Mercurial > vba-linux
changeset 23:bf9169ad4222
add SMID-oriented fast mersenne twister
| author | Robert McIntyre <rlm@mit.edu> |
|---|---|
| date | Sun, 04 Mar 2012 17:38:32 -0600 |
| parents | 8870086b716c |
| children | 59790d015f25 |
| files | src/SFMT/CHANGE-LOG.txt src/SFMT/LICENSE.txt src/SFMT/README.txt src/SFMT/SFMT-alti.h src/SFMT/SFMT-params.h src/SFMT/SFMT-params11213.h src/SFMT/SFMT-params1279.h src/SFMT/SFMT-params132049.h src/SFMT/SFMT-params19937.h src/SFMT/SFMT-params216091.h src/SFMT/SFMT-params2281.h src/SFMT/SFMT-params4253.h src/SFMT/SFMT-params44497.h src/SFMT/SFMT-params607.h src/SFMT/SFMT-params86243.h src/SFMT/SFMT-sse2.h src/SFMT/SFMT.c src/SFMT/SFMT.h src/common/lua-engine.cpp |
| diffstat | 19 files changed, 1729 insertions(+), 4 deletions(-) [+] |
line wrap: on
line diff
1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 1.2 +++ b/src/SFMT/CHANGE-LOG.txt Sun Mar 04 17:38:32 2012 -0600 1.3 @@ -0,0 +1,55 @@ 1.4 +ver 1.3.3 1.5 +------- 1.6 +change condition compile of do_recursion in SFMT.c 1.7 + 1.8 +ver 1.3.2 1.9 +------- 1.10 +bug fix to_res53_mix and genrand_res53_mix. 1.11 + 1.12 +ver 1.3.1 1.13 +------- 1.14 +gcc compile option changed form -O9 to -O3. 1.15 +add functions genrand_res53_mix and to_res53_mix. 1.16 +bug fix about definition of ALWAYS_INLINE. 1.17 +add new definition PRE_ALWAYS for MSC. 1.18 + 1.19 +ver 1.3 1.20 +------- 1.21 +bug fixed: -DONLY64 without -DBIG_ENIAN64 had been generating 1.22 +wrong sequence. 1.23 +bug fixed: There is no documentation about BIG_ENDIAN64. 1.24 +add automatic endian check by __BIG_ENDIAN__ predefined macro. 1.25 +bug fixed: change == in check.sh to = 1.26 +add SFMT-params216091.h 1.27 +add AltiVec parameter format for systems which are not osx. 1.28 +change Makefile for systems which are not osx and support AltiVec. 1.29 +change sample2 of howto-compile for Free BSD. 1.30 +change source files for BORLANDC and Visual Studio. 1.31 +change period certification code more smart. 1.32 +add params directory. 1.33 + 1.34 +ver 1.2.1 1.35 +------- 1.36 +Fix typo in SFMT-alti.c SFMT-sse2.c 1.37 +marge SFMT-alti.c and SFMT-alti.h into SFMT-alti.h 1.38 +marge SFMT-sse2.c and SFMT-sse2.h into SFMT-sse2.h 1.39 +This version is not released. 1.40 + 1.41 +ver 1.2 1.42 +------- 1.43 +Support many periods: 2^{607}, 2^{1279}, 2^{2281}, 2^{4253}, 2^{11213}, 1.44 +2^{19937}, 2^{44497}, 2^{86243}, 2^{132049} 1.45 +Fix typo in LICENSE.txt. 1.46 +Add cast to vec_perm for SFMT-alti.c, SFMT-alti64.c. 1.47 +combine source codes. 1.48 + 1.49 +ver 1.1 1.50 +------- 1.51 +The period certification method is changed from constant to function. 1.52 +The convert functions from 32-bit and 64-bit integer to double are added. 1.53 +The documentation is changed. 1.54 +Sample programs are added. 1.55 + 1.56 +ver 1.0 1.57 +------- 1.58 +The first version.
2.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 2.2 +++ b/src/SFMT/LICENSE.txt Sun Mar 04 17:38:32 2012 -0600 2.3 @@ -0,0 +1,29 @@ 2.4 +Copyright (c) 2006,2007 Mutsuo Saito, Makoto Matsumoto and Hiroshima 2.5 +University. All rights reserved. 2.6 + 2.7 +Redistribution and use in source and binary forms, with or without 2.8 +modification, are permitted provided that the following conditions are 2.9 +met: 2.10 + 2.11 + * Redistributions of source code must retain the above copyright 2.12 + notice, this list of conditions and the following disclaimer. 2.13 + * Redistributions in binary form must reproduce the above 2.14 + copyright notice, this list of conditions and the following 2.15 + disclaimer in the documentation and/or other materials provided 2.16 + with the distribution. 2.17 + * Neither the name of the Hiroshima University nor the names of 2.18 + its contributors may be used to endorse or promote products 2.19 + derived from this software without specific prior written 2.20 + permission. 2.21 + 2.22 +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 2.23 +"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 2.24 +LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 2.25 +A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 2.26 +OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 2.27 +SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 2.28 +LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 2.29 +DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 2.30 +THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 2.31 +(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 2.32 +OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
3.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 3.2 +++ b/src/SFMT/README.txt Sun Mar 04 17:38:32 2012 -0600 3.3 @@ -0,0 +1,22 @@ 3.4 + ================================================================= 3.5 + SFMT ver. 1.3.3 3.6 + SIMD oriented Fast Mersenne Twister(SFMT) 3.7 + 3.8 + Mutsuo Saito (Hiroshima University) and 3.9 + Makoto Matsumoto (Hiroshima University) 3.10 + 3.11 + Copyright (C) 2006, 2007 Mutsuo Saito, Makoto Matsumoto and Hiroshima 3.12 + University. All rights reserved. 3.13 + 3.14 + The (modified) BSD License is applied to this software, see LICENSE.txt 3.15 + ================================================================= 3.16 + 3.17 + To see documents, see html/index.html. 3.18 + 3.19 + To make test program, see html/howto-compile.html 3.20 + 3.21 + If you want to redistribute and/or change source files, see LICENSE.txt. 3.22 + 3.23 + When you change these files and redistribute them, PLEASE write your 3.24 + e-mail address in redistribution and write to contact YOU first if 3.25 + users of your changed source encounter troubles.
4.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 4.2 +++ b/src/SFMT/SFMT-alti.h Sun Mar 04 17:38:32 2012 -0600 4.3 @@ -0,0 +1,156 @@ 4.4 +/** 4.5 + * @file SFMT-alti.h 4.6 + * 4.7 + * @brief SIMD oriented Fast Mersenne Twister(SFMT) 4.8 + * pseudorandom number generator 4.9 + * 4.10 + * @author Mutsuo Saito (Hiroshima University) 4.11 + * @author Makoto Matsumoto (Hiroshima University) 4.12 + * 4.13 + * Copyright (C) 2007 Mutsuo Saito, Makoto Matsumoto and Hiroshima 4.14 + * University. All rights reserved. 4.15 + * 4.16 + * The new BSD License is applied to this software. 4.17 + * see LICENSE.txt 4.18 + */ 4.19 + 4.20 +#ifndef SFMT_ALTI_H 4.21 +#define SFMT_ALTI_H 4.22 + 4.23 +inline static vector unsigned int vec_recursion(vector unsigned int a, 4.24 + vector unsigned int b, 4.25 + vector unsigned int c, 4.26 + vector unsigned int d) 4.27 + ALWAYSINLINE; 4.28 + 4.29 +/** 4.30 + * This function represents the recursion formula in AltiVec and BIG ENDIAN. 4.31 + * @param a a 128-bit part of the interal state array 4.32 + * @param b a 128-bit part of the interal state array 4.33 + * @param c a 128-bit part of the interal state array 4.34 + * @param d a 128-bit part of the interal state array 4.35 + * @return output 4.36 + */ 4.37 +inline static vector unsigned int vec_recursion(vector unsigned int a, 4.38 + vector unsigned int b, 4.39 + vector unsigned int c, 4.40 + vector unsigned int d) { 4.41 + 4.42 + const vector unsigned int sl1 = ALTI_SL1; 4.43 + const vector unsigned int sr1 = ALTI_SR1; 4.44 +#ifdef ONLY64 4.45 + const vector unsigned int mask = ALTI_MSK64; 4.46 + const vector unsigned char perm_sl = ALTI_SL2_PERM64; 4.47 + const vector unsigned char perm_sr = ALTI_SR2_PERM64; 4.48 +#else 4.49 + const vector unsigned int mask = ALTI_MSK; 4.50 + const vector unsigned char perm_sl = ALTI_SL2_PERM; 4.51 + const vector unsigned char perm_sr = ALTI_SR2_PERM; 4.52 +#endif 4.53 + vector unsigned int v, w, x, y, z; 4.54 + x = vec_perm(a, (vector unsigned int)perm_sl, perm_sl); 4.55 + v = a; 4.56 + y = vec_sr(b, sr1); 4.57 + z = vec_perm(c, (vector unsigned int)perm_sr, perm_sr); 4.58 + w = vec_sl(d, sl1); 4.59 + z = vec_xor(z, w); 4.60 + y = vec_and(y, mask); 4.61 + v = vec_xor(v, x); 4.62 + z = vec_xor(z, y); 4.63 + z = vec_xor(z, v); 4.64 + return z; 4.65 +} 4.66 + 4.67 +/** 4.68 + * This function fills the internal state array with pseudorandom 4.69 + * integers. 4.70 + */ 4.71 +inline static void gen_rand_all(void) { 4.72 + int i; 4.73 + vector unsigned int r, r1, r2; 4.74 + 4.75 + r1 = sfmt[N - 2].s; 4.76 + r2 = sfmt[N - 1].s; 4.77 + for (i = 0; i < N - POS1; i++) { 4.78 + r = vec_recursion(sfmt[i].s, sfmt[i + POS1].s, r1, r2); 4.79 + sfmt[i].s = r; 4.80 + r1 = r2; 4.81 + r2 = r; 4.82 + } 4.83 + for (; i < N; i++) { 4.84 + r = vec_recursion(sfmt[i].s, sfmt[i + POS1 - N].s, r1, r2); 4.85 + sfmt[i].s = r; 4.86 + r1 = r2; 4.87 + r2 = r; 4.88 + } 4.89 +} 4.90 + 4.91 +/** 4.92 + * This function fills the user-specified array with pseudorandom 4.93 + * integers. 4.94 + * 4.95 + * @param array an 128-bit array to be filled by pseudorandom numbers. 4.96 + * @param size number of 128-bit pesudorandom numbers to be generated. 4.97 + */ 4.98 +inline static void gen_rand_array(w128_t *array, int size) { 4.99 + int i, j; 4.100 + vector unsigned int r, r1, r2; 4.101 + 4.102 + r1 = sfmt[N - 2].s; 4.103 + r2 = sfmt[N - 1].s; 4.104 + for (i = 0; i < N - POS1; i++) { 4.105 + r = vec_recursion(sfmt[i].s, sfmt[i + POS1].s, r1, r2); 4.106 + array[i].s = r; 4.107 + r1 = r2; 4.108 + r2 = r; 4.109 + } 4.110 + for (; i < N; i++) { 4.111 + r = vec_recursion(sfmt[i].s, array[i + POS1 - N].s, r1, r2); 4.112 + array[i].s = r; 4.113 + r1 = r2; 4.114 + r2 = r; 4.115 + } 4.116 + /* main loop */ 4.117 + for (; i < size - N; i++) { 4.118 + r = vec_recursion(array[i - N].s, array[i + POS1 - N].s, r1, r2); 4.119 + array[i].s = r; 4.120 + r1 = r2; 4.121 + r2 = r; 4.122 + } 4.123 + for (j = 0; j < 2 * N - size; j++) { 4.124 + sfmt[j].s = array[j + size - N].s; 4.125 + } 4.126 + for (; i < size; i++) { 4.127 + r = vec_recursion(array[i - N].s, array[i + POS1 - N].s, r1, r2); 4.128 + array[i].s = r; 4.129 + sfmt[j++].s = r; 4.130 + r1 = r2; 4.131 + r2 = r; 4.132 + } 4.133 +} 4.134 + 4.135 +#ifndef ONLY64 4.136 +#if defined(__APPLE__) 4.137 +#define ALTI_SWAP (vector unsigned char) \ 4.138 + (4, 5, 6, 7, 0, 1, 2, 3, 12, 13, 14, 15, 8, 9, 10, 11) 4.139 +#else 4.140 +#define ALTI_SWAP {4, 5, 6, 7, 0, 1, 2, 3, 12, 13, 14, 15, 8, 9, 10, 11} 4.141 +#endif 4.142 +/** 4.143 + * This function swaps high and low 32-bit of 64-bit integers in user 4.144 + * specified array. 4.145 + * 4.146 + * @param array an 128-bit array to be swaped. 4.147 + * @param size size of 128-bit array. 4.148 + */ 4.149 +inline static void swap(w128_t *array, int size) { 4.150 + int i; 4.151 + const vector unsigned char perm = ALTI_SWAP; 4.152 + 4.153 + for (i = 0; i < size; i++) { 4.154 + array[i].s = vec_perm(array[i].s, (vector unsigned int)perm, perm); 4.155 + } 4.156 +} 4.157 +#endif 4.158 + 4.159 +#endif
5.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 5.2 +++ b/src/SFMT/SFMT-params.h Sun Mar 04 17:38:32 2012 -0600 5.3 @@ -0,0 +1,97 @@ 5.4 +#ifndef SFMT_PARAMS_H 5.5 +#define SFMT_PARAMS_H 5.6 + 5.7 +#if !defined(MEXP) 5.8 +#ifdef __GNUC__ 5.9 + #warning "MEXP is not defined. I assume MEXP is 19937." 5.10 +#endif 5.11 + #define MEXP 19937 5.12 +#endif 5.13 +/*----------------- 5.14 + BASIC DEFINITIONS 5.15 + -----------------*/ 5.16 +/** Mersenne Exponent. The period of the sequence 5.17 + * is a multiple of 2^MEXP-1. 5.18 + * #define MEXP 19937 */ 5.19 +/** SFMT generator has an internal state array of 128-bit integers, 5.20 + * and N is its size. */ 5.21 +#define N (MEXP / 128 + 1) 5.22 +/** N32 is the size of internal state array when regarded as an array 5.23 + * of 32-bit integers.*/ 5.24 +#define N32 (N * 4) 5.25 +/** N64 is the size of internal state array when regarded as an array 5.26 + * of 64-bit integers.*/ 5.27 +#define N64 (N * 2) 5.28 + 5.29 +/*---------------------- 5.30 + the parameters of SFMT 5.31 + following definitions are in paramsXXXX.h file. 5.32 + ----------------------*/ 5.33 +/** the pick up position of the array. 5.34 +#define POS1 122 5.35 +*/ 5.36 + 5.37 +/** the parameter of shift left as four 32-bit registers. 5.38 +#define SL1 18 5.39 + */ 5.40 + 5.41 +/** the parameter of shift left as one 128-bit register. 5.42 + * The 128-bit integer is shifted by (SL2 * 8) bits. 5.43 +#define SL2 1 5.44 +*/ 5.45 + 5.46 +/** the parameter of shift right as four 32-bit registers. 5.47 +#define SR1 11 5.48 +*/ 5.49 + 5.50 +/** the parameter of shift right as one 128-bit register. 5.51 + * The 128-bit integer is shifted by (SL2 * 8) bits. 5.52 +#define SR2 1 5.53 +*/ 5.54 + 5.55 +/** A bitmask, used in the recursion. These parameters are introduced 5.56 + * to break symmetry of SIMD. 5.57 +#define MSK1 0xdfffffefU 5.58 +#define MSK2 0xddfecb7fU 5.59 +#define MSK3 0xbffaffffU 5.60 +#define MSK4 0xbffffff6U 5.61 +*/ 5.62 + 5.63 +/** These definitions are part of a 128-bit period certification vector. 5.64 +#define PARITY1 0x00000001U 5.65 +#define PARITY2 0x00000000U 5.66 +#define PARITY3 0x00000000U 5.67 +#define PARITY4 0xc98e126aU 5.68 +*/ 5.69 + 5.70 +#if MEXP == 607 5.71 + #include "SFMT-params607.h" 5.72 +#elif MEXP == 1279 5.73 + #include "SFMT-params1279.h" 5.74 +#elif MEXP == 2281 5.75 + #include "SFMT-params2281.h" 5.76 +#elif MEXP == 4253 5.77 + #include "SFMT-params4253.h" 5.78 +#elif MEXP == 11213 5.79 + #include "SFMT-params11213.h" 5.80 +#elif MEXP == 19937 5.81 + #include "SFMT-params19937.h" 5.82 +#elif MEXP == 44497 5.83 + #include "SFMT-params44497.h" 5.84 +#elif MEXP == 86243 5.85 + #include "SFMT-params86243.h" 5.86 +#elif MEXP == 132049 5.87 + #include "SFMT-params132049.h" 5.88 +#elif MEXP == 216091 5.89 + #include "SFMT-params216091.h" 5.90 +#else 5.91 +#ifdef __GNUC__ 5.92 + #error "MEXP is not valid." 5.93 + #undef MEXP 5.94 +#else 5.95 + #undef MEXP 5.96 +#endif 5.97 + 5.98 +#endif 5.99 + 5.100 +#endif /* SFMT_PARAMS_H */
6.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 6.2 +++ b/src/SFMT/SFMT-params11213.h Sun Mar 04 17:38:32 2012 -0600 6.3 @@ -0,0 +1,46 @@ 6.4 +#ifndef SFMT_PARAMS11213_H 6.5 +#define SFMT_PARAMS11213_H 6.6 + 6.7 +#define POS1 68 6.8 +#define SL1 14 6.9 +#define SL2 3 6.10 +#define SR1 7 6.11 +#define SR2 3 6.12 +#define MSK1 0xeffff7fbU 6.13 +#define MSK2 0xffffffefU 6.14 +#define MSK3 0xdfdfbfffU 6.15 +#define MSK4 0x7fffdbfdU 6.16 +#define PARITY1 0x00000001U 6.17 +#define PARITY2 0x00000000U 6.18 +#define PARITY3 0xe8148000U 6.19 +#define PARITY4 0xd0c7afa3U 6.20 + 6.21 + 6.22 +/* PARAMETERS FOR ALTIVEC */ 6.23 +#if defined(__APPLE__) /* For OSX */ 6.24 + #define ALTI_SL1 (vector unsigned int)(SL1, SL1, SL1, SL1) 6.25 + #define ALTI_SR1 (vector unsigned int)(SR1, SR1, SR1, SR1) 6.26 + #define ALTI_MSK (vector unsigned int)(MSK1, MSK2, MSK3, MSK4) 6.27 + #define ALTI_MSK64 \ 6.28 + (vector unsigned int)(MSK2, MSK1, MSK4, MSK3) 6.29 + #define ALTI_SL2_PERM \ 6.30 + (vector unsigned char)(3,21,21,21,7,0,1,2,11,4,5,6,15,8,9,10) 6.31 + #define ALTI_SL2_PERM64 \ 6.32 + (vector unsigned char)(3,4,5,6,7,29,29,29,11,12,13,14,15,0,1,2) 6.33 + #define ALTI_SR2_PERM \ 6.34 + (vector unsigned char)(5,6,7,0,9,10,11,4,13,14,15,8,19,19,19,12) 6.35 + #define ALTI_SR2_PERM64 \ 6.36 + (vector unsigned char)(13,14,15,0,1,2,3,4,19,19,19,8,9,10,11,12) 6.37 +#else /* For OTHER OSs(Linux?) */ 6.38 + #define ALTI_SL1 {SL1, SL1, SL1, SL1} 6.39 + #define ALTI_SR1 {SR1, SR1, SR1, SR1} 6.40 + #define ALTI_MSK {MSK1, MSK2, MSK3, MSK4} 6.41 + #define ALTI_MSK64 {MSK2, MSK1, MSK4, MSK3} 6.42 + #define ALTI_SL2_PERM {3,21,21,21,7,0,1,2,11,4,5,6,15,8,9,10} 6.43 + #define ALTI_SL2_PERM64 {3,4,5,6,7,29,29,29,11,12,13,14,15,0,1,2} 6.44 + #define ALTI_SR2_PERM {5,6,7,0,9,10,11,4,13,14,15,8,19,19,19,12} 6.45 + #define ALTI_SR2_PERM64 {13,14,15,0,1,2,3,4,19,19,19,8,9,10,11,12} 6.46 +#endif /* For OSX */ 6.47 +#define IDSTR "SFMT-11213:68-14-3-7-3:effff7fb-ffffffef-dfdfbfff-7fffdbfd" 6.48 + 6.49 +#endif /* SFMT_PARAMS11213_H */
7.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 7.2 +++ b/src/SFMT/SFMT-params1279.h Sun Mar 04 17:38:32 2012 -0600 7.3 @@ -0,0 +1,46 @@ 7.4 +#ifndef SFMT_PARAMS1279_H 7.5 +#define SFMT_PARAMS1279_H 7.6 + 7.7 +#define POS1 7 7.8 +#define SL1 14 7.9 +#define SL2 3 7.10 +#define SR1 5 7.11 +#define SR2 1 7.12 +#define MSK1 0xf7fefffdU 7.13 +#define MSK2 0x7fefcfffU 7.14 +#define MSK3 0xaff3ef3fU 7.15 +#define MSK4 0xb5ffff7fU 7.16 +#define PARITY1 0x00000001U 7.17 +#define PARITY2 0x00000000U 7.18 +#define PARITY3 0x00000000U 7.19 +#define PARITY4 0x20000000U 7.20 + 7.21 + 7.22 +/* PARAMETERS FOR ALTIVEC */ 7.23 +#if defined(__APPLE__) /* For OSX */ 7.24 + #define ALTI_SL1 (vector unsigned int)(SL1, SL1, SL1, SL1) 7.25 + #define ALTI_SR1 (vector unsigned int)(SR1, SR1, SR1, SR1) 7.26 + #define ALTI_MSK (vector unsigned int)(MSK1, MSK2, MSK3, MSK4) 7.27 + #define ALTI_MSK64 \ 7.28 + (vector unsigned int)(MSK2, MSK1, MSK4, MSK3) 7.29 + #define ALTI_SL2_PERM \ 7.30 + (vector unsigned char)(3,21,21,21,7,0,1,2,11,4,5,6,15,8,9,10) 7.31 + #define ALTI_SL2_PERM64 \ 7.32 + (vector unsigned char)(3,4,5,6,7,29,29,29,11,12,13,14,15,0,1,2) 7.33 + #define ALTI_SR2_PERM \ 7.34 + (vector unsigned char)(7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14) 7.35 + #define ALTI_SR2_PERM64 \ 7.36 + (vector unsigned char)(15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14) 7.37 +#else /* For OTHER OSs(Linux?) */ 7.38 + #define ALTI_SL1 {SL1, SL1, SL1, SL1} 7.39 + #define ALTI_SR1 {SR1, SR1, SR1, SR1} 7.40 + #define ALTI_MSK {MSK1, MSK2, MSK3, MSK4} 7.41 + #define ALTI_MSK64 {MSK2, MSK1, MSK4, MSK3} 7.42 + #define ALTI_SL2_PERM {3,21,21,21,7,0,1,2,11,4,5,6,15,8,9,10} 7.43 + #define ALTI_SL2_PERM64 {3,4,5,6,7,29,29,29,11,12,13,14,15,0,1,2} 7.44 + #define ALTI_SR2_PERM {7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14} 7.45 + #define ALTI_SR2_PERM64 {15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14} 7.46 +#endif /* For OSX */ 7.47 +#define IDSTR "SFMT-1279:7-14-3-5-1:f7fefffd-7fefcfff-aff3ef3f-b5ffff7f" 7.48 + 7.49 +#endif /* SFMT_PARAMS1279_H */
8.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 8.2 +++ b/src/SFMT/SFMT-params132049.h Sun Mar 04 17:38:32 2012 -0600 8.3 @@ -0,0 +1,46 @@ 8.4 +#ifndef SFMT_PARAMS132049_H 8.5 +#define SFMT_PARAMS132049_H 8.6 + 8.7 +#define POS1 110 8.8 +#define SL1 19 8.9 +#define SL2 1 8.10 +#define SR1 21 8.11 +#define SR2 1 8.12 +#define MSK1 0xffffbb5fU 8.13 +#define MSK2 0xfb6ebf95U 8.14 +#define MSK3 0xfffefffaU 8.15 +#define MSK4 0xcff77fffU 8.16 +#define PARITY1 0x00000001U 8.17 +#define PARITY2 0x00000000U 8.18 +#define PARITY3 0xcb520000U 8.19 +#define PARITY4 0xc7e91c7dU 8.20 + 8.21 + 8.22 +/* PARAMETERS FOR ALTIVEC */ 8.23 +#if defined(__APPLE__) /* For OSX */ 8.24 + #define ALTI_SL1 (vector unsigned int)(SL1, SL1, SL1, SL1) 8.25 + #define ALTI_SR1 (vector unsigned int)(SR1, SR1, SR1, SR1) 8.26 + #define ALTI_MSK (vector unsigned int)(MSK1, MSK2, MSK3, MSK4) 8.27 + #define ALTI_MSK64 \ 8.28 + (vector unsigned int)(MSK2, MSK1, MSK4, MSK3) 8.29 + #define ALTI_SL2_PERM \ 8.30 + (vector unsigned char)(1,2,3,23,5,6,7,0,9,10,11,4,13,14,15,8) 8.31 + #define ALTI_SL2_PERM64 \ 8.32 + (vector unsigned char)(1,2,3,4,5,6,7,31,9,10,11,12,13,14,15,0) 8.33 + #define ALTI_SR2_PERM \ 8.34 + (vector unsigned char)(7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14) 8.35 + #define ALTI_SR2_PERM64 \ 8.36 + (vector unsigned char)(15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14) 8.37 +#else /* For OTHER OSs(Linux?) */ 8.38 + #define ALTI_SL1 {SL1, SL1, SL1, SL1} 8.39 + #define ALTI_SR1 {SR1, SR1, SR1, SR1} 8.40 + #define ALTI_MSK {MSK1, MSK2, MSK3, MSK4} 8.41 + #define ALTI_MSK64 {MSK2, MSK1, MSK4, MSK3} 8.42 + #define ALTI_SL2_PERM {1,2,3,23,5,6,7,0,9,10,11,4,13,14,15,8} 8.43 + #define ALTI_SL2_PERM64 {1,2,3,4,5,6,7,31,9,10,11,12,13,14,15,0} 8.44 + #define ALTI_SR2_PERM {7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14} 8.45 + #define ALTI_SR2_PERM64 {15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14} 8.46 +#endif /* For OSX */ 8.47 +#define IDSTR "SFMT-132049:110-19-1-21-1:ffffbb5f-fb6ebf95-fffefffa-cff77fff" 8.48 + 8.49 +#endif /* SFMT_PARAMS132049_H */
9.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 9.2 +++ b/src/SFMT/SFMT-params19937.h Sun Mar 04 17:38:32 2012 -0600 9.3 @@ -0,0 +1,46 @@ 9.4 +#ifndef SFMT_PARAMS19937_H 9.5 +#define SFMT_PARAMS19937_H 9.6 + 9.7 +#define POS1 122 9.8 +#define SL1 18 9.9 +#define SL2 1 9.10 +#define SR1 11 9.11 +#define SR2 1 9.12 +#define MSK1 0xdfffffefU 9.13 +#define MSK2 0xddfecb7fU 9.14 +#define MSK3 0xbffaffffU 9.15 +#define MSK4 0xbffffff6U 9.16 +#define PARITY1 0x00000001U 9.17 +#define PARITY2 0x00000000U 9.18 +#define PARITY3 0x00000000U 9.19 +#define PARITY4 0x13c9e684U 9.20 + 9.21 + 9.22 +/* PARAMETERS FOR ALTIVEC */ 9.23 +#if defined(__APPLE__) /* For OSX */ 9.24 + #define ALTI_SL1 (vector unsigned int)(SL1, SL1, SL1, SL1) 9.25 + #define ALTI_SR1 (vector unsigned int)(SR1, SR1, SR1, SR1) 9.26 + #define ALTI_MSK (vector unsigned int)(MSK1, MSK2, MSK3, MSK4) 9.27 + #define ALTI_MSK64 \ 9.28 + (vector unsigned int)(MSK2, MSK1, MSK4, MSK3) 9.29 + #define ALTI_SL2_PERM \ 9.30 + (vector unsigned char)(1,2,3,23,5,6,7,0,9,10,11,4,13,14,15,8) 9.31 + #define ALTI_SL2_PERM64 \ 9.32 + (vector unsigned char)(1,2,3,4,5,6,7,31,9,10,11,12,13,14,15,0) 9.33 + #define ALTI_SR2_PERM \ 9.34 + (vector unsigned char)(7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14) 9.35 + #define ALTI_SR2_PERM64 \ 9.36 + (vector unsigned char)(15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14) 9.37 +#else /* For OTHER OSs(Linux?) */ 9.38 + #define ALTI_SL1 {SL1, SL1, SL1, SL1} 9.39 + #define ALTI_SR1 {SR1, SR1, SR1, SR1} 9.40 + #define ALTI_MSK {MSK1, MSK2, MSK3, MSK4} 9.41 + #define ALTI_MSK64 {MSK2, MSK1, MSK4, MSK3} 9.42 + #define ALTI_SL2_PERM {1,2,3,23,5,6,7,0,9,10,11,4,13,14,15,8} 9.43 + #define ALTI_SL2_PERM64 {1,2,3,4,5,6,7,31,9,10,11,12,13,14,15,0} 9.44 + #define ALTI_SR2_PERM {7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14} 9.45 + #define ALTI_SR2_PERM64 {15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14} 9.46 +#endif /* For OSX */ 9.47 +#define IDSTR "SFMT-19937:122-18-1-11-1:dfffffef-ddfecb7f-bffaffff-bffffff6" 9.48 + 9.49 +#endif /* SFMT_PARAMS19937_H */
10.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 10.2 +++ b/src/SFMT/SFMT-params216091.h Sun Mar 04 17:38:32 2012 -0600 10.3 @@ -0,0 +1,46 @@ 10.4 +#ifndef SFMT_PARAMS216091_H 10.5 +#define SFMT_PARAMS216091_H 10.6 + 10.7 +#define POS1 627 10.8 +#define SL1 11 10.9 +#define SL2 3 10.10 +#define SR1 10 10.11 +#define SR2 1 10.12 +#define MSK1 0xbff7bff7U 10.13 +#define MSK2 0xbfffffffU 10.14 +#define MSK3 0xbffffa7fU 10.15 +#define MSK4 0xffddfbfbU 10.16 +#define PARITY1 0xf8000001U 10.17 +#define PARITY2 0x89e80709U 10.18 +#define PARITY3 0x3bd2b64bU 10.19 +#define PARITY4 0x0c64b1e4U 10.20 + 10.21 + 10.22 +/* PARAMETERS FOR ALTIVEC */ 10.23 +#if defined(__APPLE__) /* For OSX */ 10.24 + #define ALTI_SL1 (vector unsigned int)(SL1, SL1, SL1, SL1) 10.25 + #define ALTI_SR1 (vector unsigned int)(SR1, SR1, SR1, SR1) 10.26 + #define ALTI_MSK (vector unsigned int)(MSK1, MSK2, MSK3, MSK4) 10.27 + #define ALTI_MSK64 \ 10.28 + (vector unsigned int)(MSK2, MSK1, MSK4, MSK3) 10.29 + #define ALTI_SL2_PERM \ 10.30 + (vector unsigned char)(3,21,21,21,7,0,1,2,11,4,5,6,15,8,9,10) 10.31 + #define ALTI_SL2_PERM64 \ 10.32 + (vector unsigned char)(3,4,5,6,7,29,29,29,11,12,13,14,15,0,1,2) 10.33 + #define ALTI_SR2_PERM \ 10.34 + (vector unsigned char)(7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14) 10.35 + #define ALTI_SR2_PERM64 \ 10.36 + (vector unsigned char)(15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14) 10.37 +#else /* For OTHER OSs(Linux?) */ 10.38 + #define ALTI_SL1 {SL1, SL1, SL1, SL1} 10.39 + #define ALTI_SR1 {SR1, SR1, SR1, SR1} 10.40 + #define ALTI_MSK {MSK1, MSK2, MSK3, MSK4} 10.41 + #define ALTI_MSK64 {MSK2, MSK1, MSK4, MSK3} 10.42 + #define ALTI_SL2_PERM {3,21,21,21,7,0,1,2,11,4,5,6,15,8,9,10} 10.43 + #define ALTI_SL2_PERM64 {3,4,5,6,7,29,29,29,11,12,13,14,15,0,1,2} 10.44 + #define ALTI_SR2_PERM {7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14} 10.45 + #define ALTI_SR2_PERM64 {15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14} 10.46 +#endif /* For OSX */ 10.47 +#define IDSTR "SFMT-216091:627-11-3-10-1:bff7bff7-bfffffff-bffffa7f-ffddfbfb" 10.48 + 10.49 +#endif /* SFMT_PARAMS216091_H */
11.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 11.2 +++ b/src/SFMT/SFMT-params2281.h Sun Mar 04 17:38:32 2012 -0600 11.3 @@ -0,0 +1,46 @@ 11.4 +#ifndef SFMT_PARAMS2281_H 11.5 +#define SFMT_PARAMS2281_H 11.6 + 11.7 +#define POS1 12 11.8 +#define SL1 19 11.9 +#define SL2 1 11.10 +#define SR1 5 11.11 +#define SR2 1 11.12 +#define MSK1 0xbff7ffbfU 11.13 +#define MSK2 0xfdfffffeU 11.14 +#define MSK3 0xf7ffef7fU 11.15 +#define MSK4 0xf2f7cbbfU 11.16 +#define PARITY1 0x00000001U 11.17 +#define PARITY2 0x00000000U 11.18 +#define PARITY3 0x00000000U 11.19 +#define PARITY4 0x41dfa600U 11.20 + 11.21 + 11.22 +/* PARAMETERS FOR ALTIVEC */ 11.23 +#if defined(__APPLE__) /* For OSX */ 11.24 + #define ALTI_SL1 (vector unsigned int)(SL1, SL1, SL1, SL1) 11.25 + #define ALTI_SR1 (vector unsigned int)(SR1, SR1, SR1, SR1) 11.26 + #define ALTI_MSK (vector unsigned int)(MSK1, MSK2, MSK3, MSK4) 11.27 + #define ALTI_MSK64 \ 11.28 + (vector unsigned int)(MSK2, MSK1, MSK4, MSK3) 11.29 + #define ALTI_SL2_PERM \ 11.30 + (vector unsigned char)(1,2,3,23,5,6,7,0,9,10,11,4,13,14,15,8) 11.31 + #define ALTI_SL2_PERM64 \ 11.32 + (vector unsigned char)(1,2,3,4,5,6,7,31,9,10,11,12,13,14,15,0) 11.33 + #define ALTI_SR2_PERM \ 11.34 + (vector unsigned char)(7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14) 11.35 + #define ALTI_SR2_PERM64 \ 11.36 + (vector unsigned char)(15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14) 11.37 +#else /* For OTHER OSs(Linux?) */ 11.38 + #define ALTI_SL1 {SL1, SL1, SL1, SL1} 11.39 + #define ALTI_SR1 {SR1, SR1, SR1, SR1} 11.40 + #define ALTI_MSK {MSK1, MSK2, MSK3, MSK4} 11.41 + #define ALTI_MSK64 {MSK2, MSK1, MSK4, MSK3} 11.42 + #define ALTI_SL2_PERM {1,2,3,23,5,6,7,0,9,10,11,4,13,14,15,8} 11.43 + #define ALTI_SL2_PERM64 {1,2,3,4,5,6,7,31,9,10,11,12,13,14,15,0} 11.44 + #define ALTI_SR2_PERM {7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14} 11.45 + #define ALTI_SR2_PERM64 {15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14} 11.46 +#endif /* For OSX */ 11.47 +#define IDSTR "SFMT-2281:12-19-1-5-1:bff7ffbf-fdfffffe-f7ffef7f-f2f7cbbf" 11.48 + 11.49 +#endif /* SFMT_PARAMS2281_H */
12.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 12.2 +++ b/src/SFMT/SFMT-params4253.h Sun Mar 04 17:38:32 2012 -0600 12.3 @@ -0,0 +1,46 @@ 12.4 +#ifndef SFMT_PARAMS4253_H 12.5 +#define SFMT_PARAMS4253_H 12.6 + 12.7 +#define POS1 17 12.8 +#define SL1 20 12.9 +#define SL2 1 12.10 +#define SR1 7 12.11 +#define SR2 1 12.12 +#define MSK1 0x9f7bffffU 12.13 +#define MSK2 0x9fffff5fU 12.14 +#define MSK3 0x3efffffbU 12.15 +#define MSK4 0xfffff7bbU 12.16 +#define PARITY1 0xa8000001U 12.17 +#define PARITY2 0xaf5390a3U 12.18 +#define PARITY3 0xb740b3f8U 12.19 +#define PARITY4 0x6c11486dU 12.20 + 12.21 + 12.22 +/* PARAMETERS FOR ALTIVEC */ 12.23 +#if defined(__APPLE__) /* For OSX */ 12.24 + #define ALTI_SL1 (vector unsigned int)(SL1, SL1, SL1, SL1) 12.25 + #define ALTI_SR1 (vector unsigned int)(SR1, SR1, SR1, SR1) 12.26 + #define ALTI_MSK (vector unsigned int)(MSK1, MSK2, MSK3, MSK4) 12.27 + #define ALTI_MSK64 \ 12.28 + (vector unsigned int)(MSK2, MSK1, MSK4, MSK3) 12.29 + #define ALTI_SL2_PERM \ 12.30 + (vector unsigned char)(1,2,3,23,5,6,7,0,9,10,11,4,13,14,15,8) 12.31 + #define ALTI_SL2_PERM64 \ 12.32 + (vector unsigned char)(1,2,3,4,5,6,7,31,9,10,11,12,13,14,15,0) 12.33 + #define ALTI_SR2_PERM \ 12.34 + (vector unsigned char)(7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14) 12.35 + #define ALTI_SR2_PERM64 \ 12.36 + (vector unsigned char)(15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14) 12.37 +#else /* For OTHER OSs(Linux?) */ 12.38 + #define ALTI_SL1 {SL1, SL1, SL1, SL1} 12.39 + #define ALTI_SR1 {SR1, SR1, SR1, SR1} 12.40 + #define ALTI_MSK {MSK1, MSK2, MSK3, MSK4} 12.41 + #define ALTI_MSK64 {MSK2, MSK1, MSK4, MSK3} 12.42 + #define ALTI_SL2_PERM {1,2,3,23,5,6,7,0,9,10,11,4,13,14,15,8} 12.43 + #define ALTI_SL2_PERM64 {1,2,3,4,5,6,7,31,9,10,11,12,13,14,15,0} 12.44 + #define ALTI_SR2_PERM {7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14} 12.45 + #define ALTI_SR2_PERM64 {15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14} 12.46 +#endif /* For OSX */ 12.47 +#define IDSTR "SFMT-4253:17-20-1-7-1:9f7bffff-9fffff5f-3efffffb-fffff7bb" 12.48 + 12.49 +#endif /* SFMT_PARAMS4253_H */
13.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 13.2 +++ b/src/SFMT/SFMT-params44497.h Sun Mar 04 17:38:32 2012 -0600 13.3 @@ -0,0 +1,46 @@ 13.4 +#ifndef SFMT_PARAMS44497_H 13.5 +#define SFMT_PARAMS44497_H 13.6 + 13.7 +#define POS1 330 13.8 +#define SL1 5 13.9 +#define SL2 3 13.10 +#define SR1 9 13.11 +#define SR2 3 13.12 +#define MSK1 0xeffffffbU 13.13 +#define MSK2 0xdfbebfffU 13.14 +#define MSK3 0xbfbf7befU 13.15 +#define MSK4 0x9ffd7bffU 13.16 +#define PARITY1 0x00000001U 13.17 +#define PARITY2 0x00000000U 13.18 +#define PARITY3 0xa3ac4000U 13.19 +#define PARITY4 0xecc1327aU 13.20 + 13.21 + 13.22 +/* PARAMETERS FOR ALTIVEC */ 13.23 +#if defined(__APPLE__) /* For OSX */ 13.24 + #define ALTI_SL1 (vector unsigned int)(SL1, SL1, SL1, SL1) 13.25 + #define ALTI_SR1 (vector unsigned int)(SR1, SR1, SR1, SR1) 13.26 + #define ALTI_MSK (vector unsigned int)(MSK1, MSK2, MSK3, MSK4) 13.27 + #define ALTI_MSK64 \ 13.28 + (vector unsigned int)(MSK2, MSK1, MSK4, MSK3) 13.29 + #define ALTI_SL2_PERM \ 13.30 + (vector unsigned char)(3,21,21,21,7,0,1,2,11,4,5,6,15,8,9,10) 13.31 + #define ALTI_SL2_PERM64 \ 13.32 + (vector unsigned char)(3,4,5,6,7,29,29,29,11,12,13,14,15,0,1,2) 13.33 + #define ALTI_SR2_PERM \ 13.34 + (vector unsigned char)(5,6,7,0,9,10,11,4,13,14,15,8,19,19,19,12) 13.35 + #define ALTI_SR2_PERM64 \ 13.36 + (vector unsigned char)(13,14,15,0,1,2,3,4,19,19,19,8,9,10,11,12) 13.37 +#else /* For OTHER OSs(Linux?) */ 13.38 + #define ALTI_SL1 {SL1, SL1, SL1, SL1} 13.39 + #define ALTI_SR1 {SR1, SR1, SR1, SR1} 13.40 + #define ALTI_MSK {MSK1, MSK2, MSK3, MSK4} 13.41 + #define ALTI_MSK64 {MSK2, MSK1, MSK4, MSK3} 13.42 + #define ALTI_SL2_PERM {3,21,21,21,7,0,1,2,11,4,5,6,15,8,9,10} 13.43 + #define ALTI_SL2_PERM64 {3,4,5,6,7,29,29,29,11,12,13,14,15,0,1,2} 13.44 + #define ALTI_SR2_PERM {5,6,7,0,9,10,11,4,13,14,15,8,19,19,19,12} 13.45 + #define ALTI_SR2_PERM64 {13,14,15,0,1,2,3,4,19,19,19,8,9,10,11,12} 13.46 +#endif /* For OSX */ 13.47 +#define IDSTR "SFMT-44497:330-5-3-9-3:effffffb-dfbebfff-bfbf7bef-9ffd7bff" 13.48 + 13.49 +#endif /* SFMT_PARAMS44497_H */
14.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 14.2 +++ b/src/SFMT/SFMT-params607.h Sun Mar 04 17:38:32 2012 -0600 14.3 @@ -0,0 +1,46 @@ 14.4 +#ifndef SFMT_PARAMS607_H 14.5 +#define SFMT_PARAMS607_H 14.6 + 14.7 +#define POS1 2 14.8 +#define SL1 15 14.9 +#define SL2 3 14.10 +#define SR1 13 14.11 +#define SR2 3 14.12 +#define MSK1 0xfdff37ffU 14.13 +#define MSK2 0xef7f3f7dU 14.14 +#define MSK3 0xff777b7dU 14.15 +#define MSK4 0x7ff7fb2fU 14.16 +#define PARITY1 0x00000001U 14.17 +#define PARITY2 0x00000000U 14.18 +#define PARITY3 0x00000000U 14.19 +#define PARITY4 0x5986f054U 14.20 + 14.21 + 14.22 +/* PARAMETERS FOR ALTIVEC */ 14.23 +#if defined(__APPLE__) /* For OSX */ 14.24 + #define ALTI_SL1 (vector unsigned int)(SL1, SL1, SL1, SL1) 14.25 + #define ALTI_SR1 (vector unsigned int)(SR1, SR1, SR1, SR1) 14.26 + #define ALTI_MSK (vector unsigned int)(MSK1, MSK2, MSK3, MSK4) 14.27 + #define ALTI_MSK64 \ 14.28 + (vector unsigned int)(MSK2, MSK1, MSK4, MSK3) 14.29 + #define ALTI_SL2_PERM \ 14.30 + (vector unsigned char)(3,21,21,21,7,0,1,2,11,4,5,6,15,8,9,10) 14.31 + #define ALTI_SL2_PERM64 \ 14.32 + (vector unsigned char)(3,4,5,6,7,29,29,29,11,12,13,14,15,0,1,2) 14.33 + #define ALTI_SR2_PERM \ 14.34 + (vector unsigned char)(5,6,7,0,9,10,11,4,13,14,15,8,19,19,19,12) 14.35 + #define ALTI_SR2_PERM64 \ 14.36 + (vector unsigned char)(13,14,15,0,1,2,3,4,19,19,19,8,9,10,11,12) 14.37 +#else /* For OTHER OSs(Linux?) */ 14.38 + #define ALTI_SL1 {SL1, SL1, SL1, SL1} 14.39 + #define ALTI_SR1 {SR1, SR1, SR1, SR1} 14.40 + #define ALTI_MSK {MSK1, MSK2, MSK3, MSK4} 14.41 + #define ALTI_MSK64 {MSK2, MSK1, MSK4, MSK3} 14.42 + #define ALTI_SL2_PERM {3,21,21,21,7,0,1,2,11,4,5,6,15,8,9,10} 14.43 + #define ALTI_SL2_PERM64 {3,4,5,6,7,29,29,29,11,12,13,14,15,0,1,2} 14.44 + #define ALTI_SR2_PERM {5,6,7,0,9,10,11,4,13,14,15,8,19,19,19,12} 14.45 + #define ALTI_SR2_PERM64 {13,14,15,0,1,2,3,4,19,19,19,8,9,10,11,12} 14.46 +#endif /* For OSX */ 14.47 +#define IDSTR "SFMT-607:2-15-3-13-3:fdff37ff-ef7f3f7d-ff777b7d-7ff7fb2f" 14.48 + 14.49 +#endif /* SFMT_PARAMS607_H */
15.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 15.2 +++ b/src/SFMT/SFMT-params86243.h Sun Mar 04 17:38:32 2012 -0600 15.3 @@ -0,0 +1,46 @@ 15.4 +#ifndef SFMT_PARAMS86243_H 15.5 +#define SFMT_PARAMS86243_H 15.6 + 15.7 +#define POS1 366 15.8 +#define SL1 6 15.9 +#define SL2 7 15.10 +#define SR1 19 15.11 +#define SR2 1 15.12 +#define MSK1 0xfdbffbffU 15.13 +#define MSK2 0xbff7ff3fU 15.14 +#define MSK3 0xfd77efffU 15.15 +#define MSK4 0xbf9ff3ffU 15.16 +#define PARITY1 0x00000001U 15.17 +#define PARITY2 0x00000000U 15.18 +#define PARITY3 0x00000000U 15.19 +#define PARITY4 0xe9528d85U 15.20 + 15.21 + 15.22 +/* PARAMETERS FOR ALTIVEC */ 15.23 +#if defined(__APPLE__) /* For OSX */ 15.24 + #define ALTI_SL1 (vector unsigned int)(SL1, SL1, SL1, SL1) 15.25 + #define ALTI_SR1 (vector unsigned int)(SR1, SR1, SR1, SR1) 15.26 + #define ALTI_MSK (vector unsigned int)(MSK1, MSK2, MSK3, MSK4) 15.27 + #define ALTI_MSK64 \ 15.28 + (vector unsigned int)(MSK2, MSK1, MSK4, MSK3) 15.29 + #define ALTI_SL2_PERM \ 15.30 + (vector unsigned char)(25,25,25,25,3,25,25,25,7,0,1,2,11,4,5,6) 15.31 + #define ALTI_SL2_PERM64 \ 15.32 + (vector unsigned char)(7,25,25,25,25,25,25,25,15,0,1,2,3,4,5,6) 15.33 + #define ALTI_SR2_PERM \ 15.34 + (vector unsigned char)(7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14) 15.35 + #define ALTI_SR2_PERM64 \ 15.36 + (vector unsigned char)(15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14) 15.37 +#else /* For OTHER OSs(Linux?) */ 15.38 + #define ALTI_SL1 {SL1, SL1, SL1, SL1} 15.39 + #define ALTI_SR1 {SR1, SR1, SR1, SR1} 15.40 + #define ALTI_MSK {MSK1, MSK2, MSK3, MSK4} 15.41 + #define ALTI_MSK64 {MSK2, MSK1, MSK4, MSK3} 15.42 + #define ALTI_SL2_PERM {25,25,25,25,3,25,25,25,7,0,1,2,11,4,5,6} 15.43 + #define ALTI_SL2_PERM64 {7,25,25,25,25,25,25,25,15,0,1,2,3,4,5,6} 15.44 + #define ALTI_SR2_PERM {7,0,1,2,11,4,5,6,15,8,9,10,17,12,13,14} 15.45 + #define ALTI_SR2_PERM64 {15,0,1,2,3,4,5,6,17,8,9,10,11,12,13,14} 15.46 +#endif /* For OSX */ 15.47 +#define IDSTR "SFMT-86243:366-6-7-19-1:fdbffbff-bff7ff3f-fd77efff-bf9ff3ff" 15.48 + 15.49 +#endif /* SFMT_PARAMS86243_H */
16.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 16.2 +++ b/src/SFMT/SFMT-sse2.h Sun Mar 04 17:38:32 2012 -0600 16.3 @@ -0,0 +1,121 @@ 16.4 +/** 16.5 + * @file SFMT-sse2.h 16.6 + * @brief SIMD oriented Fast Mersenne Twister(SFMT) for Intel SSE2 16.7 + * 16.8 + * @author Mutsuo Saito (Hiroshima University) 16.9 + * @author Makoto Matsumoto (Hiroshima University) 16.10 + * 16.11 + * @note We assume LITTLE ENDIAN in this file 16.12 + * 16.13 + * Copyright (C) 2006, 2007 Mutsuo Saito, Makoto Matsumoto and Hiroshima 16.14 + * University. All rights reserved. 16.15 + * 16.16 + * The new BSD License is applied to this software, see LICENSE.txt 16.17 + */ 16.18 + 16.19 +#ifndef SFMT_SSE2_H 16.20 +#define SFMT_SSE2_H 16.21 + 16.22 +PRE_ALWAYS static __m128i mm_recursion(__m128i *a, __m128i *b, __m128i c, 16.23 + __m128i d, __m128i mask) ALWAYSINLINE; 16.24 + 16.25 +/** 16.26 + * This function represents the recursion formula. 16.27 + * @param a a 128-bit part of the interal state array 16.28 + * @param b a 128-bit part of the interal state array 16.29 + * @param c a 128-bit part of the interal state array 16.30 + * @param d a 128-bit part of the interal state array 16.31 + * @param mask 128-bit mask 16.32 + * @return output 16.33 + */ 16.34 +PRE_ALWAYS static __m128i mm_recursion(__m128i *a, __m128i *b, 16.35 + __m128i c, __m128i d, __m128i mask) { 16.36 + __m128i v, x, y, z; 16.37 + 16.38 + x = _mm_load_si128(a); 16.39 + y = _mm_srli_epi32(*b, SR1); 16.40 + z = _mm_srli_si128(c, SR2); 16.41 + v = _mm_slli_epi32(d, SL1); 16.42 + z = _mm_xor_si128(z, x); 16.43 + z = _mm_xor_si128(z, v); 16.44 + x = _mm_slli_si128(x, SL2); 16.45 + y = _mm_and_si128(y, mask); 16.46 + z = _mm_xor_si128(z, x); 16.47 + z = _mm_xor_si128(z, y); 16.48 + return z; 16.49 +} 16.50 + 16.51 +/** 16.52 + * This function fills the internal state array with pseudorandom 16.53 + * integers. 16.54 + */ 16.55 +inline static void gen_rand_all(void) { 16.56 + int i; 16.57 + __m128i r, r1, r2, mask; 16.58 + mask = _mm_set_epi32(MSK4, MSK3, MSK2, MSK1); 16.59 + 16.60 + r1 = _mm_load_si128(&sfmt[N - 2].si); 16.61 + r2 = _mm_load_si128(&sfmt[N - 1].si); 16.62 + for (i = 0; i < N - POS1; i++) { 16.63 + r = mm_recursion(&sfmt[i].si, &sfmt[i + POS1].si, r1, r2, mask); 16.64 + _mm_store_si128(&sfmt[i].si, r); 16.65 + r1 = r2; 16.66 + r2 = r; 16.67 + } 16.68 + for (; i < N; i++) { 16.69 + r = mm_recursion(&sfmt[i].si, &sfmt[i + POS1 - N].si, r1, r2, mask); 16.70 + _mm_store_si128(&sfmt[i].si, r); 16.71 + r1 = r2; 16.72 + r2 = r; 16.73 + } 16.74 +} 16.75 + 16.76 +/** 16.77 + * This function fills the user-specified array with pseudorandom 16.78 + * integers. 16.79 + * 16.80 + * @param array an 128-bit array to be filled by pseudorandom numbers. 16.81 + * @param size number of 128-bit pesudorandom numbers to be generated. 16.82 + */ 16.83 +inline static void gen_rand_array(w128_t *array, int size) { 16.84 + int i, j; 16.85 + __m128i r, r1, r2, mask; 16.86 + mask = _mm_set_epi32(MSK4, MSK3, MSK2, MSK1); 16.87 + 16.88 + r1 = _mm_load_si128(&sfmt[N - 2].si); 16.89 + r2 = _mm_load_si128(&sfmt[N - 1].si); 16.90 + for (i = 0; i < N - POS1; i++) { 16.91 + r = mm_recursion(&sfmt[i].si, &sfmt[i + POS1].si, r1, r2, mask); 16.92 + _mm_store_si128(&array[i].si, r); 16.93 + r1 = r2; 16.94 + r2 = r; 16.95 + } 16.96 + for (; i < N; i++) { 16.97 + r = mm_recursion(&sfmt[i].si, &array[i + POS1 - N].si, r1, r2, mask); 16.98 + _mm_store_si128(&array[i].si, r); 16.99 + r1 = r2; 16.100 + r2 = r; 16.101 + } 16.102 + /* main loop */ 16.103 + for (; i < size - N; i++) { 16.104 + r = mm_recursion(&array[i - N].si, &array[i + POS1 - N].si, r1, r2, 16.105 + mask); 16.106 + _mm_store_si128(&array[i].si, r); 16.107 + r1 = r2; 16.108 + r2 = r; 16.109 + } 16.110 + for (j = 0; j < 2 * N - size; j++) { 16.111 + r = _mm_load_si128(&array[j + size - N].si); 16.112 + _mm_store_si128(&sfmt[j].si, r); 16.113 + } 16.114 + for (; i < size; i++) { 16.115 + r = mm_recursion(&array[i - N].si, &array[i + POS1 - N].si, r1, r2, 16.116 + mask); 16.117 + _mm_store_si128(&array[i].si, r); 16.118 + _mm_store_si128(&sfmt[j++].si, r); 16.119 + r1 = r2; 16.120 + r2 = r; 16.121 + } 16.122 +} 16.123 + 16.124 +#endif
17.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 17.2 +++ b/src/SFMT/SFMT.c Sun Mar 04 17:38:32 2012 -0600 17.3 @@ -0,0 +1,620 @@ 17.4 +/** 17.5 + * @file SFMT.c 17.6 + * @brief SIMD oriented Fast Mersenne Twister(SFMT) 17.7 + * 17.8 + * @author Mutsuo Saito (Hiroshima University) 17.9 + * @author Makoto Matsumoto (Hiroshima University) 17.10 + * 17.11 + * Copyright (C) 2006,2007 Mutsuo Saito, Makoto Matsumoto and Hiroshima 17.12 + * University. All rights reserved. 17.13 + * 17.14 + * The new BSD License is applied to this software, see LICENSE.txt 17.15 + */ 17.16 +#include <string.h> 17.17 +#include <assert.h> 17.18 +#include "SFMT.h" 17.19 +#include "SFMT-params.h" 17.20 + 17.21 +#if defined(__BIG_ENDIAN__) && !defined(__amd64) && !defined(BIG_ENDIAN64) 17.22 +#define BIG_ENDIAN64 1 17.23 +#endif 17.24 +#if defined(HAVE_ALTIVEC) && !defined(BIG_ENDIAN64) 17.25 +#define BIG_ENDIAN64 1 17.26 +#endif 17.27 +#if defined(ONLY64) && !defined(BIG_ENDIAN64) 17.28 + #if defined(__GNUC__) 17.29 + #error "-DONLY64 must be specified with -DBIG_ENDIAN64" 17.30 + #endif 17.31 +#undef ONLY64 17.32 +#endif 17.33 +/*------------------------------------------------------ 17.34 + 128-bit SIMD data type for Altivec, SSE2 or standard C 17.35 + ------------------------------------------------------*/ 17.36 +#if defined(HAVE_ALTIVEC) 17.37 + #if !defined(__APPLE__) 17.38 + #include <altivec.h> 17.39 + #endif 17.40 +/** 128-bit data structure */ 17.41 +union W128_T { 17.42 + vector unsigned int s; 17.43 + uint32_t u[4]; 17.44 +}; 17.45 +/** 128-bit data type */ 17.46 +typedef union W128_T w128_t; 17.47 + 17.48 +#elif defined(HAVE_SSE2) 17.49 + #include <emmintrin.h> 17.50 + 17.51 +/** 128-bit data structure */ 17.52 +union W128_T { 17.53 + __m128i si; 17.54 + uint32_t u[4]; 17.55 +}; 17.56 +/** 128-bit data type */ 17.57 +typedef union W128_T w128_t; 17.58 + 17.59 +#else 17.60 + 17.61 +/** 128-bit data structure */ 17.62 +struct W128_T { 17.63 + uint32_t u[4]; 17.64 +}; 17.65 +/** 128-bit data type */ 17.66 +typedef struct W128_T w128_t; 17.67 + 17.68 +#endif 17.69 + 17.70 +/*-------------------------------------- 17.71 + FILE GLOBAL VARIABLES 17.72 + internal state, index counter and flag 17.73 + --------------------------------------*/ 17.74 +/** the 128-bit internal state array */ 17.75 +static w128_t sfmt[N]; 17.76 +/** the 32bit integer pointer to the 128-bit internal state array */ 17.77 +static uint32_t *psfmt32 = &sfmt[0].u[0]; 17.78 +#if !defined(BIG_ENDIAN64) || defined(ONLY64) 17.79 +/** the 64bit integer pointer to the 128-bit internal state array */ 17.80 +static uint64_t *psfmt64 = (uint64_t *)&sfmt[0].u[0]; 17.81 +#endif 17.82 +/** index counter to the 32-bit internal state array */ 17.83 +static int idx; 17.84 +/** a flag: it is 0 if and only if the internal state is not yet 17.85 + * initialized. */ 17.86 +static int initialized = 0; 17.87 +/** a parity check vector which certificate the period of 2^{MEXP} */ 17.88 +static uint32_t parity[4] = {PARITY1, PARITY2, PARITY3, PARITY4}; 17.89 + 17.90 +/*---------------- 17.91 + STATIC FUNCTIONS 17.92 + ----------------*/ 17.93 +inline static int idxof(int i); 17.94 +inline static void rshift128(w128_t *out, w128_t const *in, int shift); 17.95 +inline static void lshift128(w128_t *out, w128_t const *in, int shift); 17.96 +inline static void gen_rand_all(void); 17.97 +inline static void gen_rand_array(w128_t *array, int size); 17.98 +inline static uint32_t func1(uint32_t x); 17.99 +inline static uint32_t func2(uint32_t x); 17.100 +static void period_certification(void); 17.101 +#if defined(BIG_ENDIAN64) && !defined(ONLY64) 17.102 +inline static void swap(w128_t *array, int size); 17.103 +#endif 17.104 + 17.105 +#if defined(HAVE_ALTIVEC) 17.106 + #include "SFMT-alti.h" 17.107 +#elif defined(HAVE_SSE2) 17.108 + #include "SFMT-sse2.h" 17.109 +#endif 17.110 + 17.111 +/** 17.112 + * This function simulate a 64-bit index of LITTLE ENDIAN 17.113 + * in BIG ENDIAN machine. 17.114 + */ 17.115 +#ifdef ONLY64 17.116 +inline static int idxof(int i) { 17.117 + return i ^ 1; 17.118 +} 17.119 +#else 17.120 +inline static int idxof(int i) { 17.121 + return i; 17.122 +} 17.123 +#endif 17.124 +/** 17.125 + * This function simulates SIMD 128-bit right shift by the standard C. 17.126 + * The 128-bit integer given in in is shifted by (shift * 8) bits. 17.127 + * This function simulates the LITTLE ENDIAN SIMD. 17.128 + * @param out the output of this function 17.129 + * @param in the 128-bit data to be shifted 17.130 + * @param shift the shift value 17.131 + */ 17.132 +#ifdef ONLY64 17.133 +inline static void rshift128(w128_t *out, w128_t const *in, int shift) { 17.134 + uint64_t th, tl, oh, ol; 17.135 + 17.136 + th = ((uint64_t)in->u[2] << 32) | ((uint64_t)in->u[3]); 17.137 + tl = ((uint64_t)in->u[0] << 32) | ((uint64_t)in->u[1]); 17.138 + 17.139 + oh = th >> (shift * 8); 17.140 + ol = tl >> (shift * 8); 17.141 + ol |= th << (64 - shift * 8); 17.142 + out->u[0] = (uint32_t)(ol >> 32); 17.143 + out->u[1] = (uint32_t)(ol & 0xffffffff); 17.144 + out->u[2] = (uint32_t)(oh >> 32); 17.145 + out->u[3] = (uint32_t)(oh & 0xffffffff); 17.146 +} 17.147 +#else 17.148 +inline static void rshift128(w128_t *out, w128_t const *in, int shift) { 17.149 + uint64_t th, tl, oh, ol; 17.150 + 17.151 + th = ((uint64_t)in->u[3] << 32) | ((uint64_t)in->u[2]); 17.152 + tl = ((uint64_t)in->u[1] << 32) | ((uint64_t)in->u[0]); 17.153 + 17.154 + oh = th >> (shift * 8); 17.155 + ol = tl >> (shift * 8); 17.156 + ol |= th << (64 - shift * 8); 17.157 + out->u[1] = (uint32_t)(ol >> 32); 17.158 + out->u[0] = (uint32_t)(ol & 0xffffffff); 17.159 + out->u[3] = (uint32_t)(oh >> 32); 17.160 + out->u[2] = (uint32_t)(oh & 0xffffffff); 17.161 +} 17.162 +#endif 17.163 +/** 17.164 + * This function simulates SIMD 128-bit left shift by the standard C. 17.165 + * The 128-bit integer given in in is shifted by (shift * 8) bits. 17.166 + * This function simulates the LITTLE ENDIAN SIMD. 17.167 + * @param out the output of this function 17.168 + * @param in the 128-bit data to be shifted 17.169 + * @param shift the shift value 17.170 + */ 17.171 +#ifdef ONLY64 17.172 +inline static void lshift128(w128_t *out, w128_t const *in, int shift) { 17.173 + uint64_t th, tl, oh, ol; 17.174 + 17.175 + th = ((uint64_t)in->u[2] << 32) | ((uint64_t)in->u[3]); 17.176 + tl = ((uint64_t)in->u[0] << 32) | ((uint64_t)in->u[1]); 17.177 + 17.178 + oh = th << (shift * 8); 17.179 + ol = tl << (shift * 8); 17.180 + oh |= tl >> (64 - shift * 8); 17.181 + out->u[0] = (uint32_t)(ol >> 32); 17.182 + out->u[1] = (uint32_t)(ol & 0xffffffff); 17.183 + out->u[2] = (uint32_t)(oh >> 32); 17.184 + out->u[3] = (uint32_t)(oh & 0xffffffff); 17.185 +} 17.186 +#else 17.187 +inline static void lshift128(w128_t *out, w128_t const *in, int shift) { 17.188 + uint64_t th, tl, oh, ol; 17.189 + 17.190 + th = ((uint64_t)in->u[3] << 32) | ((uint64_t)in->u[2]); 17.191 + tl = ((uint64_t)in->u[1] << 32) | ((uint64_t)in->u[0]); 17.192 + 17.193 + oh = th << (shift * 8); 17.194 + ol = tl << (shift * 8); 17.195 + oh |= tl >> (64 - shift * 8); 17.196 + out->u[1] = (uint32_t)(ol >> 32); 17.197 + out->u[0] = (uint32_t)(ol & 0xffffffff); 17.198 + out->u[3] = (uint32_t)(oh >> 32); 17.199 + out->u[2] = (uint32_t)(oh & 0xffffffff); 17.200 +} 17.201 +#endif 17.202 + 17.203 +/** 17.204 + * This function represents the recursion formula. 17.205 + * @param r output 17.206 + * @param a a 128-bit part of the internal state array 17.207 + * @param b a 128-bit part of the internal state array 17.208 + * @param c a 128-bit part of the internal state array 17.209 + * @param d a 128-bit part of the internal state array 17.210 + */ 17.211 +#if (!defined(HAVE_ALTIVEC)) && (!defined(HAVE_SSE2)) 17.212 +#ifdef ONLY64 17.213 +inline static void do_recursion(w128_t *r, w128_t *a, w128_t *b, w128_t *c, 17.214 + w128_t *d) { 17.215 + w128_t x; 17.216 + w128_t y; 17.217 + 17.218 + lshift128(&x, a, SL2); 17.219 + rshift128(&y, c, SR2); 17.220 + r->u[0] = a->u[0] ^ x.u[0] ^ ((b->u[0] >> SR1) & MSK2) ^ y.u[0] 17.221 + ^ (d->u[0] << SL1); 17.222 + r->u[1] = a->u[1] ^ x.u[1] ^ ((b->u[1] >> SR1) & MSK1) ^ y.u[1] 17.223 + ^ (d->u[1] << SL1); 17.224 + r->u[2] = a->u[2] ^ x.u[2] ^ ((b->u[2] >> SR1) & MSK4) ^ y.u[2] 17.225 + ^ (d->u[2] << SL1); 17.226 + r->u[3] = a->u[3] ^ x.u[3] ^ ((b->u[3] >> SR1) & MSK3) ^ y.u[3] 17.227 + ^ (d->u[3] << SL1); 17.228 +} 17.229 +#else 17.230 +inline static void do_recursion(w128_t *r, w128_t *a, w128_t *b, w128_t *c, 17.231 + w128_t *d) { 17.232 + w128_t x; 17.233 + w128_t y; 17.234 + 17.235 + lshift128(&x, a, SL2); 17.236 + rshift128(&y, c, SR2); 17.237 + r->u[0] = a->u[0] ^ x.u[0] ^ ((b->u[0] >> SR1) & MSK1) ^ y.u[0] 17.238 + ^ (d->u[0] << SL1); 17.239 + r->u[1] = a->u[1] ^ x.u[1] ^ ((b->u[1] >> SR1) & MSK2) ^ y.u[1] 17.240 + ^ (d->u[1] << SL1); 17.241 + r->u[2] = a->u[2] ^ x.u[2] ^ ((b->u[2] >> SR1) & MSK3) ^ y.u[2] 17.242 + ^ (d->u[2] << SL1); 17.243 + r->u[3] = a->u[3] ^ x.u[3] ^ ((b->u[3] >> SR1) & MSK4) ^ y.u[3] 17.244 + ^ (d->u[3] << SL1); 17.245 +} 17.246 +#endif 17.247 +#endif 17.248 + 17.249 +#if (!defined(HAVE_ALTIVEC)) && (!defined(HAVE_SSE2)) 17.250 +/** 17.251 + * This function fills the internal state array with pseudorandom 17.252 + * integers. 17.253 + */ 17.254 +inline static void gen_rand_all(void) { 17.255 + int i; 17.256 + w128_t *r1, *r2; 17.257 + 17.258 + r1 = &sfmt[N - 2]; 17.259 + r2 = &sfmt[N - 1]; 17.260 + for (i = 0; i < N - POS1; i++) { 17.261 + do_recursion(&sfmt[i], &sfmt[i], &sfmt[i + POS1], r1, r2); 17.262 + r1 = r2; 17.263 + r2 = &sfmt[i]; 17.264 + } 17.265 + for (; i < N; i++) { 17.266 + do_recursion(&sfmt[i], &sfmt[i], &sfmt[i + POS1 - N], r1, r2); 17.267 + r1 = r2; 17.268 + r2 = &sfmt[i]; 17.269 + } 17.270 +} 17.271 + 17.272 +/** 17.273 + * This function fills the user-specified array with pseudorandom 17.274 + * integers. 17.275 + * 17.276 + * @param array an 128-bit array to be filled by pseudorandom numbers. 17.277 + * @param size number of 128-bit pseudorandom numbers to be generated. 17.278 + */ 17.279 +inline static void gen_rand_array(w128_t *array, int size) { 17.280 + int i, j; 17.281 + w128_t *r1, *r2; 17.282 + 17.283 + r1 = &sfmt[N - 2]; 17.284 + r2 = &sfmt[N - 1]; 17.285 + for (i = 0; i < N - POS1; i++) { 17.286 + do_recursion(&array[i], &sfmt[i], &sfmt[i + POS1], r1, r2); 17.287 + r1 = r2; 17.288 + r2 = &array[i]; 17.289 + } 17.290 + for (; i < N; i++) { 17.291 + do_recursion(&array[i], &sfmt[i], &array[i + POS1 - N], r1, r2); 17.292 + r1 = r2; 17.293 + r2 = &array[i]; 17.294 + } 17.295 + for (; i < size - N; i++) { 17.296 + do_recursion(&array[i], &array[i - N], &array[i + POS1 - N], r1, r2); 17.297 + r1 = r2; 17.298 + r2 = &array[i]; 17.299 + } 17.300 + for (j = 0; j < 2 * N - size; j++) { 17.301 + sfmt[j] = array[j + size - N]; 17.302 + } 17.303 + for (; i < size; i++, j++) { 17.304 + do_recursion(&array[i], &array[i - N], &array[i + POS1 - N], r1, r2); 17.305 + r1 = r2; 17.306 + r2 = &array[i]; 17.307 + sfmt[j] = array[i]; 17.308 + } 17.309 +} 17.310 +#endif 17.311 + 17.312 +#if defined(BIG_ENDIAN64) && !defined(ONLY64) && !defined(HAVE_ALTIVEC) 17.313 +inline static void swap(w128_t *array, int size) { 17.314 + int i; 17.315 + uint32_t x, y; 17.316 + 17.317 + for (i = 0; i < size; i++) { 17.318 + x = array[i].u[0]; 17.319 + y = array[i].u[2]; 17.320 + array[i].u[0] = array[i].u[1]; 17.321 + array[i].u[2] = array[i].u[3]; 17.322 + array[i].u[1] = x; 17.323 + array[i].u[3] = y; 17.324 + } 17.325 +} 17.326 +#endif 17.327 +/** 17.328 + * This function represents a function used in the initialization 17.329 + * by init_by_array 17.330 + * @param x 32-bit integer 17.331 + * @return 32-bit integer 17.332 + */ 17.333 +static uint32_t func1(uint32_t x) { 17.334 + return (x ^ (x >> 27)) * (uint32_t)1664525UL; 17.335 +} 17.336 + 17.337 +/** 17.338 + * This function represents a function used in the initialization 17.339 + * by init_by_array 17.340 + * @param x 32-bit integer 17.341 + * @return 32-bit integer 17.342 + */ 17.343 +static uint32_t func2(uint32_t x) { 17.344 + return (x ^ (x >> 27)) * (uint32_t)1566083941UL; 17.345 +} 17.346 + 17.347 +/** 17.348 + * This function certificate the period of 2^{MEXP} 17.349 + */ 17.350 +static void period_certification(void) { 17.351 + int inner = 0; 17.352 + int i, j; 17.353 + uint32_t work; 17.354 + 17.355 + for (i = 0; i < 4; i++) 17.356 + inner ^= psfmt32[idxof(i)] & parity[i]; 17.357 + for (i = 16; i > 0; i >>= 1) 17.358 + inner ^= inner >> i; 17.359 + inner &= 1; 17.360 + /* check OK */ 17.361 + if (inner == 1) { 17.362 + return; 17.363 + } 17.364 + /* check NG, and modification */ 17.365 + for (i = 0; i < 4; i++) { 17.366 + work = 1; 17.367 + for (j = 0; j < 32; j++) { 17.368 + if ((work & parity[i]) != 0) { 17.369 + psfmt32[idxof(i)] ^= work; 17.370 + return; 17.371 + } 17.372 + work = work << 1; 17.373 + } 17.374 + } 17.375 +} 17.376 + 17.377 +/*---------------- 17.378 + PUBLIC FUNCTIONS 17.379 + ----------------*/ 17.380 +/** 17.381 + * This function returns the identification string. 17.382 + * The string shows the word size, the Mersenne exponent, 17.383 + * and all parameters of this generator. 17.384 + */ 17.385 +const char *get_idstring(void) { 17.386 + return IDSTR; 17.387 +} 17.388 + 17.389 +/** 17.390 + * This function returns the minimum size of array used for \b 17.391 + * fill_array32() function. 17.392 + * @return minimum size of array used for fill_array32() function. 17.393 + */ 17.394 +int get_min_array_size32(void) { 17.395 + return N32; 17.396 +} 17.397 + 17.398 +/** 17.399 + * This function returns the minimum size of array used for \b 17.400 + * fill_array64() function. 17.401 + * @return minimum size of array used for fill_array64() function. 17.402 + */ 17.403 +int get_min_array_size64(void) { 17.404 + return N64; 17.405 +} 17.406 + 17.407 +#ifndef ONLY64 17.408 +/** 17.409 + * This function generates and returns 32-bit pseudorandom number. 17.410 + * init_gen_rand or init_by_array must be called before this function. 17.411 + * @return 32-bit pseudorandom number 17.412 + */ 17.413 +uint32_t gen_rand32(void) { 17.414 + uint32_t r; 17.415 + 17.416 + assert(initialized); 17.417 + if (idx >= N32) { 17.418 + gen_rand_all(); 17.419 + idx = 0; 17.420 + } 17.421 + r = psfmt32[idx++]; 17.422 + return r; 17.423 +} 17.424 +#endif 17.425 +/** 17.426 + * This function generates and returns 64-bit pseudorandom number. 17.427 + * init_gen_rand or init_by_array must be called before this function. 17.428 + * The function gen_rand64 should not be called after gen_rand32, 17.429 + * unless an initialization is again executed. 17.430 + * @return 64-bit pseudorandom number 17.431 + */ 17.432 +uint64_t gen_rand64(void) { 17.433 +#if defined(BIG_ENDIAN64) && !defined(ONLY64) 17.434 + uint32_t r1, r2; 17.435 +#else 17.436 + uint64_t r; 17.437 +#endif 17.438 + 17.439 + assert(initialized); 17.440 + assert(idx % 2 == 0); 17.441 + 17.442 + if (idx >= N32) { 17.443 + gen_rand_all(); 17.444 + idx = 0; 17.445 + } 17.446 +#if defined(BIG_ENDIAN64) && !defined(ONLY64) 17.447 + r1 = psfmt32[idx]; 17.448 + r2 = psfmt32[idx + 1]; 17.449 + idx += 2; 17.450 + return ((uint64_t)r2 << 32) | r1; 17.451 +#else 17.452 + r = psfmt64[idx / 2]; 17.453 + idx += 2; 17.454 + return r; 17.455 +#endif 17.456 +} 17.457 + 17.458 +#ifndef ONLY64 17.459 +/** 17.460 + * This function generates pseudorandom 32-bit integers in the 17.461 + * specified array[] by one call. The number of pseudorandom integers 17.462 + * is specified by the argument size, which must be at least 624 and a 17.463 + * multiple of four. The generation by this function is much faster 17.464 + * than the following gen_rand function. 17.465 + * 17.466 + * For initialization, init_gen_rand or init_by_array must be called 17.467 + * before the first call of this function. This function can not be 17.468 + * used after calling gen_rand function, without initialization. 17.469 + * 17.470 + * @param array an array where pseudorandom 32-bit integers are filled 17.471 + * by this function. The pointer to the array must be \b "aligned" 17.472 + * (namely, must be a multiple of 16) in the SIMD version, since it 17.473 + * refers to the address of a 128-bit integer. In the standard C 17.474 + * version, the pointer is arbitrary. 17.475 + * 17.476 + * @param size the number of 32-bit pseudorandom integers to be 17.477 + * generated. size must be a multiple of 4, and greater than or equal 17.478 + * to (MEXP / 128 + 1) * 4. 17.479 + * 17.480 + * @note \b memalign or \b posix_memalign is available to get aligned 17.481 + * memory. Mac OSX doesn't have these functions, but \b malloc of OSX 17.482 + * returns the pointer to the aligned memory block. 17.483 + */ 17.484 +void fill_array32(uint32_t *array, int size) { 17.485 + assert(initialized); 17.486 + assert(idx == N32); 17.487 + assert(size % 4 == 0); 17.488 + assert(size >= N32); 17.489 + 17.490 + gen_rand_array((w128_t *)array, size / 4); 17.491 + idx = N32; 17.492 +} 17.493 +#endif 17.494 + 17.495 +/** 17.496 + * This function generates pseudorandom 64-bit integers in the 17.497 + * specified array[] by one call. The number of pseudorandom integers 17.498 + * is specified by the argument size, which must be at least 312 and a 17.499 + * multiple of two. The generation by this function is much faster 17.500 + * than the following gen_rand function. 17.501 + * 17.502 + * For initialization, init_gen_rand or init_by_array must be called 17.503 + * before the first call of this function. This function can not be 17.504 + * used after calling gen_rand function, without initialization. 17.505 + * 17.506 + * @param array an array where pseudorandom 64-bit integers are filled 17.507 + * by this function. The pointer to the array must be "aligned" 17.508 + * (namely, must be a multiple of 16) in the SIMD version, since it 17.509 + * refers to the address of a 128-bit integer. In the standard C 17.510 + * version, the pointer is arbitrary. 17.511 + * 17.512 + * @param size the number of 64-bit pseudorandom integers to be 17.513 + * generated. size must be a multiple of 2, and greater than or equal 17.514 + * to (MEXP / 128 + 1) * 2 17.515 + * 17.516 + * @note \b memalign or \b posix_memalign is available to get aligned 17.517 + * memory. Mac OSX doesn't have these functions, but \b malloc of OSX 17.518 + * returns the pointer to the aligned memory block. 17.519 + */ 17.520 +void fill_array64(uint64_t *array, int size) { 17.521 + assert(initialized); 17.522 + assert(idx == N32); 17.523 + assert(size % 2 == 0); 17.524 + assert(size >= N64); 17.525 + 17.526 + gen_rand_array((w128_t *)array, size / 2); 17.527 + idx = N32; 17.528 + 17.529 +#if defined(BIG_ENDIAN64) && !defined(ONLY64) 17.530 + swap((w128_t *)array, size /2); 17.531 +#endif 17.532 +} 17.533 + 17.534 +/** 17.535 + * This function initializes the internal state array with a 32-bit 17.536 + * integer seed. 17.537 + * 17.538 + * @param seed a 32-bit integer used as the seed. 17.539 + */ 17.540 +void init_gen_rand(uint32_t seed) { 17.541 + int i; 17.542 + 17.543 + psfmt32[idxof(0)] = seed; 17.544 + for (i = 1; i < N32; i++) { 17.545 + psfmt32[idxof(i)] = 1812433253UL * (psfmt32[idxof(i - 1)] 17.546 + ^ (psfmt32[idxof(i - 1)] >> 30)) 17.547 + + i; 17.548 + } 17.549 + idx = N32; 17.550 + period_certification(); 17.551 + initialized = 1; 17.552 +} 17.553 + 17.554 +/** 17.555 + * This function initializes the internal state array, 17.556 + * with an array of 32-bit integers used as the seeds 17.557 + * @param init_key the array of 32-bit integers, used as a seed. 17.558 + * @param key_length the length of init_key. 17.559 + */ 17.560 +void init_by_array(uint32_t *init_key, int key_length) { 17.561 + int i, j, count; 17.562 + uint32_t r; 17.563 + int lag; 17.564 + int mid; 17.565 + int size = N * 4; 17.566 + 17.567 + if (size >= 623) { 17.568 + lag = 11; 17.569 + } else if (size >= 68) { 17.570 + lag = 7; 17.571 + } else if (size >= 39) { 17.572 + lag = 5; 17.573 + } else { 17.574 + lag = 3; 17.575 + } 17.576 + mid = (size - lag) / 2; 17.577 + 17.578 + memset(sfmt, 0x8b, sizeof(sfmt)); 17.579 + if (key_length + 1 > N32) { 17.580 + count = key_length + 1; 17.581 + } else { 17.582 + count = N32; 17.583 + } 17.584 + r = func1(psfmt32[idxof(0)] ^ psfmt32[idxof(mid)] 17.585 + ^ psfmt32[idxof(N32 - 1)]); 17.586 + psfmt32[idxof(mid)] += r; 17.587 + r += key_length; 17.588 + psfmt32[idxof(mid + lag)] += r; 17.589 + psfmt32[idxof(0)] = r; 17.590 + 17.591 + count--; 17.592 + for (i = 1, j = 0; (j < count) && (j < key_length); j++) { 17.593 + r = func1(psfmt32[idxof(i)] ^ psfmt32[idxof((i + mid) % N32)] 17.594 + ^ psfmt32[idxof((i + N32 - 1) % N32)]); 17.595 + psfmt32[idxof((i + mid) % N32)] += r; 17.596 + r += init_key[j] + i; 17.597 + psfmt32[idxof((i + mid + lag) % N32)] += r; 17.598 + psfmt32[idxof(i)] = r; 17.599 + i = (i + 1) % N32; 17.600 + } 17.601 + for (; j < count; j++) { 17.602 + r = func1(psfmt32[idxof(i)] ^ psfmt32[idxof((i + mid) % N32)] 17.603 + ^ psfmt32[idxof((i + N32 - 1) % N32)]); 17.604 + psfmt32[idxof((i + mid) % N32)] += r; 17.605 + r += i; 17.606 + psfmt32[idxof((i + mid + lag) % N32)] += r; 17.607 + psfmt32[idxof(i)] = r; 17.608 + i = (i + 1) % N32; 17.609 + } 17.610 + for (j = 0; j < N32; j++) { 17.611 + r = func2(psfmt32[idxof(i)] + psfmt32[idxof((i + mid) % N32)] 17.612 + + psfmt32[idxof((i + N32 - 1) % N32)]); 17.613 + psfmt32[idxof((i + mid) % N32)] ^= r; 17.614 + r -= i; 17.615 + psfmt32[idxof((i + mid + lag) % N32)] ^= r; 17.616 + psfmt32[idxof(i)] = r; 17.617 + i = (i + 1) % N32; 17.618 + } 17.619 + 17.620 + idx = N32; 17.621 + period_certification(); 17.622 + initialized = 1; 17.623 +}
18.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 18.2 +++ b/src/SFMT/SFMT.h Sun Mar 04 17:38:32 2012 -0600 18.3 @@ -0,0 +1,165 @@ 18.4 +/** 18.5 + * @file SFMT.h 18.6 + * 18.7 + * @brief SIMD oriented Fast Mersenne Twister(SFMT) pseudorandom 18.8 + * number generator 18.9 + * 18.10 + * @author Mutsuo Saito (Hiroshima University) 18.11 + * @author Makoto Matsumoto (Hiroshima University) 18.12 + * 18.13 + * Copyright (C) 2006, 2007 Mutsuo Saito, Makoto Matsumoto and Hiroshima 18.14 + * University. All rights reserved. 18.15 + * 18.16 + * The new BSD License is applied to this software. 18.17 + * see LICENSE.txt 18.18 + * 18.19 + * @note We assume that your system has inttypes.h. If your system 18.20 + * doesn't have inttypes.h, you have to typedef uint32_t and uint64_t, 18.21 + * and you have to define PRIu64 and PRIx64 in this file as follows: 18.22 + * @verbatim 18.23 + typedef unsigned int uint32_t 18.24 + typedef unsigned long long uint64_t 18.25 + #define PRIu64 "llu" 18.26 + #define PRIx64 "llx" 18.27 +@endverbatim 18.28 + * uint32_t must be exactly 32-bit unsigned integer type (no more, no 18.29 + * less), and uint64_t must be exactly 64-bit unsigned integer type. 18.30 + * PRIu64 and PRIx64 are used for printf function to print 64-bit 18.31 + * unsigned int and 64-bit unsigned int in hexadecimal format. 18.32 + */ 18.33 + 18.34 +#ifndef SFMT_H 18.35 +#define SFMT_H 18.36 + 18.37 +#include <stdio.h> 18.38 + 18.39 +#if defined(__STDC_VERSION__) && (__STDC_VERSION__ >= 199901L) 18.40 + #include <inttypes.h> 18.41 +#elif defined(_MSC_VER) || defined(__BORLANDC__) 18.42 + typedef unsigned int uint32_t; 18.43 + typedef unsigned __int64 uint64_t; 18.44 + #define inline __inline 18.45 +#else 18.46 + #include <inttypes.h> 18.47 + #if defined(__GNUC__) 18.48 + #define inline __inline__ 18.49 + #endif 18.50 +#endif 18.51 + 18.52 +#ifndef PRIu64 18.53 + #if defined(_MSC_VER) || defined(__BORLANDC__) 18.54 + #define PRIu64 "I64u" 18.55 + #define PRIx64 "I64x" 18.56 + #else 18.57 + #define PRIu64 "llu" 18.58 + #define PRIx64 "llx" 18.59 + #endif 18.60 +#endif 18.61 + 18.62 +#if defined(__GNUC__) 18.63 +#define ALWAYSINLINE __attribute__((always_inline)) 18.64 +#else 18.65 +#define ALWAYSINLINE 18.66 +#endif 18.67 + 18.68 +#if defined(_MSC_VER) 18.69 + #if _MSC_VER >= 1200 18.70 + #define PRE_ALWAYS __forceinline 18.71 + #else 18.72 + #define PRE_ALWAYS inline 18.73 + #endif 18.74 +#else 18.75 + #define PRE_ALWAYS inline 18.76 +#endif 18.77 + 18.78 +#if defined(__cplusplus) || defined(c_plusplus) 18.79 +//extern "C" { 18.80 +#endif 18.81 + 18.82 +uint32_t gen_rand32(void); 18.83 +uint64_t gen_rand64(void); 18.84 +void fill_array32(uint32_t *array, int size); 18.85 +void fill_array64(uint64_t *array, int size); 18.86 +void init_gen_rand(uint32_t seed); 18.87 +void init_by_array(uint32_t *init_key, int key_length); 18.88 +const char *get_idstring(void); 18.89 +int get_min_array_size32(void); 18.90 +int get_min_array_size64(void); 18.91 + 18.92 +#if defined(__cplusplus) || defined(c_plusplus) 18.93 +//} // end extern 18.94 +#endif 18.95 + 18.96 +/* These real versions are due to Isaku Wada */ 18.97 +/** generates a random number on [0,1]-real-interval */ 18.98 +inline static double to_real1(uint32_t v) 18.99 +{ 18.100 + return v * (1.0/4294967295.0); 18.101 + /* divided by 2^32-1 */ 18.102 +} 18.103 + 18.104 +/** generates a random number on [0,1]-real-interval */ 18.105 +inline static double genrand_real1(void) 18.106 +{ 18.107 + return to_real1(gen_rand32()); 18.108 +} 18.109 + 18.110 +/** generates a random number on [0,1)-real-interval */ 18.111 +inline static double to_real2(uint32_t v) 18.112 +{ 18.113 + return v * (1.0/4294967296.0); 18.114 + /* divided by 2^32 */ 18.115 +} 18.116 + 18.117 +/** generates a random number on [0,1)-real-interval */ 18.118 +inline static double genrand_real2(void) 18.119 +{ 18.120 + return to_real2(gen_rand32()); 18.121 +} 18.122 + 18.123 +/** generates a random number on (0,1)-real-interval */ 18.124 +inline static double to_real3(uint32_t v) 18.125 +{ 18.126 + return (((double)v) + 0.5)*(1.0/4294967296.0); 18.127 + /* divided by 2^32 */ 18.128 +} 18.129 + 18.130 +/** generates a random number on (0,1)-real-interval */ 18.131 +inline static double genrand_real3(void) 18.132 +{ 18.133 + return to_real3(gen_rand32()); 18.134 +} 18.135 +/** These real versions are due to Isaku Wada */ 18.136 + 18.137 +/** generates a random number on [0,1) with 53-bit resolution*/ 18.138 +inline static double to_res53(uint64_t v) 18.139 +{ 18.140 + return v * (1.0/18446744073709551616.0L); 18.141 +} 18.142 + 18.143 +/** generates a random number on [0,1) with 53-bit resolution from two 18.144 + * 32 bit integers */ 18.145 +inline static double to_res53_mix(uint32_t x, uint32_t y) 18.146 +{ 18.147 + return to_res53(x | ((uint64_t)y << 32)); 18.148 +} 18.149 + 18.150 +/** generates a random number on [0,1) with 53-bit resolution 18.151 + */ 18.152 +inline static double genrand_res53(void) 18.153 +{ 18.154 + return to_res53(gen_rand64()); 18.155 +} 18.156 + 18.157 +/** generates a random number on [0,1) with 53-bit resolution 18.158 + using 32bit integer. 18.159 + */ 18.160 +inline static double genrand_res53_mix(void) 18.161 +{ 18.162 + uint32_t x, y; 18.163 + 18.164 + x = gen_rand32(); 18.165 + y = gen_rand32(); 18.166 + return to_res53_mix(x, y); 18.167 +} 18.168 +#endif
19.1 --- a/src/common/lua-engine.cpp Sun Mar 04 17:35:30 2012 -0600 19.2 +++ b/src/common/lua-engine.cpp Sun Mar 04 17:38:32 2012 -0600 19.3 @@ -49,10 +49,10 @@ 19.4 19.5 extern "C" 19.6 { 19.7 -#include "../lua/src/lua.h" 19.8 -#include "../lua/src/lauxlib.h" 19.9 -#include "../lua/src/lualib.h" 19.10 -#include "../lua/src/lstate.h" 19.11 +#include "../lua/lua.h" 19.12 +#include "../lua/lauxlib.h" 19.13 +#include "../lua/lualib.h" 19.14 +#include "../lua/lstate.h" 19.15 } 19.16 #include "vbalua.h" 19.17