HW Lockin Emulation Cost

This repo include scripts for measuring emulation cost in the HW lockin paper.

Env Setup

  • Python 3.11
  • Install dependecies using ./requirements.txt

Scripts

The following two scripts measures the latency and throughput of quantization-aware/pruning-aware GEMMs and LLM inference.

  1. Quantization emulation cost:
python profile-q-int.py
  1. Pruning emulation cost
python profile-p.py