sp+$bias+$frame+0]
std $nlob,[%sp+$bias+$frame+8]
std $nloc,[%sp+$bias+$frame+16]
add $j,8,$j
std $nlod,[%sp+$bias+$frame+24]
ldd [$ap_l+$j],$alo ! load a[j] in double format
ldd [$ap_h+$j],$ahi
ldd [$np_l+$j],$nlo ! load n[j] in double format
ldd [$np_h+$j],$nhi
fmuld $alo,$ba,$aloa
fmuld $nlo,$na,$nloa
fmuld $alo,$bb,$alob
fmuld $nlo,$nb,$nlob
fmuld $alo,$bc,$aloc
ldx [%sp+$bias+$frame+0],%o0
faddd $aloa,$nloa,$nloa
fmuld $nlo,$nc,$nloc
ldx [%sp+$bias+$frame+8],%o1
fmuld $alo,$bd,$alod
ldx [%sp+$bias+$frame+16],%o2
faddd $alob,$nlob,$nlob
fmuld $nlo,$nd,$nlod
ldx [%sp+$bias+$frame+24],%o3
fmuld $ahi,$ba,$ahia
srlx %o0,16,%o7
faddd $aloc,$nloc,$nloc
fmuld $nhi,$na,$nhia
add %o7,%o1,%o1
fmuld $ahi,$bb,$ahib
srlx %o1,16,%o7
faddd $alod,$nlod,$nlod
fmuld $nhi,$nb,$nhib
add %o7,%o2,%o2
fmuld $ahi,$bc,$ahic
srlx %o2,16,%o7
faddd $ahia,$nhia,$nhia
fmuld $nhi,$nc,$nhic
add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
! why?
and %o0,$mask,%o0
fmuld $ahi,$bd,$ahid
and %o1,$mask,%o1
and %o2,$mask,%o2
faddd $ahib,$nhib,$nhib
fmuld $nhi,$nd,$nhid
sllx %o1,16,%o1
faddd $dota,$nloa,$nloa
sllx %o2,32,%o2
faddd $dotb,$nlob,$nlob
sllx %o3,48,%o7
or %o1,%o0,%o0
faddd $ahic,$nhic,$dota ! $nhic
or %o2,%o0,%o0
faddd $ahid,$nhid,$dotb ! $nhid
or %o7,%o0,%o0 ! 64-bit result
ldx [$tp],%o7
faddd $nloc,$nhia,$nloc
addcc %o7,%o0,%o0
! end-of-why?
faddd $nlod,$nhib,$nlod
srlx %o3,16,%g1 ! 34-bit carry
fdtox $nloa,$nloa
bcs,a %xcc,.+8
add %g1,1,%g1
fdtox $nlob,$nlob
fdtox $nloc,$nloc
fdtox $nlod,$nlod
std $nloa,[%sp+$bias+$frame+0]
std $nlob,[%sp+$bias+$frame+8]
addcc $j,8,$j
std $nloc,[%sp+$bias+$frame+16]
bz,pn %icc,.Linnerskip
std $nlod,[%sp+$bias+$frame+24]
ba .Linner
nop
.align 32
.Linner:
ldd [$ap_l+$j],$alo ! load a[j] in double format
ldd [$ap_h+$j],$ahi
ldd [$np_l+$j],$nlo ! load n[j] in double format
ldd [$np_h+$j],$nhi
fmuld $alo,$ba,$aloa
fmuld $nlo,$na,$nloa
fmuld $alo,$bb,$alob
fmuld $nlo,$nb,$nlob
fmuld $alo,$bc,$aloc
ldx [%sp+$bias+$frame+0],%o0
faddd $aloa,$nloa,$nloa
fmuld $nlo,$nc,$nloc
ldx [%sp+$bias+$frame+8],%o1
fmuld $alo,$bd,$alod
ldx [%sp+$bias+$frame+16],%o2
faddd $alob,$nlob,$nlob
fmuld $nlo,$nd,$nlod
ldx [%sp+$bias+$frame+24],%o3
fmuld $ahi,$ba,$ahia
srlx %o0,16,%o7
faddd $aloc,$nloc,$nloc
fmuld $nhi,$na,$nhia
add %o7,%o1,%o1
fmuld $ahi,$bb,$ahib
srlx %o1,16,%o7
faddd $alod,$nlod,$nlod
fmuld $nhi,$nb,$nhib
add %o7,%o2,%o2
fmuld $ahi,$bc,$ahic
srlx %o2,16,%o7
faddd $ahia,$nhia,$nhia
fmuld $nhi,$nc,$nhic
add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
and %o0,$mask,%o0
fmuld $ahi,$bd,$ahid
and %o1,$mask,%o1
and %o2,$mask,%o2
faddd $ahib,$nhib,$nhib
fmuld $nhi,$nd,$nhid
sllx %o1,16,%o1
faddd $dota,$nloa,$nloa
sllx %o2,32,%o2
faddd $dotb,$nlob,$nlob
sllx %o3,48,%o7
or %o1,%o0,%o0
faddd $ahic,$nhic,$dota ! $nhic
or %o2,%o0,%o0
faddd $ahid,$nhid,$dotb ! $nhid
or %o7,%o0,%o0 ! 64-bit result
faddd $nloc,$nhia,$nloc
addcc %g1,%o0,%o0
ldx [$tp+8],%o7 ! tp[j]
faddd $nlod,$nhib,$nlod
srlx %o3,16,%g1 ! 34-bit carry
fdtox $nloa,$nloa
bcs,a %xcc,.+8
add %g1,1,%g1
fdtox $nlob,$nlob
addcc %o7,%o0,%o0
fdtox $nloc,$nloc
bcs,a %xcc,.+8
add %g1,1,%g1
stx %o0,[$tp] ! tp[j-1]
fdtox $nlod,$nlod
std $nloa,[%sp+$bias+$frame+0]
std $nlob,[%sp+$bias+$frame+8]
std $nloc,[%sp+$bias+$frame+16]
addcc $j,8,$j
std $nlod,[%sp+$bias+$frame+24]
bnz,pt %icc,.Linner
add $tp,8,$tp
.Linnerskip:
fdtox $dota,$dota
fdtox $dotb,$dotb
ldx [%sp+$bias+$frame+0],%o0
ldx [%sp+$bias+$frame+8],%o1
ldx [%sp+$bias+$frame+16],%o2
ldx [%sp+$bias+$frame+24],%o3
srlx %o0,16,%o7
std $dota,[%sp+$bias+$frame+32]
add %o7,%o1,%o1
std $dotb,[%sp+$bias+$frame+40]
srlx %o1,16,%o7
add %o7,%o2,%o2
srlx %o2,16,%o7
add %o7,%o3,%o3 ! %o3.%o2[0..15].%o1[0..15].%o0[0..15]
and %o0,$mask,%o0
and %o1,$mask,%o1
and %o2,$mask,%o2
sllx %o1,16,%o1
sllx %o2,32,%o2
sllx %o3,48,%o7
or %o1,%o0,%o0
or %o2,%o0,%o0
ldx [%sp+$bias+$frame+32],%o4
or %o7,%o0,%o0 ! 64-bit result
ldx [%sp+$bias+$frame+40],%o5
addcc %g1,%o0,%o0
ldx [$tp+8],%o7 ! tp[j]
srlx %o3,16,%g1 ! 34-bit carry
bcs,a %xcc,.+8
add %g1,1,%g1
addcc %o7,%o0,%o0
bcs,a %xcc,.+8
add %g1,1,%g1
stx %o0,[$tp] ! tp[j-1]
add $tp,8,$tp
srlx %o4,16,%o7
add %o7,%o5,%o5
and %o4,$mask,%o4
sllx %o5,16,%o7
or %o7,%o4,%o4
addcc %g1,%o4,%o4
srlx %o5,48,%g1
bcs,a %xcc,.+8
add %g1,1,%g1
addcc $carry,%o4,%o4
stx %o4,[$tp] ! tp[num-1]
mov %g1,$carry
bcs,a %xcc,.+8
add $carry,1,$carry
addcc $i,8,$i
bnz %icc,.Louter
nop
add $tp,8,$tp ! adjust tp to point at the end
orn %g0,%g0,%g4
sub %g0,$num,%o7 ! n=-num
ba .Lsub
subcc %g0,%g0,%g0 ! clear %icc.c
.align 32
.Lsub:
ldx [$tp+%o7],%o0
add $np,%o7,%g1
ld [%g1+0],%o2
ld [%g1+4],%o3
srlx %o0,32,%o1
subccc %o0,%o2,%o2
add $rp,%o7,%g1
subccc %o1,%o3,%o3
st %o2,[%g1+0]
add %o7,8,%o7
brnz,pt %o7,.Lsub
st %o3,[%g1+4]
subc $carry,0,%g4
sub %g0,$num,%o7 ! n=-num
ba .Lcopy
nop
.align 32
.Lcopy:
ldx [$tp+%o7],%o0
add $rp,%o7,%g1
ld [%g1+0],%o2
ld [%g1+4],%o3
stx %g0,[$tp+%o7]
and %o0,%g4,%o0
srlx %o0,32,%o1
andn %o2,%g4,%o2
andn %o3,%g4,%o3
or %o2,%o0,%o0
or %o3,%o1,%o1
st %o0,[%g1+0]
add %o7,8,%o7
brnz,pt %o7,.Lcopy
st %o1,[%g1+4]
sub %g0,$num,%o7 ! n=-num
.Lzap:
stx %g0,[$ap_l+%o7]
stx %g0,[$ap_h+%o7]
stx %g0,[$np_l+%o7]
stx %g0,[$np_h+%o7]
add %o7,8,%o7
brnz,pt %o7,.Lzap
nop
ldx [%sp+$bias+$frame+48],%o7
wr %g0,%o7,%asi ! restore %asi
mov 1,%i0
.Lret:
ret
restore
.type $fname,#function
.size $fname,(.-$fname)
.asciz "Montgomery Multipltication for UltraSPARC, CRYPTOGAMS by <appro\@openssl.org>"
.align 32
___
$code =~ s/\`([^\`]*)\`/eval($1)/gem;
# Below substitution makes it possible to compile without demanding
# VIS extentions on command line, e.g. -xarch=v9 vs. -xarch=v9a. I
# dare to do this, because VIS capability is detected at run-time now
# and this routine is not called on CPU not capable to execute it. Do
# note that fzeros is not the only VIS dependency! Another dependency
# is implicit and is just _a_ numerical value loaded to %asi register,
# which assembler can't recognize as VIS specific...
$code =~ s/fzeros\s+%f([0-9]+)/
sprintf(".word\t0x%x\t! fzeros %%f%d",0x81b00c20|($1<<25),$1)
/gem;
print $code;
# flush
close STDOUT;

242
drivers/builtin_openssl2/crypto/bn/asm/via-mont.pl
View File

@ -1,242 +0,0 @@
#!/usr/bin/env perl
#
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
#
# Wrapper around 'rep montmul', VIA-specific instruction accessing
# PadLock Montgomery Multiplier. The wrapper is designed as drop-in
# replacement for OpenSSL bn_mul_mont [first implemented in 0.9.9].
#
# Below are interleaved outputs from 'openssl speed rsa dsa' for 4
# different software configurations on 1.5GHz VIA Esther processor.
# Lines marked with "software integer" denote performance of hand-
# coded integer-only assembler found in OpenSSL 0.9.7. "Software SSE2"
# refers to hand-coded SSE2 Montgomery multiplication procedure found
# OpenSSL 0.9.9. "Hardware VIA SDK" refers to padlock_pmm routine from
# Padlock SDK 2.0.1 available for download from VIA, which naturally
# utilizes the magic 'repz montmul' instruction. And finally "hardware
# this" refers to *this* implementation which also uses 'repz montmul'
#
# sign verify sign/s verify/s
# rsa 512 bits 0.001720s 0.000140s 581.4 7149.7 software integer
# rsa 512 bits 0.000690s 0.000086s 1450.3 11606.0 software SSE2
# rsa 512 bits 0.006136s 0.000201s 163.0 4974.5 hardware VIA SDK
# rsa 512 bits 0.000712s 0.000050s 1404.9 19858.5 hardware this
#
# rsa 1024 bits 0.008518s 0.000413s 117.4 2420.8 software integer
# rsa 1024 bits 0.004275s 0.000277s 233.9 3609.7 software SSE2
# rsa 1024 bits 0.012136s 0.000260s 82.4 3844.5 hardware VIA SDK
# rsa 1024 bits 0.002522s 0.000116s 396.5 8650.9 hardware this
#
# rsa 2048 bits 0.050101s 0.001371s 20.0 729.6 software integer
# rsa 2048 bits 0.030273s 0.001008s 33.0 991.9 software SSE2
# rsa 2048 bits 0.030833s 0.000976s 32.4 1025.1 hardware VIA SDK
# rsa 2048 bits 0.011879s 0.000342s 84.2 2921.7 hardware this
#
# rsa 4096 bits 0.327097s 0.004859s 3.1 205.8 software integer
# rsa 4096 bits 0.229318s 0.003859s 4.4 259.2 software SSE2
# rsa 4096 bits 0.233953s 0.003274s 4.3 305.4 hardware VIA SDK
# rsa 4096 bits 0.070493s 0.001166s 14.2 857.6 hardware this
#
# dsa 512 bits 0.001342s 0.001651s 745.2 605.7 software integer
# dsa 512 bits 0.000844s 0.000987s 1185.3 1013.1 software SSE2
# dsa 512 bits 0.001902s 0.002247s 525.6 444.9 hardware VIA SDK
# dsa 512 bits 0.000458s 0.000524s 2182.2 1909.1 hardware this
#
# dsa 1024 bits 0.003964s 0.004926s 252.3 203.0 software integer
# dsa 1024 bits 0.002686s 0.003166s 372.3 315.8 software SSE2
# dsa 1024 bits 0.002397s 0.002823s 417.1 354.3 hardware VIA SDK
# dsa 1024 bits 0.000978s 0.001170s 1022.2 855.0 hardware this
#
# dsa 2048 bits 0.013280s 0.016518s 75.3 60.5 software integer
# dsa 2048 bits 0.009911s 0.011522s 100.9 86.8 software SSE2
# dsa 2048 bits 0.009542s 0.011763s 104.8 85.0 hardware VIA SDK
# dsa 2048 bits 0.002884s 0.003352s 346.8 298.3 hardware this
#
# To give you some other reference point here is output for 2.4GHz P4
# running hand-coded SSE2 bn_mul_mont found in 0.9.9, i.e. "software
# SSE2" in above terms.
#
# rsa 512 bits 0.000407s 0.000047s 2454.2 21137.0
# rsa 1024 bits 0.002426s 0.000141s 412.1 7100.0
# rsa 2048 bits 0.015046s 0.000491s 66.5 2034.9
# rsa 4096 bits 0.109770s 0.002379s 9.1 420.3
# dsa 512 bits 0.000438s 0.000525s 2281.1 1904.1
# dsa 1024 bits 0.001346s 0.001595s 742.7 627.0
# dsa 2048 bits 0.004745s 0.005582s 210.7 179.1
#
# Conclusions:
# - VIA SDK leaves a *lot* of room for improvement (which this
# implementation successfully fills:-);
# - 'rep montmul' gives up to >3x performance improvement depending on
# key length;
# - in terms of absolute performance it delivers approximately as much
# as modern out-of-order 32-bit cores [again, for longer keys].
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
push(@INC,"${dir}","${dir}../../perlasm");
require "x86asm.pl";
&asm_init($ARGV[0],"via-mont.pl");
# int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, const BN_ULONG *np,const BN_ULONG *n0, int num);
$func="bn_mul_mont_padlock";
$pad=16*1; # amount of reserved bytes on top of every vector
# stack layout
$mZeroPrime=&DWP(0,"esp"); # these are specified by VIA
$A=&DWP(4,"esp");
$B=&DWP(8,"esp");
$T=&DWP(12,"esp");
$M=&DWP(16,"esp");
$scratch=&DWP(20,"esp");
$rp=&DWP(24,"esp"); # these are mine
$sp=&DWP(28,"esp");
# &DWP(32,"esp") # 32 byte scratch area
# &DWP(64+(4*$num+$pad)*0,"esp") # padded tp[num]
# &DWP(64+(4*$num+$pad)*1,"esp") # padded copy of ap[num]
# &DWP(64+(4*$num+$pad)*2,"esp") # padded copy of bp[num]
# &DWP(64+(4*$num+$pad)*3,"esp") # padded copy of np[num]
# Note that SDK suggests to unconditionally allocate 2K per vector. This
# has quite an impact on performance. It naturally depends on key length,
# but to give an example 1024 bit private RSA key operations suffer >30%
# penalty. I allocate only as much as actually required...
&function_begin($func);
&xor ("eax","eax");
&mov ("ecx",&wparam(5)); # num
# meet VIA's limitations for num [note that the specification
# expresses them in bits, while we work with amount of 32-bit words]
&test ("ecx",3);
&jnz (&label("leave")); # num % 4 != 0
&cmp ("ecx",8);
&jb (&label("leave")); # num < 8
&cmp ("ecx",1024);
&ja (&label("leave")); # num > 1024
&pushf ();
&cld ();
&mov ("edi",&wparam(0)); # rp
&mov ("eax",&wparam(1)); # ap
&mov ("ebx",&wparam(2)); # bp
&mov ("edx",&wparam(3)); # np
&mov ("esi",&wparam(4)); # n0
&mov ("esi",&DWP(0,"esi")); # *n0
&lea ("ecx",&DWP($pad,"","ecx",4)); # ecx becomes vector size in bytes
&lea ("ebp",&DWP(64,"","ecx",4)); # allocate 4 vectors + 64 bytes
&neg ("ebp");
&add ("ebp","esp");
&and ("ebp",-64); # align to cache-line
&xchg ("ebp","esp"); # alloca
&mov ($rp,"edi"); # save rp
&mov ($sp,"ebp"); # save esp
&mov ($mZeroPrime,"esi");
&lea ("esi",&DWP(64,"esp")); # tp
&mov ($T,"esi");
&lea ("edi",&DWP(32,"esp")); # scratch area
&mov ($scratch,"edi");
&mov ("esi","eax");
&lea ("ebp",&DWP(-$pad,"ecx"));
&shr ("ebp",2); # restore original num value in ebp
&xor ("eax","eax");
&mov ("ecx","ebp");
&lea ("ecx",&DWP((32+$pad)/4,"ecx"));# padded tp + scratch
&data_byte(0xf3,0xab); # rep stosl, bzero
&mov ("ecx","ebp");
&lea ("edi",&DWP(64+$pad,"esp","ecx",4));# pointer to ap copy
&mov ($A,"edi");
&data_byte(0xf3,0xa5); # rep movsl, memcpy
&mov ("ecx",$pad/4);
&data_byte(0xf3,0xab); # rep stosl, bzero pad
# edi points at the end of padded ap copy...
&mov ("ecx","ebp");
&mov ("esi","ebx");
&mov ($B,"edi");
&data_byte(0xf3,0xa5); # rep movsl, memcpy
&mov ("ecx",$pad/4);
&data_byte(0xf3,0xab); # rep stosl, bzero pad
# edi points at the end of padded bp copy...
&mov ("ecx","ebp");
&mov ("esi","edx");
&mov ($M,"edi");
&data_byte(0xf3,0xa5); # rep movsl, memcpy
&mov ("ecx",$pad/4);
&data_byte(0xf3,0xab); # rep stosl, bzero pad
# edi points at the end of padded np copy...
# let magic happen...
&mov ("ecx","ebp");
&mov ("esi","esp");
&shl ("ecx",5); # convert word counter to bit counter
&align (4);
&data_byte(0xf3,0x0f,0xa6,0xc0);# rep montmul
&mov ("ecx","ebp");
&lea ("esi",&DWP(64,"esp")); # tp
# edi still points at the end of padded np copy...
&neg ("ebp");
&lea ("ebp",&DWP(-$pad,"edi","ebp",4)); # so just "rewind"
&mov ("edi",$rp); # restore rp
&xor ("edx","edx"); # i=0 and clear CF
&set_label("sub",8);
&mov ("eax",&DWP(0,"esi","edx",4));
&sbb ("eax",&DWP(0,"ebp","edx",4));
&mov (&DWP(0,"edi","edx",4),"eax"); # rp[i]=tp[i]-np[i]
&lea ("edx",&DWP(1,"edx")); # i++
&loop (&label("sub")); # doesn't affect CF!
&mov ("eax",&DWP(0,"esi","edx",4)); # upmost overflow bit
&sbb ("eax",0);
&and ("esi","eax");
&not ("eax");
&mov ("ebp","edi");
&and ("ebp","eax");
&or ("esi","ebp"); # tp=carry?tp:rp
&mov ("ecx","edx"); # num
&xor ("edx","edx"); # i=0
&set_label("copy",8);
&mov ("eax",&DWP(0,"esi","edx",4));
&mov (&DWP(64,"esp","edx",4),"ecx"); # zap tp
&mov (&DWP(0,"edi","edx",4),"eax");
&lea ("edx",&DWP(1,"edx")); # i++
&loop (&label("copy"));
&mov ("ebp",$sp);
&xor ("eax","eax");
&mov ("ecx",64/4);
&mov ("edi","esp"); # zap frame including scratch area
&data_byte(0xf3,0xab); # rep stosl, bzero
# zap copies of ap, bp and np
&lea ("edi",&DWP(64+$pad,"esp","edx",4));# pointer to ap
&lea ("ecx",&DWP(3*$pad/4,"edx","edx",2));
&data_byte(0xf3,0xab); # rep stosl, bzero
&mov ("esp","ebp");
&inc ("eax"); # signal "done"
&popf ();
&set_label("leave");
&function_end($func);
&asciz("Padlock Montgomery Multiplication, CRYPTOGAMS by <appro\@openssl.org>");
&asm_finish();

313
drivers/builtin_openssl2/crypto/bn/asm/x86-gf2m.pl
View File

@ -1,313 +0,0 @@
#!/usr/bin/env perl
#
# ====================================================================
# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
#
# May 2011
#
# The module implements bn_GF2m_mul_2x2 polynomial multiplication used
# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for
# the time being... Except that it has three code paths: pure integer
# code suitable for any x86 CPU, MMX code suitable for PIII and later
# and PCLMULQDQ suitable for Westmere and later. Improvement varies
# from one benchmark and µ-arch to another. Below are interval values
# for 163- and 571-bit ECDH benchmarks relative to compiler-generated
# code:
#
# PIII 16%-30%
# P4 12%-12%
# Opteron 18%-40%
# Core2 19%-44%
# Atom 38%-64%
# Westmere 53%-121%(PCLMULQDQ)/20%-32%(MMX)
# Sandy Bridge 72%-127%(PCLMULQDQ)/27%-23%(MMX)
#
# Note that above improvement coefficients are not coefficients for
# bn_GF2m_mul_2x2 itself. For example 120% ECDH improvement is result
# of bn_GF2m_mul_2x2 being >4x faster. As it gets faster, benchmark
# is more and more dominated by other subroutines, most notably by
# BN_GF2m_mod[_mul]_arr...
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
push(@INC,"${dir}","${dir}../../perlasm");
require "x86asm.pl";
&asm_init($ARGV[0],$0,$x86only = $ARGV[$#ARGV] eq "386");
$sse2=0;
for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
&external_label("OPENSSL_ia32cap_P") if ($sse2);
$a="eax";
$b="ebx";
($a1,$a2,$a4)=("ecx","edx","ebp");
$R="mm0";
@T=("mm1","mm2");
($A,$B,$B30,$B31)=("mm2","mm3","mm4","mm5");
@i=("esi","edi");
if (!$x86only) {
&function_begin_B("_mul_1x1_mmx");
&sub ("esp",32+4);
&mov ($a1,$a);
&lea ($a2,&DWP(0,$a,$a));
&and ($a1,0x3fffffff);
&lea ($a4,&DWP(0,$a2,$a2));
&mov (&DWP(0*4,"esp"),0);
&and ($a2,0x7fffffff);
&movd ($A,$a);
&movd ($B,$b);
&mov (&DWP(1*4,"esp"),$a1); # a1
&xor ($a1,$a2); # a1^a2
&pxor ($B31,$B31);
&pxor ($B30,$B30);
&mov (&DWP(2*4,"esp"),$a2); # a2
&xor ($a2,$a4); # a2^a4
&mov (&DWP(3*4,"esp"),$a1); # a1^a2
&pcmpgtd($B31,$A); # broadcast 31st bit
&paddd ($A,$A); # $A<<=1
&xor ($a1,$a2); # a1^a4=a1^a2^a2^a4
&mov (&DWP(4*4,"esp"),$a4); # a4
&xor ($a4,$a2); # a2=a4^a2^a4
&pand ($B31,$B);
&pcmpgtd($B30,$A); # broadcast 30th bit
&mov (&DWP(5*4,"esp"),$a1); # a1^a4
&xor ($a4,$a1); # a1^a2^a4
&psllq ($B31,31);
&pand ($B30,$B);
&mov (&DWP(6*4,"esp"),$a2); # a2^a4
&mov (@i[0],0x7);
&mov (&DWP(7*4,"esp"),$a4); # a1^a2^a4
&mov ($a4,@i[0]);
&and (@i[0],$b);
&shr ($b,3);
&mov (@i[1],$a4);
&psllq ($B30,30);
&and (@i[1],$b);
&shr ($b,3);
&movd ($R,&DWP(0,"esp",@i[0],4));
&mov (@i[0],$a4);
&and (@i[0],$b);
&shr ($b,3);
for($n=1;$n<9;$n++) {
&movd (@T[1],&DWP(0,"esp",@i[1],4));
&mov (@i[1],$a4);
&psllq (@T[1],3*$n);
&and (@i[1],$b);
&shr ($b,3);
&pxor ($R,@T[1]);
push(@i,shift(@i)); push(@T,shift(@T));
}
&movd (@T[1],&DWP(0,"esp",@i[1],4));
&pxor ($R,$B30);
&psllq (@T[1],3*$n++);
&pxor ($R,@T[1]);
&movd (@T[0],&DWP(0,"esp",@i[0],4));
&pxor ($R,$B31);
&psllq (@T[0],3*$n);
&add ("esp",32+4);
&pxor ($R,@T[0]);
&ret ();
&function_end_B("_mul_1x1_mmx");
}
($lo,$hi)=("eax","edx");
@T=("ecx","ebp");
&function_begin_B("_mul_1x1_ialu");
&sub ("esp",32+4);
&mov ($a1,$a);
&lea ($a2,&DWP(0,$a,$a));
&lea ($a4,&DWP(0,"",$a,4));
&and ($a1,0x3fffffff);
&lea (@i[1],&DWP(0,$lo,$lo));
&sar ($lo,31); # broadcast 31st bit
&mov (&DWP(0*4,"esp"),0);
&and ($a2,0x7fffffff);
&mov (&DWP(1*4,"esp"),$a1); # a1
&xor ($a1,$a2); # a1^a2
&mov (&DWP(2*4,"esp"),$a2); # a2
&xor ($a2,$a4); # a2^a4
&mov (&DWP(3*4,"esp"),$a1); # a1^a2
&xor ($a1,$a2); # a1^a4=a1^a2^a2^a4
&mov (&DWP(4*4,"esp"),$a4); # a4
&xor ($a4,$a2); # a2=a4^a2^a4
&mov (&DWP(5*4,"esp"),$a1); # a1^a4
&xor ($a4,$a1); # a1^a2^a4
&sar (@i[1],31); # broardcast 30th bit
&and ($lo,$b);
&mov (&DWP(6*4,"esp"),$a2); # a2^a4
&and (@i[1],$b);
&mov (&DWP(7*4,"esp"),$a4); # a1^a2^a4
&mov ($hi,$lo);
&shl ($lo,31);
&mov (@T[0],@i[1]);
&shr ($hi,1);
&mov (@i[0],0x7);
&shl (@i[1],30);
&and (@i[0],$b);
&shr (@T[0],2);
&xor ($lo,@i[1]);
&shr ($b,3);
&mov (@i[1],0x7); # 5-byte instruction!?
&and (@i[1],$b);
&shr ($b,3);
&xor ($hi,@T[0]);
&xor ($lo,&DWP(0,"esp",@i[0],4));
&mov (@i[0],0x7);
&and (@i[0],$b);
&shr ($b,3);
for($n=1;$n<9;$n++) {
&mov (@T[1],&DWP(0,"esp",@i[1],4));
&mov (@i[1],0x7);
&mov (@T[0],@T[1]);
&shl (@T[1],3*$n);
&and (@i[1],$b);
&shr (@T[0],32-3*$n);
&xor ($lo,@T[1]);
&shr ($b,3);
&xor ($hi,@T[0]);
push(@i,shift(@i)); push(@T,shift(@T));
}
&mov (@T[1],&DWP(0,"esp",@i[1],4));
&mov (@T[0],@T[1]);
&shl (@T[1],3*$n);
&mov (@i[1],&DWP(0,"esp",@i[0],4));
&shr (@T[0],32-3*$n); $n++;
&mov (@i[0],@i[1]);
&xor ($lo,@T[1]);
&shl (@i[1],3*$n);
&xor ($hi,@T[0]);
&shr (@i[0],32-3*$n);
&xor ($lo,@i[1]);
&xor ($hi,@i[0]);
&add ("esp",32+4);
&ret ();
&function_end_B("_mul_1x1_ialu");
# void bn_GF2m_mul_2x2(BN_ULONG *r, BN_ULONG a1, BN_ULONG a0, BN_ULONG b1, BN_ULONG b0);
&function_begin_B("bn_GF2m_mul_2x2");
if (!$x86only) {
&picmeup("edx","OPENSSL_ia32cap_P");
&mov ("eax",&DWP(0,"edx"));
&mov ("edx",&DWP(4,"edx"));
&test ("eax",1<<23); # check MMX bit
&jz (&label("ialu"));
if ($sse2) {
&test ("eax",1<<24); # check FXSR bit
&jz (&label("mmx"));
&test ("edx",1<<1); # check PCLMULQDQ bit
&jz (&label("mmx"));
&movups ("xmm0",&QWP(8,"esp"));
&shufps ("xmm0","xmm0",0b10110001);
&pclmulqdq ("xmm0","xmm0",1);
&mov ("eax",&DWP(4,"esp"));
&movups (&QWP(0,"eax"),"xmm0");
&ret ();
&set_label("mmx",16);
}
&push ("ebp");
&push ("ebx");
&push ("esi");
&push ("edi");
&mov ($a,&wparam(1));
&mov ($b,&wparam(3));
&call ("_mul_1x1_mmx"); # a1·b1
&movq ("mm7",$R);
&mov ($a,&wparam(2));
&mov ($b,&wparam(4));
&call ("_mul_1x1_mmx"); # a0·b0
&movq ("mm6",$R);
&mov ($a,&wparam(1));
&mov ($b,&wparam(3));
&xor ($a,&wparam(2));
&xor ($b,&wparam(4));
&call ("_mul_1x1_mmx"); # (a0+a1)·(b0+b1)
&pxor ($R,"mm7");
&mov ($a,&wparam(0));
&pxor ($R,"mm6"); # (a0+a1)·(b0+b1)-a1·b1-a0·b0
&movq ($A,$R);
&psllq ($R,32);
&pop ("edi");
&psrlq ($A,32);
&pop ("esi");
&pxor ($R,"mm6");
&pop ("ebx");
&pxor ($A,"mm7");
&movq (&QWP(0,$a),$R);
&pop ("ebp");
&movq (&QWP(8,$a),$A);
&emms ();
&ret ();
&set_label("ialu",16);
}
&push ("ebp");
&push ("ebx");
&push ("esi");
&push ("edi");
&stack_push(4+1);
&mov ($a,&wparam(1));
&mov ($b,&wparam(3));
&call ("_mul_1x1_ialu"); # a1·b1
&mov (&DWP(8,"esp"),$lo);
&mov (&DWP(12,"esp"),$hi);
&mov ($a,&wparam(2));
&mov ($b,&wparam(4));
&call ("_mul_1x1_ialu"); # a0·b0
&mov (&DWP(0,"esp"),$lo);
&mov (&DWP(4,"esp"),$hi);
&mov ($a,&wparam(1));
&mov ($b,&wparam(3));
&xor ($a,&wparam(2));
&xor ($b,&wparam(4));
&call ("_mul_1x1_ialu"); # (a0+a1)·(b0+b1)
&mov ("ebp",&wparam(0));
@r=("ebx","ecx","edi","esi");
&mov (@r[0],&DWP(0,"esp"));
&mov (@r[1],&DWP(4,"esp"));
&mov (@r[2],&DWP(8,"esp"));
&mov (@r[3],&DWP(12,"esp"));
&xor ($lo,$hi);
&xor ($hi,@r[1]);
&xor ($lo,@r[0]);
&mov (&DWP(0,"ebp"),@r[0]);
&xor ($hi,@r[2]);
&mov (&DWP(12,"ebp"),@r[3]);
&xor ($lo,@r[3]);
&stack_pop(4+1);
&xor ($hi,@r[3]);
&pop ("edi");
&xor ($lo,$hi);
&pop ("esi");
&mov (&DWP(8,"ebp"),$hi);
&pop ("ebx");
&mov (&DWP(4,"ebp"),$lo);
&pop ("ebp");
&ret ();
&function_end_B("bn_GF2m_mul_2x2");
&asciz ("GF(2^m) Multiplication for x86, CRYPTOGAMS by <appro\@openssl.org>");
&asm_finish();

593
drivers/builtin_openssl2/crypto/bn/asm/x86-mont.pl
View File

@ -1,593 +0,0 @@
#!/usr/bin/env perl
# ====================================================================
# Written by Andy Polyakov <appro@fy.chalmers.se> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
# October 2005
#
# This is a "teaser" code, as it can be improved in several ways...
# First of all non-SSE2 path should be implemented (yes, for now it
# performs Montgomery multiplication/convolution only on SSE2-capable
# CPUs such as P4, others fall down to original code). Then inner loop
# can be unrolled and modulo-scheduled to improve ILP and possibly
# moved to 128-bit XMM register bank (though it would require input
# rearrangement and/or increase bus bandwidth utilization). Dedicated
# squaring procedure should give further performance improvement...
# Yet, for being draft, the code improves rsa512 *sign* benchmark by
# 110%(!), rsa1024 one - by 70% and rsa4096 - by 20%:-)
# December 2006
#
# Modulo-scheduling SSE2 loops results in further 15-20% improvement.
# Integer-only code [being equipped with dedicated squaring procedure]
# gives ~40% on rsa512 sign benchmark...
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
push(@INC,"${dir}","${dir}../../perlasm");
require "x86asm.pl";
&asm_init($ARGV[0],$0);
$sse2=0;
for (@ARGV) { $sse2=1 if (/-DOPENSSL_IA32_SSE2/); }
&external_label("OPENSSL_ia32cap_P") if ($sse2);
&function_begin("bn_mul_mont");
$i="edx";
$j="ecx";
$ap="esi"; $tp="esi"; # overlapping variables!!!
$rp="edi"; $bp="edi"; # overlapping variables!!!
$np="ebp";
$num="ebx";
$_num=&DWP(4*0,"esp"); # stack top layout
$_rp=&DWP(4*1,"esp");
$_ap=&DWP(4*2,"esp");
$_bp=&DWP(4*3,"esp");
$_np=&DWP(4*4,"esp");
$_n0=&DWP(4*5,"esp"); $_n0q=&QWP(4*5,"esp");
$_sp=&DWP(4*6,"esp");
$_bpend=&DWP(4*7,"esp");
$frame=32; # size of above frame rounded up to 16n
&xor ("eax","eax");
&mov ("edi",&wparam(5)); # int num
&cmp ("edi",4);
&jl (&label("just_leave"));
&lea ("esi",&wparam(0)); # put aside pointer to argument block
&lea ("edx",&wparam(1)); # load ap
&mov ("ebp","esp"); # saved stack pointer!
&add ("edi",2); # extra two words on top of tp
&neg ("edi");
&lea ("esp",&DWP(-$frame,"esp","edi",4)); # alloca($frame+4*(num+2))
&neg ("edi");
# minimize cache contention by arraning 2K window between stack
# pointer and ap argument [np is also position sensitive vector,
# but it's assumed to be near ap, as it's allocated at ~same
# time].
&mov ("eax","esp");
&sub ("eax","edx");
&and ("eax",2047);
&sub ("esp","eax"); # this aligns sp and ap modulo 2048
&xor ("edx","esp");
&and ("edx",2048);
&xor ("edx",2048);
&sub ("esp","edx"); # this splits them apart modulo 4096
&and ("esp",-64); # align to cache line
################################# load argument block...
&mov ("eax",&DWP(0*4,"esi"));# BN_ULONG *rp
&mov ("ebx",&DWP(1*4,"esi"));# const BN_ULONG *ap
&mov ("ecx",&DWP(2*4,"esi"));# const BN_ULONG *bp
&mov ("edx",&DWP(3*4,"esi"));# const BN_ULONG *np
&mov ("esi",&DWP(4*4,"esi"));# const BN_ULONG *n0
#&mov ("edi",&DWP(5*4,"esi"));# int num
&mov ("esi",&DWP(0,"esi")); # pull n0[0]
&mov ($_rp,"eax"); # ... save a copy of argument block
&mov ($_ap,"ebx");
&mov ($_bp,"ecx");
&mov ($_np,"edx");
&mov ($_n0,"esi");
&lea ($num,&DWP(-3,"edi")); # num=num-1 to assist modulo-scheduling
#&mov ($_num,$num); # redundant as $num is not reused
&mov ($_sp,"ebp"); # saved stack pointer!
if($sse2) {
$acc0="mm0"; # mmx register bank layout
$acc1="mm1";
$car0="mm2";
$car1="mm3";
$mul0="mm4";
$mul1="mm5";
$temp="mm6";
$mask="mm7";
&picmeup("eax","OPENSSL_ia32cap_P");
&bt (&DWP(0,"eax"),26);
&jnc (&label("non_sse2"));
&mov ("eax",-1);
&movd ($mask,"eax"); # mask 32 lower bits
&mov ($ap,$_ap); # load input pointers
&mov ($bp,$_bp);
&mov ($np,$_np);
&xor ($i,$i); # i=0
&xor ($j,$j); # j=0
&movd ($mul0,&DWP(0,$bp)); # bp[0]
&movd ($mul1,&DWP(0,$ap)); # ap[0]
&movd ($car1,&DWP(0,$np)); # np[0]
&pmuludq($mul1,$mul0); # ap[0]*bp[0]
&movq ($car0,$mul1);
&movq ($acc0,$mul1); # I wish movd worked for
&pand ($acc0,$mask); # inter-register transfers
&pmuludq($mul1,$_n0q); # *=n0
&pmuludq($car1,$mul1); # "t[0]"*np[0]*n0
&paddq ($car1,$acc0);
&movd ($acc1,&DWP(4,$np)); # np[1]
&movd ($acc0,&DWP(4,$ap)); # ap[1]
&psrlq ($car0,32);
&psrlq ($car1,32);
&inc ($j); # j++
&set_label("1st",16);
&pmuludq($acc0,$mul0); # ap[j]*bp[0]
&pmuludq($acc1,$mul1); # np[j]*m1
&paddq ($car0,$acc0); # +=c0
&paddq ($car1,$acc1); # +=c1
&movq ($acc0,$car0);
&pand ($acc0,$mask);
&movd ($acc1,&DWP(4,$np,$j,4)); # np[j+1]
&paddq ($car1,$acc0); # +=ap[j]*bp[0];
&movd ($acc0,&DWP(4,$ap,$j,4)); # ap[j+1]
&psrlq ($car0,32);
&movd (&DWP($frame-4,"esp",$j,4),$car1); # tp[j-1]=
&psrlq ($car1,32);
&lea ($j,&DWP(1,$j));
&cmp ($j,$num);
&jl (&label("1st"));
&pmuludq($acc0,$mul0); # ap[num-1]*bp[0]
&pmuludq($acc1,$mul1); # np[num-1]*m1
&paddq ($car0,$acc0); # +=c0
&paddq ($car1,$acc1); # +=c1
&movq ($acc0,$car0);
&pand ($acc0,$mask);
&paddq ($car1,$acc0); # +=ap[num-1]*bp[0];
&movd (&DWP($frame-4,"esp",$j,4),$car1); # tp[num-2]=
&psrlq ($car0,32);
&psrlq ($car1,32);
&paddq ($car1,$car0);
&movq (&QWP($frame,"esp",$num,4),$car1); # tp[num].tp[num-1]
&inc ($i); # i++
&set_label("outer");
&xor ($j,$j); # j=0
&movd ($mul0,&DWP(0,$bp,$i,4)); # bp[i]
&movd ($mul1,&DWP(0,$ap)); # ap[0]
&movd ($temp,&DWP($frame,"esp")); # tp[0]
&movd ($car1,&DWP(0,$np)); # np[0]
&pmuludq($mul1,$mul0); # ap[0]*bp[i]
&paddq ($mul1,$temp); # +=tp[0]
&movq ($acc0,$mul1);
&movq ($car0,$mul1);
&pand ($acc0,$mask);
&pmuludq($mul1,$_n0q); # *=n0
&pmuludq($car1,$mul1);
&paddq ($car1,$acc0);
&movd ($temp,&DWP($frame+4,"esp")); # tp[1]
&movd ($acc1,&DWP(4,$np)); # np[1]
&movd ($acc0,&DWP(4,$ap)); # ap[1]
&psrlq ($car0,32);
&psrlq ($car1,32);
&paddq ($car0,$temp); # +=tp[1]
&inc ($j); # j++
&dec ($num);
&set_label("inner");
&pmuludq($acc0,$mul0); # ap[j]*bp[i]
&pmuludq($acc1,$mul1); # np[j]*m1
&paddq ($car0,$acc0); # +=c0
&paddq ($car1,$acc1); # +=c1
&movq ($acc0,$car0);
&movd ($temp,&DWP($frame+4,"esp",$j,4));# tp[j+1]
&pand ($acc0,$mask);
&movd ($acc1,&DWP(4,$np,$j,4)); # np[j+1]
&paddq ($car1,$acc0); # +=ap[j]*bp[i]+tp[j]
&movd ($acc0,&DWP(4,$ap,$j,4)); # ap[j+1]
&psrlq ($car0,32);
&movd (&DWP($frame-4,"esp",$j,4),$car1);# tp[j-1]=
&psrlq ($car1,32);
&paddq ($car0,$temp); # +=tp[j+1]
&dec ($num);
&lea ($j,&DWP(1,$j)); # j++
&jnz (&label("inner"));
&mov ($num,$j);
&pmuludq($acc0,$mul0); # ap[num-1]*bp[i]
&pmuludq($acc1,$mul1); # np[num-1]*m1
&paddq ($car0,$acc0); # +=c0
&paddq ($car1,$acc1); # +=c1
&movq ($acc0,$car0);
&pand ($acc0,$mask);
&paddq ($car1,$acc0); # +=ap[num-1]*bp[i]+tp[num-1]
&movd (&DWP($frame-4,"esp",$j,4),$car1); # tp[num-2]=
&psrlq ($car0,32);
&psrlq ($car1,32);
&movd ($temp,&DWP($frame+4,"esp",$num,4)); # += tp[num]
&paddq ($car1,$car0);
&paddq ($car1,$temp);
&movq (&QWP($frame,"esp",$num,4),$car1); # tp[num].tp[num-1]
&lea ($i,&DWP(1,$i)); # i++
&cmp ($i,$num);
&jle (&label("outer"));
&emms (); # done with mmx bank
&jmp (&label("common_tail"));
&set_label("non_sse2",16);
}
if (0) {
&mov ("esp",$_sp);
&xor ("eax","eax"); # signal "not fast enough [yet]"
&jmp (&label("just_leave"));
# While the below code provides competitive performance for
# all key lengthes on modern Intel cores, it's still more
# than 10% slower for 4096-bit key elsewhere:-( "Competitive"
# means compared to the original integer-only assembler.
# 512-bit RSA sign is better by ~40%, but that's about all
# one can say about all CPUs...
} else {
$inp="esi"; # integer path uses these registers differently
$word="edi";
$carry="ebp";
&mov ($inp,$_ap);
&lea ($carry,&DWP(1,$num));
&mov ($word,$_bp);
&xor ($j,$j); # j=0
&mov ("edx",$inp);
&and ($carry,1); # see if num is even
&sub ("edx",$word); # see if ap==bp
&lea ("eax",&DWP(4,$word,$num,4)); # &bp[num]
&or ($carry,"edx");
&mov ($word,&DWP(0,$word)); # bp[0]
&jz (&label("bn_sqr_mont"));
&mov ($_bpend,"eax");
&mov ("eax",&DWP(0,$inp));
&xor ("edx","edx");
&set_label("mull",16);
&mov ($carry,"edx");
&mul ($word); # ap[j]*bp[0]
&add ($carry,"eax");
&lea ($j,&DWP(1,$j));
&adc ("edx",0);
&mov ("eax",&DWP(0,$inp,$j,4)); # ap[j+1]
&cmp ($j,$num);
&mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]=
&jl (&label("mull"));
&mov ($carry,"edx");
&mul ($word); # ap[num-1]*bp[0]
&mov ($word,$_n0);
&add ("eax",$carry);
&mov ($inp,$_np);
&adc ("edx",0);
&imul ($word,&DWP($frame,"esp")); # n0*tp[0]
&mov (&DWP($frame,"esp",$num,4),"eax"); # tp[num-1]=
&xor ($j,$j);
&mov (&DWP($frame+4,"esp",$num,4),"edx"); # tp[num]=
&mov (&DWP($frame+8,"esp",$num,4),$j); # tp[num+1]=
&mov ("eax",&DWP(0,$inp)); # np[0]
&mul ($word); # np[0]*m
&add ("eax",&DWP($frame,"esp")); # +=tp[0]
&mov ("eax",&DWP(4,$inp)); # np[1]
&adc ("edx",0);
&inc ($j);
&jmp (&label("2ndmadd"));
&set_label("1stmadd",16);
&mov ($carry,"edx");
&mul ($word); # ap[j]*bp[i]
&add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j]
&lea ($j,&DWP(1,$j));
&adc ("edx",0);
&add ($carry,"eax");
&mov ("eax",&DWP(0,$inp,$j,4)); # ap[j+1]
&adc ("edx",0);
&cmp ($j,$num);
&mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]=
&jl (&label("1stmadd"));
&mov ($carry,"edx");
&mul ($word); # ap[num-1]*bp[i]
&add ("eax",&DWP($frame,"esp",$num,4)); # +=tp[num-1]
&mov ($word,$_n0);
&adc ("edx",0);
&mov ($inp,$_np);
&add ($carry,"eax");
&adc ("edx",0);
&imul ($word,&DWP($frame,"esp")); # n0*tp[0]
&xor ($j,$j);
&add ("edx",&DWP($frame+4,"esp",$num,4)); # carry+=tp[num]
&mov (&DWP($frame,"esp",$num,4),$carry); # tp[num-1]=
&adc ($j,0);
&mov ("eax",&DWP(0,$inp)); # np[0]
&mov (&DWP($frame+4,"esp",$num,4),"edx"); # tp[num]=
&mov (&DWP($frame+8,"esp",$num,4),$j); # tp[num+1]=
&mul ($word); # np[0]*m
&add ("eax",&DWP($frame,"esp")); # +=tp[0]
&mov ("eax",&DWP(4,$inp)); # np[1]
&adc ("edx",0);
&mov ($j,1);
&set_label("2ndmadd",16);
&mov ($carry,"edx");
&mul ($word); # np[j]*m
&add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j]
&lea ($j,&DWP(1,$j));
&adc ("edx",0);
&add ($carry,"eax");
&mov ("eax",&DWP(0,$inp,$j,4)); # np[j+1]
&adc ("edx",0);
&cmp ($j,$num);
&mov (&DWP($frame-8,"esp",$j,4),$carry); # tp[j-1]=
&jl (&label("2ndmadd"));
&mov ($carry,"edx");
&mul ($word); # np[j]*m
&add ($carry,&DWP($frame,"esp",$num,4)); # +=tp[num-1]
&adc ("edx",0);
&add ($carry,"eax");
&adc ("edx",0);
&mov (&DWP($frame-4,"esp",$num,4),$carry); # tp[num-2]=
&xor ("eax","eax");
&mov ($j,$_bp); # &bp[i]
&add ("edx",&DWP($frame+4,"esp",$num,4)); # carry+=tp[num]
&adc ("eax",&DWP($frame+8,"esp",$num,4)); # +=tp[num+1]
&lea ($j,&DWP(4,$j));
&mov (&DWP($frame,"esp",$num,4),"edx"); # tp[num-1]=
&cmp ($j,$_bpend);
&mov (&DWP($frame+4,"esp",$num,4),"eax"); # tp[num]=
&je (&label("common_tail"));
&mov ($word,&DWP(0,$j)); # bp[i+1]
&mov ($inp,$_ap);
&mov ($_bp,$j); # &bp[++i]
&xor ($j,$j);
&xor ("edx","edx");
&mov ("eax",&DWP(0,$inp));
&jmp (&label("1stmadd"));
&set_label("bn_sqr_mont",16);
$sbit=$num;
&mov ($_num,$num);
&mov ($_bp,$j); # i=0
&mov ("eax",$word); # ap[0]
&mul ($word); # ap[0]*ap[0]
&mov (&DWP($frame,"esp"),"eax"); # tp[0]=
&mov ($sbit,"edx");
&shr ("edx",1);
&and ($sbit,1);
&inc ($j);
&set_label("sqr",16);
&mov ("eax",&DWP(0,$inp,$j,4)); # ap[j]
&mov ($carry,"edx");
&mul ($word); # ap[j]*ap[0]
&add ("eax",$carry);
&lea ($j,&DWP(1,$j));
&adc ("edx",0);
&lea ($carry,&DWP(0,$sbit,"eax",2));
&shr ("eax",31);
&cmp ($j,$_num);
&mov ($sbit,"eax");
&mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]=
&jl (&label("sqr"));
&mov ("eax",&DWP(0,$inp,$j,4)); # ap[num-1]
&mov ($carry,"edx");
&mul ($word); # ap[num-1]*ap[0]
&add ("eax",$carry);
&mov ($word,$_n0);
&adc ("edx",0);
&mov ($inp,$_np);
&lea ($carry,&DWP(0,$sbit,"eax",2));
&imul ($word,&DWP($frame,"esp")); # n0*tp[0]
&shr ("eax",31);
&mov (&DWP($frame,"esp",$j,4),$carry); # tp[num-1]=
&lea ($carry,&DWP(0,"eax","edx",2));
&mov ("eax",&DWP(0,$inp)); # np[0]
&shr ("edx",31);
&mov (&DWP($frame+4,"esp",$j,4),$carry); # tp[num]=
&mov (&DWP($frame+8,"esp",$j,4),"edx"); # tp[num+1]=
&mul ($word); # np[0]*m
&add ("eax",&DWP($frame,"esp")); # +=tp[0]
&mov ($num,$j);
&adc ("edx",0);
&mov ("eax",&DWP(4,$inp)); # np[1]
&mov ($j,1);
&set_label("3rdmadd",16);
&mov ($carry,"edx");
&mul ($word); # np[j]*m
&add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j]
&adc ("edx",0);
&add ($carry,"eax");
&mov ("eax",&DWP(4,$inp,$j,4)); # np[j+1]
&adc ("edx",0);
&mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j-1]=
&mov ($carry,"edx");
&mul ($word); # np[j+1]*m
&add ($carry,&DWP($frame+4,"esp",$j,4)); # +=tp[j+1]
&lea ($j,&DWP(2,$j));
&adc ("edx",0);
&add ($carry,"eax");
&mov ("eax",&DWP(0,$inp,$j,4)); # np[j+2]
&adc ("edx",0);
&cmp ($j,$num);
&mov (&DWP($frame-8,"esp",$j,4),$carry); # tp[j]=
&jl (&label("3rdmadd"));
&mov ($carry,"edx");
&mul ($word); # np[j]*m
&add ($carry,&DWP($frame,"esp",$num,4)); # +=tp[num-1]
&adc ("edx",0);
&add ($carry,"eax");
&adc ("edx",0);
&mov (&DWP($frame-4,"esp",$num,4),$carry); # tp[num-2]=
&mov ($j,$_bp); # i
&xor ("eax","eax");
&mov ($inp,$_ap);
&add ("edx",&DWP($frame+4,"esp",$num,4)); # carry+=tp[num]
&adc ("eax",&DWP($frame+8,"esp",$num,4)); # +=tp[num+1]
&mov (&DWP($frame,"esp",$num,4),"edx"); # tp[num-1]=
&cmp ($j,$num);
&mov (&DWP($frame+4,"esp",$num,4),"eax"); # tp[num]=
&je (&label("common_tail"));
&mov ($word,&DWP(4,$inp,$j,4)); # ap[i]
&lea ($j,&DWP(1,$j));
&mov ("eax",$word);
&mov ($_bp,$j); # ++i
&mul ($word); # ap[i]*ap[i]
&add ("eax",&DWP($frame,"esp",$j,4)); # +=tp[i]
&adc ("edx",0);
&mov (&DWP($frame,"esp",$j,4),"eax"); # tp[i]=
&xor ($carry,$carry);
&cmp ($j,$num);
&lea ($j,&DWP(1,$j));
&je (&label("sqrlast"));
&mov ($sbit,"edx"); # zaps $num
&shr ("edx",1);
&and ($sbit,1);
&set_label("sqradd",16);
&mov ("eax",&DWP(0,$inp,$j,4)); # ap[j]
&mov ($carry,"edx");
&mul ($word); # ap[j]*ap[i]
&add ("eax",$carry);
&lea ($carry,&DWP(0,"eax","eax"));
&adc ("edx",0);
&shr ("eax",31);
&add ($carry,&DWP($frame,"esp",$j,4)); # +=tp[j]
&lea ($j,&DWP(1,$j));
&adc ("eax",0);
&add ($carry,$sbit);
&adc ("eax",0);
&cmp ($j,$_num);
&mov (&DWP($frame-4,"esp",$j,4),$carry); # tp[j]=
&mov ($sbit,"eax");
&jle (&label("sqradd"));
&mov ($carry,"edx");
&add ("edx","edx");
&shr ($carry,31);
&add ("edx",$sbit);
&adc ($carry,0);
&set_label("sqrlast");
&mov ($word,$_n0);
&mov ($inp,$_np);
&imul ($word,&DWP($frame,"esp")); # n0*tp[0]
&add ("edx",&DWP($frame,"esp",$j,4)); # +=tp[num]
&mov ("eax",&DWP(0,$inp)); # np[0]
&adc ($carry,0);
&mov (&DWP($frame,"esp",$j,4),"edx"); # tp[num]=
&mov (&DWP($frame+4,"esp",$j,4),$carry); # tp[num+1]=
&mul ($word); # np[0]*m
&add ("eax",&DWP($frame,"esp")); # +=tp[0]
&lea ($num,&DWP(-1,$j));
&adc ("edx",0);
&mov ($j,1);
&mov ("eax",&DWP(4,$inp)); # np[1]
&jmp (&label("3rdmadd"));
}
&set_label("common_tail",16);
&mov ($np,$_np); # load modulus pointer
&mov ($rp,$_rp); # load result pointer
&lea ($tp,&DWP($frame,"esp")); # [$ap and $bp are zapped]
&mov ("eax",&DWP(0,$tp)); # tp[0]
&mov ($j,$num); # j=num-1
&xor ($i,$i); # i=0 and clear CF!
&set_label("sub",16);
&sbb ("eax",&DWP(0,$np,$i,4));
&mov (&DWP(0,$rp,$i,4),"eax"); # rp[i]=tp[i]-np[i]
&dec ($j); # doesn't affect CF!
&mov ("eax",&DWP(4,$tp,$i,4)); # tp[i+1]
&lea ($i,&DWP(1,$i)); # i++
&jge (&label("sub"));
&sbb ("eax",0); # handle upmost overflow bit
&and ($tp,"eax");
&not ("eax");
&mov ($np,$rp);
&and ($np,"eax");
&or ($tp,$np); # tp=carry?tp:rp
&set_label("copy",16); # copy or in-place refresh
&mov ("eax",&DWP(0,$tp,$num,4));
&mov (&DWP(0,$rp,$num,4),"eax"); # rp[i]=tp[i]
&mov (&DWP($frame,"esp",$num,4),$j); # zap temporary vector
&dec ($num);
&jge (&label("copy"));
&mov ("esp",$_sp); # pull saved stack pointer
&mov ("eax",1);
&set_label("just_leave");
&function_end("bn_mul_mont");
&asciz("Montgomery Multiplication for x86, CRYPTOGAMS by <appro\@openssl.org>");
&asm_finish();

28
drivers/builtin_openssl2/crypto/bn/asm/x86.pl
View File

@ -1,28 +0,0 @@
#!/usr/local/bin/perl
push(@INC,"perlasm","../../perlasm");
require "x86asm.pl";
require("x86/mul_add.pl");
require("x86/mul.pl");
require("x86/sqr.pl");
require("x86/div.pl");
require("x86/add.pl");
require("x86/sub.pl");
require("x86/comba.pl");
&asm_init($ARGV[0],$0);
&bn_mul_add_words("bn_mul_add_words");
&bn_mul_words("bn_mul_words");
&bn_sqr_words("bn_sqr_words");
&bn_div_words("bn_div_words");
&bn_add_words("bn_add_words");
&bn_sub_words("bn_sub_words");
&bn_mul_comba("bn_mul_comba8",8);
&bn_mul_comba("bn_mul_comba4",4);
&bn_sqr_comba("bn_sqr_comba8",8);
&bn_sqr_comba("bn_sqr_comba4",4);
&asm_finish();

76
drivers/builtin_openssl2/crypto/bn/asm/x86/add.pl
View File

@ -1,76 +0,0 @@
#!/usr/local/bin/perl
# x86 assember
sub bn_add_words
{
local($name)=@_;
&function_begin($name,"");
&comment("");
$a="esi";
$b="edi";
$c="eax";
$r="ebx";
$tmp1="ecx";
$tmp2="edx";
$num="ebp";
&mov($r,&wparam(0)); # get r
&mov($a,&wparam(1)); # get a
&mov($b,&wparam(2)); # get b
&mov($num,&wparam(3)); # get num
&xor($c,$c); # clear carry
&and($num,0xfffffff8); # num / 8
&jz(&label("aw_finish"));
&set_label("aw_loop",0);
for ($i=0; $i<8; $i++)
{
&comment("Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov($tmp2,&DWP($i*4,$b,"",0)); # *b
&add($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&add($tmp1,$tmp2);
&adc($c,0);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
}
&comment("");
&add($a,32);
&add($b,32);
&add($r,32);
&sub($num,8);
&jnz(&label("aw_loop"));
&set_label("aw_finish",0);
&mov($num,&wparam(3)); # get num
&and($num,7);
&jz(&label("aw_end"));
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov($tmp2,&DWP($i*4,$b,"",0));# *b
&add($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&add($tmp1,$tmp2);
&adc($c,0);
&dec($num) if ($i != 6);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *a
&jz(&label("aw_end")) if ($i != 6);
}
&set_label("aw_end",0);
# &mov("eax",$c); # $c is "eax"
&function_end($name);
}
1;

277
drivers/builtin_openssl2/crypto/bn/asm/x86/comba.pl
View File

@ -1,277 +0,0 @@
#!/usr/local/bin/perl
# x86 assember
sub mul_add_c
{
local($a,$ai,$b,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
# pos == -1 if eax and edx are pre-loaded, 0 to load from next
# words, and 1 if load return value
&comment("mul a[$ai]*b[$bi]");
# "eax" and "edx" will always be pre-loaded.
# &mov("eax",&DWP($ai*4,$a,"",0)) ;
# &mov("edx",&DWP($bi*4,$b,"",0));
&mul("edx");
&add($c0,"eax");
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # laod next a
&mov("eax",&wparam(0)) if $pos > 0; # load r[]
###
&adc($c1,"edx");
&mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 0; # laod next b
&mov("edx",&DWP(($nb)*4,$b,"",0)) if $pos == 1; # laod next b
###
&adc($c2,0);
# is pos > 1, it means it is the last loop
&mov(&DWP($i*4,"eax","",0),$c0) if $pos > 0; # save r[];
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # laod next a
}
sub sqr_add_c
{
local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
# pos == -1 if eax and edx are pre-loaded, 0 to load from next
# words, and 1 if load return value
&comment("sqr a[$ai]*a[$bi]");
# "eax" and "edx" will always be pre-loaded.
# &mov("eax",&DWP($ai*4,$a,"",0)) ;
# &mov("edx",&DWP($bi*4,$b,"",0));
if ($ai == $bi)
{ &mul("eax");}
else
{ &mul("edx");}
&add($c0,"eax");
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a
###
&adc($c1,"edx");
&mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos == 1) && ($na != $nb);
###
&adc($c2,0);
# is pos > 1, it means it is the last loop
&mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[];
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b
}
sub sqr_add_c2
{
local($r,$a,$ai,$bi,$c0,$c1,$c2,$pos,$i,$na,$nb)=@_;
# pos == -1 if eax and edx are pre-loaded, 0 to load from next
# words, and 1 if load return value
&comment("sqr a[$ai]*a[$bi]");
# "eax" and "edx" will always be pre-loaded.
# &mov("eax",&DWP($ai*4,$a,"",0)) ;
# &mov("edx",&DWP($bi*4,$a,"",0));
if ($ai == $bi)
{ &mul("eax");}
else
{ &mul("edx");}
&add("eax","eax");
###
&adc("edx","edx");
###
&adc($c2,0);
&add($c0,"eax");
&adc($c1,"edx");
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 0; # load next a
&mov("eax",&DWP(($na)*4,$a,"",0)) if $pos == 1; # load next b
&adc($c2,0);
&mov(&DWP($i*4,$r,"",0),$c0) if $pos > 0; # save r[];
&mov("edx",&DWP(($nb)*4,$a,"",0)) if ($pos <= 1) && ($na != $nb);
###
}
sub bn_mul_comba
{
local($name,$num)=@_;
local($a,$b,$c0,$c1,$c2);
local($i,$as,$ae,$bs,$be,$ai,$bi);
local($tot,$end);
&function_begin_B($name,"");
$c0="ebx";
$c1="ecx";
$c2="ebp";
$a="esi";
$b="edi";
$as=0;
$ae=0;
$bs=0;
$be=0;
$tot=$num+$num-1;
&push("esi");
&mov($a,&wparam(1));
&push("edi");
&mov($b,&wparam(2));
&push("ebp");
&push("ebx");
&xor($c0,$c0);
&mov("eax",&DWP(0,$a,"",0)); # load the first word
&xor($c1,$c1);
&mov("edx",&DWP(0,$b,"",0)); # load the first second
for ($i=0; $i<$tot; $i++)
{
$ai=$as;
$bi=$bs;
$end=$be+1;
&comment("################## Calculate word $i");
for ($j=$bs; $j<$end; $j++)
{
&xor($c2,$c2) if ($j == $bs);
if (($j+1) == $end)
{
$v=1;
$v=2 if (($i+1) == $tot);
}
else
{ $v=0; }
if (($j+1) != $end)
{
$na=($ai-1);
$nb=($bi+1);
}
else
{
$na=$as+($i < ($num-1));
$nb=$bs+($i >= ($num-1));
}
#printf STDERR "[$ai,$bi] -> [$na,$nb]\n";
&mul_add_c($a,$ai,$b,$bi,$c0,$c1,$c2,$v,$i,$na,$nb);
if ($v)
{
&comment("saved r[$i]");
# &mov("eax",&wparam(0));
# &mov(&DWP($i*4,"eax","",0),$c0);
($c0,$c1,$c2)=($c1,$c2,$c0);
}
$ai--;
$bi++;
}
$as++ if ($i < ($num-1));
$ae++ if ($i >= ($num-1));
$bs++ if ($i >= ($num-1));
$be++ if ($i < ($num-1));
}
&comment("save r[$i]");
# &mov("eax",&wparam(0));
&mov(&DWP($i*4,"eax","",0),$c0);
&pop("ebx");
&pop("ebp");
&pop("edi");
&pop("esi");
&ret();
&function_end_B($name);
}
sub bn_sqr_comba
{
local($name,$num)=@_;
local($r,$a,$c0,$c1,$c2)=@_;
local($i,$as,$ae,$bs,$be,$ai,$bi);
local($b,$tot,$end,$half);
&function_begin_B($name,"");
$c0="ebx";
$c1="ecx";
$c2="ebp";
$a="esi";
$r="edi";
&push("esi");
&push("edi");
&push("ebp");
&push("ebx");
&mov($r,&wparam(0));
&mov($a,&wparam(1));
&xor($c0,$c0);
&xor($c1,$c1);
&mov("eax",&DWP(0,$a,"",0)); # load the first word
$as=0;
$ae=0;
$bs=0;
$be=0;
$tot=$num+$num-1;
for ($i=0; $i<$tot; $i++)
{
$ai=$as;
$bi=$bs;
$end=$be+1;
&comment("############### Calculate word $i");
for ($j=$bs; $j<$end; $j++)
{
&xor($c2,$c2) if ($j == $bs);
if (($ai-1) < ($bi+1))
{
$v=1;
$v=2 if ($i+1) == $tot;
}
else
{ $v=0; }
if (!$v)
{
$na=$ai-1;
$nb=$bi+1;
}
else
{
$na=$as+($i < ($num-1));
$nb=$bs+($i >= ($num-1));
}
if ($ai == $bi)
{
&sqr_add_c($r,$a,$ai,$bi,
$c0,$c1,$c2,$v,$i,$na,$nb);
}
else
{
&sqr_add_c2($r,$a,$ai,$bi,
$c0,$c1,$c2,$v,$i,$na,$nb);
}
if ($v)
{
&comment("saved r[$i]");
#&mov(&DWP($i*4,$r,"",0),$c0);
($c0,$c1,$c2)=($c1,$c2,$c0);
last;
}
$ai--;
$bi++;
}
$as++ if ($i < ($num-1));
$ae++ if ($i >= ($num-1));
$bs++ if ($i >= ($num-1));
$be++ if ($i < ($num-1));
}
&mov(&DWP($i*4,$r,"",0),$c0);
&pop("ebx");
&pop("ebp");
&pop("edi");
&pop("esi");
&ret();
&function_end_B($name);
}
1;

15
drivers/builtin_openssl2/crypto/bn/asm/x86/div.pl
View File

@ -1,15 +0,0 @@
#!/usr/local/bin/perl
# x86 assember
sub bn_div_words
{
local($name)=@_;
&function_begin($name,"");
&mov("edx",&wparam(0)); #
&mov("eax",&wparam(1)); #
&mov("ebx",&wparam(2)); #
&div("ebx");
&function_end($name);
}
1;

77
drivers/builtin_openssl2/crypto/bn/asm/x86/mul.pl
View File

@ -1,77 +0,0 @@
#!/usr/local/bin/perl
# x86 assember
sub bn_mul_words
{
local($name)=@_;
&function_begin($name,"");
&comment("");
$Low="eax";
$High="edx";
$a="ebx";
$w="ecx";
$r="edi";
$c="esi";
$num="ebp";
&xor($c,$c); # clear carry
&mov($r,&wparam(0)); #
&mov($a,&wparam(1)); #
&mov($num,&wparam(2)); #
&mov($w,&wparam(3)); #
&and($num,0xfffffff8); # num / 8
&jz(&label("mw_finish"));
&set_label("mw_loop",0);
for ($i=0; $i<32; $i+=4)
{
&comment("Round $i");
&mov("eax",&DWP($i,$a,"",0)); # *a
&mul($w); # *a * w
&add("eax",$c); # L(t)+=c
# XXX
&adc("edx",0); # H(t)+=carry
&mov(&DWP($i,$r,"",0),"eax"); # *r= L(t);
&mov($c,"edx"); # c= H(t);
}
&comment("");
&add($a,32);
&add($r,32);
&sub($num,8);
&jz(&label("mw_finish"));
&jmp(&label("mw_loop"));
&set_label("mw_finish",0);
&mov($num,&wparam(2)); # get num
&and($num,7);
&jnz(&label("mw_finish2"));
&jmp(&label("mw_end"));
&set_label("mw_finish2",1);
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov("eax",&DWP($i*4,$a,"",0));# *a
&mul($w); # *a * w
&add("eax",$c); # L(t)+=c
# XXX
&adc("edx",0); # H(t)+=carry
&mov(&DWP($i*4,$r,"",0),"eax");# *r= L(t);
&mov($c,"edx"); # c= H(t);
&dec($num) if ($i != 7-1);
&jz(&label("mw_end")) if ($i != 7-1);
}
&set_label("mw_end",0);
&mov("eax",$c);
&function_end($name);
}
1;

87
drivers/builtin_openssl2/crypto/bn/asm/x86/mul_add.pl
View File

@ -1,87 +0,0 @@
#!/usr/local/bin/perl
# x86 assember
sub bn_mul_add_words
{
local($name)=@_;
&function_begin($name,"");
&comment("");
$Low="eax";
$High="edx";
$a="ebx";
$w="ebp";
$r="edi";
$c="esi";
&xor($c,$c); # clear carry
&mov($r,&wparam(0)); #
&mov("ecx",&wparam(2)); #
&mov($a,&wparam(1)); #
&and("ecx",0xfffffff8); # num / 8
&mov($w,&wparam(3)); #
&push("ecx"); # Up the stack for a tmp variable
&jz(&label("maw_finish"));
&set_label("maw_loop",0);
&mov(&swtmp(0),"ecx"); #
for ($i=0; $i<32; $i+=4)
{
&comment("Round $i");
&mov("eax",&DWP($i,$a,"",0)); # *a
&mul($w); # *a * w
&add("eax",$c); # L(t)+= *r
&mov($c,&DWP($i,$r,"",0)); # L(t)+= *r
&adc("edx",0); # H(t)+=carry
&add("eax",$c); # L(t)+=c
&adc("edx",0); # H(t)+=carry
&mov(&DWP($i,$r,"",0),"eax"); # *r= L(t);
&mov($c,"edx"); # c= H(t);
}
&comment("");
&mov("ecx",&swtmp(0)); #
&add($a,32);
&add($r,32);
&sub("ecx",8);
&jnz(&label("maw_loop"));
&set_label("maw_finish",0);
&mov("ecx",&wparam(2)); # get num
&and("ecx",7);
&jnz(&label("maw_finish2")); # helps branch prediction
&jmp(&label("maw_end"));
&set_label("maw_finish2",1);
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov("eax",&DWP($i*4,$a,"",0));# *a
&mul($w); # *a * w
&add("eax",$c); # L(t)+=c
&mov($c,&DWP($i*4,$r,"",0)); # L(t)+= *r
&adc("edx",0); # H(t)+=carry
&add("eax",$c);
&adc("edx",0); # H(t)+=carry
&dec("ecx") if ($i != 7-1);
&mov(&DWP($i*4,$r,"",0),"eax"); # *r= L(t);
&mov($c,"edx"); # c= H(t);
&jz(&label("maw_end")) if ($i != 7-1);
}
&set_label("maw_end",0);
&mov("eax",$c);
&pop("ecx"); # clear variable from
&function_end($name);
}
1;

60
drivers/builtin_openssl2/crypto/bn/asm/x86/sqr.pl
View File

@ -1,60 +0,0 @@
#!/usr/local/bin/perl
# x86 assember
sub bn_sqr_words
{
local($name)=@_;
&function_begin($name,"");
&comment("");
$r="esi";
$a="edi";
$num="ebx";
&mov($r,&wparam(0)); #
&mov($a,&wparam(1)); #
&mov($num,&wparam(2)); #
&and($num,0xfffffff8); # num / 8
&jz(&label("sw_finish"));
&set_label("sw_loop",0);
for ($i=0; $i<32; $i+=4)
{
&comment("Round $i");
&mov("eax",&DWP($i,$a,"",0)); # *a
# XXX
&mul("eax"); # *a * *a
&mov(&DWP($i*2,$r,"",0),"eax"); #
&mov(&DWP($i*2+4,$r,"",0),"edx");#
}
&comment("");
&add($a,32);
&add($r,64);
&sub($num,8);
&jnz(&label("sw_loop"));
&set_label("sw_finish",0);
&mov($num,&wparam(2)); # get num
&and($num,7);
&jz(&label("sw_end"));
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov("eax",&DWP($i*4,$a,"",0)); # *a
# XXX
&mul("eax"); # *a * *a
&mov(&DWP($i*8,$r,"",0),"eax"); #
&dec($num) if ($i != 7-1);
&mov(&DWP($i*8+4,$r,"",0),"edx");
&jz(&label("sw_end")) if ($i != 7-1);
}
&set_label("sw_end",0);
&function_end($name);
}
1;

76
drivers/builtin_openssl2/crypto/bn/asm/x86/sub.pl
View File

@ -1,76 +0,0 @@
#!/usr/local/bin/perl
# x86 assember
sub bn_sub_words
{
local($name)=@_;
&function_begin($name,"");
&comment("");
$a="esi";
$b="edi";
$c="eax";
$r="ebx";
$tmp1="ecx";
$tmp2="edx";
$num="ebp";
&mov($r,&wparam(0)); # get r
&mov($a,&wparam(1)); # get a
&mov($b,&wparam(2)); # get b
&mov($num,&wparam(3)); # get num
&xor($c,$c); # clear carry
&and($num,0xfffffff8); # num / 8
&jz(&label("aw_finish"));
&set_label("aw_loop",0);
for ($i=0; $i<8; $i++)
{
&comment("Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov($tmp2,&DWP($i*4,$b,"",0)); # *b
&sub($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&sub($tmp1,$tmp2);
&adc($c,0);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *r
}
&comment("");
&add($a,32);
&add($b,32);
&add($r,32);
&sub($num,8);
&jnz(&label("aw_loop"));
&set_label("aw_finish",0);
&mov($num,&wparam(3)); # get num
&and($num,7);
&jz(&label("aw_end"));
for ($i=0; $i<7; $i++)
{
&comment("Tail Round $i");
&mov($tmp1,&DWP($i*4,$a,"",0)); # *a
&mov($tmp2,&DWP($i*4,$b,"",0));# *b
&sub($tmp1,$c);
&mov($c,0);
&adc($c,$c);
&sub($tmp1,$tmp2);
&adc($c,0);
&dec($num) if ($i != 6);
&mov(&DWP($i*4,$r,"",0),$tmp1); # *a
&jz(&label("aw_end")) if ($i != 6);
}
&set_label("aw_end",0);
# &mov("eax",$c); # $c is "eax"
&function_end($name);
}
1;

108
drivers/builtin_openssl2/crypto/bn/asm/x86_64-gcc.c
View File

@ -55,7 +55,7 @@
* machine.
*/
# ifdef _WIN64
# if defined(_WIN64) || !defined(__LP64__)
# define BN_ULONG unsigned long long
# else
# define BN_ULONG unsigned long
@ -63,7 +63,6 @@
# undef mul
# undef mul_add
# undef sqr
/*-
* "m"(a), "+m"(r) is the way to favor DirectPath µ-code;
@ -99,8 +98,8 @@
: "cc"); \
(r)=carry, carry=high; \
} while (0)
# define sqr(r0,r1,a) \
# undef sqr
# define sqr(r0,r1,a) \
asm ("mulq %2" \
: "=a"(r0),"=d"(r1) \
: "a"(a) \
@ -204,20 +203,22 @@ BN_ULONG bn_div_words(BN_ULONG h, BN_ULONG l, BN_ULONG d)
BN_ULONG bn_add_words(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
int n)
{
BN_ULONG ret = 0, i = 0;
BN_ULONG ret;
size_t i = 0;
if (n <= 0)
return 0;
asm volatile (" subq %2,%2 \n"
asm volatile (" subq %0,%0 \n" /* clear carry */
" jmp 1f \n"
".p2align 4 \n"
"1: movq (%4,%2,8),%0 \n"
" adcq (%5,%2,8),%0 \n"
" movq %0,(%3,%2,8) \n"
" leaq 1(%2),%2 \n"
" lea 1(%2),%2 \n"
" loop 1b \n"
" sbbq %0,%0 \n":"=&a" (ret), "+c"(n),
"=&r"(i)
" sbbq %0,%0 \n":"=&r" (ret), "+c"(n),
"+r"(i)
:"r"(rp), "r"(ap), "r"(bp)
:"cc", "memory");
@ -228,20 +229,22 @@ BN_ULONG bn_add_words(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
BN_ULONG bn_sub_words(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp,
int n)
{
BN_ULONG ret = 0, i = 0;
BN_ULONG ret;
size_t i = 0;
if (n <= 0)
return 0;
asm volatile (" subq %2,%2 \n"
asm volatile (" subq %0,%0 \n" /* clear borrow */
" jmp 1f \n"
".p2align 4 \n"
"1: movq (%4,%2,8),%0 \n"
" sbbq (%5,%2,8),%0 \n"
" movq %0,(%3,%2,8) \n"
" leaq 1(%2),%2 \n"
" lea 1(%2),%2 \n"
" loop 1b \n"
" sbbq %0,%0 \n":"=&a" (ret), "+c"(n),
"=&r"(i)
" sbbq %0,%0 \n":"=&r" (ret), "+c"(n),
"+r"(i)
:"r"(rp), "r"(ap), "r"(bp)
:"cc", "memory");
@ -313,55 +316,58 @@ BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
*/
# if 0
/* original macros are kept for reference purposes */
# define mul_add_c(a,b,c0,c1,c2) { \
BN_ULONG ta=(a),tb=(b); \
t1 = ta * tb; \
t2 = BN_UMULT_HIGH(ta,tb); \
c0 += t1; t2 += (c0<t1)?1:0; \
c1 += t2; c2 += (c1<t2)?1:0; \
}
# define mul_add_c(a,b,c0,c1,c2) do { \
BN_ULONG ta = (a), tb = (b); \
BN_ULONG lo, hi; \
BN_UMULT_LOHI(lo,hi,ta,tb); \
c0 += lo; hi += (c0<lo)?1:0; \
c1 += hi; c2 += (c1<hi)?1:0; \
} while(0)
# define mul_add_c2(a,b,c0,c1,c2) { \
BN_ULONG ta=(a),tb=(b),t0; \
t1 = BN_UMULT_HIGH(ta,tb); \
t0 = ta * tb; \
c0 += t0; t2 = t1+((c0<t0)?1:0);\
c1 += t2; c2 += (c1<t2)?1:0; \
c0 += t0; t1 += (c0<t0)?1:0; \
c1 += t1; c2 += (c1<t1)?1:0; \
}
# define mul_add_c2(a,b,c0,c1,c2) do { \
BN_ULONG ta = (a), tb = (b); \
BN_ULONG lo, hi, tt; \
BN_UMULT_LOHI(lo,hi,ta,tb); \
c0 += lo; tt = hi+((c0<lo)?1:0); \
c1 += tt; c2 += (c1<tt)?1:0; \
c0 += lo; hi += (c0<lo)?1:0; \
c1 += hi; c2 += (c1<hi)?1:0; \
} while(0)
# define sqr_add_c(a,i,c0,c1,c2) do { \
BN_ULONG ta = (a)[i]; \
BN_ULONG lo, hi; \
BN_UMULT_LOHI(lo,hi,ta,ta); \
c0 += lo; hi += (c0<lo)?1:0; \
c1 += hi; c2 += (c1<hi)?1:0; \
} while(0)
# else
# define mul_add_c(a,b,c0,c1,c2) do { \
# define mul_add_c(a,b,c0,c1,c2) do { \
BN_ULONG t1,t2; \
asm ("mulq %3" \
: "=a"(t1),"=d"(t2) \
: "a"(a),"m"(b) \
: "cc"); \
asm ("addq %2,%0; adcq %3,%1" \
: "+r"(c0),"+d"(t2) \
: "a"(t1),"g"(0) \
: "cc"); \
asm ("addq %2,%0; adcq %3,%1" \
: "+r"(c1),"+r"(c2) \
: "d"(t2),"g"(0) \
: "cc"); \
asm ("addq %3,%0; adcq %4,%1; adcq %5,%2" \
: "+r"(c0),"+r"(c1),"+r"(c2) \
: "r"(t1),"r"(t2),"g"(0) \
: "cc"); \
} while (0)
# define sqr_add_c(a,i,c0,c1,c2) do { \
# define sqr_add_c(a,i,c0,c1,c2) do { \
BN_ULONG t1,t2; \
asm ("mulq %2" \
: "=a"(t1),"=d"(t2) \
: "a"(a[i]) \
: "cc"); \
asm ("addq %2,%0; adcq %3,%1" \
: "+r"(c0),"+d"(t2) \
: "a"(t1),"g"(0) \
: "cc"); \
asm ("addq %2,%0; adcq %3,%1" \
: "+r"(c1),"+r"(c2) \
: "d"(t2),"g"(0) \
: "cc"); \
asm ("addq %3,%0; adcq %4,%1; adcq %5,%2" \
: "+r"(c0),"+r"(c1),"+r"(c2) \
: "r"(t1),"r"(t2),"g"(0) \
: "cc"); \
} while (0)
# define mul_add_c2(a,b,c0,c1,c2) do { \
# define mul_add_c2(a,b,c0,c1,c2) do { \
BN_ULONG t1,t2; \
asm ("mulq %3" \
: "=a"(t1),"=d"(t2) \
: "a"(a),"m"(b) \
@ -382,7 +388,6 @@ BN_ULONG bn_sub_words(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n)
void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
{
BN_ULONG t1, t2;
BN_ULONG c1, c2, c3;
c1 = 0;
@ -486,7 +491,6 @@ void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
{
BN_ULONG t1, t2;
BN_ULONG c1, c2, c3;
c1 = 0;
@ -526,7 +530,6 @@ void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
{
BN_ULONG t1, t2;
BN_ULONG c1, c2, c3;
c1 = 0;
@ -602,7 +605,6 @@ void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
{
BN_ULONG t1, t2;
BN_ULONG c1, c2, c3;
c1 = 0;

390
drivers/builtin_openssl2/crypto/bn/asm/x86_64-gf2m.pl
View File

@ -1,390 +0,0 @@
#!/usr/bin/env perl
#
# ====================================================================
# Written by Andy Polyakov <appro@openssl.org> for the OpenSSL
# project. The module is, however, dual licensed under OpenSSL and
# CRYPTOGAMS licenses depending on where you obtain it. For further
# details see http://www.openssl.org/~appro/cryptogams/.
# ====================================================================
#
# May 2011
#
# The module implements bn_GF2m_mul_2x2 polynomial multiplication used
# in bn_gf2m.c. It's kind of low-hanging mechanical port from C for
# the time being... Except that it has two code paths: code suitable
# for any x86_64 CPU and PCLMULQDQ one suitable for Westmere and
# later. Improvement varies from one benchmark and µ-arch to another.
# Vanilla code path is at most 20% faster than compiler-generated code
# [not very impressive], while PCLMULQDQ - whole 85%-160% better on
# 163- and 571-bit ECDH benchmarks on Intel CPUs. Keep in mind that
# these coefficients are not ones for bn_GF2m_mul_2x2 itself, as not
# all CPU time is burnt in it...
$flavour = shift;
$output = shift;
if ($flavour =~ /\./) { $output = $flavour; undef $flavour; }
$win64=0; $win64=1 if ($flavour =~ /[nm]asm|mingw64/ || $output =~ /\.asm$/);
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
( $xlate="${dir}x86_64-xlate.pl" and -f $xlate ) or
( $xlate="${dir}../../perlasm/x86_64-xlate.pl" and -f $xlate) or
die "can't locate x86_64-xlate.pl";
open OUT,"| \"$^X\" $xlate $flavour $output";
*STDOUT=*OUT;
($lo,$hi)=("%rax","%rdx"); $a=$lo;
($i0,$i1)=("%rsi","%rdi");
($t0,$t1)=("%rbx","%rcx");
($b,$mask)=("%rbp","%r8");
($a1,$a2,$a4,$a8,$a12,$a48)=map("%r$_",(9..15));
($R,$Tx)=("%xmm0","%xmm1");
$code.=<<___;
.text
.type _mul_1x1,\@abi-omnipotent
.align 16
_mul_1x1:
sub \$128+8,%rsp
mov \$-1,$a1
lea ($a,$a),$i0
shr \$3,$a1
lea (,$a,4),$i1
and $a,$a1 # a1=a&0x1fffffffffffffff
lea (,$a,8),$a8
sar \$63,$a # broadcast 63rd bit
lea ($a1,$a1),$a2
sar \$63,$i0 # broadcast 62nd bit
lea (,$a1,4),$a4
and $b,$a
sar \$63,$i1 # boardcast 61st bit
mov $a,$hi # $a is $lo
shl \$63,$lo
and $b,$i0
shr \$1,$hi
mov $i0,$t1
shl \$62,$i0
and $b,$i1
shr \$2,$t1
xor $i0,$lo
mov $i1,$t0
shl \$61,$i1
xor $t1,$hi
shr \$3,$t0
xor $i1,$lo
xor $t0,$hi
mov $a1,$a12
movq \$0,0(%rsp) # tab[0]=0
xor $a2,$a12 # a1^a2
mov $a1,8(%rsp) # tab[1]=a1
mov $a4,$a48
mov $a2,16(%rsp) # tab[2]=a2
xor $a8,$a48 # a4^a8
mov $a12,24(%rsp) # tab[3]=a1^a2
xor $a4,$a1
mov $a4,32(%rsp) # tab[4]=a4
xor $a4,$a2
mov $a1,40(%rsp) # tab[5]=a1^a4
xor $a4,$a12
mov $a2,48(%rsp) # tab[6]=a2^a4
xor $a48,$a1 # a1^a4^a4^a8=a1^a8
mov $a12,56(%rsp) # tab[7]=a1^a2^a4
xor $a48,$a2 # a2^a4^a4^a8=a1^a8
mov $a8,64(%rsp) # tab[8]=a8
xor $a48,$a12 # a1^a2^a4^a4^a8=a1^a2^a8
mov $a1,72(%rsp) # tab[9]=a1^a8
xor $a4,$a1 # a1^a8^a4
mov $a2,80(%rsp) # tab[10]=a2^a8
xor $a4,$a2 # a2^a8^a4
mov $a12,88(%rsp) # tab[11]=a1^a2^a8
xor $a4,$a12 # a1^a2^a8^a4
mov $a48,96(%rsp) # tab[12]=a4^a8
mov $mask,$i0
mov $a1,104(%rsp) # tab[13]=a1^a4^a8
and $b,$i0
mov $a2,112(%rsp) # tab[14]=a2^a4^a8
shr \$4,$b
mov $a12,120(%rsp) # tab[15]=a1^a2^a4^a8
mov $mask,$i1
and $b,$i1
shr \$4,$b
movq (%rsp,$i0,8),$R # half of calculations is done in SSE2
mov $mask,$i0
and $b,$i0
shr \$4,$b
___
for ($n=1;$n<8;$n++) {
$code.=<<___;
mov (%rsp,$i1,8),$t1
mov $mask,$i1
mov $t1,$t0
shl \$`8*$n-4`,$t1
and $b,$i1
movq (%rsp,$i0,8),$Tx
shr \$`64-(8*$n-4)`,$t0
xor $t1,$lo
pslldq \$$n,$Tx
mov $mask,$i0
shr \$4,$b
xor $t0,$hi
and $b,$i0
shr \$4,$b
pxor $Tx,$R
___
}
$code.=<<___;
mov (%rsp,$i1,8),$t1
mov $t1,$t0
shl \$`8*$n-4`,$t1
movq $R,$i0
shr \$`64-(8*$n-4)`,$t0
xor $t1,$lo
psrldq \$8,$R
xor $t0,$hi
movq $R,$i1
xor $i0,$lo
xor $i1,$hi
add \$128+8,%rsp
ret
.Lend_mul_1x1:
.size _mul_1x1,.-_mul_1x1
___
($rp,$a1,$a0,$b1,$b0) = $win64? ("%rcx","%rdx","%r8", "%r9","%r10") : # Win64 order
("%rdi","%rsi","%rdx","%rcx","%r8"); # Unix order
$code.=<<___;
.extern OPENSSL_ia32cap_P
.globl bn_GF2m_mul_2x2
.type bn_GF2m_mul_2x2,\@abi-omnipotent
.align 16
bn_GF2m_mul_2x2:
mov OPENSSL_ia32cap_P(%rip),%rax
bt \$33,%rax
jnc .Lvanilla_mul_2x2
movq $a1,%xmm0
movq $b1,%xmm1
movq $a0,%xmm2
___
$code.=<<___ if ($win64);
movq 40(%rsp),%xmm3
___
$code.=<<___ if (!$win64);
movq $b0,%xmm3
___
$code.=<<___;
movdqa %xmm0,%xmm4
movdqa %xmm1,%xmm5
pclmulqdq \$0,%xmm1,%xmm0 # a1·b1
pxor %xmm2,%xmm4
pxor %xmm3,%xmm5
pclmulqdq \$0,%xmm3,%xmm2 # a0·b0
pclmulqdq \$0,%xmm5,%xmm4 # (a0+a1)·(b0+b1)
xorps %xmm0,%xmm4
xorps %xmm2,%xmm4 # (a0+a1)·(b0+b1)-a0·b0-a1·b1
movdqa %xmm4,%xmm5
pslldq \$8,%xmm4
psrldq \$8,%xmm5
pxor %xmm4,%xmm2
pxor %xmm5,%xmm0
movdqu %xmm2,0($rp)
movdqu %xmm0,16($rp)
ret
.align 16
.Lvanilla_mul_2x2:
lea -8*17(%rsp),%rsp
___
$code.=<<___ if ($win64);
mov `8*17+40`(%rsp),$b0
mov %rdi,8*15(%rsp)
mov %rsi,8*16(%rsp)
___
$code.=<<___;
mov %r14,8*10(%rsp)
mov %r13,8*11(%rsp)
mov %r12,8*12(%rsp)
mov %rbp,8*13(%rsp)
mov %rbx,8*14(%rsp)
.Lbody_mul_2x2:
mov $rp,32(%rsp) # save the arguments
mov $a1,40(%rsp)
mov $a0,48(%rsp)
mov $b1,56(%rsp)
mov $b0,64(%rsp)
mov \$0xf,$mask
mov $a1,$a
mov $b1,$b
call _mul_1x1 # a1·b1
mov $lo,16(%rsp)
mov $hi,24(%rsp)
mov 48(%rsp),$a
mov 64(%rsp),$b
call _mul_1x1 # a0·b0
mov $lo,0(%rsp)
mov $hi,8(%rsp)
mov 40(%rsp),$a
mov 56(%rsp),$b
xor 48(%rsp),$a
xor 64(%rsp),$b
call _mul_1x1 # (a0+a1)·(b0+b1)
___
@r=("%rbx","%rcx","%rdi","%rsi");
$code.=<<___;
mov 0(%rsp),@r[0]
mov 8(%rsp),@r[1]
mov 16(%rsp),@r[2]
mov 24(%rsp),@r[3]
mov 32(%rsp),%rbp
xor $hi,$lo
xor @r[1],$hi
xor @r[0],$lo
mov @r[0],0(%rbp)
xor @r[2],$hi
mov @r[3],24(%rbp)
xor @r[3],$lo
xor @r[3],$hi
xor $hi,$lo
mov $hi,16(%rbp)
mov $lo,8(%rbp)
mov 8*10(%rsp),%r14
mov 8*11(%rsp),%r13
mov 8*12(%rsp),%r12
mov 8*13(%rsp),%rbp
mov 8*14(%rsp),%rbx
___
$code.=<<___ if ($win64);
mov 8*15(%rsp),%rdi
mov 8*16(%rsp),%rsi
___
$code.=<<___;
lea 8*17(%rsp),%rsp
ret
.Lend_mul_2x2:
.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2
.asciz "GF(2^m) Multiplication for x86_64, CRYPTOGAMS by <appro\@openssl.org>"
.align 16
___
# EXCEPTION_DISPOSITION handler (EXCEPTION_RECORD *rec,ULONG64 frame,
# CONTEXT *context,DISPATCHER_CONTEXT *disp)
if ($win64) {
$rec="%rcx";
$frame="%rdx";
$context="%r8";
$disp="%r9";
$code.=<<___;
.extern __imp_RtlVirtualUnwind
.type se_handler,\@abi-omnipotent
.align 16
se_handler:
push %rsi
push %rdi
push %rbx
push %rbp
push %r12
push %r13
push %r14
push %r15
pushfq
sub \$64,%rsp
mov 152($context),%rax # pull context->Rsp
mov 248($context),%rbx # pull context->Rip
lea .Lbody_mul_2x2(%rip),%r10
cmp %r10,%rbx # context->Rip<"prologue" label
jb .Lin_prologue
mov 8*10(%rax),%r14 # mimic epilogue
mov 8*11(%rax),%r13
mov 8*12(%rax),%r12
mov 8*13(%rax),%rbp
mov 8*14(%rax),%rbx
mov 8*15(%rax),%rdi
mov 8*16(%rax),%rsi
mov %rbx,144($context) # restore context->Rbx
mov %rbp,160($context) # restore context->Rbp
mov %rsi,168($context) # restore context->Rsi
mov %rdi,176($context) # restore context->Rdi
mov %r12,216($context) # restore context->R12
mov %r13,224($context) # restore context->R13
mov %r14,232($context) # restore context->R14
.Lin_prologue:
lea 8*17(%rax),%rax
mov %rax,152($context) # restore context->Rsp
mov 40($disp),%rdi # disp->ContextRecord
mov $context,%rsi # context
mov \$154,%ecx # sizeof(CONTEXT)
.long 0xa548f3fc # cld; rep movsq
mov $disp,%rsi
xor %rcx,%rcx # arg1, UNW_FLAG_NHANDLER
mov 8(%rsi),%rdx # arg2, disp->ImageBase
mov 0(%rsi),%r8 # arg3, disp->ControlPc
mov 16(%rsi),%r9 # arg4, disp->FunctionEntry
mov 40(%rsi),%r10 # disp->ContextRecord
lea 56(%rsi),%r11 # &disp->HandlerData
lea 24(%rsi),%r12 # &disp->EstablisherFrame
mov %r10,32(%rsp) # arg5
mov %r11,40(%rsp) # arg6
mov %r12,48(%rsp) # arg7
mov %rcx,56(%rsp) # arg8, (NULL)
call *__imp_RtlVirtualUnwind(%rip)
mov \$1,%eax # ExceptionContinueSearch
add \$64,%rsp
popfq
pop %r15
pop %r14
pop %r13
pop %r12
pop %rbp
pop %rbx
pop %rdi
pop %rsi
ret
.size se_handler,.-se_handler
.section .pdata
.align 4
.rva _mul_1x1
.rva .Lend_mul_1x1
.rva .LSEH_info_1x1
.rva .Lvanilla_mul_2x2
.rva .Lend_mul_2x2
.rva .LSEH_info_2x2
.section .xdata
.align 8
.LSEH_info_1x1:
.byte 0x01,0x07,0x02,0x00
.byte 0x07,0x01,0x11,0x00 # sub rsp,128+8
.LSEH_info_2x2:
.byte 9,0,0,0
.rva se_handler
___
}
$code =~ s/\`([^\`]*)\`/eval($1)/gem;
print $code;
close STDOUT;

1681
drivers/builtin_openssl2/crypto/bn/asm/x86_64-mont.pl
View File

File diff suppressed because it is too large Load Diff

1186
drivers/builtin_openssl2/crypto/bn/asm/x86_64-mont5.pl
View File

File diff suppressed because it is too large Load Diff

229
drivers/builtin_openssl2/crypto/bn/bn_asm.c
View File

@ -489,121 +489,144 @@ BN_ULONG bn_sub_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
* c=(c2,c1,c0)
*/
/*
* Keep in mind that carrying into high part of multiplication result
* can not overflow, because it cannot be all-ones.
*/
# ifdef BN_LLONG
# define mul_add_c(a,b,c0,c1,c2) \
t=(BN_ULLONG)a*b; \
t1=(BN_ULONG)Lw(t); \
t2=(BN_ULONG)Hw(t); \
c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
/*
* Keep in mind that additions to multiplication result can not
* overflow, because its high half cannot be all-ones.
*/
# define mul_add_c(a,b,c0,c1,c2) do { \
BN_ULONG hi; \
BN_ULLONG t = (BN_ULLONG)(a)*(b); \
t += c0; /* no carry */ \
c0 = (BN_ULONG)Lw(t); \
hi = (BN_ULONG)Hw(t); \
c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
} while(0)
# define mul_add_c2(a,b,c0,c1,c2) \
t=(BN_ULLONG)a*b; \
tt=(t+t)&BN_MASK; \
if (tt < t) c2++; \
t1=(BN_ULONG)Lw(tt); \
t2=(BN_ULONG)Hw(tt); \
c0=(c0+t1)&BN_MASK2; \
if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
# define mul_add_c2(a,b,c0,c1,c2) do { \
BN_ULONG hi; \
BN_ULLONG t = (BN_ULLONG)(a)*(b); \
BN_ULLONG tt = t+c0; /* no carry */ \
c0 = (BN_ULONG)Lw(tt); \
hi = (BN_ULONG)Hw(tt); \
c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
t += c0; /* no carry */ \
c0 = (BN_ULONG)Lw(t); \
hi = (BN_ULONG)Hw(t); \
c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
} while(0)
# define sqr_add_c(a,i,c0,c1,c2) \
t=(BN_ULLONG)a[i]*a[i]; \
t1=(BN_ULONG)Lw(t); \
t2=(BN_ULONG)Hw(t); \
c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
# define sqr_add_c(a,i,c0,c1,c2) do { \
BN_ULONG hi; \
BN_ULLONG t = (BN_ULLONG)a[i]*a[i]; \
t += c0; /* no carry */ \
c0 = (BN_ULONG)Lw(t); \
hi = (BN_ULONG)Hw(t); \
c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
} while(0)
# define sqr_add_c2(a,i,j,c0,c1,c2) \
mul_add_c2((a)[i],(a)[j],c0,c1,c2)
# elif defined(BN_UMULT_LOHI)
/*
* Keep in mind that additions to hi can not overflow, because
* the high word of a multiplication result cannot be all-ones.
*/
# define mul_add_c(a,b,c0,c1,c2) do { \
BN_ULONG ta = (a), tb = (b); \
BN_ULONG lo, hi; \
BN_UMULT_LOHI(lo,hi,ta,tb); \
c0 += lo; hi += (c0<lo)?1:0; \
c1 += hi; c2 += (c1<hi)?1:0; \
} while(0)
# define mul_add_c(a,b,c0,c1,c2) { \
BN_ULONG ta=(a),tb=(b); \
BN_UMULT_LOHI(t1,t2,ta,tb); \
c0 += t1; t2 += (c0<t1)?1:0; \
c1 += t2; c2 += (c1<t2)?1:0; \
}
# define mul_add_c2(a,b,c0,c1,c2) do { \
BN_ULONG ta = (a), tb = (b); \
BN_ULONG lo, hi, tt; \
BN_UMULT_LOHI(lo,hi,ta,tb); \
c0 += lo; tt = hi+((c0<lo)?1:0); \
c1 += tt; c2 += (c1<tt)?1:0; \
c0 += lo; hi += (c0<lo)?1:0; \
c1 += hi; c2 += (c1<hi)?1:0; \
} while(0)
# define mul_add_c2(a,b,c0,c1,c2) { \
BN_ULONG ta=(a),tb=(b),t0; \
BN_UMULT_LOHI(t0,t1,ta,tb); \
c0 += t0; t2 = t1+((c0<t0)?1:0);\
c1 += t2; c2 += (c1<t2)?1:0; \
c0 += t0; t1 += (c0<t0)?1:0; \
c1 += t1; c2 += (c1<t1)?1:0; \
}
# define sqr_add_c(a,i,c0,c1,c2) { \
BN_ULONG ta=(a)[i]; \
BN_UMULT_LOHI(t1,t2,ta,ta); \
c0 += t1; t2 += (c0<t1)?1:0; \
c1 += t2; c2 += (c1<t2)?1:0; \
}
# define sqr_add_c(a,i,c0,c1,c2) do { \
BN_ULONG ta = (a)[i]; \
BN_ULONG lo, hi; \
BN_UMULT_LOHI(lo,hi,ta,ta); \
c0 += lo; hi += (c0<lo)?1:0; \
c1 += hi; c2 += (c1<hi)?1:0; \
} while(0)
# define sqr_add_c2(a,i,j,c0,c1,c2) \
mul_add_c2((a)[i],(a)[j],c0,c1,c2)
# elif defined(BN_UMULT_HIGH)
/*
* Keep in mind that additions to hi can not overflow, because
* the high word of a multiplication result cannot be all-ones.
*/
# define mul_add_c(a,b,c0,c1,c2) do { \
BN_ULONG ta = (a), tb = (b); \
BN_ULONG lo = ta * tb; \
BN_ULONG hi = BN_UMULT_HIGH(ta,tb); \
c0 += lo; hi += (c0<lo)?1:0; \
c1 += hi; c2 += (c1<hi)?1:0; \
} while(0)
# define mul_add_c(a,b,c0,c1,c2) { \
BN_ULONG ta=(a),tb=(b); \
t1 = ta * tb; \
t2 = BN_UMULT_HIGH(ta,tb); \
c0 += t1; t2 += (c0<t1)?1:0; \
c1 += t2; c2 += (c1<t2)?1:0; \
}
# define mul_add_c2(a,b,c0,c1,c2) do { \
BN_ULONG ta = (a), tb = (b), tt; \
BN_ULONG lo = ta * tb; \
BN_ULONG hi = BN_UMULT_HIGH(ta,tb); \
c0 += lo; tt = hi + ((c0<lo)?1:0); \
c1 += tt; c2 += (c1<tt)?1:0; \
c0 += lo; hi += (c0<lo)?1:0; \
c1 += hi; c2 += (c1<hi)?1:0; \
} while(0)
# define mul_add_c2(a,b,c0,c1,c2) { \
BN_ULONG ta=(a),tb=(b),t0; \
t1 = BN_UMULT_HIGH(ta,tb); \
t0 = ta * tb; \
c0 += t0; t2 = t1+((c0<t0)?1:0);\
c1 += t2; c2 += (c1<t2)?1:0; \
c0 += t0; t1 += (c0<t0)?1:0; \
c1 += t1; c2 += (c1<t1)?1:0; \
}
# define sqr_add_c(a,i,c0,c1,c2) { \
BN_ULONG ta=(a)[i]; \
t1 = ta * ta; \
t2 = BN_UMULT_HIGH(ta,ta); \
c0 += t1; t2 += (c0<t1)?1:0; \
c1 += t2; c2 += (c1<t2)?1:0; \
}
# define sqr_add_c(a,i,c0,c1,c2) do { \
BN_ULONG ta = (a)[i]; \
BN_ULONG lo = ta * ta; \
BN_ULONG hi = BN_UMULT_HIGH(ta,ta); \
c0 += lo; hi += (c0<lo)?1:0; \
c1 += hi; c2 += (c1<hi)?1:0; \
} while(0)
# define sqr_add_c2(a,i,j,c0,c1,c2) \
mul_add_c2((a)[i],(a)[j],c0,c1,c2)
# else /* !BN_LLONG */
# define mul_add_c(a,b,c0,c1,c2) \
t1=LBITS(a); t2=HBITS(a); \
bl=LBITS(b); bh=HBITS(b); \
mul64(t1,t2,bl,bh); \
c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
/*
* Keep in mind that additions to hi can not overflow, because
* the high word of a multiplication result cannot be all-ones.
*/
# define mul_add_c(a,b,c0,c1,c2) do { \
BN_ULONG lo = LBITS(a), hi = HBITS(a); \
BN_ULONG bl = LBITS(b), bh = HBITS(b); \
mul64(lo,hi,bl,bh); \
c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
} while(0)
# define mul_add_c2(a,b,c0,c1,c2) \
t1=LBITS(a); t2=HBITS(a); \
bl=LBITS(b); bh=HBITS(b); \
mul64(t1,t2,bl,bh); \
if (t2 & BN_TBIT) c2++; \
t2=(t2+t2)&BN_MASK2; \
if (t1 & BN_TBIT) t2++; \
t1=(t1+t1)&BN_MASK2; \
c0=(c0+t1)&BN_MASK2; \
if ((c0 < t1) && (((++t2)&BN_MASK2) == 0)) c2++; \
c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
# define mul_add_c2(a,b,c0,c1,c2) do { \
BN_ULONG tt; \
BN_ULONG lo = LBITS(a), hi = HBITS(a); \
BN_ULONG bl = LBITS(b), bh = HBITS(b); \
mul64(lo,hi,bl,bh); \
tt = hi; \
c0 = (c0+lo)&BN_MASK2; if (c0<lo) tt++; \
c1 = (c1+tt)&BN_MASK2; if (c1<tt) c2++; \
c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
} while(0)
# define sqr_add_c(a,i,c0,c1,c2) \
sqr64(t1,t2,(a)[i]); \
c0=(c0+t1)&BN_MASK2; if ((c0) < t1) t2++; \
c1=(c1+t2)&BN_MASK2; if ((c1) < t2) c2++;
# define sqr_add_c(a,i,c0,c1,c2) do { \
BN_ULONG lo, hi; \
sqr64(lo,hi,(a)[i]); \
c0 = (c0+lo)&BN_MASK2; if (c0<lo) hi++; \
c1 = (c1+hi)&BN_MASK2; if (c1<hi) c2++; \
} while(0)
# define sqr_add_c2(a,i,j,c0,c1,c2) \
mul_add_c2((a)[i],(a)[j],c0,c1,c2)
@ -611,12 +634,6 @@ BN_ULONG bn_sub_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b,
void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
{
# ifdef BN_LLONG
BN_ULLONG t;
# else
BN_ULONG bl, bh;
# endif
BN_ULONG t1, t2;
BN_ULONG c1, c2, c3;
c1 = 0;
@ -720,12 +737,6 @@ void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
{
# ifdef BN_LLONG
BN_ULLONG t;
# else
BN_ULONG bl, bh;
# endif
BN_ULONG t1, t2;
BN_ULONG c1, c2, c3;
c1 = 0;
@ -765,12 +776,6 @@ void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b)
void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
{
# ifdef BN_LLONG
BN_ULLONG t, tt;
# else
BN_ULONG bl, bh;
# endif
BN_ULONG t1, t2;
BN_ULONG c1, c2, c3;
c1 = 0;
@ -846,12 +851,6 @@ void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a)
void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a)
{
# ifdef BN_LLONG
BN_ULLONG t, tt;
# else
BN_ULONG bl, bh;
# endif
BN_ULONG t1, t2;
BN_ULONG c1, c2, c3;
c1 = 0;

0
drivers/builtin_openssl2/crypto/bn/bn_const.c Executable file → Normal file
View File

320
drivers/builtin_openssl2/crypto/bn/bn_exp.c
View File

@ -123,6 +123,17 @@
# ifndef alloca
# define alloca(s) __builtin_alloca((s))
# endif
#elif defined(__sun)
# include <alloca.h>
#endif
#include "rsaz_exp.h"
#undef SPARC_T4_MONT
#if defined(OPENSSL_BN_ASM_MONT) && (defined(__sparc__) || defined(__sparc))
# include "sparc_arch.h"
extern unsigned int OPENSSL_sparcv9cap_P[];
# define SPARC_T4_MONT
#endif
/* maximum precomputation table size for *variable* sliding windows */
@ -476,6 +487,23 @@ int BN_mod_exp_mont(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
wstart = bits - 1; /* The top bit of the window */
wend = 0; /* The bottom bit of the window */
#if 1 /* by Shay Gueron's suggestion */
j = m->top; /* borrow j */
if (m->d[j - 1] & (((BN_ULONG)1) << (BN_BITS2 - 1))) {
if (bn_wexpand(r, j) == NULL)
goto err;
/* 2^(top*BN_BITS2) - m */
r->d[0] = (0 - m->d[0]) & BN_MASK2;
for (i = 1; i < j; i++)
r->d[i] = (~m->d[i]) & BN_MASK2;
r->top = j;
/*
* Upper words will be zero if the corresponding words of 'm' were
* 0xfff[...], so decrement r->top accordingly.
*/
bn_correct_top(r);
} else
#endif
if (!BN_to_montgomery(r, BN_value_one(), mont, ctx))
goto err;
for (;;) {
@ -527,6 +555,17 @@ int BN_mod_exp_mont(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
if (wstart < 0)
break;
}
#if defined(SPARC_T4_MONT)
if (OPENSSL_sparcv9cap_P[0] & (SPARCV9_VIS3 | SPARCV9_PREFER_FPU)) {
j = mont->N.top; /* borrow j */
val[0]->d[0] = 1; /* borrow val[0] */
for (i = 1; i < j; i++)
val[0]->d[i] = 0;
val[0]->top = j;
if (!BN_mod_mul_montgomery(rr, r, val[0], mont, ctx))
goto err;
} else
#endif
if (!BN_from_montgomery(rr, r, mont, ctx))
goto err;
ret = 1;
@ -538,6 +577,27 @@ int BN_mod_exp_mont(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
return (ret);
}
#if defined(SPARC_T4_MONT)
static BN_ULONG bn_get_bits(const BIGNUM *a, int bitpos)
{
BN_ULONG ret = 0;
int wordpos;
wordpos = bitpos / BN_BITS2;
bitpos %= BN_BITS2;
if (wordpos >= 0 && wordpos < a->top) {
ret = a->d[wordpos] & BN_MASK2;
if (bitpos) {
ret >>= bitpos;
if (++wordpos < a->top)
ret |= a->d[wordpos] << (BN_BITS2 - bitpos);
}
}
return ret & BN_MASK2;
}
#endif
/*
* BN_mod_exp_mont_consttime() stores the precomputed powers in a specific
* layout so that accessing any of these table values shows the same access
@ -644,6 +704,9 @@ int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
int powerbufLen = 0;
unsigned char *powerbuf = NULL;
BIGNUM tmp, am;
#if defined(SPARC_T4_MONT)
unsigned int t4 = 0;
#endif
bn_check_top(a);
bn_check_top(p);
@ -683,21 +746,62 @@ int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
goto err;
}
#ifdef RSAZ_ENABLED
/*
* If the size of the operands allow it, perform the optimized
* RSAZ exponentiation. For further information see
* crypto/bn/rsaz_exp.c and accompanying assembly modules.
*/
if ((16 == a->top) && (16 == p->top) && (BN_num_bits(m) == 1024)
&& rsaz_avx2_eligible()) {
if (NULL == bn_wexpand(rr, 16))
goto err;
RSAZ_1024_mod_exp_avx2(rr->d, a->d, p->d, m->d, mont->RR.d,
mont->n0[0]);
rr->top = 16;
rr->neg = 0;
bn_correct_top(rr);
ret = 1;
goto err;
} else if ((8 == a->top) && (8 == p->top) && (BN_num_bits(m) == 512)) {
if (NULL == bn_wexpand(rr, 8))
goto err;
RSAZ_512_mod_exp(rr->d, a->d, p->d, m->d, mont->n0[0], mont->RR.d);
rr->top = 8;
rr->neg = 0;
bn_correct_top(rr);
ret = 1;
goto err;
}
#endif
/* Get the window size to use with size of p. */
window = BN_window_bits_for_ctime_exponent_size(bits);
#if defined(OPENSSL_BN_ASM_MONT5)
if (window == 6 && bits <= 1024)
window = 5; /* ~5% improvement of 2048-bit RSA sign */
#if defined(SPARC_T4_MONT)
if (window >= 5 && (top & 15) == 0 && top <= 64 &&
(OPENSSL_sparcv9cap_P[1] & (CFR_MONTMUL | CFR_MONTSQR)) ==
(CFR_MONTMUL | CFR_MONTSQR) && (t4 = OPENSSL_sparcv9cap_P[0]))
window = 5;
else
#endif
#if defined(OPENSSL_BN_ASM_MONT5)
if (window >= 5) {
window = 5; /* ~5% improvement for RSA2048 sign, and even
* for RSA4096 */
/* reserve space for mont->N.d[] copy */
powerbufLen += top * sizeof(mont->N.d[0]);
}
#endif
(void)0;
/*
* Allocate a buffer large enough to hold all of the pre-computed powers
* of am, am itself and tmp.
*/
numPowers = 1 << window;
powerbufLen = sizeof(m->d[0]) * (top * numPowers +
((2 * top) >
numPowers ? (2 * top) : numPowers));
powerbufLen += sizeof(m->d[0]) * (top * numPowers +
((2 * top) >
numPowers ? (2 * top) : numPowers));
#ifdef alloca
if (powerbufLen < 3072)
powerbufFree =
@ -727,15 +831,17 @@ int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
tmp.flags = am.flags = BN_FLG_STATIC_DATA;
/* prepare a^0 in Montgomery domain */
#if 1
#if 1 /* by Shay Gueron's suggestion */
if (m->d[top - 1] & (((BN_ULONG)1) << (BN_BITS2 - 1))) {
/* 2^(top*BN_BITS2) - m */
tmp.d[0] = (0 - m->d[0]) & BN_MASK2;
for (i = 1; i < top; i++)
tmp.d[i] = (~m->d[i]) & BN_MASK2;
tmp.top = top;
} else
#endif
if (!BN_to_montgomery(&tmp, BN_value_one(), mont, ctx))
goto err;
#else
tmp.d[0] = (0 - m->d[0]) & BN_MASK2; /* 2^(top*BN_BITS2) - m */
for (i = 1; i < top; i++)
tmp.d[i] = (~m->d[i]) & BN_MASK2;
tmp.top = top;
#endif
/* prepare a^1 in Montgomery domain */
if (a->neg || BN_ucmp(a, m) >= 0) {
@ -746,6 +852,138 @@ int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
} else if (!BN_to_montgomery(&am, a, mont, ctx))
goto err;
#if defined(SPARC_T4_MONT)
if (t4) {
typedef int (*bn_pwr5_mont_f) (BN_ULONG *tp, const BN_ULONG *np,
const BN_ULONG *n0, const void *table,
int power, int bits);
int bn_pwr5_mont_t4_8(BN_ULONG *tp, const BN_ULONG *np,
const BN_ULONG *n0, const void *table,
int power, int bits);
int bn_pwr5_mont_t4_16(BN_ULONG *tp, const BN_ULONG *np,
const BN_ULONG *n0, const void *table,
int power, int bits);
int bn_pwr5_mont_t4_24(BN_ULONG *tp, const BN_ULONG *np,
const BN_ULONG *n0, const void *table,
int power, int bits);
int bn_pwr5_mont_t4_32(BN_ULONG *tp, const BN_ULONG *np,
const BN_ULONG *n0, const void *table,
int power, int bits);
static const bn_pwr5_mont_f pwr5_funcs[4] = {
bn_pwr5_mont_t4_8, bn_pwr5_mont_t4_16,
bn_pwr5_mont_t4_24, bn_pwr5_mont_t4_32
};
bn_pwr5_mont_f pwr5_worker = pwr5_funcs[top / 16 - 1];
typedef int (*bn_mul_mont_f) (BN_ULONG *rp, const BN_ULONG *ap,
const void *bp, const BN_ULONG *np,
const BN_ULONG *n0);
int bn_mul_mont_t4_8(BN_ULONG *rp, const BN_ULONG *ap, const void *bp,
const BN_ULONG *np, const BN_ULONG *n0);
int bn_mul_mont_t4_16(BN_ULONG *rp, const BN_ULONG *ap,
const void *bp, const BN_ULONG *np,
const BN_ULONG *n0);
int bn_mul_mont_t4_24(BN_ULONG *rp, const BN_ULONG *ap,
const void *bp, const BN_ULONG *np,
const BN_ULONG *n0);
int bn_mul_mont_t4_32(BN_ULONG *rp, const BN_ULONG *ap,
const void *bp, const BN_ULONG *np,
const BN_ULONG *n0);
static const bn_mul_mont_f mul_funcs[4] = {
bn_mul_mont_t4_8, bn_mul_mont_t4_16,
bn_mul_mont_t4_24, bn_mul_mont_t4_32
};
bn_mul_mont_f mul_worker = mul_funcs[top / 16 - 1];
void bn_mul_mont_vis3(BN_ULONG *rp, const BN_ULONG *ap,
const void *bp, const BN_ULONG *np,
const BN_ULONG *n0, int num);
void bn_mul_mont_t4(BN_ULONG *rp, const BN_ULONG *ap,
const void *bp, const BN_ULONG *np,
const BN_ULONG *n0, int num);
void bn_mul_mont_gather5_t4(BN_ULONG *rp, const BN_ULONG *ap,
const void *table, const BN_ULONG *np,
const BN_ULONG *n0, int num, int power);
void bn_flip_n_scatter5_t4(const BN_ULONG *inp, size_t num,
void *table, size_t power);
void bn_gather5_t4(BN_ULONG *out, size_t num,
void *table, size_t power);
void bn_flip_t4(BN_ULONG *dst, BN_ULONG *src, size_t num);
BN_ULONG *np = mont->N.d, *n0 = mont->n0;
int stride = 5 * (6 - (top / 16 - 1)); /* multiple of 5, but less
* than 32 */
/*
* BN_to_montgomery can contaminate words above .top [in
* BN_DEBUG[_DEBUG] build]...
*/
for (i = am.top; i < top; i++)
am.d[i] = 0;
for (i = tmp.top; i < top; i++)
tmp.d[i] = 0;
bn_flip_n_scatter5_t4(tmp.d, top, powerbuf, 0);
bn_flip_n_scatter5_t4(am.d, top, powerbuf, 1);
if (!(*mul_worker) (tmp.d, am.d, am.d, np, n0) &&
!(*mul_worker) (tmp.d, am.d, am.d, np, n0))
bn_mul_mont_vis3(tmp.d, am.d, am.d, np, n0, top);
bn_flip_n_scatter5_t4(tmp.d, top, powerbuf, 2);
for (i = 3; i < 32; i++) {
/* Calculate a^i = a^(i-1) * a */
if (!(*mul_worker) (tmp.d, tmp.d, am.d, np, n0) &&
!(*mul_worker) (tmp.d, tmp.d, am.d, np, n0))
bn_mul_mont_vis3(tmp.d, tmp.d, am.d, np, n0, top);
bn_flip_n_scatter5_t4(tmp.d, top, powerbuf, i);
}
/* switch to 64-bit domain */
np = alloca(top * sizeof(BN_ULONG));
top /= 2;
bn_flip_t4(np, mont->N.d, top);
bits--;
for (wvalue = 0, i = bits % 5; i >= 0; i--, bits--)
wvalue = (wvalue << 1) + BN_is_bit_set(p, bits);
bn_gather5_t4(tmp.d, top, powerbuf, wvalue);
/*
* Scan the exponent one window at a time starting from the most
* significant bits.
*/
while (bits >= 0) {
if (bits < stride)
stride = bits + 1;
bits -= stride;
wvalue = bn_get_bits(p, bits + 1);
if ((*pwr5_worker) (tmp.d, np, n0, powerbuf, wvalue, stride))
continue;
/* retry once and fall back */
if ((*pwr5_worker) (tmp.d, np, n0, powerbuf, wvalue, stride))
continue;
bits += stride - 5;
wvalue >>= stride - 5;
wvalue &= 31;
bn_mul_mont_t4(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont_t4(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont_t4(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont_t4(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont_t4(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont_gather5_t4(tmp.d, tmp.d, powerbuf, np, n0, top,
wvalue);
}
bn_flip_t4(tmp.d, tmp.d, top);
top *= 2;
/* back to 32-bit domain */
tmp.top = top;
bn_correct_top(&tmp);
OPENSSL_cleanse(np, top * sizeof(BN_ULONG));
} else
#endif
#if defined(OPENSSL_BN_ASM_MONT5)
if (window == 5 && top > 1) {
/*
@ -764,8 +1002,15 @@ int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
void bn_scatter5(const BN_ULONG *inp, size_t num,
void *table, size_t power);
void bn_gather5(BN_ULONG *out, size_t num, void *table, size_t power);
void bn_power5(BN_ULONG *rp, const BN_ULONG *ap,
const void *table, const BN_ULONG *np,
const BN_ULONG *n0, int num, int power);
int bn_get_bits5(const BN_ULONG *ap, int off);
int bn_from_montgomery(BN_ULONG *rp, const BN_ULONG *ap,
const BN_ULONG *not_used, const BN_ULONG *np,
const BN_ULONG *n0, int num);
BN_ULONG *np = mont->N.d, *n0 = mont->n0;
BN_ULONG *n0 = mont->n0, *np;
/*
* BN_to_montgomery can contaminate words above .top [in
@ -776,6 +1021,12 @@ int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
for (i = tmp.top; i < top; i++)
tmp.d[i] = 0;
/*
* copy mont->N.d[] to improve cache locality
*/
for (np = am.d + top, i = 0; i < top; i++)
np[i] = mont->N.d[i];
bn_scatter5(tmp.d, top, powerbuf, 0);
bn_scatter5(am.d, am.top, powerbuf, 1);
bn_mul_mont(tmp.d, am.d, am.d, np, n0, top);
@ -822,20 +1073,34 @@ int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
* Scan the exponent one window at a time starting from the most
* significant bits.
*/
while (bits >= 0) {
for (wvalue = 0, i = 0; i < 5; i++, bits--)
wvalue = (wvalue << 1) + BN_is_bit_set(p, bits);
if (top & 7)
while (bits >= 0) {
for (wvalue = 0, i = 0; i < 5; i++, bits--)
wvalue = (wvalue << 1) + BN_is_bit_set(p, bits);
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont_gather5(tmp.d, tmp.d, powerbuf, np, n0, top, wvalue);
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont(tmp.d, tmp.d, tmp.d, np, n0, top);
bn_mul_mont_gather5(tmp.d, tmp.d, powerbuf, np, n0, top,
wvalue);
} else {
while (bits >= 0) {
wvalue = bn_get_bits5(p->d, bits - 4);
bits -= 5;
bn_power5(tmp.d, tmp.d, powerbuf, np, n0, top, wvalue);
}
}
ret = bn_from_montgomery(tmp.d, tmp.d, NULL, np, n0, top);
tmp.top = top;
bn_correct_top(&tmp);
if (ret) {
if (!BN_copy(rr, &tmp))
ret = 0;
goto err; /* non-zero ret means it's not error */
}
} else
#endif
{
@ -901,6 +1166,15 @@ int BN_mod_exp_mont_consttime(BIGNUM *rr, const BIGNUM *a, const BIGNUM *p,
}
/* Convert the final result from montgomery to standard format */
#if defined(SPARC_T4_MONT)
if (OPENSSL_sparcv9cap_P[0] & (SPARCV9_VIS3 | SPARCV9_PREFER_FPU)) {
am.d[0] = 1; /* borrow am */
for (i = 1; i < top; i++)
am.d[i] = 0;
if (!BN_mod_mul_montgomery(rr, &tmp, &am, mont, ctx))
goto err;
} else
#endif
if (!BN_from_montgomery(rr, &tmp, mont, ctx))
goto err;
ret = 1;

3
drivers/builtin_openssl2/crypto/bn/bn_gf2m.c
View File

@ -450,8 +450,7 @@ int BN_GF2m_mod_arr(BIGNUM *r, const BIGNUM *a, const int p[])
d0 = p[k] % BN_BITS2;
d1 = BN_BITS2 - d0;
z[n] ^= (zz << d0);
tmp_ulong = zz >> d1;
if (d0 && tmp_ulong)
if (d0 && (tmp_ulong = zz >> d1))
z[n + 1] ^= tmp_ulong;
}

27
drivers/builtin_openssl2/crypto/bn/bn_lcl.h
View File

@ -204,6 +204,24 @@ extern "C" {
# define BN_MUL_LOW_RECURSIVE_SIZE_NORMAL (32)/* 32 */
# define BN_MONT_CTX_SET_SIZE_WORD (64)/* 32 */
/*
* 2011-02-22 SMS. In various places, a size_t variable or a type cast to
* size_t was used to perform integer-only operations on pointers. This
* failed on VMS with 64-bit pointers (CC /POINTER_SIZE = 64) because size_t
* is still only 32 bits. What's needed in these cases is an integer type
* with the same size as a pointer, which size_t is not certain to be. The
* only fix here is VMS-specific.
*/
# if defined(OPENSSL_SYS_VMS)
# if __INITIAL_POINTER_SIZE == 64
# define PTR_SIZE_INT long long
# else /* __INITIAL_POINTER_SIZE == 64 */
# define PTR_SIZE_INT int
# endif /* __INITIAL_POINTER_SIZE == 64 [else] */
# elif !defined(PTR_SIZE_INT) /* defined(OPENSSL_SYS_VMS) */
# define PTR_SIZE_INT size_t
# endif /* defined(OPENSSL_SYS_VMS) [else] */
# if !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM) && !defined(PEDANTIC)
/*
* BN_UMULT_HIGH section.
@ -295,6 +313,15 @@ unsigned __int64 _umul128(unsigned __int64 a, unsigned __int64 b,
: "r"(a), "r"(b));
# endif
# endif
# elif defined(__aarch64__) && defined(SIXTY_FOUR_BIT_LONG)
# if defined(__GNUC__) && __GNUC__>=2
# define BN_UMULT_HIGH(a,b) ({ \
register BN_ULONG ret; \
asm ("umulh %0,%1,%2" \
: "=r"(ret) \
: "r"(a), "r"(b)); \
ret; })
# endif
# endif /* cpu */
# endif /* OPENSSL_NO_ASM */

119
drivers/builtin_openssl2/crypto/bn/bn_prime.pl
View File

@ -1,119 +0,0 @@
#!/usr/local/bin/perl
# bn_prime.pl
$num=2048;
$num=$ARGV[0] if ($#ARGV >= 0);
push(@primes,2);
$p=1;
loop: while ($#primes < $num-1)
{
$p+=2;
$s=int(sqrt($p));
for ($i=0; defined($primes[$i]) && $primes[$i]<=$s; $i++)
{
next loop if (($p%$primes[$i]) == 0);
}
push(@primes,$p);
}
# print <<"EOF";
# /* Auto generated by bn_prime.pl */
# /* Copyright (C) 1995-1997 Eric Young (eay\@mincom.oz.au).
# * All rights reserved.
# * Copyright remains Eric Young's, and as such any Copyright notices in
# * the code are not to be removed.
# * See the COPYRIGHT file in the SSLeay distribution for more details.
# */
#
# EOF
print <<\EOF;
/* Auto generated by bn_prime.pl */
/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
* All rights reserved.
*
* This package is an SSL implementation written
* by Eric Young (eay@cryptsoft.com).
* The implementation was written so as to conform with Netscapes SSL.
*
* This library is free for commercial and non-commercial use as long as
* the following conditions are aheared to. The following conditions
* apply to all code found in this distribution, be it the RC4, RSA,
* lhash, DES, etc., code; not just the SSL code. The SSL documentation
* included with this distribution is covered by the same copyright terms
* except that the holder is Tim Hudson (tjh@cryptsoft.com).
*
* Copyright remains Eric Young's, and as such any Copyright notices in
* the code are not to be removed.
* If this package is used in a product, Eric Young should be given attribution
* as the author of the parts of the library used.
* This can be in the form of a textual message at program startup or
* in documentation (online or textual) provided with the package.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* "This product includes cryptographic software written by
* Eric Young (eay@cryptsoft.com)"
* The word 'cryptographic' can be left out if the rouines from the library
* being used are not cryptographic related :-).
* 4. If you include any Windows specific code (or a derivative thereof) from
* the apps directory (application code) you must include an acknowledgement:
* "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
*
* THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* The licence and distribution terms for any publically available version or
* derivative of this code cannot be changed. i.e. this code cannot simply be
* copied and put under another distribution licence
* [including the GNU Public Licence.]
*/
EOF
for ($i=0; $i <= $#primes; $i++)
{
if ($primes[$i] > 256)
{
$eight=$i;
last;
}
}
printf "#ifndef EIGHT_BIT\n";
printf "#define NUMPRIMES %d\n",$num;
printf "typedef unsigned short prime_t;\n";
printf "#else\n";
printf "#define NUMPRIMES %d\n",$eight;
printf "typedef unsigned char prime_t;\n";
printf "#endif\n";
print "static const prime_t primes[NUMPRIMES]=\n\t{\n\t";
$init=0;
for ($i=0; $i <= $#primes; $i++)
{
printf "\n#ifndef EIGHT_BIT\n\t" if ($primes[$i] > 256) && !($init++);
printf("\n\t") if (($i%8) == 0) && ($i != 0);
printf("%4d,",$primes[$i]);
}
print "\n#endif\n\t};\n";

2137
drivers/builtin_openssl2/crypto/bn/bntest.c
View File

File diff suppressed because it is too large Load Diff

42
drivers/builtin_openssl2/crypto/bn/divtest.c
View File

@ -1,42 +0,0 @@
#include <openssl/bn.h>
#include <openssl/rand.h>
static int Rand(n)
{
unsigned char x[2];
RAND_pseudo_bytes(x, 2);
return (x[0] + 2 * x[1]);
}
static void bug(char *m, BIGNUM *a, BIGNUM *b)
{
printf("%s!\na=", m);
BN_print_fp(stdout, a);
printf("\nb=");
BN_print_fp(stdout, b);
printf("\n");
fflush(stdout);
}
main()
{
BIGNUM *a = BN_new(), *b = BN_new(), *c = BN_new(), *d = BN_new(),
*C = BN_new(), *D = BN_new();
BN_RECP_CTX *recp = BN_RECP_CTX_new();
BN_CTX *ctx = BN_CTX_new();
for (;;) {
BN_pseudo_rand(a, Rand(), 0, 0);
BN_pseudo_rand(b, Rand(), 0, 0);
if (BN_is_zero(b))
continue;
BN_RECP_CTX_set(recp, b, ctx);
if (BN_div(C, D, a, b, ctx) != 1)
bug("BN_div failed", a, b);
if (BN_div_recp(c, d, a, recp, ctx) != 1)
bug("BN_div_recp failed", a, b);
else if (BN_cmp(c, C) != 0 || BN_cmp(c, C) != 0)
bug("mismatch", a, b);
}
}

313
drivers/builtin_openssl2/crypto/bn/exptest.c
View File

@ -1,313 +0,0 @@
/* crypto/bn/exptest.c */
/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
* All rights reserved.
*
* This package is an SSL implementation written
* by Eric Young (eay@cryptsoft.com).
* The implementation was written so as to conform with Netscapes SSL.
*
* This library is free for commercial and non-commercial use as long as
* the following conditions are aheared to. The following conditions
* apply to all code found in this distribution, be it the RC4, RSA,
* lhash, DES, etc., code; not just the SSL code. The SSL documentation
* included with this distribution is covered by the same copyright terms
* except that the holder is Tim Hudson (tjh@cryptsoft.com).
*
* Copyright remains Eric Young's, and as such any Copyright notices in
* the code are not to be removed.
* If this package is used in a product, Eric Young should be given attribution
* as the author of the parts of the library used.
* This can be in the form of a textual message at program startup or
* in documentation (online or textual) provided with the package.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. All advertising materials mentioning features or use of this software
* must display the following acknowledgement:
* "This product includes cryptographic software written by
* Eric Young (eay@cryptsoft.com)"
* The word 'cryptographic' can be left out if the rouines from the library
* being used are not cryptographic related :-).
* 4. If you include any Windows specific code (or a derivative thereof) from
* the apps directory (application code) you must include an acknowledgement:
* "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
*
* THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* The licence and distribution terms for any publically available version or
* derivative of this code cannot be changed. i.e. this code cannot simply be
* copied and put under another distribution licence
* [including the GNU Public Licence.]
*/
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "../e_os.h"
#include <openssl/bio.h>
#include <openssl/bn.h>
#include <openssl/rand.h>
#include <openssl/err.h>
#define NUM_BITS (BN_BITS*2)
static const char rnd_seed[] =
"string to make the random number generator think it has entropy";
/*
* Test that r == 0 in test_exp_mod_zero(). Returns one on success,
* returns zero and prints debug output otherwise.
*/
static int a_is_zero_mod_one(const char *method, const BIGNUM *r,
const BIGNUM *a) {
if (!BN_is_zero(r)) {
fprintf(stderr, "%s failed:\n", method);
fprintf(stderr, "a ** 0 mod 1 = r (should be 0)\n");
fprintf(stderr, "a = ");
BN_print_fp(stderr, a);
fprintf(stderr, "\nr = ");
BN_print_fp(stderr, r);
fprintf(stderr, "\n");
return 0;
}
return 1;
}
/*
* test_exp_mod_zero tests that x**0 mod 1 == 0. It returns zero on success.
*/
static int test_exp_mod_zero()
{
BIGNUM a, p, m;
BIGNUM r;
BN_ULONG one_word = 1;
BN_CTX *ctx = BN_CTX_new();
int ret = 1, failed = 0;
BN_init(&m);
BN_one(&m);
BN_init(&a);
BN_one(&a);
BN_init(&p);
BN_zero(&p);
BN_init(&r);
if (!BN_rand(&a, 1024, 0, 0))
goto err;
if (!BN_mod_exp(&r, &a, &p, &m, ctx))
goto err;
if (!a_is_zero_mod_one("BN_mod_exp", &r, &a))
failed = 1;
if (!BN_mod_exp_recp(&r, &a, &p, &m, ctx))
goto err;
if (!a_is_zero_mod_one("BN_mod_exp_recp", &r, &a))
failed = 1;
if (!BN_mod_exp_simple(&r, &a, &p, &m, ctx))
goto err;
if (!a_is_zero_mod_one("BN_mod_exp_simple", &r, &a))
failed = 1;
if (!BN_mod_exp_mont(&r, &a, &p, &m, ctx, NULL))
goto err;
if (!a_is_zero_mod_one("BN_mod_exp_mont", &r, &a))
failed = 1;
if (!BN_mod_exp_mont_consttime(&r, &a, &p, &m, ctx, NULL)) {
goto err;
}
if (!a_is_zero_mod_one("BN_mod_exp_mont_consttime", &r, &a))
failed = 1;
/*
* A different codepath exists for single word multiplication
* in non-constant-time only.
*/
if (!BN_mod_exp_mont_word(&r, one_word, &p, &m, ctx, NULL))
goto err;
if (!BN_is_zero(&r)) {
fprintf(stderr, "BN_mod_exp_mont_word failed:\n");
fprintf(stderr, "1 ** 0 mod 1 = r (should be 0)\n");
fprintf(stderr, "r = ");
BN_print_fp(stderr, &r);
fprintf(stderr, "\n");
return 0;
}
ret = failed;
err:
BN_free(&r);
BN_free(&a);
BN_free(&p);
BN_free(&m);
BN_CTX_free(ctx);
return ret;
}
int main(int argc, char *argv[])
{
BN_CTX *ctx;
BIO *out = NULL;
int i, ret;
unsigned char c;
BIGNUM *r_mont, *r_mont_const, *r_recp, *r_simple, *a, *b, *m;
RAND_seed(rnd_seed, sizeof rnd_seed); /* or BN_rand may fail, and we
* don't even check its return
* value (which we should) */
ERR_load_BN_strings();
ctx = BN_CTX_new();
if (ctx == NULL)
EXIT(1);
r_mont = BN_new();
r_mont_const = BN_new();
r_recp = BN_new();
r_simple = BN_new();
a = BN_new();
b = BN_new();
m = BN_new();
if ((r_mont == NULL) || (r_recp == NULL) || (a == NULL) || (b == NULL))
goto err;
out = BIO_new(BIO_s_file());
if (out == NULL)
EXIT(1);
BIO_set_fp(out, stdout, BIO_NOCLOSE);
for (i = 0; i < 200; i++) {
RAND_bytes(&c, 1);
c = (c % BN_BITS) - BN_BITS2;
BN_rand(a, NUM_BITS + c, 0, 0);
RAND_bytes(&c, 1);
c = (c % BN_BITS) - BN_BITS2;
BN_rand(b, NUM_BITS + c, 0, 0);
RAND_bytes(&c, 1);
c = (c % BN_BITS) - BN_BITS2;
BN_rand(m, NUM_BITS + c, 0, 1);
BN_mod(a, a, m, ctx);
BN_mod(b, b, m, ctx);
ret = BN_mod_exp_mont(r_mont, a, b, m, ctx, NULL);
if (ret <= 0) {
printf("BN_mod_exp_mont() problems\n");
ERR_print_errors(out);
EXIT(1);
}
ret = BN_mod_exp_recp(r_recp, a, b, m, ctx);
if (ret <= 0) {
printf("BN_mod_exp_recp() problems\n");
ERR_print_errors(out);
EXIT(1);
}
ret = BN_mod_exp_simple(r_simple, a, b, m, ctx);
if (ret <= 0) {
printf("BN_mod_exp_simple() problems\n");
ERR_print_errors(out);
EXIT(1);
}
ret = BN_mod_exp_mont_consttime(r_mont_const, a, b, m, ctx, NULL);
if (ret <= 0) {
printf("BN_mod_exp_mont_consttime() problems\n");
ERR_print_errors(out);
EXIT(1);
}
if (BN_cmp(r_simple, r_mont) == 0
&& BN_cmp(r_simple, r_recp) == 0
&& BN_cmp(r_simple, r_mont_const) == 0) {
printf(".");
fflush(stdout);
} else {
if (BN_cmp(r_simple, r_mont) != 0)
printf("\nsimple and mont results differ\n");
if (BN_cmp(r_simple, r_mont_const) != 0)
printf("\nsimple and mont const time results differ\n");
if (BN_cmp(r_simple, r_recp) != 0)
printf("\nsimple and recp results differ\n");
printf("a (%3d) = ", BN_num_bits(a));
BN_print(out, a);
printf("\nb (%3d) = ", BN_num_bits(b));
BN_print(out, b);
printf("\nm (%3d) = ", BN_num_bits(m));
BN_print(out, m);
printf("\nsimple =");
BN_print(out, r_simple);
printf("\nrecp =");
BN_print(out, r_recp);
printf("\nmont =");
BN_print(out, r_mont);
printf("\nmont_ct =");
BN_print(out, r_mont_const);
printf("\n");
EXIT(1);
}
}
BN_free(r_mont);
BN_free(r_mont_const);
BN_free(r_recp);
BN_free(r_simple);
BN_free(a);
BN_free(b);
BN_free(m);
BN_CTX_free(ctx);
ERR_remove_thread_state(NULL);
CRYPTO_mem_leaks(out);
BIO_free(out);
printf("\n");
if (test_exp_mod_zero() != 0)
goto err;
printf("done\n");
EXIT(0);
err:
ERR_load_crypto_strings();
ERR_print_errors(out);
#ifdef OPENSSL_SYS_NETWARE
printf("ERROR\n");
#endif
EXIT(1);
return (1);
}

346
drivers/builtin_openssl2/crypto/bn/rsaz_exp.c Normal file
View File

@ -0,0 +1,346 @@
/*****************************************************************************
* *
* Copyright (c) 2012, Intel Corporation *
* *
* All rights reserved. *
* *
* Redistribution and use in source and binary forms, with or without *
* modification, are permitted provided that the following conditions are *
* met: *
* *
* * Redistributions of source code must retain the above copyright *
* notice, this list of conditions and the following disclaimer. *
* *
* * Redistributions in binary form must reproduce the above copyright *
* notice, this list of conditions and the following disclaimer in the *
* documentation and/or other materials provided with the *
* distribution. *
* *
* * Neither the name of the Intel Corporation nor the names of its *
* contributors may be used to endorse or promote products derived from *
* this software without specific prior written permission. *
* *
* *
* THIS SOFTWARE IS PROVIDED BY INTEL CORPORATION ""AS IS"" AND ANY *
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE *
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR *
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL CORPORATION OR *
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, *
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, *
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR *
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF *
* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING *
* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS *
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. *
* *
******************************************************************************
* Developers and authors: *
* Shay Gueron (1, 2), and Vlad Krasnov (1) *
* (1) Intel Corporation, Israel Development Center, Haifa, Israel *
* (2) University of Haifa, Israel *
*****************************************************************************/
#include "rsaz_exp.h"
#ifdef RSAZ_ENABLED
/*
* See crypto/bn/asm/rsaz-avx2.pl for further details.
*/
void rsaz_1024_norm2red_avx2(void *red, const void *norm);
void rsaz_1024_mul_avx2(void *ret, const void *a, const void *b,
const void *n, BN_ULONG k);
void rsaz_1024_sqr_avx2(void *ret, const void *a, const void *n, BN_ULONG k,
int cnt);
void rsaz_1024_scatter5_avx2(void *tbl, const void *val, int i);
void rsaz_1024_gather5_avx2(void *val, const void *tbl, int i);
void rsaz_1024_red2norm_avx2(void *norm, const void *red);
#if defined(__GNUC__)
# define ALIGN64 __attribute__((aligned(64)))
#elif defined(_MSC_VER)
# define ALIGN64 __declspec(align(64))
#elif defined(__SUNPRO_C)
# define ALIGN64
# pragma align 64(one,two80)
#else
/* not fatal, might hurt performance a little */
# define ALIGN64
#endif
ALIGN64 static const BN_ULONG one[40] = {
1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
};
ALIGN64 static const BN_ULONG two80[40] = {
0, 0, 1 << 22, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0
};
void RSAZ_1024_mod_exp_avx2(BN_ULONG result_norm[16],
const BN_ULONG base_norm[16],
const BN_ULONG exponent[16],
const BN_ULONG m_norm[16], const BN_ULONG RR[16],
BN_ULONG k0)
{
unsigned char storage[320 * 3 + 32 * 9 * 16 + 64]; /* 5.5KB */
unsigned char *p_str = storage + (64 - ((size_t)storage % 64));
unsigned char *a_inv, *m, *result;
unsigned char *table_s = p_str + 320 * 3;
unsigned char *R2 = table_s; /* borrow */
int index;
int wvalue;
if ((((size_t)p_str & 4095) + 320) >> 12) {
result = p_str;
a_inv = p_str + 320;
m = p_str + 320 * 2; /* should not cross page */
} else {
m = p_str; /* should not cross page */
result = p_str + 320;
a_inv = p_str + 320 * 2;
}
rsaz_1024_norm2red_avx2(m, m_norm);
rsaz_1024_norm2red_avx2(a_inv, base_norm);
rsaz_1024_norm2red_avx2(R2, RR);
rsaz_1024_mul_avx2(R2, R2, R2, m, k0);
rsaz_1024_mul_avx2(R2, R2, two80, m, k0);
/* table[0] = 1 */
rsaz_1024_mul_avx2(result, R2, one, m, k0);
/* table[1] = a_inv^1 */
rsaz_1024_mul_avx2(a_inv, a_inv, R2, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 0);
rsaz_1024_scatter5_avx2(table_s, a_inv, 1);
/* table[2] = a_inv^2 */
rsaz_1024_sqr_avx2(result, a_inv, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 2);
#if 0
/* this is almost 2x smaller and less than 1% slower */
for (index = 3; index < 32; index++) {
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, index);
}
#else
/* table[4] = a_inv^4 */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 4);
/* table[8] = a_inv^8 */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 8);
/* table[16] = a_inv^16 */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 16);
/* table[17] = a_inv^17 */
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 17);
/* table[3] */
rsaz_1024_gather5_avx2(result, table_s, 2);
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 3);
/* table[6] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 6);
/* table[12] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 12);
/* table[24] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 24);
/* table[25] */
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 25);
/* table[5] */
rsaz_1024_gather5_avx2(result, table_s, 4);
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 5);
/* table[10] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 10);
/* table[20] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 20);
/* table[21] */
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 21);
/* table[7] */
rsaz_1024_gather5_avx2(result, table_s, 6);
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 7);
/* table[14] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 14);
/* table[28] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 28);
/* table[29] */
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 29);
/* table[9] */
rsaz_1024_gather5_avx2(result, table_s, 8);
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 9);
/* table[18] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 18);
/* table[19] */
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 19);
/* table[11] */
rsaz_1024_gather5_avx2(result, table_s, 10);
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 11);
/* table[22] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 22);
/* table[23] */
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 23);
/* table[13] */
rsaz_1024_gather5_avx2(result, table_s, 12);
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 13);
/* table[26] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 26);
/* table[27] */
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 27);
/* table[15] */
rsaz_1024_gather5_avx2(result, table_s, 14);
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 15);
/* table[30] */
rsaz_1024_sqr_avx2(result, result, m, k0, 1);
rsaz_1024_scatter5_avx2(table_s, result, 30);
/* table[31] */
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
rsaz_1024_scatter5_avx2(table_s, result, 31);
#endif
/* load first window */
p_str = (unsigned char *)exponent;
wvalue = p_str[127] >> 3;
rsaz_1024_gather5_avx2(result, table_s, wvalue);
index = 1014;
while (index > -1) { /* loop for the remaining 127 windows */
rsaz_1024_sqr_avx2(result, result, m, k0, 5);
wvalue = *((unsigned short *)&p_str[index / 8]);
wvalue = (wvalue >> (index % 8)) & 31;
index -= 5;
rsaz_1024_gather5_avx2(a_inv, table_s, wvalue); /* borrow a_inv */
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
}
/* square four times */
rsaz_1024_sqr_avx2(result, result, m, k0, 4);
wvalue = p_str[0] & 15;
rsaz_1024_gather5_avx2(a_inv, table_s, wvalue); /* borrow a_inv */
rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
/* from Montgomery */
rsaz_1024_mul_avx2(result, result, one, m, k0);
rsaz_1024_red2norm_avx2(result_norm, result);
OPENSSL_cleanse(storage, sizeof(storage));
}
/*
* See crypto/bn/rsaz-x86_64.pl for further details.
*/
void rsaz_512_mul(void *ret, const void *a, const void *b, const void *n,
BN_ULONG k);
void rsaz_512_mul_scatter4(void *ret, const void *a, const void *n,
BN_ULONG k, const void *tbl, unsigned int power);
void rsaz_512_mul_gather4(void *ret, const void *a, const void *tbl,
const void *n, BN_ULONG k, unsigned int power);
void rsaz_512_mul_by_one(void *ret, const void *a, const void *n, BN_ULONG k);
void rsaz_512_sqr(void *ret, const void *a, const void *n, BN_ULONG k,
int cnt);
void rsaz_512_scatter4(void *tbl, const BN_ULONG *val, int power);
void rsaz_512_gather4(BN_ULONG *val, const void *tbl, int power);
void RSAZ_512_mod_exp(BN_ULONG result[8],
const BN_ULONG base[8], const BN_ULONG exponent[8],
const BN_ULONG m[8], BN_ULONG k0, const BN_ULONG RR[8])
{
unsigned char storage[16 * 8 * 8 + 64 * 2 + 64]; /* 1.2KB */
unsigned char *table = storage + (64 - ((size_t)storage % 64));
BN_ULONG *a_inv = (BN_ULONG *)(table + 16 * 8 * 8);
BN_ULONG *temp = (BN_ULONG *)(table + 16 * 8 * 8 + 8 * 8);
unsigned char *p_str = (unsigned char *)exponent;
int index;
unsigned int wvalue;
/* table[0] = 1_inv */
temp[0] = 0 - m[0];
temp[1] = ~m[1];
temp[2] = ~m[2];
temp[3] = ~m[3];
temp[4] = ~m[4];
temp[5] = ~m[5];
temp[6] = ~m[6];
temp[7] = ~m[7];
rsaz_512_scatter4(table, temp, 0);
/* table [1] = a_inv^1 */
rsaz_512_mul(a_inv, base, RR, m, k0);
rsaz_512_scatter4(table, a_inv, 1);
/* table [2] = a_inv^2 */
rsaz_512_sqr(temp, a_inv, m, k0, 1);
rsaz_512_scatter4(table, temp, 2);
for (index = 3; index < 16; index++)
rsaz_512_mul_scatter4(temp, a_inv, m, k0, table, index);
/* load first window */
wvalue = p_str[63];
rsaz_512_gather4(temp, table, wvalue >> 4);
rsaz_512_sqr(temp, temp, m, k0, 4);
rsaz_512_mul_gather4(temp, temp, table, m, k0, wvalue & 0xf);
for (index = 62; index >= 0; index--) {
wvalue = p_str[index];
rsaz_512_sqr(temp, temp, m, k0, 4);
rsaz_512_mul_gather4(temp, temp, table, m, k0, wvalue >> 4);
rsaz_512_sqr(temp, temp, m, k0, 4);
rsaz_512_mul_gather4(temp, temp, table, m, k0, wvalue & 0x0f);
}
/* from Montgomery */
rsaz_512_mul_by_one(result, temp, m, k0);
OPENSSL_cleanse(storage, sizeof(storage));
}
#else
# if defined(PEDANTIC) || defined(__DECC) || defined(__clang__)
static void *dummy = &dummy;
# endif
#endif

68
drivers/builtin_openssl2/crypto/bn/rsaz_exp.h Normal file
View File

@ -0,0 +1,68 @@
/*****************************************************************************
* *
* Copyright (c) 2012, Intel Corporation *
* *
* All rights reserved. *
* *
* Redistribution and use in source and binary forms, with or without *
* modification, are permitted provided that the following conditions are *
* met: *
* *
* * Redistributions of source code must retain the above copyright *
* notice, this list of conditions and the following disclaimer. *
* *
* * Redistributions in binary form must reproduce the above copyright *
* notice, this list of conditions and the following disclaimer in the *
* documentation and/or other materials provided with the *
* distribution. *
* *
* * Neither the name of the Intel Corporation nor the names of its *
* contributors may be used to endorse or promote products derived from *
* this software without specific prior written permission. *
* *
* *
* THIS SOFTWARE IS PROVIDED BY INTEL CORPORATION ""AS IS"" AND ANY *
* EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE *
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR *
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL CORPORATION OR *
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, *
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, *
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR *
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF *
* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING *
* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS *
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. *
* *
******************************************************************************
* Developers and authors: *
* Shay Gueron (1, 2), and Vlad Krasnov (1) *
* (1) Intel Corporation, Israel Development Center, Haifa, Israel *
* (2) University of Haifa, Israel *
*****************************************************************************/
#ifndef RSAZ_EXP_H
# define RSAZ_EXP_H
# undef RSAZ_ENABLED
# if defined(OPENSSL_BN_ASM_MONT) && \
(defined(__x86_64) || defined(__x86_64__) || \
defined(_M_AMD64) || defined(_M_X64))
# define RSAZ_ENABLED
# include <openssl/bn.h>
void RSAZ_1024_mod_exp_avx2(BN_ULONG result[16],
const BN_ULONG base_norm[16],
const BN_ULONG exponent[16],
const BN_ULONG m_norm[16], const BN_ULONG RR[16],
BN_ULONG k0);
int rsaz_avx2_eligible();
void RSAZ_512_mod_exp(BN_ULONG result[8],
const BN_ULONG base_norm[8], const BN_ULONG exponent[8],
const BN_ULONG m_norm[8], BN_ULONG k0,
const BN_ULONG RR[8]);
# endif
#endif

11
drivers/builtin_openssl2/crypto/buffer/buf_str.c
View File

@ -61,6 +61,15 @@
#include <limits.h>
#include <openssl/buffer.h>
size_t BUF_strnlen(const char *str, size_t maxlen)
{
const char *p;
for (p = str; maxlen-- != 0 && *p != '\0'; ++p) ;
return p - str;
}
char *BUF_strdup(const char *str)
{
if (str == NULL)
@ -75,6 +84,8 @@ char *BUF_strndup(const char *str, size_t siz)
if (str == NULL)
return NULL;
siz = BUF_strnlen(str, siz);
if (siz >= INT_MAX)
return NULL;

1138
drivers/builtin_openssl2/crypto/camellia/asm/cmll-x86.pl
View File

File diff suppressed because it is too large Load Diff

1081
drivers/builtin_openssl2/crypto/camellia/asm/cmll-x86_64.pl
View File

File diff suppressed because it is too large Load Diff

177
drivers/builtin_openssl2/crypto/cast/asm/cast-586.pl
View File

@ -1,177 +0,0 @@
#!/usr/local/bin/perl
# define for pentium pro friendly version
$ppro=1;
$0 =~ m/(.*[\/\\])[^\/\\]+$/; $dir=$1;
push(@INC,"${dir}","${dir}../../perlasm");
require "x86asm.pl";
require "cbc.pl";
&asm_init($ARGV[0],"cast-586.pl",$ARGV[$#ARGV] eq "386");
$CAST_ROUNDS=16;
$L="edi";
$R="esi";
$K="ebp";
$tmp1="ecx";
$tmp2="ebx";
$tmp3="eax";
$tmp4="edx";
$S1="CAST_S_table0";
$S2="CAST_S_table1";
$S3="CAST_S_table2";
$S4="CAST_S_table3";
@F1=("add","xor","sub");
@F2=("xor","sub","add");
@F3=("sub","add","xor");
&CAST_encrypt("CAST_encrypt",1);
&CAST_encrypt("CAST_decrypt",0);
&cbc("CAST_cbc_encrypt","CAST_encrypt","CAST_decrypt",1,4,5,3,-1,-1);
&asm_finish();
sub CAST_encrypt {
local($name,$enc)=@_;
local($win_ex)=<<"EOF";
EXTERN _CAST_S_table0:DWORD
EXTERN _CAST_S_table1:DWORD
EXTERN _CAST_S_table2:DWORD
EXTERN _CAST_S_table3:DWORD
EOF
&main::external_label(
"CAST_S_table0",
"CAST_S_table1",
"CAST_S_table2",
"CAST_S_table3",
);
&function_begin_B($name,$win_ex);
&comment("");
&push("ebp");
&push("ebx");
&mov($tmp2,&wparam(0));
&mov($K,&wparam(1));
&push("esi");
&push("edi");
&comment("Load the 2 words");
&mov($L,&DWP(0,$tmp2,"",0));
&mov($R,&DWP(4,$tmp2,"",0));
&comment('Get short key flag');
&mov($tmp3,&DWP(128,$K,"",0));
if($enc) {
&push($tmp3);
} else {
&or($tmp3,$tmp3);
&jnz(&label('cast_dec_skip'));
}
&xor($tmp3, $tmp3);
# encrypting part
if ($enc) {
&E_CAST( 0,$S,$L,$R,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 1,$S,$R,$L,$K,@F2,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 2,$S,$L,$R,$K,@F3,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 3,$S,$R,$L,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 4,$S,$L,$R,$K,@F2,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 5,$S,$R,$L,$K,@F3,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 6,$S,$L,$R,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 7,$S,$R,$L,$K,@F2,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 8,$S,$L,$R,$K,@F3,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 9,$S,$R,$L,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST(10,$S,$L,$R,$K,@F2,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST(11,$S,$R,$L,$K,@F3,$tmp1,$tmp2,$tmp3,$tmp4);
&comment('test short key flag');
&pop($tmp4);
&or($tmp4,$tmp4);
&jnz(&label('cast_enc_done'));
&E_CAST(12,$S,$L,$R,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST(13,$S,$R,$L,$K,@F2,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST(14,$S,$L,$R,$K,@F3,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST(15,$S,$R,$L,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
} else {
&E_CAST(15,$S,$L,$R,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST(14,$S,$R,$L,$K,@F3,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST(13,$S,$L,$R,$K,@F2,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST(12,$S,$R,$L,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
&set_label('cast_dec_skip');
&E_CAST(11,$S,$L,$R,$K,@F3,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST(10,$S,$R,$L,$K,@F2,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 9,$S,$L,$R,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 8,$S,$R,$L,$K,@F3,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 7,$S,$L,$R,$K,@F2,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 6,$S,$R,$L,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 5,$S,$L,$R,$K,@F3,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 4,$S,$R,$L,$K,@F2,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 3,$S,$L,$R,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 2,$S,$R,$L,$K,@F3,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 1,$S,$L,$R,$K,@F2,$tmp1,$tmp2,$tmp3,$tmp4);
&E_CAST( 0,$S,$R,$L,$K,@F1,$tmp1,$tmp2,$tmp3,$tmp4);
}
&set_label('cast_enc_done') if $enc;
# Why the nop? - Ben 17/1/99
&nop();
&mov($tmp3,&wparam(0));
&mov(&DWP(4,$tmp3,"",0),$L);
&mov(&DWP(0,$tmp3,"",0),$R);
&function_end($name);
}
sub E_CAST {
local($i,$S,$L,$R,$K,$OP1,$OP2,$OP3,$tmp1,$tmp2,$tmp3,$tmp4)=@_;
# Ri needs to have 16 pre added.
&comment("round $i");
&mov( $tmp4, &DWP($i*8,$K,"",1));
&mov( $tmp1, &DWP($i*8+4,$K,"",1));
&$OP1( $tmp4, $R);
&rotl( $tmp4, &LB($tmp1));
if ($ppro) {
&mov( $tmp2, $tmp4); # B
&xor( $tmp1, $tmp1);
&movb( &LB($tmp1), &HB($tmp4)); # A
&and( $tmp2, 0xff);
&shr( $tmp4, 16); #
&xor( $tmp3, $tmp3);
} else {
&mov( $tmp2, $tmp4); # B
&movb( &LB($tmp1), &HB($tmp4)); # A # BAD BAD BAD
&shr( $tmp4, 16); #
&and( $tmp2, 0xff);
}
&movb( &LB($tmp3), &HB($tmp4)); # C # BAD BAD BAD
&and( $tmp4, 0xff); # D
&mov( $tmp1, &DWP($S1,"",$tmp1,4));
&mov( $tmp2, &DWP($S2,"",$tmp2,4));
&$OP2( $tmp1, $tmp2);
&mov( $tmp2, &DWP($S3,"",$tmp3,4));
&$OP3( $tmp1, $tmp2);
&mov( $tmp2, &DWP($S4,"",$tmp4,4));
&$OP1( $tmp1, $tmp2);
# XXX
&xor( $L, $tmp1);
# XXX
}

2
drivers/builtin_openssl2/crypto/cast/cast_lcl.h
View File

@ -152,6 +152,8 @@
#if defined(OPENSSL_SYS_WIN32) && defined(_MSC_VER)
# define ROTL(a,n) (_lrotl(a,n))
#elif defined(PEDANTIC)
# define ROTL(a,n) ((((a)<<(n))&0xffffffffL)|((a)>>((32-(n))&31)))
#else
# define ROTL(a,n) ((((a)<<(n))&0xffffffffL)|((a)>>(32-(n))))
#endif

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