help Needed for Parallel processing of Modelsim Simulation

I am working on FPGA project, which simulation takes require time of one day, the constraint of time on simulation is making my progress slow on this project. In this regard can any one tell me how can i process my design simulation on parallel PCs, can i used multiple PCs to cut short the processing of simulation, if yes kindly give me some guideline how can i do it or from where i can get information regarding Parallel processing of Modelsim.

if it is functional simulation then you can increase the frequency. instead of using signals use variable, this will help you to mininmize the memory. If you are using any RAM or FIFO choose minimum memory. Increase your system RAM. Default time unit of a simulator is picosecond make it as nanosecond.This may increase your simulation speed.

I am not sure what you are simulating, but I simulated a whole OMAP2430 for 1 second of video data and it took me less than that.
First of all, if you have such a CPU intensive simulation, you should think about purchasing multi core PC, increase the memory, de-frag your hard drive, install and use the Linux64 bit version of the MS etc.

As far as I know MS does not support tread simulation over a network.

Re: help Needed for Parallel processing of Modelsim Simulati

irshan said:

I am working on FPGA project, which simulation takes require time of one day, the constraint of time on simulation is making my progress slow on this project. In this regard can any one tell me how can i process my design simulation on parallel PCs, can i used multiple PCs to cut short the processing of simulation, if yes kindly give me some guideline how can i do it or from where i can get information regarding Parallel processing of Modelsim.

Click to expand...

That' strange.
Maybe your PC is invected by a Virus or your modelsim version is buggy

Re: help Needed for Parallel processing of Modelsim Simulati

Do this Linux64 version of the MS support "cluster computing" and "grid-computing".....?

i am wanted to make cluster or grid of my lab computers in order to reduce simulation time, the information is required if any version of modelsim support cluster or grid computing.

Added after 1 hours 30 minutes:

I am simulating the DDR2 Memory Controller design generated by the MIG tool which is a part of the Xilinx Core Generator tool. The design is generated for a Virtex-4 FPGA. MIG also creates a simulation file sim.do which can directly be run in ModelSim. The version of ModelSim I'm using is the ModelSim XE III (Xilinx Edition) Starter vesion. The simulation runs such that 1,000,000 picosec are run in approx. 2 min. The waveform does display, but the init_complete signal (for memory initialization) is asserted after more than 20 hours of simulation, and then the write and read operations begin. I want to know whether this simulation time can be reduced. My PC is a Core2 Duo system. I have also tried this on a laptop with Centrino processor.

Re: help Needed for Parallel processing of Modelsim Simulati

Ok, that is one whole different story, you have MAJOR issue with your setup, I did DDR2 simulation many times and it took me about an hour or so with my P4-2.8GHz 'old' computer.

I will dig out some data about what you need to do to get rid of the initial time problem in simulation of DDR2.

Could you please send the following information:
-Your ISE version
-Your Modelsim version
- Your MIG version

Also, did you perform a re-compile of all Xilinx libraries recently? Did you install the latest service pack for your ISE? These are all items that speed up the simulation tremendously since MIG is changing all the time.

Cheers,
/Farhad

The problem is most likely with initial calibration of the DDR2 core. The core generator should have an option to reduce the calibation effort for simulation purposes, the A.ltera IP core at least has.

If ModelSim had a distributed computing option (I'm not aware of), it surely isn't available with a free Xilinx version.

Re: help Needed for Parallel processing of Modelsim Simulati

I will appreciate any data or suggestion that will resolve the initial time problem in simulation of DDR2 and It will be a great relief, if i can do this simulation in an hour or so.
I am using following softwares
Xilinx ISE 9.2i version with service pack 4,
Memory Interface Generator (MIG) version 2.0,
and ModelSim XE III (Xilinx Edition) Starter vesion

I did perform the compilation of xilinx file recently.

Re: help Needed for Parallel processing of Modelsim Simulati

Hi Irshan,
I am afraid I had to remove all my old files at my work PC when I left the company, I had this message from Xilinx long time ago:

"there is a size limit beyond which the license causes the system to slow down. If this is the case you should have seen a message when the simulation started something like "expect performance to be severely affected"."

Click to expand...

Check the log for the simulation, if you have this error, then it means you need to have a Modelsim SE or even PE instead of XE.

But in general Modelsim XE and SDRAM simulation is a BIG no-no, I don't know anyone who has ever made it to work.

your problem is not your system, but the SW you use, see if you can get the license for Modelsim SE and do the same simulation, I bet it will take you about an hour or so (max).

Here is another comment from Xilinx help board:

a) change away from ModelSim XE. I'm using Veritak which is very reasonably priced. It has no size restrictions and runs reasonably fast. The author is very responsive to bug reports and suggestions for improvements.

b) If you're using a V5, I created a simple functional model of the user interface that is much smaller/simpler for testing user logic. It does *not* exersize the DDR2 pins, it is geared to providing a test module for the user logic. You can get the current version a:

**broken link removed**p

Note that this is a relatively new design, so it does not yet have a lot of testing. So please be semi-cautious when using it & let me know if you find any bugs!

Click to expand...