问题：
为什么CPLEX在不同的机器上输出不同的结果？
东门宜
当我跑的时候。使用CPLEX的NET 4应用程序，我在不同的机器上得到不同的输出。在我的开发机器上，CPLEX输出一个结果（异常并卡在某个大值上），在所有其他机器上，结果都可以。
首先，我认为它与操作系统有关，因为我的开发机器上同时有视窗7 x64和视窗8 x64，所以我尝试在两个系统上运行应用程序。结果是一样的——有缺陷。
然后我试着在两台不同的台式机上运行，效果很好。我甚至在虚拟机内部进行了尝试，在我的开发机器上安装了Windows7Guest，在所有这些情况下，结果都是正常的。
如何解决这个问题？我如何调试正在发生的事情？任何建议都会有帮助。
其他信息：
-
我的开发机器是华硕N73N笔记本电脑，配有英特尔酷睿i7-2670QM CPU@2.20GHz，4核，8逻辑处理器和8GB内存。
- 使用的CPLEX： v12.4
以下是我开发机器的错误输出：
IBM ILOG CPLEX Optimization Studio Commercial Edition
Tried aggregator 1 time.
MIP Presolve eliminated 125 rows and 0 columns.
MIP Presolve modified 15000 coefficients.
Reduced MIP has 28795 rows, 14520 columns, and 965648 nonzeros.
Reduced MIP has 14520 binaries, 0 generals, 0 SOSs, and 0 indicators.
Probing time = 0.05 sec.
Tried aggregator 1 time.
Presolve time = 1.39 sec.
Found feasible solution after 1.48 sec. Objective = 4.5736e+016
Probing time = 0.03 sec.
Clique table members: 14520.
MIP emphasis: balance optimality and feasibility.
MIP search method: dynamic search.
Parallel mode: deterministic, using up to 8 threads.
Root relaxation solution time = 2.70 sec.
Nodes Cuts/
Node Left Objective IInf Best Integer Best Bound ItCnt Gap
* 0+ 0 4.57357e+016 1870 ---
* 0+ 0 2.93779e+016 1870 ---
0 0 1.54596e+015 811 2.93779e+016 1.54596e+015 1870 94.74%
* 0+ 0 5.78471e+015 1.54596e+015 2258 73.28%
0 0 1.59388e+015 942 5.78471e+015 Cuts: 377 2258 72.45%
0 0 1.65120e+015 813 5.78471e+015 ZeroHalf: 216 2590 71.46%
0 0 1.67178e+015 968 5.78471e+015 ZeroHalf: 285 2857 71.10%
0 0 1.71285e+015 957 5.78471e+015 ZeroHalf: 123 3241 70.39%
0 0 1.73687e+015 991 5.78471e+015 ZeroHalf: 148 3642 69.97%
0 0 1.76181e+015 997 5.78471e+015 ZeroHalf: 50 4048 69.54%
0 0 1.77762e+015 976 5.78471e+015 ZeroHalf: 23 4344 69.27%
0 0 1.78664e+015 977 5.78471e+015 ZeroHalf: 22 4601 69.11%
0 0 1.80055e+015 994 5.78471e+015 ZeroHalf: 75 4943 68.87%
0 0 1.80615e+015 993 5.78471e+015 ZeroHalf: 9 5143 68.78%
0 0 1.81464e+015 996 5.78471e+015 ZeroHalf: 33 5440 68.63%
0 0 1.82294e+015 991 5.78471e+015 Cuts: 27 5765 68.49%
0 0 1.83208e+015 989 5.78471e+015 Cuts: 18 6104 68.33%
0 0 1.84944e+015 946 5.78471e+015 Cuts: 182 6698 68.03%
0 0 1.88828e+015 956 5.78471e+015 Cuts: 269 7432 67.36%
0 0 1.89984e+015 971 5.78471e+015 Cuts: 118 7822 67.16%
0 0 1.91068e+015 985 5.78471e+015 Cuts: 45 8155 66.97%
0 0 1.92274e+015 980 5.78471e+015 Cuts: 40 8597 66.76%
0 0 1.94969e+015 977 5.78471e+015 Cuts: 121 9295 66.30%
0 0 1.95749e+015 970 5.78471e+015 Cuts: 69 9682 66.16%
0 0 1.96689e+015 978 5.78471e+015 Cuts: 42 10093 66.00%
0 0 1.97224e+015 980 5.78471e+015 Cuts: 30 10392 65.91%
0 0 1.97649e+015 978 5.78471e+015 Cuts: 13 10622 65.83%
0 0 1.98557e+015 976 5.78471e+015 Cuts: 50 10876 65.68%
0 0 1.99259e+015 987 5.78471e+015 Cuts: 46 11183 65.55%
0 0 2.00116e+015 987 5.78471e+015 Cuts: 77 11494 65.41%
0 0 2.01149e+015 990 5.78471e+015 Cuts: 37 11921 65.23%
0 0 2.01969e+015 976 5.78471e+015 Cuts: 60 12286 65.09%
0 0 2.02492e+015 966 5.78471e+015 Cuts: 43 12578 65.00%
0 0 2.03316e+015 969 5.78471e+015 Cuts: 84 12881 64.85%
0 0 2.04175e+015 971 5.78471e+015 Cuts: 52 13174 64.70%
0 0 2.05273e+015 964 5.78471e+015 Cuts: 68 13649 64.51%
0 0 2.05591e+015 966 5.78471e+015 Cuts: 23 13902 64.46%
0 0 2.06170e+015 964 5.78471e+015 Cuts: 43 14235 64.36%
0 0 2.06574e+015 971 5.78471e+015 Cuts: 46 14481 64.29%
0 0 2.07500e+015 952 5.78471e+015 Cuts: 47 14890 64.13%
0 0 2.08393e+015 941 5.78471e+015 Cuts: 53 15280 63.98%
0 0 2.09065e+015 948 5.78471e+015 Cuts: 36 15601 63.86%
0 0 2.09803e+015 955 5.78471e+015 Cuts: 104 16015 63.73%
0 0 2.10316e+015 944 5.78471e+015 Cuts: 63 16384 63.64%
0 0 2.11460e+015 948 5.78471e+015 Cuts: 76 16846 63.45%
0 0 2.12942e+015 956 5.78471e+015 Cuts: 97 17366 63.19%
0 0 2.16151e+015 931 5.78471e+015 Cuts: 327 18262 62.63%
0 0 2.17502e+015 930 5.78471e+015 Cuts: 120 18714 62.40%
0 0 2.27173e+015 884 5.78471e+015 Cuts: 385 20411 60.73%
0 0 2.29553e+015 899 5.78471e+015 Cuts: 134 21117 60.32%
0 0 2.30977e+015 920 5.78471e+015 Cuts: 71 21621 60.07%
0 0 2.31979e+015 909 5.78471e+015 Cuts: 38 22064 59.90%
0 0 2.32293e+015 911 5.78471e+015 Cuts: 33 22333 59.84%
0 0 2.32804e+015 908 5.78471e+015 Cuts: 60 22682 59.76%
0 0 2.33324e+015 906 5.78471e+015 Cuts: 43 22991 59.67%
0 0 2.34725e+015 909 5.78471e+015 Cuts: 22 23332 59.42%
0 0 2.35327e+015 903 5.78471e+015 Cuts: 72 23571 59.32%
0 0 2.36485e+015 889 5.78471e+015 Cuts: 61 23938 59.12%
0 0 2.37310e+015 884 5.78471e+015 Cuts: 45 24319 58.98%
0 0 2.39039e+015 885 5.78471e+015 Cuts: 79 24958 58.68%
0 0 2.39677e+015 901 5.78471e+015 Cuts: 56 25186 58.57%
0 0 2.40149e+015 885 5.78471e+015 Cuts: 41 25442 58.49%
0 0 2.40325e+015 883 5.78471e+015 Cuts: 60 25617 58.46%
0 0 2.41812e+015 894 5.78471e+015 Cuts: 84 26070 58.20%
0 0 2.42607e+015 903 5.78471e+015 Cuts: 68 26474 58.06%
0 0 2.52725e+015 844 5.78471e+015 Cuts: 524 28147 56.31%
0 0 2.55516e+015 835 5.78471e+015 Cuts: 196 28727 55.83%
0 0 3.23302e+015 654 5.78471e+015 Cuts: 889 30777 44.11%
0 0 3.32795e+015 604 5.78471e+015 Cuts: 337 31390 42.47%
0 0 3.37780e+015 686 5.78471e+015 Cuts: 212 31791 41.61%
0 0 3.41202e+015 701 5.78471e+015 Cuts: 121 32148 41.02%
0 0 3.52260e+015 654 5.78471e+015 Cuts: 398 33114 39.10%
0 0 3.57604e+015 639 5.78471e+015 Cuts: 244 33613 38.18%
0 0 3.61288e+015 625 5.78471e+015 Cuts: 235 34098 37.54%
0 0 3.67035e+015 615 5.78471e+015 Cuts: 163 34491 36.55%
0 0 3.73087e+015 663 5.78471e+015 Cuts: 157 35105 35.50%
0 0 3.75834e+015 668 5.78471e+015 Cuts: 144 35530 35.03%
0 0 3.77753e+015 662 5.78471e+015 Cuts: 93 35899 34.70%
0 0 3.81446e+015 660 5.78471e+015 Cuts: 82 36310 34.06%
0 0 3.83284e+015 648 5.78471e+015 Cuts: 76 36608 33.74%
0 0 3.84779e+015 656 5.78471e+015 Cuts: 58 36845 33.48%
0 0 3.85941e+015 639 5.78471e+015 Cuts: 67 37084 33.28%
0 0 3.87033e+015 649 5.78471e+015 Cuts: 40 37284 33.09%
0 0 3.87967e+015 662 5.78471e+015 Cuts: 32 37422 32.93%
0 0 3.88305e+015 666 5.78471e+015 Cuts: 24 37577 32.87%
0 0 3.89084e+015 654 5.78471e+015 Cuts: 59 37771 32.74%
0 0 3.89489e+015 654 5.78471e+015 Cuts: 38 37968 32.67%
0 0 3.90160e+015 669 5.78471e+015 Cuts: 55 38172 32.55%
0 0 3.90440e+015 663 5.78471e+015 Cuts: 40 38329 32.50%
0 0 3.90794e+015 660 5.78471e+015 Cuts: 27 38492 32.44%
0 0 3.91249e+015 661 5.78471e+015 Cuts: 31 38662 32.36%
0 0 3.91569e+015 673 5.78471e+015 Cuts: 32 38814 32.31%
0 2 3.91569e+015 673 5.78471e+015 3.91569e+015 38814 32.31%
Elapsed real time = 312.08 sec. (tree size = 0.01 MB, solutions = 3)
1 3 4.19433e+015 728 5.78471e+015 3.91569e+015 41427 32.31%
2 2 cutoff 5.78471e+015 3.91569e+015 43784 32.31%
3 3 4.35193e+015 730 5.78471e+015 3.91569e+015 45337 32.31%
4 2 cutoff 5.78471e+015 3.91569e+015 46000 32.31%
5 3 4.52240e+015 741 5.78471e+015 3.91569e+015 47554 32.31%
6 2 cutoff 5.78471e+015 3.91569e+015 49263 32.31%
7 3 4.66098e+015 799 5.78471e+015 3.91569e+015 51056 32.31%
8 2 cutoff 5.78471e+015 3.91569e+015 51943 32.31%
9 3 4.85025e+015 807 5.78471e+015 3.91569e+015 52422 32.31%
12 2 cutoff 5.78471e+015 3.91569e+015 53363 32.31%
Elapsed real time = 350.88 sec. (tree size = 0.01 MB, solutions = 3)
15 3 5.59278e+015 896 5.78471e+015 3.91569e+015 54625 32.31%
Clique cuts applied: 1901
Implied bound cuts applied: 160
Zero-half cuts applied: 150
Root node processing (before b&c):
Real time = 310.11
Parallel b&c, 8 threads:
Real time = 39.39
Sync time (average) = 0.00
Wait time (average) = 0.00
-------
Total (root+branch&cut) = 349.50 sec.
inst\instance.txt 00:05:50.9629598 5,78471010310452E+15
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
1 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 1 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0
这是所有其他机器的正常输出：
IBM ILOG CPLEX Optimization Studio Commercial Edition
Tried aggregator 1 time.
MIP Presolve eliminated 129 rows and 0 columns.
MIP Presolve modified 15480 coefficients.
Reduced MIP has 28791 rows, 14520 columns, and 965163 nonzeros.
Reduced MIP has 14520 binaries, 0 generals, 0 SOSs, and 0 indicators.
Probing time = 0.03 sec.
Tried aggregator 1 time.
Presolve time = 1.51 sec.
Found feasible solution after 1.62 sec. Objective = 5170425.9868
Probing time = 0.05 sec.
Clique table members: 14520.
MIP emphasis: balance optimality and feasibility.
MIP search method: dynamic search.
Parallel mode: deterministic, using up to 8 threads.
Root relaxation solution time = 4.82 sec.
Nodes Cuts/
Node Left Objective IInf Best Integer Best Bound ItCnt Gap
* 0+ 0 5170425.9868 2742 ---
* 0+ 0 2176197.5073 2742 ---
0 0 963286.8006 1320 2176197.5073 963286.8006 2742 55.74%
* 0+ 0 1742629.0387 963286.8006 3146 44.72%
0 0 968918.1899 1292 1742629.0387 Cuts: 180 3146 44.40%
0 0 972621.9253 1285 1742629.0387 ZeroHalf: 36 3543 44.19%
0 0 974930.7260 1336 1742629.0387 ZeroHalf: 157 3934 44.05%
0 0 975330.5044 1344 1742629.0387 ZeroHalf: 9 4042 44.03%
0 0 976752.3324 1344 1742629.0387 ZeroHalf: 23 4298 43.95%
0 0 977523.1431 1375 1742629.0387 ZeroHalf: 13 4523 43.91%
0 0 978030.2843 1358 1742629.0387 ZeroHalf: 7 4732 43.88%
0 0 978582.8390 1290 1742629.0387 Cuts: 13 4897 43.84%
0 0 979462.4757 1394 1742629.0387 Cuts: 22 5107 43.79%
0 0 980081.0140 1361 1742629.0387 Cuts: 16 5256 43.76%
0 0 986296.0889 1453 1742629.0387 Cuts: 135 6066 43.40%
0 0 987753.3557 1412 1742629.0387 Cuts: 105 6386 43.32%
0 0 988885.7034 1442 1742629.0387 Cuts: 57 6694 43.25%
0 0 989742.4516 1434 1742629.0387 Cuts: 27 6924 43.20%
0 0 990497.2364 1434 1742629.0387 Cuts: 45 7178 43.16%
* 0+ 0 1402648.1251 990497.2364 7489 29.38%
0 0 991849.2315 1437 1402648.1251 Cuts: 288 7489 29.29%
0 0 1243733.1371 593 1402648.1251 Cuts: 2446 12636 11.33%
0 0 1276887.2546 583 1402648.1251 Cuts: 556 14078 8.97%
0 0 1296511.1088 701 1402648.1251 Cuts: 600 14813 7.57%
0 0 1305221.3720 679 1402648.1251 Cuts: 193 15488 6.95%
0 0 1309747.0795 660 1402648.1251 Cuts: 139 15848 6.62%
0 0 1318085.1042 725 1402648.1251 Cuts: 43 16281 6.03%
0 0 1323312.2600 747 1402648.1251 Cuts: 55 16728 5.66%
0 0 1325171.8791 677 1402648.1251 Cuts: 23 17033 5.52%
0 0 1326274.8424 704 1402648.1251 Cuts: 19 17265 5.44%
0 0 1327853.0467 714 1402648.1251 Cuts: 20 17519 5.33%
0 0 1328078.7180 687 1402648.1251 Cuts: 6 17621 5.32%
0 0 1328705.5414 702 1402648.1251 Cuts: 18 17768 5.27%
0 0 1329753.6276 716 1402648.1251 Cuts: 15 17988 5.20%
0 0 1330728.0832 705 1402648.1251 ZeroHalf: 5 18165 5.13%
0 0 1331208.9314 707 1402648.1251 Cuts: 15 18322 5.09%
0 0 1331935.6017 709 1402648.1251 ZeroHalf: 5 18432 5.04%
0 0 1332438.9391 738 1402648.1251 ZeroHalf: 4 18536 5.01%
0 0 1332565.6221 712 1402648.1251 Cuts: 3 18578 5.00%
0 0 1332961.2175 712 1402648.1251 ZeroHalf: 7 18653 4.97%
0 2 1332962.3523 711 1402648.1251 1332962.3523 18656 4.97%
Elapsed real time = 136.63 sec. (tree size = 0.01 MB, solutions = 4)
1 3 1355937.1387 819 1402648.1251 1332962.3523 20475 4.97%
2 2 cutoff 1402648.1251 1332962.3523 20810 4.97%
3 3 1373864.6210 889 1402648.1251 1332962.3523 21989 4.97%
5 3 1391805.4913 902 1402648.1251 1332962.3523 22812 4.97%
6 2 cutoff 1402648.1251 1332962.3523 24410 4.97%
7 1 cutoff 1402648.1251 1332962.3523 25377 4.97%
8 0 cutoff 1402648.1251 1332962.3523 26154 4.97%
Clique cuts applied: 2880
Implied bound cuts applied: 73
Zero-half cuts applied: 97
Root node processing (before b&c):
Real time = 134.80
Parallel b&c, 8 threads:
Real time = 16.44
Sync time (average) = 0.00
Wait time (average) = 0.00
-------
Total (root+branch&cut) = 151.24 sec.
inst\instance.txt 00:02:32.8137404 1402648.12514155
0 0 0 0 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
共有1个答案
解鸿运
首先，你应该绝对确定你正在解决完全相同的问题。
查看日志，CPLEX解决的问题似乎有所不同
例如，presolve在第一种情况下消除125行，在第二种情况下消除129行，在第一种情况下修改15000个系数，在第二种情况下修改15480个系数；此外，缩减后的MIP似乎有所不同：
第一种情况下为28795行、14520列和965648个非零，第二种情况下为28791行、14520列和965163个非零。
我非常确定CPLEX presolve是完全确定的，并且绝对独立于平台。所以在我看来，出于某种原因，你产生了两个不同的问题。
您可以尝试将问题导出到dev和测试机器上的LP文件，然后检查这两个文件是否相等（或者甚至尝试将测试机器上生成的.LP文件解析到开发机器上，并检查日志）
无论如何，在IBM网站上有一个关于“运行再现性”的部分
他们指出，在不同的硬件体系结构上，结果实际上可以改变，但最优解决方案目标值（当然）保持不变
然而，我真的怀疑这是您的情况（值太不同），所以我会比较导出的LP问题，看看开发机器和测试机器之间是否有任何差异。
类似资料：
-
这是代码： 如果我在我的机器（）或这里（）上尝试： 相反，这里（）： 这是不同的。这是由于机器厄普西隆？还是编译器精度标志？还是不同的评估？ 造成这种漂移的原因是什么？问题似乎出现在函数中（因为其他值似乎相同）。
-
我有相同的图像和相同大小的画布，但是输出是不同的。我想要同样的输出，该怎么做呢？
-
我在Cplex中使用Python API来解决一个线性编程问题。使用Cplex时，我的结果如下： 但随后我将LP prolem保存为LP文件，并再次使用Cplex进行求解，结果与第一个略有不同： 下面是我的功能：
-
<代码>car\U gear字段在数据库中填写为“stick”（斗杆）。在图像标记后，输出更改为“自动” 为什么结果是$car\u result1-
-
问题内容： 我正在使用具有Spring安全性的BCryptPasswordEncoder。我的期望是，对于相同的输入，我将始终获得相同的输出。但是对于相同的输入，我得到不同的输出。您可以使用以下代码段对其进行测试： 输出：$ 2a $ 10 $ cYLM.qoXpeAzcZhJ3oXRLu9Slkb61LHyWW5qJ4QKvHEMhaxZ5qCPi 输出2：$ 2a $ 10 $ KEvYX9y
-
我正在使用带有Spring Security的BCryptPasswordEncoder。我的期望是，对于相同的输入，我总是得到相同的输出。但是对于相同的输入，我得到不同的输出。您可以使用下面的代码片段对其进行测试： 输出:$2A$10$CYLM.QOXPEAZCZHJ3OXRLU9SLKB61LHYWW5QJ4QKVHEMHAXZ5QCPI 输出2:$2A$10$kevyx9yjj0f1x3wl