Running SystemC simulations on cloud-based compute clusters

Regressions are an integral part of any VLSI design process. Every time a design change is implemented in the RTL code, a complex series of regression tests are run to validate the change. The regressions serve to ensure that there are no breakages to the existing functionality of the design. As the design heads towards code freeze deadlines, the demand on the compute-servers that execute these regressions rises exponentially. Sometimes tape-outs are delayed because the regressions could not be completed on-time.

To manage peak load requirements, we have implemented a flexible regression system that can utilize “cloud-based compute clusters” (such as EC2 instances of Amazon Web Services) to provide the additional capacity. Our regression system focusses on the following key features:

  • Flexibility: Regressions can be easily switched to run on
    • Local mode
    • Cloud mode and
    • Mixed mode, involving both cloud and local machines.
  • Security: Job information transmitted to the cloud and simulation results (transmitted back from the cloud) adhere to an acceptable level of security.
  • Local Compilation: Only the executables (.exe files) are shipped to the cloud machines to ensure that the source code is never sent outside the organization.
  • End to End Encryption: All the information transmitted between the local servers and the cloud machines are compressed and encrypted
  • Ease of use: The user could seamlessly pick the available cloud machines from the machine-list file.
  • Automatic polling for job completion: Determines the status of the jobs. The polling frequency is controllable by the user.

  • For more information about this project, or if you are interested any cloud based EDA services, please contact us