Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

Synopsys on Multiprocessor(SMP) machines

Status
Not open for further replies.

fiber7

Member level 1
Member level 1
Joined
Mar 14, 2002
Messages
32
Helped
0
Reputation
0
Reaction score
0
Trophy points
1,286
Activity points
286
Hello everybody.

I was wondering if there is a way to take full advantage of SMP with Synopsys stuff on Linux.

Is there maybe an option to set the number of processors to be used?

Would be also nice to find out if is possible to use also it in a cluster.

Bye!
 

From my experience from Synopsys I guess that the application software (i.e. Design Compiler) has nothing to do with SMP. The OS (Linux) has to deal with it. So, IMHO, if you find an SMP aware Linux distribution you may be able to see speed improvements.
 

Hi geconom.

It is required that the OS support the SMP architecture but it is not enough.

The application must be written so that work can be parallelized i.e. more than one process(fork)is exectued at least.

I personally tried DC on an SMP and it doesn't use other processors.

Maybe DC doesn't support this kind of architecture but I think it would be a great improovement.
Think about using OpenMOSIX and DC.
You could high reduce the synth & sim time.


Anyway if someone knows something more on this argument please join us!

Cheers
 

Yes, this is what I was expecting. I have been using Synopsys, I have
heard nothing about SMP in its current or even in its future releases.
I have also heard nothing about other EDA software and SMP, except
rumors about Aldec's Riviera and Cadence's IC. But I have never tested
any of them. You may want to check their announcements.
For Synopsys, you have to stay with any improvement the SMP aware
OS has to offer. I guess!
 

Some products from Avant! (now belong to synopsys) at least Hspice that I have ever used can support SMP. I think many of them may also support SMP too.
 

When both of OS and Applications supports SMP, you can run the app on a SMP machine.
 

Here is my results about synopsys DC run on SMP machines.
I have a VIO PIII 866MHz 1GB SDRAM and a dual PIII 666MHZ 512MB SDRAM I tried to run DC 2001.8 for both WIN2000 Advance Server, WinXT and Linux (7.2) platform (all support SMP). The design is fairly small (ARM core). The result is:

+ DC on Windows is slow any more likely to crash.
+ DC on Linux runs stable and lots faster.
+ DC on SMP Linux machines run faster even run several jobs than on the single MP machine.
+ I also can not get two DC run the same time on Windows. They all crash appearantly.

Conclusion:
+ There may have some bug on 2001.8
+ Does not matter the DC actually supports multiple threads and SMP platform, SMP OS definitely does a good job (i.e Linux).
+ Linux will be the OS of choice for EDA!
:oops:
 

Some ralative information:

Synopsys, as from version 1999.10 and on, includes the Automated Chip Synthesis (ACS) with Design Compiler. This tool can be used to partition your design and perform different synthesis tasks in different Solaris or Linux machines in the same network. For example, a cluster of Linux machines. I do not know if ACS can be applied to SMP machnines also but I will check!
 

Some products of Synopsys from ex-Avant! support SMP such as Hspice, Star_RCXt, Hercules.... But You won't get a free luch, it require an extra license for every pairs of CPU use, 1-2 CPUs 1 license, 3-4CPUs 2 licenses, 5-6 CPUs 3 license . For 6 CPUs the performance only increase 3 times. Personnaly, I think that, in simulation case, if your circuit have a lot of feedback, the performance haven't improve much. and Yes, you have to put in a switch at the line comand to tell that you are running multi-threading
 

Physical Compiler from Synopsys can use SMP. when you do phyopts, add a
option -num_cpus 2, then you can see
two cpus are used to do the optimization.

For Cadence, the SOCEncounter-Nanoroute can use multiply cpus do a fast route.

For synthesis, maybe the ACS is the only choice that you can use multiply machines.

BTW, for SMP , you need addtional license to run the Physical Compiler or Nanoroute....
 

For geconom:

Thanks A LOT! I didn't hear nothing about ACS until now.
I'm very interested to use DC in SMP or cluster Linux machines so if you have
new infos I'll be happy to listen to them!

For Maria R:

Do you know which is the switch to tell DC how many CPUs are present on the machine?

Thanks guys!!
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top