评估CPU内核的优化 - 页 3

新评论
 
Aleksey Vyazmikin:

我们的想法是，你可以得到一个更快的处理器（在eBay或阿里上买），它就会很好，但它不够快。

不，它是一个焊接的处理器。事实上，它是一个媒体播放器，用于观看各种胡言乱语的睡眠，与投影仪 相连。实验的重点 是 在 现有的最cheese的硬件上 进行测试（这也是metatrader运行的地方）。

已添加。

一个良好的开端。

code generated          
0 error(s), 0 warning(s), 344528 msec elapsed  
 
Serhii Shevchuk:

不，有一个处理器被焊接在里面。它基本上是一个媒体播放器，用于观看各种无意义的睡眠，连接到投影仪上。实验的重点 是 在 现有的最俗气的硬件（运行Metatrader）上 测试。

啊，原来如此，是8G内存的可用性让我感到困惑!

谢尔盖-谢夫丘克。

一个良好的开端。

而且，如果你把计算移到正文中，而不是移到一个函数中，就需要花几个小时来编译......

 
Aleksey Vyazmikin:

请在有4个和8个代理的FX上测试这个版本。

所以。
代理人。4.优化通道：8。结果。

2019.08.10 01:31:47.465 Core 4  pass 2 returned result 1001000.00 in 0:02:54.814
2019.08.10 01:31:48.782 Core 1  pass 0 returned result 1001000.00 in 0:02:56.136
2019.08.10 01:31:49.263 Core 2  pass 4 returned result 1001000.00 in 0:02:56.557
2019.08.10 01:31:50.412 Core 3  pass 6 returned result 1001000.00 in 0:02:57.711
2019.08.10 01:34:45.025 Core 4  pass 3 returned result 1001000.00 in 0:02:57.560
2019.08.10 01:34:45.482 Core 2  pass 5 returned result 1001000.00 in 0:02:56.218
2019.08.10 01:34:45.637 Core 1  pass 1 returned result 1001000.00 in 0:02:56.850
2019.08.10 01:34:49.330 Core 3  pass 7 returned result 1001000.00 in 0:02:58.916
2019.08.10 01:34:49.330 Tester  optimization finished, total passes 8
2019.08.10 01:34:49.341 Statistics      optimization done in 5 minutes 58 seconds
2019.08.10 01:34:49.341 Statistics      shortest pass 0:02:54.814, longest pass 0:02:58.916, average pass 0:02:56.845

代理人。8.优化通道：8。结果。

2019.08.10 01:41:03.259 Core 2  pass 1 returned result 1001000.00 in 0:04:22.641
2019.08.10 01:41:07.004 Core 8  pass 7 returned result 1001000.00 in 0:04:25.297
2019.08.10 01:41:07.715 Core 7  pass 4 returned result 1001000.00 in 0:04:26.076
2019.08.10 01:41:08.051 Core 1  pass 0 returned result 1001000.00 in 0:04:27.445
2019.08.10 01:41:09.096 Core 6  pass 5 returned result 1001000.00 in 0:04:27.458
2019.08.10 01:41:09.459 Core 4  pass 3 returned result 1001000.00 in 0:04:28.851
2019.08.10 01:41:09.695 Core 3  pass 2 returned result 1001000.00 in 0:04:29.082
2019.08.10 01:41:09.966 Core 5  pass 6 returned result 1001000.00 in 0:04:28.213
2019.08.10 01:41:09.966 Tester  optimization finished, total passes 8
2019.08.10 01:41:09.977 Statistics      optimization done in 4 minutes 29 seconds
2019.08.10 01:41:09.977 Statistics      shortest pass 0:04:22.641, longest pass 0:04:29.082, average pass 0:04:26.882

已添加。

所以你是对的--FX处理器有超线程的味道（尽管我认为它是一个完整的8核）。

而代理管理器也将其视为8核（在现有7个代理的情况下，它提供增加1个，见图片）。

代理人

P.S. 现在这很有趣!

删除了所有的代理。决定再次添加。然后--这是我不知道多少年来的第一次--我被要求增加4名代理人，而不是通常的8名。

代理商2019

因此，在测试器中与古代AMD石头有关的东西已经发生了变化，而且，是非常新鲜的变化。

P.P.S.

是的，2085年的建设也提议增加8个代理人。一个非常新鲜的创新。

2085

随着时间的推移，一切都在萎缩。以前有8名特工，现在有4名。
 
Serhii Shevchuk:

所以。
代理人。4.优化通行证：8。结果。

代理人。8.优化通道：8。结果。

已添加。

所以你是对的--FX处理器有超线程的味道（尽管我认为它是一个完整的8核）。

而代理管理器也将其视为8核（在现有7个代理的情况下，它提供增加1个，见图片）。


P.S. 现在这很有趣!

删除了所有的代理。决定再次添加。然后--这是我不知道多少年来的第一次--我被要求增加4名代理人，而不是通常的8名。

因此，在测试器中与古代AMD石头有关的东西已经发生了变化，而且，是非常新鲜的变化。

P.P.S.

是的，2085年的建设也提议增加8个代理人。这是一个相当新鲜的创新。


随着时间的推移，一切都在萎缩。以前有8名特工，现在有4名。

我不知道已经发生了对Build 2097的更新--我从7月中旬开始使用它（作为测试者），我知道这种禁用有问题的代理的方法--超交易和诸如AMD，目前所有的代理仍然可以在本地机器上使用，希望这不会改变。为什么这样做--我怀疑是为了在出售资源时对其价值进行公平评估。

不过，有趣的是，8个线程的收益非常体面--25%，而且加载所有8个代理是值得的。

在AMD的两个计算核心（ALU），负责加法和减法的操作，还有比较逻辑，可能还有1个核心（FPU），负责除法和乘法，即浮点运算，也就是说，它不是纯粹的超交易。

 

抓到一个上网本，但它比Xeron快--令人惊讶。

2019.08.05 22:37:53.817 Terminal        Windows 7 Service Pack 1 (build 7601), IE 11, UAC, Intel Atom  N570 @ 1.66 GHz, Memory: 578 / 2035 Mb, Disk: 56 / 280 Gb, GMT+3

2019.08.10 23:58:27.648 Core 1  pass 0 returned result 1001000.00 in 0:09:33.408
2019.08.10 23:58:28.363 Core 2  pass 2 returned result 1001000.00 in 0:09:34.188
2019.08.11 00:08:04.213 Core 1  pass 1 returned result 1001000.00 in 0:09:36.913
2019.08.11 00:08:05.355 Core 2  pass 3 returned result 1001000.00 in 0:09:37.257
2019.08.11 00:08:05.355 Tester  optimization finished, total passes 4
2019.08.11 00:08:05.366 Statistics      optimization done in 19 minutes 11 seconds
2019.08.11 00:08:05.366 Statistics      shortest pass 0:09:33.408, longest pass 0:09:37.257, average pass 0:09:35.441
 

有机会用FX-8350测试主机--没有4000Mhz超频。

2019.08.11 10:41:32.541 Terminal        Windows 7 Service Pack 1 (build 7601) x64, IE 10, AMD FX-8350 Eight-Core Processor , Memory: 21877 / 24533 Mb, Disk: 51 / 499 Gb, GMT+3

Tree_Brut_TestPL - 4个代理商

2019.08.11 10:57:01.270 Core 3  pass 2 returned result 1001000.00 in 0:01:21.389
2019.08.11 10:57:01.466 Core 1  pass 0 returned result 1001000.00 in 0:01:21.616
2019.08.11 10:57:01.851 Core 4  pass 6 returned result 1001000.00 in 0:01:21.950
2019.08.11 10:57:03.201 Core 2  pass 4 returned result 1001000.00 in 0:01:23.292
2019.08.11 10:58:21.943 Core 3  pass 3 returned result 1001000.00 in 0:01:20.680
2019.08.11 10:58:22.763 Core 1  pass 1 returned result 1001000.00 in 0:01:21.304
2019.08.11 10:58:23.899 Core 4  pass 7 returned result 1001000.00 in 0:01:22.056
2019.08.11 10:58:26.569 Core 2  pass 5 returned result 1001000.00 in 0:01:23.375
2019.08.11 10:58:26.569 Tester  optimization finished, total passes 8
2019.08.11 10:58:26.579 Statistics      optimization done in 2 minutes 47 seconds
2019.08.11 10:58:26.579 Statistics      shortest pass 0:01:20.680, longest pass 0:01:23.375, average pass 0:01:21.957

Tree_Brut_TestPL - 8个代理商

2019.08.11 11:11:21.820 Core 7  pass 5 returned result 1001000.00 in 0:02:03.874
2019.08.11 11:11:22.139 Core 2  pass 2 returned result 1001000.00 in 0:02:04.354
2019.08.11 11:11:24.113 Core 3  pass 6 returned result 1001000.00 in 0:02:06.141
2019.08.11 11:11:24.195 Core 4  pass 1 returned result 1001000.00 in 0:02:06.470
2019.08.11 11:11:24.394 Core 6  pass 0 returned result 1001000.00 in 0:02:06.539
2019.08.11 11:11:24.917 Core 5  pass 7 returned result 1001000.00 in 0:02:06.966
2019.08.11 11:11:28.852 Core 1  pass 3 returned result 1001000.00 in 0:02:11.027
2019.08.11 11:11:30.336 Core 8  pass 4 returned result 1001000.00 in 0:02:12.302
2019.08.11 11:11:30.336 Tester  optimization finished, total passes 8
2019.08.11 11:11:30.346 Statistics      optimization done in 2 minutes 13 seconds
2019.08.11 11:11:30.346 Statistics      shortest pass 0:02:03.874, longest pass 0:02:12.302, average pass 0:02:07.209

Tree_Brut_TestPL_F - 4个代理商

2019.08.11 11:15:56.836 Core 4  pass 0 returned result 1001000.00 in 0:02:03.360
2019.08.11 11:15:57.088 Core 3  pass 6 returned result 1001000.00 in 0:02:03.567
2019.08.11 11:15:57.744 Core 2  pass 4 returned result 1001000.00 in 0:02:04.248
2019.08.11 11:15:58.259 Core 1  pass 2 returned result 1001000.00 in 0:02:04.762
2019.08.11 11:17:53.839 Core 2  pass 5 returned result 1001000.00 in 0:01:56.106
2019.08.11 11:17:55.203 Core 3  pass 7 returned result 1001000.00 in 0:01:58.126
2019.08.11 11:17:55.210 Core 4  pass 1 returned result 1001000.00 in 0:01:58.387
2019.08.11 11:17:55.615 Core 1  pass 3 returned result 1001000.00 in 0:01:57.366
2019.08.11 11:17:55.615 Tester  optimization finished, total passes 8
2019.08.11 11:17:55.625 Statistics      optimization done in 4 minutes 03 seconds
2019.08.11 11:17:55.625 Statistics      shortest pass 0:01:56.106, longest pass 0:02:04.762, average pass 0:02:00.740

Tree_Brut_TestPL_F - 8家代理商

2019.08.11 11:24:05.758 Core 6  pass 6 returned result 1001000.00 in 0:03:30.450
2019.08.11 11:24:06.511 Core 1  pass 2 returned result 1001000.00 in 0:03:32.370
2019.08.11 11:24:07.029 Core 4  pass 3 returned result 1001000.00 in 0:03:32.860
2019.08.11 11:24:08.345 Core 2  pass 1 returned result 1001000.00 in 0:03:34.210
2019.08.11 11:24:08.447 Core 5  pass 7 returned result 1001000.00 in 0:03:33.167
2019.08.11 11:24:08.482 Core 3  pass 0 returned result 1001000.00 in 0:03:34.280
2019.08.11 11:24:08.768 Core 8  pass 4 returned result 1001000.00 in 0:03:33.688
2019.08.11 11:24:10.260 Core 7  pass 5 returned result 1001000.00 in 0:03:35.018
2019.08.11 11:24:10.260 Tester  optimization finished, total passes 8
2019.08.11 11:24:10.270 Statistics      optimization done in 3 minutes 37 seconds
2019.08.11 11:24:10.270 Statistics      shortest pass 0:03:30.450, longest pass 0:03:35.018, average pass 0:03:33.255

而使用8个代理与4个代理相比，却有明显的性能提升，尽管使用4个代理时每个代理的速度明显更高。

 

下一台电脑将是老式的Phenom II X6 1065T，CPU为2900Mhz。

2019.08.11 12:08:13.905 Terminal        Windows 7 (build 7600) x64, IE 8, AMD Phenom II X6 1065 T Processor, Memory: 5996 / 8191 Mb, Disk: 79 / 224 Gb, GMT+3

Tree_Brut_TestPL - 6个代理商

2019.08.11 12:11:06.766 Core 3  pass 0 returned result 1001000.00 in 0:02:16.413
2019.08.11 12:11:07.043 Core 2  pass 4 returned result 1001000.00 in 0:02:16.593
2019.08.11 12:11:08.079 Core 4  pass 2 returned result 1001000.00 in 0:02:17.538
2019.08.11 12:11:08.277 Core 6  pass 3 returned result 1001000.00 in 0:02:17.698
2019.08.11 12:11:16.133 Core 1  pass 1 returned result 1001000.00 in 0:02:25.720
2019.08.11 12:11:17.239 Core 5  pass 5 returned result 1001000.00 in 0:02:26.692
2019.08.11 12:11:17.239 Tester  optimization finished, total passes 6
2019.08.11 12:11:17.249 Statistics      optimization done in 2 minutes 27 seconds
2019.08.11 12:11:17.249 Statistics      shortest pass 0:02:16.413, longest pass 0:02:26.692, average pass 0:02:20.109

Tree_Brut_TestPL - 3个代理商

2019.08.11 12:16:01.529 Core 2  pass 4 returned result 1001000.00 in 0:02:17.960
2019.08.11 12:16:01.530 Core 1  pass 2 returned result 1001000.00 in 0:02:17.960
2019.08.11 12:16:01.787 Core 3  pass 0 returned result 1001000.00 in 0:02:18.219
2019.08.11 12:18:19.602 Core 2  pass 5 returned result 1001000.00 in 0:02:18.073
2019.08.11 12:18:19.630 Core 1  pass 3 returned result 1001000.00 in 0:02:18.100
2019.08.11 12:18:20.100 Core 3  pass 1 returned result 1001000.00 in 0:02:18.311
2019.08.11 12:18:20.100 Tester  optimization finished, total passes 6
2019.08.11 12:18:20.110 Statistics      optimization done in 4 minutes 37 seconds
2019.08.11 12:18:20.110 Statistics      shortest pass 0:02:17.960, longest pass 0:02:18.311, average pass 0:02:18.103

在3个代理上进行了测试，这样就可以看到与FX相比在结构上的差异--在这里我们有一个真正的6个FPU。

Tree_Brut_TestPL_F - 6个代理商

2019.08.11 12:23:16.283 Core 1  pass 0 returned result 1001000.00 in 0:03:39.626
2019.08.11 12:23:16.652 Core 5  pass 4 returned result 1001000.00 in 0:03:39.614
2019.08.11 12:23:16.861 Core 3  pass 1 returned result 1001000.00 in 0:03:40.286
2019.08.11 12:23:17.968 Core 2  pass 2 returned result 1001000.00 in 0:03:41.294
2019.08.11 12:23:30.936 Core 4  pass 5 returned result 1001000.00 in 0:03:53.860
2019.08.11 12:23:32.878 Core 6  pass 3 returned result 1001000.00 in 0:03:55.949
2019.08.11 12:23:32.878 Tester  optimization finished, total passes 6
2019.08.11 12:23:32.888 Statistics      optimization done in 3 minutes 57 seconds
2019.08.11 12:23:32.888 Statistics      shortest pass 0:03:39.614, longest pass 0:03:55.949, average pass 0:03:45.104

在核心速度上仅比FX落后20秒，考虑到1100Mhz的频率差异，这并不坏

 

我的结果（Inter Core i7-8700，3.2 GHz，16 Gb，启用超交易，有12个代理，但只有6个工作，按物理核心的数量计算）。

2019.08.11 12:35:27.825 Core 04 pass 1 returned result 1001000.00 in 0:01:10.378
2019.08.11 12:35:27.841 Core 01 pass 0 returned result 1001000.00 in 0:01:10.241
2019.08.11 12:35:28.620 Core 06 pass 2 returned result 1001000.00 in 0:01:11.130
2019.08.11 12:35:28.626 Core 03 pass 5 returned result 1001000.00 in 0:01:11.036
2019.08.11 12:35:28.704 Core 12 pass 3 returned result 1001000.00 in 0:01:11.100
2019.08.11 12:35:29.296 Core 02 pass 4 returned result 1001000.00 in 0:01:11.706
2019.08.11 12:35:29.296 Tester  optimization finished, total passes 6
2019.08.11 12:35:29.307 Statistics      optimization done in 1 minutes 13 seconds
2019.08.11 12:35:29.307 Statistics      shortest pass 0:01:10.241, longest pass 0:01:11.706, average pass 0:01:10.931
在我看来，对于超级交易，数据适合于CPU缓存是非常重要的。当处理器不需要访问主内存时，当所有的数据都在缓存中时，虚拟核心恰恰会受益。相应地，如果正在计算大的数组（几年后的真实刻度）--超交易将没有什么用。然而，当处理相对较小的数据时（根据我的经验--在1MOHLC上大约一年或一年半）--超级交易带来相当明显的性能提升。
 
Georgiy Merts:

我的结果（Inter Core i7-8700，3.2 GHz，16 Gb，启用超交易，有12个代理，但只有6个工作，按物理核心的数量计算）。

在我看来，对于超级交易来说，数据适合处理器缓存是非常重要的。当处理器不需要访问主内存时，当所有数据都在缓存中时，虚拟核心就会受益。相应地，如果正在计算大的数组（几年后的真实刻度）--超交易将没有什么用。然而，当处理相对较小的数据时（根据我的经验--在1MOHLC上大约一年或一年半）--超级交易带来相当明显的性能提升。

现在建议把Tree_Brut_TestPL和Tree_Brut_TestPL_F 的测试--如果可能的话（说明是什么样的专家顾问），然后给（给）两个选项的信息，好了，有hypertrading（所有代理）和没有--虽然没有了解目标，但也需要它或 没有。

 
Aleksey Vyazmikin:

现在我们提议把Tree_Brut_TestPL和Tree_Brut_TestPL_F 的测试--如果可能的话（说明是什么样的专家顾问），然后给（给）两个版本的信息，好了，有超交易（所有代理）和没有--虽然没有了解客观，但你需要或不需要它。

以上 -树_Brut_TestPL_F

这是第二张，配置相同。

2019.08.11 14:32:43.819 Core 07 pass 4 returned result 1001000.00 in 0:00:33.157
2019.08.11 14:32:44.209 Core 04 pass 5 returned result 1001000.00 in 0:00:33.494
2019.08.11 14:32:44.291 Core 03 pass 3 returned result 1001000.00 in 0:00:33.664
2019.08.11 14:32:44.415 Core 01 pass 1 returned result 1001000.00 in 0:00:33.846
2019.08.11 14:32:44.568 Core 02 pass 0 returned result 1001000.00 in 0:00:34.031
2019.08.11 14:32:44.683 Core 05 pass 2 returned result 1001000.00 in 0:00:34.082
2019.08.11 14:32:44.683 Tester  optimization finished, total passes 6
2019.08.11 14:32:44.693 Statistics      optimization done in 0 minutes 34 seconds
2019.08.11 14:32:44.693 Statistics      shortest pass 0:00:33.157, longest pass 0:00:34.082, average pass 0:00:33.712

运行12个虚拟核心时也是如此（没有F）。

2019.08.11 14:36:39.685 Core 05 pass 4 returned result 1001000.00 in 0:01:43.939
2019.08.11 14:36:39.843 Core 03 pass 2 returned result 1001000.00 in 0:01:44.102
2019.08.11 14:36:40.327 Core 04 pass 3 returned result 1001000.00 in 0:01:44.581
2019.08.11 14:36:40.394 Core 07 pass 5 returned result 1001000.00 in 0:01:44.650
2019.08.11 14:36:40.514 Core 10 pass 7 returned result 1001000.00 in 0:01:44.215
2019.08.11 14:36:40.707 Core 06 pass 6 returned result 1001000.00 in 0:01:44.454
2019.08.11 14:36:40.732 Core 09 pass 9 returned result 1001000.00 in 0:01:44.367
2019.08.11 14:36:40.885 Core 02 pass 1 returned result 1001000.00 in 0:01:45.143
2019.08.11 14:36:41.253 Core 01 pass 0 returned result 1001000.00 in 0:01:45.512
2019.08.11 14:36:41.727 Core 12 pass 11 returned result 1001000.00 in 0:01:45.325
2019.08.11 14:36:41.786 Core 11 pass 10 returned result 1001000.00 in 0:01:45.407
2019.08.11 14:36:41.899 Core 08 pass 8 returned result 1001000.00 in 0:01:45.563
2019.08.11 14:36:41.899 Tester  optimization finished, total passes 12
2019.08.11 14:36:41.909 Statistics      optimization done in 1 minutes 46 seconds
2019.08.11 14:36:41.909 Statistics      shortest pass 0:01:43.939, longest pass 0:01:45.563, average pass 0:01:44.771

当运行12个虚拟核心时，专家 "与F"。

2019.08.11 14:48:04.349 Core 09 pass 10 returned result 1001000.00 in 0:03:16.005
2019.08.11 14:48:06.012 Core 02 pass 0 returned result 1001000.00 in 0:03:18.194
2019.08.11 14:48:06.269 Core 11 pass 4 returned result 1001000.00 in 0:03:18.152
2019.08.11 14:48:06.902 Core 01 pass 3 returned result 1001000.00 in 0:03:18.869
2019.08.11 14:48:06.925 Core 10 pass 8 returned result 1001000.00 in 0:03:18.590
2019.08.11 14:48:06.958 Core 05 pass 5 returned result 1001000.00 in 0:03:18.816
2019.08.11 14:48:07.269 Core 07 pass 7 returned result 1001000.00 in 0:03:19.044
2019.08.11 14:48:07.460 Core 04 pass 2 returned result 1001000.00 in 0:03:19.470
2019.08.11 14:48:07.818 Core 06 pass 6 returned result 1001000.00 in 0:03:19.634
2019.08.11 14:48:08.151 Core 12 pass 11 returned result 1001000.00 in 0:03:19.777
2019.08.11 14:48:08.463 Core 03 pass 1 returned result 1001000.00 in 0:03:20.563
2019.08.11 14:48:09.072 Core 08 pass 9 returned result 1001000.00 in 0:03:20.736
2019.08.11 14:48:09.072 Tester  optimization finished, total passes 12
2019.08.11 14:48:09.083 Statistics      optimization done in 3 minutes 23 seconds
2019.08.11 14:48:09.083 Statistics      shortest pass 0:03:16.005, longest pass 0:03:20.736, average pass 0:03:18.987

我现在要关闭超级交易......

专家 "没有F"，没有超额交易，有6个通道。

2019.08.11 14:56:16.918 Core 5  pass 4 returned result 1001000.00 in 0:01:09.291
2019.08.11 14:56:16.984 Core 1  pass 1 returned result 1001000.00 in 0:01:09.452
2019.08.11 14:56:17.228 Core 3  pass 2 returned result 1001000.00 in 0:01:09.635
2019.08.11 14:56:17.797 Core 6  pass 5 returned result 1001000.00 in 0:01:10.163
2019.08.11 14:56:18.164 Core 2  pass 0 returned result 1001000.00 in 0:01:10.725
2019.08.11 14:56:18.284 Core 4  pass 3 returned result 1001000.00 in 0:01:10.675

专家 "与F "和没有超额交易，有6个通道。

2019.08.11 15:01:43.644 Core 3  pass 2 returned result 1001000.00 in 0:01:10.138
2019.08.11 15:01:43.984 Core 1  pass 0 returned result 1001000.00 in 0:01:10.494
2019.08.11 15:01:44.121 Core 5  pass 4 returned result 1001000.00 in 0:01:10.585
2019.08.11 15:01:44.310 Core 2  pass 1 returned result 1001000.00 in 0:01:10.868
2019.08.11 15:01:44.318 Core 4  pass 3 returned result 1001000.00 in 0:01:10.782
2019.08.11 15:01:44.397 Core 6  pass 5 returned result 1001000.00 in 0:01:10.859
2019.08.11 15:01:44.397 Tester  optimization finished, total passes 6
2019.08.11 15:01:44.407 Statistics      optimization done in 1 minutes 12 seconds
2019.08.11 15:01:44.407 Statistics      shortest pass 0:01:10.138, longest pass 0:01:10.868, average pass 0:01:10.621
2019.08.11 15:01:44.407 Statistics      6000 frames (2.36 Mb total, 412 bytes per frame) received
12345678910...22
新评论