UPGRADE YOUR BROWSER

We have detected your current browser version is not the latest one. Xilinx.com uses the latest web technologies to bring you the best online experience possible. Please upgrade to a Xilinx.com supported browser:Chrome, Firefox, Internet Explorer 11, Safari. Thank you!

cancel
Showing results for 
Search instead for 
Did you mean: 
19,649 Views
Registered: ‎01-03-2008

ISE 9.2.04 Multi-core processor support

Does anyone know that if the newest version of ISE can take advantage of dual or quad core processor?
I am using and PentiumD (dual core) computer, the CPU usage is around 55%, could there be any way to make it higher?
Thanks
0 Kudos
6 Replies
19,609 Views
Registered: ‎03-05-2008

Re: ISE 9.2.04 Multi-core processor support

Hi,
 
Please have a look at:
AR #24394 - 9.1i ISE - How to control which processor(s) on which Project Navigator will run?
 
Regards,
Stephan van Beek
0 Kudos
19,598 Views
Registered: ‎01-03-2008

Re: ISE 9.2.04 Multi-core processor support

Hi,
   Thanks for the help!
   So if I understand correctly, processes like xst, map and par still do not have multiprocessor support?
0 Kudos
Xilinx Employee
Xilinx Employee
19,592 Views
Registered: ‎08-13-2007

Re: ISE 9.2.04 Multi-core processor support

Correct.
 
Our synthesis and implementation tools are not currently multi-threaded. These are generally algorithms that don't lend themselves to clean partitioning between multiple processes - particularly if you want the results to be repeatable. We are aware of the reasons for user interest here.
 
That isn't to say that multiple processor machines aren't still useful
-xst/map/par are very good at taxing a single CPU. If this is your primary machine and you are targeting a larger device, it can be difficult to use this machine for anything else when you only have a single processor. Having a dual core CPU lets you concurrently use the machine for something else (simulation, email, web exploration, etc.)
-It is possible to run multiple implementation runs (e.g. par) on different processors to explore different approaches concurrently. The operating systems generally seem to load balance pretty well by automatically distributing the different par runs to unique processors. It can be useful to set the processor affinity as mentioned above because I've occasionally seen the OS swap processes between processors frequently. This approach is generally better suited for scripted runs though versus using the Project Navigator IDE. It can be a useful technique when you want to try different options (e.g. the design is difficult to meet timing) but isn't so useful if you are just trying to speed up a particular run. I've even seen users script this with some added intelligence to farm multiple jobs across the network to the available machines.
-I have also seen users pipeline their development/testing (particularly useful with the longer run times on larger Virtex-4 & 5 parts). They'll make change 1 to the design and while it implementing on the first processor, work on a separate change which they later implement. When the first run is finished, they'll test this and then move the to the 2nd when it has routed. Obviously you have to manage this process carefully (including separate directories for different revisions) and keep track of your changes.
 
I hope this helps.
 
Cheers,
bt


Message Edited by timpe on 03-05-2008 02:36 PM
0 Kudos
19,555 Views
Registered: ‎01-03-2008

Re: ISE 9.2.04 Multi-core processor support

thanks a lot!
that makes a lot of sense to me.
0 Kudos
Newbie aermeydan
Newbie
17,875 Views
Registered: ‎06-20-2008

Re: ISE 9.2.04 Multi-core processor support

Hi Timpe,

 

We want to buy a workstation for ISE simulations. Can multiple core (dual or quadro) or multiple processors provide faster solution? We are considering a workstation with 2 quad-core Xeon processors and a 64 GB 800 MHz RAM. This makes a total of 8 processor cores with roughly 8 GB of RAM for each processor core in average. How would the performance of this system compare with a single core processor system with 8 GB of RAM? The performance metric that I am mostly interested in is the time that  it takes to complete the full implementation cycle for a single design. How much of an improvement can I expect to see from the workstation? I am using ISE 9.2. Does this program support multi-threaded execution that favors a parallel processing environment?

I am using the Modelsim simulator for the post place &route simulations, can I also expect to see similar improvements in the simulation time?

 

Aermeydan       

0 Kudos
Highlighted
14,532 Views
Registered: ‎03-10-2009

Re: ISE 9.2.04 Multi-core processor support

Hi Timpe,

 

I have a very important question regarding multiprocessor machines.

I'm trying to run multiple P&R processes concurrently on an IBM server with 4 dual core processors, 32GB of RAM, running Red Hat 4.

 

For some reason, unclear to me, only one P&R can run at a normal speed.

When I run more than one process of P&R the server dies. It literally chokes the server so that nothing runs.

It seems as if the scheduler goes crazy giving processes like top 100% CPU.

The same thing happens when MAP runs or XST, whenever I run multiple processes.

I tried using affinity on the processes and the specific process is set to the correct processor but the server still dies when more than one is running.

 

The server is running Red Hat 4 with a Linux 2.6.9 kernel, and the ISE tools are 9.2.04

 

Did you encounter any problem like this? Do you have any thought about what the problem can be?

 

I'd appreciate very much anyone's help. I'm extremely bewildered with this.... :-S

0 Kudos