RE: [EXTERNAL] [sst-simulator:710] Is it impractical to attempt to use Sandia SST on Windows?

[Hammond, Simon David (-EXP)]

We have not tried a run on Windows yet but using GNU via Cygwin may work. Whilst it won't be something officially supported we may be able to answer some questions you have as you go along if needed.

We are currently looking into a SystemC wrapper to SST core, this work is underway in our research group but is not completed yet.

I'd be interested to hear from you with your progress and any issues you encounter.



Si Hammond
Sandia National Laboratories
Remote Connection


-----Original Message-----
From: John Clinton [jcbi...@gmail.com]
Sent: Wednesday, December 18, 2013 01:07 PM Mountain Standard Time
To: SST-si...@googlegroups.com
Subject: [EXTERNAL] [sst-simulator:710] Is it impractical to attempt to use Sandia SST on Windows?

I know it's not an officially supported platform but since I'm stuck using Windows (and not using Cygwin) for my project I'm wondering how bad the effort by me to adapt to Windows would be.   I assume at least someone has considered doing this but haven't found any information.

I'm also curious if there is any information out there comparing SystemC to Sandia SST with both features and types of projects that differentiate them.   My application relates to modeling various hardware and firmware in a disk drive to help evaluate new ideas and how they relate to performance and maybe power at a high level (not remotely cycle time accuracy).

Thanks,
John

--
You received this message because you are subscribed to the Google Groups "Structural Simulation Toolkit" group.
To unsubscribe from this group and stop receiving emails from it, send an email to SST-simulato...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[John Clinton]

So you don't know up front of any highly Unix specific features that are required. Related to SystemC I'm more interested in information comparing them to help evaluate the relative value of using one over another for various project types.

I'm trying to improve how the company I work for does modeling (currently quite ad hoc) and trying to decide whether SystemC or Sandia SST is the best route. I do see some features that might make SST better but the user base seems quite small along with the available resources to learn more compared to SystemC and SystemC already runs on Windows. So I'm trying to determine whether the effort to port SST to Windows is really worth it or if I should go with SystemC.

[Arun Rodrigues]

Off hand, I would guess the biggest roadblock might be the way dynamic libraries are loaded. Most of the code is C and C++, and MPI calls. That said, there probably are a number of issues I haven't thought of.

For comparison with SystemC, we don't have a formal document. Major differences:

  – Parallelism: SST had parallelism 'baked in' and focuses on parallel execution much more than SystemC.

  – Overhead: We haven't measured in a while, but when we looked at it several years ago, SystemC's process model seemed to have higher overheads than the DES at the core of SST

  – Multi-scale: SystemC evolved out of an RTL background, whereas SST was influenced more by higher level simulators like Simplescalar, gem5, etc...  This cuts both ways – there is no path to compile SST to RTL, but SST components tend to be a little more flexible and easy to connect together.

Arun

[John Clinton]

So SST explicitly loads dynamic libraries from calls in the main code?

The key differences you mention are ones of interest to me. I heard about the potential lower overhead of SST though I didn't know how that would be when compared to the lighter weight Transaction Level Modeling (TLM) SystemC introduced.

I'm assuming the SST user base is pretty small if I can't find any evidence via internet searches of anyone that has tried to port to Windows. A large, active user community is always a plus.