Getting started:

Clone the UGRAMM repository:
- git clone https://github.com/UniversalGRAMM/UGRAMM.git
Compile the source code:
- make
Above step will produce a UGRAMM binary file in the root directory.

❯ ./UGRAMM --help
[UGRAMM] allowed options =:
  -h [ --help ]         Print help messages
  --seed arg            Seed for the run
  --verbose_level arg   0: info, 1: debug, 2: trace
  --dfile arg           Device model file
  --afile arg           Application graph file
  --config arg          UGRAMM config file

Fig 1.UGRAMM inputs

UGRAMM primarily accepts three types of files:
- afile [REQUIRED]: An application-graph file in Graphviz dot format (ref). This file describes the input application to be mapped onto the device model (mapping here refers to checking if this smaller graph is a minor of the larger device-model graph).
- dfile [REQUIRED]: A device-model file in Graphviz dot format. This file represents the device model of the hardware, and the tool checks if the application graph (H) is a minor of this device model graph (G).
- config [OPTIONAL]: An optional configuration file that can be used in scenarios such as skipping the placement of specific nodes or locking certain nodes placement.

We have created a helper script (scripts/device_model_gen.py) to generate device model files for CGRA architectures such as RIKEN and ADRES.
The UGRAMM repository includes a variety of benchmarks for CGRA applications in the Kernels/ folder, featuring Stencil, Convolution, and FFT benchmarks.
While UGRAMM has been tested for the CGRA mapping process, it can also be utilized for other applications with minimal modifications. Users must ensure that the input files are in a format compatible with UGRAMM, as outlined in this documentation.
- We have designed UGRAMM with generalized features to accommodate a wide range of applications. However, if you need specific functionality or have any questions, please create an issue related to your requirements on the UGRAMM GitHub repository.

Demo:

A helper script (run_ugramm.sh) has been developed to simplify running UGRAMM. The operations of this script are described below. You can run UGRAMM using this script or directly.

./run_ugramm.sh $1 $2 $3 $4 $5

Argument information about this script:
- $1 [Seed] = 0,1,15 etc…
- $2 [NR] = 8 //Number of rows in the CGRA architecture
- $3 [NC] = 8 //Number of columns in the CGRA architecture
- $4 [ApplicationGraph] = Kernels/Conv_Balance/conv_nounroll_Balance.dot
- $5 [grammConfigFile] = config.json //Config file for the UGRAMM
This helper script will generate a specific CGRA device model file based on user input; Compile, and run UGRAMM with the generated device model file as input. It will also convert the output of UGRAMM into PNG format for easier debugging and visualization.
- To generate the device model using the external script:
```
  cd scripts && ./device_model_gen.py -NR $2 -NC $3 -Arch RIKEN && cd ..
```
- To execute UGRAMM and produce the mapping result in mapping_output.dot:
```
  make && ./UGRAMM --seed $1 --verbose_level 0 --dfile $device_model_output --afile $4 --config $5
```
- Finally, to convert the mapped output DOT files into PNG format:
```
  neato -Tpng positioned_dot_output.dot -o positioned_dot_output.png
  dot -Tpng unpositioned_dot_output.dot -o unpositioned_dot_output.png
```
- The successful mapping results will be saved in unpositioned_dot_output.png and positioned_dot_output.png.

Running UGRAMM:

The following command maps the conv_nounroll_Balance benchmark onto an 8x8 RIKEN CGRA using a seed value of 15.

./run_ugramm.sh 15 8 8 Kernels/Conv_Balance/conv_nounroll_Balance.dot config.json

After running above command the mapped output results should be present in the root directory of UGRAMM.
positioned_dot_output:
- This dot-file contains user provided co-ordinates of the device-model node cells.

Fig 1. positioned_dot_output

unpositioned_dot_output:
- This dot-file does not contains user provided co-ordinates of the device-model node cells.
- This file also displays the input application graph for which UGRAMM was executed.

Fig 2. unpositioned_dot_output

Table of Contents

Getting started:

Demo:

Running UGRAMM: