Linux版本可以使用比Windows更多的CPU内核吗？

僵尸。
获得Linux。
期。
您将获得更多VIVADO的多线程子工具，这些子工具仅限于Windows上较少的线程数。
我相信有一些文件。
V
弗拉迪斯拉夫·穆拉文
在原帖中查看解决方案



以下为原文

Zombie.
 
Get Linux. Period. You get more multi-threaded sub-tools of VIVADO that are limited to a lesser number of threads on Windows. I believe there is some doc on that.
 
V
 
Vladislav MuravinView solution in original post

嗨，
Vivado是我们当前的一代工具，与ISE相比具有许多先进功能，但它仅支持7系列FPGA。
如果您要定位7系列设备，我建议您使用Vivado。
如果您的目标是任何其他FPGA（6系列及以下），我建议您使用最新的ISE 14.7
我们最新的Vivado 2014.3支持以下操作系统：
Microsoft Windows支持•Windows 7和7 SP1 Professional（32位和64位），英语/日语•Windows 8.1 Professional（64位），英语/日语
Windows XP支持对于Vivado 2014.3版本，将不再支持Windows XP。
这在很大程度上是由于微软终止了对Windows XP的支持。
Linux支持•红帽企业工作站5.8  -  5.10（32位和64位）•红帽企业工作站6.4  -  6.5（32位和64位）•SUSE Linux Enterprise 11.1  -  11.2（32位和64位）
位）•Cent OS 6.4和6.5（64位）•Ubuntu Linux 14.04 LTS（64位）
最新的ISE 14.7支持以下操作系统：
Microsoft Windows支持•Windows XP Professional（32位和64位），英语/日语•Windows 7 Professional（32位和64位），英语/日语•Windows Server 2008（64位）
Linux支持•红帽企业工作站5（32位和64位）•红帽企业工作站6（32位和64位）•SUSE Linux Enterprise 11（32位和64位）
我们的工具在列出的操作系统上进行了测试，您可以根据您的意愿在支持的操作系统上使用它们。



以下为原文

Hi,
 
Vivado is our current generation tool which has many advanced features compared to ISE but it supports only 7-series FPGAs.
If you are targetting 7 series devices I would suggest you to use Vivado.
If you are targetting for any other FPGA (6 series and below) I would suggest you to use the latest ISE 14.7
 
The below operating systems are supported with our Latest Vivado 2014.3:
 
Microsoft Windows Support
• Windows 7 and 7 SP1 Professional (32-bit and 64-bit), English/Japanese
• Windows 8.1 Professional (64-bit), English/Japanese

Windows XP Support
For the Vivado 2014.3 release, Windows XP will no longer be supported. This is due, in large part, to Microsoft’s termination of Windows XP support.

Linux Support
• Red Hat Enterprise Workstation 5.8 - 5.10 (32-bit and 64-bit)
• Red Hat Enterprise Workstation 6.4 - 6.5 (32-bit and 64-bit)
• SUSE Linux Enterprise 11.1 - 11.2 (32-bit and 64-bit)
• Cent OS 6.4 and 6.5 (64-bit)
• Ubuntu Linux 14.04 LTS (64-bit)
 
The below operating systems are supported with latest ISE 14.7:
 
Microsoft Windows Support
• Windows XP Professional (32-bit and 64-bit), English/Japanese
• Windows 7 Professional (32-bit and 64-bit), English/Japanese
• Windows Server 2008 (64-bit)

Linux Support
• Red Hat Enterprise Workstation 5 (32-bit and 64-bit)
• Red Hat Enterprise Workstation 6 (32-bit and 64-bit)
• SUSE Linux Enterprise 11 (32-bit and 64-bit)
 
Our tools are tested on the listed Operating systems and you can use them on supported Operating systems according to your convinent.
 
 

感谢您的努力，但通用的复制粘贴电子邮件并没有回答我的问题。
我试图弄清楚哪个支持的操作系统使用给定的硬件会给我最高的性能，或者没有区别，这只是优点？



以下为原文

Thank you for the effort but that generic copy paste email does not answer my question on the slightest. I am trying to figure out which supported operating system with the given hardware would give me the highest performance, or there is no difference and it is just preferance? 

僵尸。
获得Linux。
期。
您将获得更多VIVADO的多线程子工具，这些子工具仅限于Windows上较少的线程数。
我相信有一些文件。
V
弗拉迪斯拉夫·穆拉文



以下为原文

Zombie.
 
Get Linux. Period. You get more multi-threaded sub-tools of VIVADO that are limited to a lesser number of threads on Windows. I believe there is some doc on that.
 
V
 
Vladislav Muravin

感谢muravin，这就是我一直在寻找的答案。
oyu可以指导我查看该文档或任何列出或讨论Linux版本和winodws版本的Xilinx工具之间的多线程差异的文档吗？



以下为原文

Thanks muravin, that was the answer I was looking for. Can oyu direct me towards that document or any document that lists or talks about the multi threading differences between linux version and winodws version of Xilinx tools? 

好吧，不是那么快：o）
我的错，我读的帖子太快 - 我以为你使用的是7系列，因此可以使用VIVADO。
在那里，您可以获得Linux的巨大优势。
因此，如果您要为Virtex-5进行综合，那么您将看到一些稳定性优势和LINUX的一小部分性能优势，因为您必然会使用ISE / XPS合成引擎。
你必须使用V5吗？
再次，抱歉。
弗拉德
弗拉迪斯拉夫·穆拉文



以下为原文

Well, not so fast :o)
 
My fault, I read your post too fast - I thought you are using 7-series and therefore could be using VIVADO. And there you get a huge benefit for Linux.
 
So, if you are going to synthesize for Virtex-5, then you will see some stability benefit and A LITTLE performance benefit on LINUX as you are bound to use ISE/XPS synthesis engine. Must you use V5?
 
Again, sorry about that.
Vlad
 
Vladislav Muravin

没问题，我必须将Virtex5用于飞行项目，因为Viretex6和7不符合飞行资格，但我将使用virtex 7作为开发板。



以下为原文

No problem, I have to use Virtex5 for flight projects since Viretex6 and 7 are not flight qualified, I will be using virtex 7 's for development boards though.

另一点是几乎所有新的Windows系统现在都是Windows 8.1，只有最新版本的工具才支持它。
除非您可以使用Windows 7获得新系统，否则如果您需要支持可能需要较旧版本ISE的任何早期设备或项目，您可能会更好地使用RedHat。
它可能是64位系统中的一种清洗，但有趣的是，似乎Linux使用更少的内存来为内核提供更多的空间（更少的缓存抖动）来运行更大的设计。
在Virtex5和更新的部件中，最长的运行时间通常在Map阶段，通常只有一个线程在运行，但您需要尽可能快地随机访问尽可能多的内存。
这意味着您可以使用可能拥有更少处理器但每个处理器更多L2缓存和更快前端总线的系统获胜。
已经在ISE中，可以使用多个处理器或线程进行综合，也可以使用SmartXplorer运行多个Map传递（如果您无法达到时序收敛，则非常有用）。
但是，如前所述，真正的长时间运行时间往往是多线程不可选的部分。
-  Gabor



以下为原文

One other point is that almost all new Windows systems are now Windows 8.1 which only is supported for the latest versions of the tools.  Unless you can get a new system with Windows 7, you might be better off with RedHat if you need to support any earlier devices or projects that might require an older version of ISE.
 
It may be a wash in 64-bit systems, but anecdotally it seems that Linux uses less memory for the kernal and allows you more room (less cache thrashing) to run larger designs.  In Virtex5 and newer parts, the longest run times are typically in the Map phase where there is usually just one thread running, but you need as fast as possible random access to as much memory as possible.  This means you win with a system that might have fewer processors but more L2 cache per processor and a faster front-side bus.
 
Already in ISE it's possible to use multiple processors or threads for synthesis, and also when running multiple Map passes using SmartXplorer (useful if you are having trouble reaching timing closure).  However as noted, the really long run times tend to be in the parts where multithreading is not an option.
-- Gabor

二级缓存......这很有趣。
当我参加计算机体系结构课程时，我的教师总是习惯说L1和L2缓存大小是每个指令计算的许多时钟中最大的限制因素，因为缓存未命中导致的内存访问时间可能会使超级处理器瘫痪并迫使无操作
...
我想我们正在使用20 + Mb的L2Cache和2133Mhz RAM来获得Haswell E系列cpu。
我几乎在Red Hat上销售，但有一件事仍然困扰着我。
Matlab，Xilinx和ModelSim都在linux下运行......非常适合FPGA FFT开发。
但我也使用Altium和Altium进行电路板布局只是PC ....对于其他一切，我使用Mac但没有任何表现Matlab在Mac上工作。
唯一可以运行我需要的所有东西而不需要任何虚拟化的平台是64位Windows 7.只有当Altium设计器被移植到Linux时，它才会很棒。



以下为原文

L2 cache...That is interesting. When I took computer architecture courses my instructor always used to say L1 and L2 cache sizes were the biggest limiters in many clocks per instruction calculation since memory access times due to cache miss could bring a super duper processor down to his knees and force no-ops... 
 
I guess it is good that we are getting the Haswell E series cpu's with 20+Mb of L2Cache and 2133Mhz RAM. 
 
I am pretty much sold on Red Hat but there is one thing that is still bothering me. Matlab, Xilinx and ModelSim all run under linux... perfect for FPGA FFT development. But I also do board layouts using Altium and Altium is PC only....For everything else I use a Mac but nothing expcet Matlab works on a mac. The only platform that can run everything I need without any virtualization is 64 bit Windows 7. Only if Altium designer was ported over to linux, it would be great. 

如果不是Vivado而不是ISE对于多线程肯定很差。
您应该获得最快的CPU（也可能最大化CPU缓存）和大量RAM（最多32GB，具体取决于项目大小）。
CPU提升（当只有一个核心工作时）也无助于2线程映射。
此外，我在Linux下运行时有不好的经验，我不知道如何解决它。
使用Windows运行，90％的人类经验可以帮助您。



以下为原文

If not Vivado than ISE is definately poor for multi-threading.
You should get the fastest CPU you can get (possibly maximize CPU cache also) and much RAM (up to 32GB depending on project size). Also CPU boosting (when only one core works) will not help with 2-threads mapping.
 
Also I have bad expirience with running under Linux and I dont know how to fix it.
Run with windows and 90% human expirience helps you.

你是对的，ISE只有2个线程用于映射，4个用于PAR。
然而，一些高利用率的FPGA导致ISE进入oods。
VIVADO最多有8个，每个都值几杆，所以是的，需要一个重型洗衣机，最好是1个以上。
至于人类经验，这是个人偏好的问题，就是这样。
人们太习惯用于Windows了。
在Linux中，您可以使用gui执行所有操作，也可以编写运行脚本并以批处理方式运行它。我们的环境是，您键入“create_new_run”缩写为“cnr”，按Enter键并继续工作;
合成完成后，您会收到一封电子邮件。
这在Linux上要容易得多，而且在Windows上设置起来要困难得多。
弗拉德
弗拉迪斯拉夫·穆拉文



以下为原文

You are right, ISE had only 2 threads for mapping and 4 for PAR. Yet some high-utilization FPGAs causes ISE to go into the oods.
 
VIVADO has up to 8 and each one is a few Gig worth, so yes, one needs to have a heavy duty washer and prefereably, more than 1.
 
As for human experience, this is a question of personal preference, that's all it is. People are simply too used for Windows. In Linux, you can either do all that in gui or you can script your run and run it in batch mode.. Our environment is, you type "create_new_run" abbreviated "cnr", hit enter, and keep working; you get an email once the synthesis is finished. This is much easier to do on Linux and much harder to set up on Windows.
 
Vlad
Vladislav Muravin