README.md 3.28 KB
Newer Older
Long Chung Chan's avatar
Long Chung Chan committed
1 2
# Partitioning FPGA-Optimized Systolic Arrays

LongChan's avatar
LongChan committed
3
We provide a fast optimization algorithm and a step-to-step guide on how to generate the dataset for a specific board and topologies to be used by our optimization tool.
Long Chung Chan's avatar
Long Chung Chan committed
4 5 6 7 8

>    *Long Chung Chan, Gurshaant Singh Malik and Nachiket Kapre*
>    [**"Partitioning FPGA-Optimized Systolic Arrays for Fun and Profit"**](),
>    2019 International Conference on Field-Programmable Technology 

LongChan's avatar
LongChan committed
9
### TODO
Long Chung Chan's avatar
Long Chung Chan committed
10 11 12 13 14 15 16 17 18
- [] Add a public link to the paper
- [] Basic Demo Flow
- [] Step-by-step guide on
  - [] their own data set
  - [] running the optimzation algo on the generated dataset
- [] What each files/directories are responsible for
- [] Changes done to ScaleSim
- [] Explanation of the testing result

LongChan's avatar
LongChan committed
19 20
### Demos
The following demos use pre-generated datasets and topologies that can be found in:
Long Chung Chan's avatar
Long Chung Chan committed
21

LongChan's avatar
LongChan committed
22 23 24 25 26 27 28 29 30 31 32 33 34
- [~/optimization_algo/topologies](~/optimization_algo/topologies/) contains all the topologies descriping their respective CNN structures
- [~/optimization_algo/data_source](~/optimization_algo/data_source/) contains all the cycle-accurate data generated using [SCALE sim]()

The instruction below will do a sweep run on each of the following networks:
   - FasterRCNN
   - Mobilenet
   - Yolo tiny
   - Googlenet
   - Alexnet
   - AlphaGoZero
   - NCF_rec
   - Resnet_50_v1

LongChan's avatar
LongChan committed
35
To obtain individual optimization result for a specific network and a specfic number of partition, please refer to the section below.
LongChan's avatar
LongChan committed
36 37

To get optimization result with 
LongChan's avatar
LongChan committed
38
1. Covariance Matrix Adaptation Evolution Strategy (CMA-es)
LongChan's avatar
LongChan committed
39 40 41 42 43
    ```bash
    # cd optimization_algo/scripts
    # ./sweep_nets_cma.sh
    ```

LongChan's avatar
LongChan committed
44
1. Genetic Algorithm (GA)
LongChan's avatar
LongChan committed
45 46
    ```bash
    # cd optimization_algo/scripts
LongChan's avatar
LongChan committed
47
    # ./sweep_nets_ga.sh
LongChan's avatar
LongChan committed
48 49
    ```

LongChan's avatar
LongChan committed
50 51 52 53 54 55 56 57 58
1. CMA
    ```bash
    # cd optimization_algo/scripts
    # ./sweep_nets_cma.sh
    ```

2. Brute Force
    ```bash
    # cd optimization_algo/scripts
LongChan's avatar
LongChan committed
59
    # ./sweep_nets_brute.sh
LongChan's avatar
LongChan committed
60
    ```
Long Chung Chan's avatar
Long Chung Chan committed
61

LongChan's avatar
LongChan committed
62 63
#### Custom topologies
#### Custom target board
Long Chung Chan's avatar
Long Chung Chan committed
64

LongChan's avatar
LongChan committed
65
### Repo breakdown
Long Chung Chan's avatar
Long Chung Chan committed
66

LongChan's avatar
LongChan committed
67
### Changes made to Scale-sim
Long Chung Chan's avatar
Long Chung Chan committed
68

LongChan's avatar
LongChan committed
69
### Interpretation to the test result
Long Chung Chan's avatar
Long Chung Chan committed
70

LongChan's avatar
LongChan committed
71
## License
Long Chung Chan's avatar
Long Chung Chan committed
72 73
This tool is distributed under MIT license.

LongChan's avatar
LongChan committed
74
Copyright (c) 2019 Long Chung Chan, Gurshaant Singh Malik, Nachiket Kapre
Long Chung Chan's avatar
Long Chung Chan committed
75 76 77 78 79 80 81 82 83 84 85 86 87 88

<div style="text-align: justify;"> 
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
<br><br>
</div>

<div style="text-align: justify;"> 
<b>The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.</b>
<br><br>
</div>

<div style="text-align: justify;"> 
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 </div>