simmani/esp-tools-prebuilt

Simmani Runtime Power Modeling

This repo contains Simmani's core implementation as well as its simple examples.

Preliminaries

• Power modeling is implemented with Python, which resides in simmani.
• All Scala code is for example designs generated by Chisel:
  • GCD: the Euclide's algorithm for the greatest common divisor (GCD) of two numbers. A.k.a. hello world of hardware.
  • Stack: hardware implementation of stack.
  • Risc: simple RISC machine with flip-flop-based memories.
  • RiscSRAM: simple RISC machine with SRAMs.
  • mini: RISC-V mini, a 3-stage RISC-V pipeline.
• Each command is executed as follows:

  scons --<design> <command>
  

  <design> is one of examples available above, which is mini by default. For example, if you want to generate verilog for GCD, run: scons --GCD rtl-v.
• All CAD tools were run with HAMMER. Due to the NDA, we cannot release code in tools/hammer-cad-plugins and tools/hammer-adept-plugins.
• Instead, we provide pre-generated power traces in power. RISC-V mini's power traces are available as a tarball due to their sizes.

Prerequisites

• SBT
• Python3
• Python packages: scons, numpy, scipy, matplotlib, scikit-learn

  pip3 install [--user] <package>
  

• Synopsys VCS

Step 1: Getting Started

git clone https://github.com/Simmani/simmani.git
cd simmani
./setup.sh # initialize submodules

Step 2: Generating Verilog

scons --<design> rtl-v

Step 3: Running RTL Simulation for VCDs

scons --<design> run-testers

Step 4: Training Power Models with Pre-generated Power Traces

scons --<design> simmani-train

Step 5: Test Power Models with Pre-generated Power Traces

scons --<design> simmani-test

Power training and test results are available in power/<design>/simmani.

Publications

• Donggyu Kim, Jerry Zhao, Jonathan Bachrach, and Krste Asanović, "Simmani: Runtime Power Modeling for Arbitrary RTL with Automatic Signal Selection", In proceedings of the 52nd IEEE/ACM International Symposium on Microarchitecture (MICRO'19), Columbus, OH, October 2019.
• Donggyu Kim, "FPGA-Accelerated Evaluation and Verification of RTL Designs", Ph.D. Thesis, University of California, Berkeley, May 2019.