Description

This artifact describes how to set up the compilation infrastructure of HERON and how to run the workloads described in Section 6.2. Concretely, this guide provides instructions to: • Set up the experimental environment of HERON. • Run experiments to demonstrate the optimization ability of HERON as shown in Figure 6, Figure 7, Figure 8, and Figure 10. • Visualization of the search spaces as shown in Figure 11. • Run experiments to demonstrate the effectiveness of CGA as shown in Figure 12 and Figure 13.

Instructions

General Installation

Software Dependencies:

TVM : ddfa1da691bacbb0018b53fca8409c5cfd6d LLVM 8.0.0

Hardware Dependencies:

NVIDIA GPUs with TensorCore extension

General Installation:

First, install TVM following the instructions in https://tvm.apache.org/docs/install/from_source. Then, install Heron using following commands: unzip Heron_main.zip mv Heron_main Heron cd Heron source env.sh pip install -r requirements.txt

Experimental Installation

Installation:

Run source env.sh before evaluation.

Work Flows:

1. To replicate the experiments in Figure 6, please run the following commands to reproduce the performance of a specific operator(e.g. GEMM for the following commands): cd figure6 python run.py -c ./configs/gemm.json

2. To compare the results of CGA, SA, GA, and RAND as shown in Figure 12. Please run the following commands: cd figure12/figure12b ./run.sh

Evaluation:

For 1, please compare the results of Heron and other baselines (e.g., PyTorch). For 2, please compare the generated graph with FIG.12.

Provenance

The code is produced by ourselves.

License