SIMPLE-MAGIC

SIMPLE MAGIC: Synthesis and In-memory MaPping of Logic Execution for Memristor Aided loGIC

Dependencies

In order to use SIMPLE-MAGIC, you will need a Linux machine with:

Python 2.7
Z3: Run the following commands to install it:

git clone --recursive https://github.com/Z3Prover/z3.git
cd z3
python scripts/mk_make.py --python
cd build
make
make install

ABC Synthesis Tool
The following Python packages (for graphic description of the solution):

apt-get install python-tk

matplotlib

apt-get install python-matplotlib

OpenCV

Manual

Configure: in the file simple_conf.cfg you will find the following content:

[input_output]
input_path=cm138a.blif
; input_format can get one of the values: verilog, blif
input_format=blif
output_path=cm138a_output

[abc]
abc_dir_path=/home/adi/abc/alanmi-abc-eac02745facf

[Z3]
Z3_path=/usr/bin/z3

Change the parameters according to your needs.

Run:

python simple_main.py

Parse: When Z3 finishes running and output_path is created, run:

python convert_gates_2_array.py -f output_path

The memory array will be printed, for example:

In addition, two images will be created:

Graphic description of the logic execution in the memory. The table drawn in this image represents the memory array. Each circle is a gate port (input or output). Inputs are denoted with the letters A-C and outputs are denoted with E. the number following the letter identifies the gate. Each clock cycle is marked with a different color.
Clock cycle legend.

Executing using docker

Project consist of two docker Images

dockerfiles/DockerfileCV (name = simple-magic-cv) - with OpenCV3 (Required for OpenCV based visualisation Only)
dockerfiles/Dockerfile (name = simple-magic) - Lightweight without OpenCV

simple-magic-cv and simple-magic are derived from OpenCV3+Pytho2.7 and Ubuntu18:04 Images respectively. It has verified installation of ABC Synthesis Tool + Z3 Solver. Pull docker images using following commands.

docker pull goreganesh007/simple-magic-cv
docker pull goreganesh007/simple-magic

Note : To modify docker image, refer Readme file in dockerfiles folder

Once you download images execute following commands to create command shortcuts. It will create two shotcut commnad

smagiccv - Execute commands in goreganesh007/simple-magic-cv
smagic - Execute commands in goreganesh007/simple-magic

#To create alias of docker command (smagic) Read comments in the file to learn more
dockerfiles\windowsSource.bat # for windows 
source linuxSource.sh # for linux

You can use shotcuts as - smagic <command_to_run_> or smagiccv <command_to_run_>

For example:

Note : No need to configure simple_conf.cfg default cconfiguration will work fine

smagic python simple_main.py  # Will create final netlist file 

smagiccv python convert_gates_2_array.py -f <SynthesizedNetlistName> # Will create OpenCV3 visualisation image

Visualising crossbar computation using SVGs - MagicPrinter.py

Basic idea is create SVG file of crossbar configuration for each clock, and then use browser ( index.html ) to browse through each clock sequencially.

Input to the script is *.smagic file which lists all the dummy assembly instructions to run in cross bar. Specification of file is as follows

smagic file Specification

Directives

.GRID <Rows> <Cols>

.GRID directive initiliases cross bar with defined rows and columns

Instructions

Each cell is defined as R<-RowNo->C<-CellNo->

#Instruction Template
#<INPUT|OUTPUT> <CellName>
#<NOT|NOR> <OUTPUTCell> <ListofInputCells>

# Define Input in R1C1
INPUT R1C1

# Define OUTPUT in R1C4
OUTPUT R1C4

# R1C3 = Not R1C2
NOT R1C3 R1C2

# R1C3 = R1C2 NOR R1C1
NOR R1C3 R1C2 R1C1

#You can also write multiple instructions per row 
#It represents Multiple computations in same clock cycle 
NOR R1C3 R1C2 R1C1 | NOR R2C3 R2C2 R2C1 | ...

# Optionally you can also mention tag for only output cell
NOR R1C3(I1) R1C2 R1C1

To render example code, run following command

smagic python MagicPrinter.py --input MagicPrinterExample.smagic

Open index.html file in borwser and use right and left arrow to navigate

Creating smagic (dummy assembly) file from SMT file - CompileToSmagic.py

This script accepts Z3 constraint file and outputs smagic file indicating computation in each clock

smagic python CompileToSmagic.py --input < Input_SMT_Filename > --output < Output_smagic_filename >

Example

# Default input file : outputNorInv.v_python_Z3input.txt
# Default output file : main.smagic
smagic python CompileToSmagic.py

Quick Run

docker pull goreganesh007/simple-magic
dockerfiles\windowsSource.bat
# Modify full_adder_1bit.v verilog file default module FullBitAddere
smagic python simple_main.py
smagic python CompileToSmagic.py
smagic python MagicPrinter.py
# Browse Index.html

ganeshgore / SIMPLE-MAGIC