diff options
Diffstat (limited to 'SFMT/html/howto-compile.html')
| -rw-r--r-- | SFMT/html/howto-compile.html | 493 |
1 files changed, 493 insertions, 0 deletions
diff --git a/SFMT/html/howto-compile.html b/SFMT/html/howto-compile.html new file mode 100644 index 0000000..8d08d1e --- /dev/null +++ b/SFMT/html/howto-compile.html @@ -0,0 +1,493 @@ +<?xml version="1.0" encoding="UTF-8" ?> +<!DOCTYPE html + PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" + "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd"> +<html xmlns="http://www.w3.org/1999/xhtml"> + <head> + <meta http-equiv="Content-Type" content="text/html" /> + <title>How to compile SFMT</title> + <style type="text/css"> + BLOCKQUOTE {background-color:#a0ffa0; + padding-left: 1em;} + </style> + </head> + <body> + <h2> How to compile SFMT </h2> + + <p> + This document explains how to compile SFMT for users who + are using UNIX like systems (for example Linux, Free BSD, + cygwin, osx, etc) on terminal. I can't help those who use IDE + (Integrated Development Environment,) please see your IDE's help + to use SIMD feature of your CPU. + </p> + + <h3>1. First Step: Compile test programs using Makefile.</h3> + <h4>1-1. Compile standard C test program.</h4> + <p> + Check if SFMT.c and Makefile are in your current directory. + If not, <strong>cd</strong> to the directory where they exist. + Then, type + </p> + <blockquote> + <pre>make std</pre> + </blockquote> + <p> + If it causes an error, try to type + </p> + <blockquote> + <pre>cc -DSFMT_MEXP=19937 -o test-std-M19937 test.c SFMT.c</pre> + </blockquote> + <p> + or try to type + </p> + <blockquote> + <pre>gcc -DSFMT_MEXP=19937 -o test-std-M19937 test.c SFMT.c</pre> + </blockquote> + <p> + If success, then check the test program. Type + </p> + <blockquote> + <pre>./test-std-M19937 -b32</pre> + </blockquote> + <p> + You will see many random numbers displayed on your screen. + If you want to check these random numbers are correct output, + redirect output to a file and <strong>diff</strong> it with + <strong>SFMT.19937.out.txt</strong>, like this:</p> + <blockquote> + <pre>./test-std-M19937 -b32 > foo.txt +diff -w foo.txt SFMT.19937.out.txt</pre> + </blockquote> + <p> + Silence means they are the same because <strong>diff</strong> + reports the difference of two file. + </p> + <p> + If you want to know the generation speed of SFMT, type + </p> + <blockquote> + <pre>./test-std-M19937 -s</pre> + </blockquote> + <p> + It is very slow. To make it fast, compile it + with <strong>-O3</strong> option. If your compiler is gcc, you + should specify <strong>-fno-strict-aliasing</strong> option + with <strong>-O3</strong>. type + </p> + <blockquote> + <pre>gcc -O3 -fno-strict-aliasing -DSFMT_MEXP=19937 -o test-std-M19937 test.c SFMT.c +./test-std-M19937 -s</pre> + </blockquote> + + <h4>1-2. Compile SSE2 test program.</h4> + <p> + If your CPU supports SSE2 and you can use gcc version 3.4 or later, + you can make test-sse2-Mxxx. To do this, type + </p> + <blockquote> + <pre>make sse2</pre> + </blockquote> + <p>or type</p> + <blockquote> + <pre>gcc -O3 -msse2 -fno-strict-aliasing -DHAVE_SSE2=1 -DSFMT_MEXP=19937 -o test-sse2-M19937 test.c SFMT.c</pre> + </blockquote> + <p>If everything works well,</p> + <blockquote> + <pre>./test-sse2-M19937 -s</pre> + </blockquote> + <p>will show much shorter time than <strong>test-std-M19937 -s</strong>.</p> + + <!--h4>1-3. Compile AltiVec test program.</h4> + <p> + If you are using Macintosh computer with PowerPC G4 or G5, and + your gcc version is later 3.3, you can make test-alti-M19937. To + do this, type + </p> + <blockquote> + <pre>make osx-alti</pre> + </blockquote> + <p>or type</p> + <blockquote> + <pre>gcc -O3 -faltivec -fno-strict-aliasing -DHAVE_ALTIVEC=1 -DSFMT_MEXP=19937 -o test-alti-M19937 test.c</pre> + </blockquote> + <p>If everything works well,</p> + <blockquote> + <pre>./test-alti-M19937 -s</pre> + </blockquote> + <p>shows much shorter time than <strong>test-std-M19937 -s</strong>.</p> + <p>If you are using a CPU which supports AltiVec under Linux, use + <strong>alti</strong> instead of <strong>osx-alti</strong>.</p--> + + <h4>1-4. Compile and check output automatically.</h4> + <p> + To make test program and check 32-bit output + automatically for all supported MEXPs of SFMT, type + </p> + <blockquote> + <pre>make std-check</pre> + </blockquote> + <!--p> + To check test program optimized for 64bit output of big endian CPU, type + </p> + <blockquote> + <pre>make big-check</pre> + </blockquote--> + <p> + To check test program optimized for SSE2, type + </p> + <blockquote> + <pre>make sse2-check</pre> + </blockquote> + <!--p> + To check test program optimized for OSX AltiVec, type + </p> + <blockquote> + <pre>make osx-alti-check</pre> + </blockquote> + <p> + To check test program optimized for OSX AltiVec and 64bit output, type + </p> + <blockquote> + <pre>make osx-altibig-check</pre> + </blockquote--> + <p> + These commands may take some time. + </p> + + <h3>2. Second Step: Use SFMT pseudorandom number generator with + your C program.</h3> + <h4>2-1. Use sequential call and static link.</h4> + <p> + Here is a very simple program <strong>sample1.c</strong> which + calculates PI using Monte-Carlo method. + </p> + <blockquote> + <pre> +#include <stdio.h> +#include <stdlib.h> +#include "SFMT.h" + +int main(int argc, char* argv[]) { + int i, cnt, seed; + double x, y, pi; + const int NUM = 10000; + sfmt_t sfmt; + + if (argc >= 2) { + seed = strtol(argv[1], NULL, 10); + } else { + seed = 12345; + } + cnt = 0; + sfmt_init_gen_rand(&sfmt, seed); + for (i = 0; i < NUM; i++) { + x = sfmt_genrand_res53(&sfmt); + y = sfmt_genrand_res53(&sfmt); + if (x * x + y * y < 1.0) { + cnt++; + } + } + pi = (double)cnt / NUM * 4; + printf("%lf\n", pi); + return 0; +} + </pre> + </blockquote> + <p>To compile <strong>sample1.c</strong> with SFMT.c with the period of + 2<sup>607</sup>, type</p> + <blockquote> + <pre>gcc -O3 -DSFMT_MEXP=607 -o sample1 SFMT.c sample1.c</pre> + </blockquote> + <!--p>If your CPU is BIG ENDIAN you need to type</p> + <blockquote> + <pre>gcc -DSFMT_MEXP=607 -DBIG_ENDIAN64 -o sample1 SFMT.c sample1.c</pre> + </blockquote> + <p>because genrand_res53() uses gen_rand64().</p--> + <p>If your CPU supports SSE2 and you want to use optimized SFMT for + SSE2, type</p> + <blockquote> + <pre>gcc -O3 -msse2 -DHAVE_SSE2 -DSFMT_MEXP=607 -o sample1 SFMT.c sample1.c</pre> + </blockquote> + <!--p>If your CPU supports AltiVec and you want to use optimized SFMT + for AltiVec, type</p> + <blockquote> + <pre>gcc -faltivec -DBIG_ENDIAN64 -DHAVE_ALTIVEC -DSFMT_MEXP=607 -o sample1 SFMT.c sample1.c</pre> + </blockquote--> + + <h4>2-2. Use block call and static link.</h4> + <p> + Here is <strong>sample2.c</strong> which modifies sample1.c. + The block call <strong>fill_array64</strong> is much faster than + sequential call, but it needs an aligned memory. The standard function + to get an aligned memory is <strong>posix_memalign</strong>, but + it isn't usable in every OS. + </p> + <blockquote> + <pre> +#include <stdio.h> +#define _XOPEN_SOURCE 600 +#include <stdlib.h> +#include "SFMT.h" + +int main(int argc, char* argv[]) { + int i, j, cnt, seed; + double x, y, pi; + const int NUM = 10000; + const int R_SIZE = 2 * NUM; + int size; + uint64_t *array; + sfmt_t sfmt; + + if (argc >= 2) { + seed = strtol(argv[1], NULL, 10); + } else { + seed = 12345; + } + size = sfmt_get_min_array_size64(&sfmt); + if (size < R_SIZE) { + size = R_SIZE; + } +#if defined(__APPLE__) || \ + (defined(__FreeBSD__) && __FreeBSD__ >= 3 && __FreeBSD__ <= 6) + printf("malloc used\n"); + array = malloc(sizeof(double) * size); + if (array == NULL) { + printf("can't allocate memory.\n"); + return 1; + } +#elif defined(_POSIX_C_SOURCE) + printf("posix_memalign used\n"); + if (posix_memalign((void **)&array, 16, sizeof(double) * size) != 0) { + printf("can't allocate memory.\n"); + return 1; + } +#elif defined(__GNUC__) && (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 3)) + printf("memalign used\n"); + array = memalign(16, sizeof(double) * size); + if (array == NULL) { + printf("can't allocate memory.\n"); + return 1; + } +#else /* in this case, gcc doesn't support SSE2 */ + printf("malloc used\n"); + array = malloc(sizeof(double) * size); + if (array == NULL) { + printf("can't allocate memory.\n"); + return 1; + } +#endif + cnt = 0; + j = 0; + sfmt_init_gen_rand(&sfmt, seed); + sfmt_fill_array64(&sfmt, array, size); + for (i = 0; i < NUM; i++) { + x = sfmt_to_res53(array[j++]); + y = sfmt_to_res53(array[j++]); + if (x * x + y * y < 1.0) { + cnt++; + } + } + free(array); + pi = (double)cnt / NUM * 4; + printf("%lf\n", pi); + return 0; +} + </pre> + </blockquote> + <p>To compile <strong>sample2.c</strong> with SFMT.c with the period of + 2<sup>2281</sup>, type</p> + <blockquote> + <pre>gcc -O3 -DSFMT_MEXP=2281 -o sample2 SFMT.c sample2.c</pre> + </blockquote> + <!--p>or </p> + <blockquote> + <pre>gcc -DSFMT_MEXP=2281 -DBIG_ENDIAN64 -o sample2 SFMT.c sample2.c</pre> + </blockquote --> + <p>If your CPU supports SSE2 and you want to use optimized SFMT for + SSE2, type</p> + <blockquote> + <pre>gcc -O3 -msse2 -DHAVE_SSE2 -DSFMT_MEXP=2281 -o sample2 SFMT.c sample2.c</pre> + </blockquote> + <!--p>If your CPU supports AltiVec and you want to use optimized SFMT + for AltiVec, type</p> + <blockquote> + <pre>gcc -faltivec -DHAVE_ALTIVEC -DSFMT_MEXP=2281 -DBIG_ENDIAN64 -o sample2 SFMT.c sample2.c</pre> + </blockquote> + <p>or type</p> + <blockquote> + <pre>gcc -faltivec -DHAVE_ALTIVEC -DBIG_ENDIAN64 -DONLY64 -DSFMT_MEXP=2281 -o sample2 SFMT.c sample2.c</pre> + </blockquote> + <p>The effect of the option -DONLY64 is: + When -DONLY64 option is used, the executive file can generate + 64-bit integers faster but 32-bit output is not supported. + </p--> + <!--h4>2-3. Use sequential call and inline functions.</h4> + <p> + Here is <strong>sample3.c</strong> which modifies sample1.c. + This is very similar to sample1.c. The difference is only one line. + It include <strong>"SFMT.c"</strong> instead of <strong>"SFMT.h" + </strong>. + </p> + <blockquote> + <pre> +#include <stdio.h> +#include <stdlib.h> +#include "SFMT.c" + +int main(int argc, char* argv[]) { + int i, cnt, seed; + double x, y, pi; + const int NUM = 10000; + + if (argc >= 2) { + seed = strtol(argv[1], NULL, 10); + } else { + seed = 12345; + } + cnt = 0; + init_gen_rand(seed); + for (i = 0; i < NUM; i++) { + x = genrand_res53(); + y = genrand_res53(); + if (x * x + y * y < 1.0) { + cnt++; + } + } + pi = (double)cnt / NUM * 4; + printf("%lf\n", pi); + return 0; +} + </pre> + </blockquote> + <p>To compile <strong>sample3.c</strong>, type</p> + <blockquote> + <pre>gcc -DSFMT_MEXP=1279 -o sample3 sample3.c</pre> + </blockquote> + <p> or </p> + <blockquote> + <pre>gcc -DSFMT_MEXP=1279 -DBIG_ENDIAN64 -o sample3 sample3.c</pre> + </blockquote> + <p>If your CPU supports SSE2 and you want to use optimized SFMT for + SSE2, then type</p> + <blockquote> + <pre>gcc -msse2 -DHAVE_SSE2 -DSFMT_MEXP=1279 -o sample3 sample3.c</pre> + </blockquote> + <p>If your CPU supports AltiVec and you want to use optimized SFMT + for AltiVec, type</p> + <blockquote> + <pre>gcc -faltivec -DHAVE_ALTIVEC -DBIG_ENDIAN64 -DSFMT_MEXP=1279 -o sample3 sample3.c</pre> + </blockquote> + <p>or type</p> + <blockquote> + <pre>gcc -faltivec -DHAVE_ALTIVEC -DBIG_ENDIAN64 -DONLY64 -DSFMT_MEXP=1279 -o sample3 sample3.c</pre> + </blockquote--> + + <h4>2-4. Initialize SFMT using sfmt_init_by_array function.</h4> + <p> + Here is <strong>sample4.c</strong> which modifies sample1.c. + The 32-bit integer seed can only make 2<sup>32</sup> kinds of + initial state, to avoid this problem, SFMT + provides <strong>sfmt_init_by_array</strong> function. This sample + uses sfmt_init_by_array function which initialize the internal state + array with an array of 32-bit. The size of an array can be + larger than the internal state array and all elements of the + array are used for initialization, but too large array is + wasteful. + </p> + <blockquote> + <pre> +#include <stdio.h> +#include <string.h> +#include "SFMT.h" + +int main(int argc, char* argv[]) { + int i, cnt, seed_cnt; + double x, y, pi; + const int NUM = 10000; + uint32_t seeds[100]; + sfmt_t sfmt; + + if (argc >= 2) { + seed_cnt = 0; + for (i = 0; (i < 100) && (i < strlen(argv[1])); i++) { + seeds[i] = argv[1][i]; + seed_cnt++; + } + } else { + seeds[0] = 12345; + seed_cnt = 1; + } + cnt = 0; + sfmt_init_by_array(&sfmt, seeds, seed_cnt); + for (i = 0; i < NUM; i++) { + x = sfmt_genrand_res53(&sfmt); + y = sfmt_genrand_res53(&sfmt); + if (x * x + y * y < 1.0) { + cnt++; + } + } + pi = (double)cnt / NUM * 4; + printf("%lf\n", pi); + return 0; +} + </pre> + </blockquote> + <p>To compile <strong>sample4.c</strong>, type</p> + <blockquote> + <pre>gcc -O3 -DSFMT_MEXP=19937 -o sample4 SFMT.c sample4.c</pre> + </blockquote> + <!--p>or</p> + <blockquote> + <pre>gcc -DSFMT_MEXP=19937 -DBIG_ENDIAN64 -o sample4 SFMT.c sample4.c</pre> + </blockquote--> + <p>Now, seed can be a string. Like this:</p> + <blockquote> + <pre>./sample4 your-full-name</pre> + </blockquote> + <h3>Appendix: C preprocessor definitions</h3> + <p> + Here is a list of C preprocessor definitions that users can + specify to control code generation. These macros must be set + just after -D compiler option. + </p> + <dl> + <dt>SFMT_MEXP</dt> + <dd>This macro is required. This macro means Mersenne exponent + and the period of generated code will be 2<sup>SFMT_MEXP</sup>-1. + SFMT_MEXP must be one of 607, 1279, 2281, 4253, 11213, 19937, + 44497, 86243, 132049, 216091. + </dd> + <dt>HAVE_SSE2</dt> + <dd>This is optional. If this macro is specified, optimized code + for SSE2 will be generated.</dd> + <dt>HAVE_ALTIVEC</dt> + <dd>This is optional. If this macro is specified, optimized code + for AltiVec will be generated. This macro automatically turns on + BIG_ENDIAN64 macro. <b>This macro of SFMT ver. 1.4 is not tested + at all.</b></dd> + <dt>BIG_ENDIAN64</dt> + <dd>This macro is required when your CPU is BIG ENDIAN and you + use 64-bit output. If __BIG_ENDIAN__ macro is defined, this macro + is automatically turned on. GCC defines __BIG_ENDIAN__ macro on + BIG ENDIAN CPUs. <b>This macro of SFMT ver. 1.4 is not tested + at all.</b></dd> + <dt>ONLY64</dt> + <dd>This macro is optional. If this macro is specified, + optimized code for 64-bit output for BIG ENDIAN CPUs will be + generated and code for 32-bit output won't be + generated. BIG_ENDIAN64 macro must be specified with this macro + by user or automatically. <b>This macro of SFMT ver. 1.4 is not tested + at all.</b></dd> + </dl> + <table border="1" align="center"> + <tr><td></td><td>32-bit output</td><td>LITTLE ENDIAN 64-bit output</td> + <td>BIG ENDIAN 64-bit output</td></tr> + <tr><td>required</td><td>SFMT_MEXP</td><td>SFMT_MEXP</td><td>SFMT_MEXP, + <strong>BIG_ENDIAN64</strong></td></tr> + <tr><td>optional</td><td>HAVE_SSE2, + HAVE_ALTIVEC</td><td>HAVE_SSE2</td><td>HAVE_ALTIVEC, ONLY64</td> + </tr> + </table> + </body> +</html> |