数学研发论坛

 找回密码
 欢迎注册
楼主: 〇〇

[讨论] mathe能否提供更新的windows版gmp库

[复制链接]
发表于 2010-9-9 10:12:01 | 显示全部楼层
10# 〇〇
你好像没有make install,
毋因群疑而阻独见  毋任己意而废人言
毋私小惠而伤大体  毋借公论以快私情
发表于 2010-9-9 10:22:15 | 显示全部楼层
你要静态还是动态的,要不要支持C++,要不要开断言,我可以拷贝你
毋因群疑而阻独见  毋任己意而废人言
毋私小惠而伤大体  毋借公论以快私情
 楼主| 发表于 2010-9-10 05:47:53 | 显示全部楼层
都要呢...
毋因群疑而阻独见  毋任己意而废人言
毋私小惠而伤大体  毋借公论以快私情
发表于 2010-9-10 11:34:28 | 显示全部楼层
13# 〇〇
在GCC4.5.1下编译出来的。
--enable-cxx --enable-assert
gmp5.0.1-static.7z (274.32 KB, 下载次数: 26)
毋因群疑而阻独见  毋任己意而废人言
毋私小惠而伤大体  毋借公论以快私情
 楼主| 发表于 2010-9-11 10:29:35 | 显示全部楼层
谢谢
毋因群疑而阻独见  毋任己意而废人言
毋私小惠而伤大体  毋借公论以快私情
发表于 2010-9-11 20:35:37 | 显示全部楼层
是在四核机器下编译的, build system type  .... core2-pc-mingw32
毋因群疑而阻独见  毋任己意而废人言
毋私小惠而伤大体  毋借公论以快私情
 楼主| 发表于 2010-10-20 21:03:35 | 显示全部楼层

  1. /*William,
  2. Here is a simple program that  computes PI (using GMP library) to compare the performance differences for "Cilk V2 cilk_spawn", "OpenMP 3.0 task"  or serially.  The program was modified  from the source found at http://gmplib.org/list-archives/gmp-discuss/2010-April/004117.html and any mistakes are my own.


  3. /* Pi computation using Chudnovsky's algortithm.

  4. * Copyright 2002, 2005 Hanhong Xue (macroxue at yahoo dot com)

  5. * Slightly modified 2005 by Torbjorn Granlund to allow more than 2G
  6.    digits to be computed.

  7. * Modified 2010 by David Carver (dcarver at tacc dot utexas dot edu) to
  8.     demonstrate a parallel recursive version of the gmp-chudnovsky program
  9.     using the Intel's C++ Compiler XE 12.0 with either "Cilk V2" or "OpenMP 3.0 task".

  10.    \$ icc -V
  11.    Intel(R) C Compiler XE for applications running on Intel(R) 64, Version 12.0.0 Beta Build 20100512
  12.    Copyright (C) 1985-2010 Intel Corporation.  All rights reserved.


  13.    To compile for serial execution
  14.    \$ icc -O3 -o pgmp-chudnovsky pgmp-chudnovsky.c -lgmp -limf

  15.    and run:

  16.    \$ ./pgmp-chudnovsky 100000000 0
  17.    #terms=7051366 depth=24 nprocs=12 SERIAL
  18.    bs         cputime = 207.030  wallclock = 207.060
  19.    total      cputime = 241.660  wallclock = 241.694
  20.       P size=248778666 digits (2.487787)
  21.       Q size=248778659 digits (2.487787)


  22.    To compile for parallel Cilk V2 execution:
  23.    \$ icc -DOCILK -O3 -o pgmp-chudnovsky pgmp-chudnovsky.c -lgmp -limf

  24.    and you may specify the number of workers with the Cilk V2 environment variable CILK_NPROC

  25.    \$ export CILK_NWORKERS=12
  26.    \$ ./pgmp-chudnovsky 100000000 0
  27.    #terms=7051366 depth=24 nprocs=12 CILK_NWORKERS=12
  28.    bs         cputime = 639.680  wallclock = 53.331
  29.    total      cputime = 674.790  wallclock = 88.448
  30.       P size=248778666 digits (2.487787)
  31.       Q size=248778659 digits (2.487787)


  32.    To compile for a simple parallel OpenMP 3.0 Task execution:
  33.    \$ icc -openmp -O3 -o pgmp-chudnovsky pgmp-chudnovsky.c -lgmp -limf

  34.    and you may specify the number of threads with the OpenMP environment variable OMP_NUM_THREADS

  35.    \$ export OMP_NUM_THREADS=12
  36.    \$ ./pgmp-chudnovsky 100000000 0
  37.    #terms=7051366 depth=24 nprocs=12 OMP_NUM_THREADS=12
  38.    bs         cputime = 710.110  wallclock = 59.223
  39.    total      cputime = 745.390  wallclock = 94.340
  40.       P size=248778666 digits (2.487787)
  41.       Q size=248778659 digits (2.487787)


  42.    To get help run the program with no options:

  43.    \$ ./pgmp-chudnovsky

  44.    Syntax: ./pgmp-chudnovsky <digits> <option>
  45.       <digits> digits of pi to output
  46.       <option> 0 - just run (default)
  47.                1 - output decimal digits to stdout


  48. * Redistribution and use in source and binary forms,with or without
  49. * modification,are permitted provided that the following conditions are met:
  50. * 1. Redistributions of source code must retain the above copyright notice,
  51. * this list of conditions and the following disclaimer.
  52. * 2. Redistributions in binary form must reproduce the above copyright notice,
  53. * this list of conditions and the following disclaimer in the documentation
  54. * and/or other materials provided with the distribution.
  55. *
  56. * THIS SOFTWARE IS PROVIDED BY THE AUTHORS ``AS IS'' AND ANY EXPRESS OR
  57. * IMPLIED WARRANTIES,INCLUDING,BUT NOT LIMITED TO,THE IMPLIED WARRANTIES OF
  58. * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO
  59. * EVENT SHALL THE AUTHORS BE LIABLE FOR ANY DIRECT,INDIRECT,INCIDENTAL,
  60. * SPECIAL,EXEMPLARY,OR CONSEQUENTIAL DAMAGES (INCLUDING,BUT NOT LIMITED TO,
  61. * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,DATA,OR PROFITS;
  62. * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
  63. * WHETHER IN CONTRACT,STRICT LIABILITY,OR TORT (INCLUDING NEGLIGENCE OR
  64. * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,EVEN IF
  65. * ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  66. */

  67. #include <assert.h>
  68. #include <math.h>
  69. #include <stdio.h>
  70. #include <stdlib.h>
  71. #include <string.h>
  72. #include <time.h>
  73. //#include <sys/time.h>
  74. //#include <sys/sysinfo.h>

  75. #ifdef OCILK
  76. #include <cilk/cilk.h>
  77. #elif _OPENMP
  78. #include <omp.h>
  79. #endif /* OCILK */
  80. #include <gmp.h>

  81. #define A   13591409
  82. #define B   545140134
  83. #define C   640320
  84. #define D   12

  85. #define BITS_PER_DIGIT   3.32192809488736234787
  86. #define DIGITS_PER_ITER  14.1816474627254776555
  87. #define DOUBLE_PREC      53

  88. ////////////////////////////////////////////////////////////////////////////
  89. /*
  90. double wall_clock()
  91. {
  92.         struct timeval timeval;
  93.         (void) gettimeofday (&timeval,NULL);
  94.         return (timeval.tv_sec + (timeval.tv_usec / 1000000.0));
  95. }
  96. */
  97. ////////////////////////////////////////////////////////////////////////////

  98. /* binary splitting */
  99. void
  100. bs(unsigned long a,unsigned long b,unsigned gflag,unsigned level,mpz_t pstack1,mpz_t qstack1,mpz_t gstack1)
  101. {
  102.   unsigned long mid;
  103.   mpz_t pstack2,qstack2,gstack2;

  104. if (b-a==1) {

  105.     /*
  106.       g(b-1,b) = (6b-5)(2b-1)(6b-1)
  107.       p(b-1,b) = b^3 * C^3 / 24
  108.       q(b-1,b) = (-1)^b*g(b-1,b)*(A+Bb).
  109.     */

  110.     mpz_set_ui(pstack1,b);
  111.     mpz_mul_ui(pstack1,pstack1,b);
  112.     mpz_mul_ui(pstack1,pstack1,b);
  113.     mpz_mul_ui(pstack1,pstack1,(C/24)*(C/24));
  114.     mpz_mul_ui(pstack1,pstack1,C*24);

  115.     mpz_set_ui(gstack1,2*b-1);
  116.     mpz_mul_ui(gstack1,gstack1,6*b-1);
  117.     mpz_mul_ui(gstack1,gstack1,6*b-5);

  118.     mpz_set_ui(qstack1,b);
  119.     mpz_mul_ui(qstack1,qstack1,B);
  120.     mpz_add_ui(qstack1,qstack1,A);
  121.     mpz_mul   (qstack1,qstack1,gstack1);
  122.     if (b%2)
  123.       mpz_neg(qstack1,qstack1);

  124.   } else {

  125.     /*
  126.       p(a,b) = p(a,m) * p(m,b)
  127.       g(a,b) = g(a,m) * g(m,b)
  128.       q(a,b) = q(a,m) * p(m,b) + q(m,b) * g(a,m)
  129.     */

  130.     mid = a+(b-a)*0.5224;     /* tuning parameter */

  131.     mpz_init(pstack2);
  132.     mpz_init(qstack2);
  133.     mpz_init(gstack2);

  134. #ifdef OCILK
  135.     cilk_spawn bs(mid,b,gflag,level+1,pstack2,qstack2,gstack2);
  136.     bs(a,mid,1,level+1,pstack1,qstack1,gstack1);
  137.     cilk_sync;
  138. #elif _OPENMP
  139.     #pragma omp task firstprivate(mid,b,gflag,level) shared(pstack2,qstack2,gstack2)
  140.        bs(mid,b,gflag,level+1,pstack2,qstack2,gstack2);
  141. #pragma omp task firstprivate(mid,a,gflag,level) shared(pstack1,qstack1,gstack1)
  142.        bs(a,mid,1,level+1,pstack1,qstack1,gstack1);
  143.     #pragma omp taskwait
  144. #else /* SERIAL */
  145.      bs(mid,b,gflag,level+1,pstack2,qstack2,gstack2);
  146.      bs(a,mid,1,level+1,pstack1,qstack1,gstack1);
  147. #endif /* OCILK */

  148.     mpz_mul(pstack1,pstack1,pstack2);
  149.     mpz_mul(qstack1,qstack1,pstack2);
  150.     mpz_mul(qstack2,qstack2,gstack1);
  151.     mpz_add(qstack1,qstack1,qstack2);

  152.     if (gflag) {
  153.       mpz_mul(gstack1,gstack1,gstack2);
  154.     }

  155.     mpz_clear(pstack2);
  156.     mpz_clear(qstack2);
  157.     mpz_clear(gstack2);
  158.   }
  159. }

  160. int
  161. main(int argc,char *argv[])
  162. {
  163.   mpf_t  pi,qi;
  164.   mpz_t  pstack,qstack,gstack;
  165.   unsigned long terms,d=100,out=0,depth,psize,qsize;
  166.   clock_t begin,end;
  167.   double wbegin,wend;
  168.   char *prog_name;

  169.   prog_name = argv[0];

  170.   if (argc==1) {
  171.     fprintf(stderr,"\nSyntax: %s <digits> <option>\n",prog_name);
  172.     fprintf(stderr,"      <digits> digits of pi to output\n");
  173.     fprintf(stderr,"      <option> 0 - just run (default)\n");
  174.     fprintf(stderr,"               1 - output decimal digits to stdout\n");
  175.     exit(1);
  176.   }
  177.   if (argc>1)
  178.     d = strtoul(argv[1],0,0);
  179.   if (argc>2)
  180.     out = atoi(argv[2]);

  181.   begin = clock();
  182.   //wbegin = wall_clock();

  183.   terms = d/DIGITS_PER_ITER;
  184.   depth = 0;
  185.   while ((1L<<depth)<terms)
  186.     depth++;
  187.   depth++;

  188.   mpz_init(pstack);
  189.   mpz_init(qstack);
  190.   mpz_init(gstack);

  191.   /* begin binary splitting process */
  192.   if (terms<=0) {
  193.     mpz_set_ui(pstack,1);
  194.     mpz_set_ui(qstack,0);
  195.     mpz_set_ui(gstack,1);
  196.   } else {

  197. #ifdef OCILK
  198.     fprintf(stderr,"#terms=%ld depth=%ld nprocs=%d CILK_NWORKERS=%d\n",terms,depth,1/*get_nprocs()*/,
  199.       __cilkrts_get_nworkers());
  200.     bs(0,terms,0,0,pstack,qstack,gstack);
  201. #elif _OPENMP
  202.     #pragma omp parallel
  203.       #pragma omp single nowait
  204.       {
  205.         fprintf(stderr,"#terms=%ld depth=%ld nprocs=%d OMP_NUM_THREADS=%d\n",terms,depth,1/*get_nprocs()*/,
  206.           omp_get_num_threads());
  207.         bs(0,terms,0,0,pstack,qstack,gstack);
  208.       }
  209. #else /* SERIAL */
  210.     fprintf(stderr,"#terms=%ld depth=%ld nprocs=%d SERIAL\n",terms,depth,1/*get_nprocs()*/);
  211.     bs(0,terms,0,0,pstack,qstack,gstack);
  212. #endif /* OCILK */

  213.   }

  214.   end = clock();
  215.   //wend = wall_clock();
  216.   fprintf(stderr,"bs         cputime = %6.3f  wallclock = %6.3f\n",
  217.     (double)(end-begin)/CLOCKS_PER_SEC,/*(wend-wbegin)*/1);
  218.   fflush(stderr);

  219.   mpz_clear(gstack);

  220.   /* prepare to convert integers to floats */

  221.   mpf_set_default_prec((long int)(d*BITS_PER_DIGIT+16));

  222.   /*
  223.           p*(C/D)*sqrt(C)
  224.     pi = -----------------
  225.              (q+A*p)
  226.   */

  227.   psize = mpz_sizeinbase(pstack,10);
  228.   qsize = mpz_sizeinbase(qstack,10);

  229.   mpz_addmul_ui(qstack,pstack,A);
  230.   mpz_mul_ui(pstack,pstack,C/D);

  231.   mpf_init(pi);
  232.   mpf_set_z(pi,pstack);
  233.   mpz_clear(pstack);

  234.   mpf_init(qi);
  235.   mpf_set_z(qi,qstack);
  236.   mpz_clear(qstack);

  237.   /* final step */

  238.   mpf_div(qi,pi,qi);
  239.   mpf_sqrt_ui(pi,C);
  240.   mpf_mul(qi,qi,pi);
  241.   mpf_clear(pi);

  242.   end = clock();
  243.   //wend = wall_clock();

  244.   /* output Pi and timing statistics */

  245.   fprintf(stderr,"total      cputime = %6.3f  wallclock = %6.3f\n",
  246.     (double)(end-begin)/CLOCKS_PER_SEC,1/*(wend-wbegin)*/);
  247. fflush(stderr);

  248.   printf("   P size=%ld digits (%f)\n"
  249.          "   Q size=%ld digits (%f)\n",
  250.          psize,(double)psize/d,qsize,(double)qsize/d);

  251.   if (out&1)  {
  252.     printf("pi(0,%ld)=\n",terms);
  253.     mpf_out_str(stdout,10,d+2,qi);
  254.     printf("\n");
  255.   }

  256.   /* free float resources */

  257.   mpf_clear(qi);

  258.   exit (0);
  259. }



复制代码
D:\lt\dl\gmp421>cl chudnovsky.cpp -I .\include  /link D:\lt\dl\gmp421\lib\Win32\Release\gmp.lib
Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 13.10.3077 for 80x86
Copyright (C) Microsoft Corporation 1984-2002. All rights reserved.

chudnovsky.cpp
Microsoft (R) Incremental Linker Version 7.10.3077
Copyright (C) Microsoft Corporation.  All rights reserved.

/out:chudnovsky.exe
D:\lt\dl\gmp421\lib\Win32\Release\gmp.lib
chudnovsky.obj
LINK : warning LNK4098: defaultlib 'LIBCMT' conflicts with use of other libs; use /NODEFAULTLIB:library
gmp.lib(get_str.obj) : error LNK2019: unresolved external symbol __ftol2_sse referenced in function ___gmpf_get_str
chudnovsky.exe : fatal error LNK1120: 1 unresolved externals
毋因群疑而阻独见  毋任己意而废人言
毋私小惠而伤大体  毋借公论以快私情
 楼主| 发表于 2010-10-21 10:30:51 | 显示全部楼层
自己利用
http://www.mpir.org/

http://www.tortall.net/projects/yasm/wiki/Download
做出来了
configure
make
to_gmp
毋因群疑而阻独见  毋任己意而废人言
毋私小惠而伤大体  毋借公论以快私情
 楼主| 发表于 2010-10-21 10:42:03 | 显示全部楼层
gpi被去掉了计时,用timer代替

Timer.rar

1.75 KB, 下载次数: 0, 下载积分: 金币 -1 枚, 经验 1 点, 下载 1 次

mpir.part2.rar

187.21 KB, 下载次数: 0, 下载积分: 金币 -1 枚, 经验 1 点, 下载 1 次

毋因群疑而阻独见  毋任己意而废人言
毋私小惠而伤大体  毋借公论以快私情
 楼主| 发表于 2010-10-21 10:42:35 | 显示全部楼层
512限制

mpir.part1.rar

500 KB, 下载次数: 0, 下载积分: 金币 -1 枚, 经验 1 点, 下载 1 次

毋因群疑而阻独见  毋任己意而废人言
毋私小惠而伤大体  毋借公论以快私情
您需要登录后才可以回帖 登录 | 欢迎注册

本版积分规则

小黑屋|手机版|数学研发网 ( 苏ICP备07505100号 )

GMT+8, 2019-8-21 13:47 , Processed in 0.084185 second(s), 17 queries .

Powered by Discuz! X3.4

© 2001-2017 Comsenz Inc.

快速回复 返回顶部 返回列表