Skip to content

Latest commit

 

History

History
132 lines (91 loc) · 5.3 KB

README.md

File metadata and controls

132 lines (91 loc) · 5.3 KB
Raw

% Hybrogen Hybrid Code Generation, fast and small as the hydrogen element % HP Charles

Introduction

Installation Dependency

  • Grammar use ANTLR4 with python4 backend

    • pip3 install antlr4-python3-runtime==4.7.2

  • Compilation need a postresql database. To install postgresql and configure it you can use this commands :

    • sudo apt install postgresql

  • Install a python postgresql connector:

    • sudo apt install python3-psycopg2

  • Initialize the database

    • sudo -i -u postgres
    • createdb hybrogen
    • createuser --pwprompt hybrogen # the default password in the code is "hybrogen"
    • Under psql prompt grant all privileges on database hybrogen to hybrogen;

    It is a good practice to not create the database under your own username. (The database could be multiuser accessible)

  • Qemu build need ninja

    • sudo apt install ninja-build

Hybrogen installation

Create cross-compilers, debugger and emulator for supported platforms

  • Clone HybroGen in a directory to extract the source files (with git or fetch / tar)

  • git clone [email protected]:CEA-LIST/HybroGen.git

  • or

    • wget https://github.com/CEA-LIST/HybroGen/archive/refs/tags/v4.0.tar.gz
    • tar xf v4.0.tar.gz

  • Choose a target directory to install the release e.g. /opt/H4.0/

  • For each platforms riscv, aarch64, powerpc, cxram-linux

    Run ./GenCrossTools.py -a <platform> -p /opt/H4.0/ -w /opt/H4.0/tmp

    This command will generate the cross-compiler environment (gcc, gdb, qemu, linux-headers). This command could take some time to run.

    • Run ./GenCrossTools.py -a riscv -p /opt/H4.0/ -w /opt/H4.0/tmp -s
      • This will generate the shell environment (csh like or bash like)
      • A full installation could be :
./GenCrossTools.py -a aarch64 -p /opt/H4.0/ -w /opt/H4.0/tmp
./GenCrossTools.py -a aarch64 -p /opt/H4.0/ -w /opt/H4.0/tmp -s

./GenCrossTools.py -a riscv   -p /opt/H4.0/ -w /opt/H4.0/tmp
./GenCrossTools.py -a riscv   -p /opt/H4.0/ -w /opt/H4.0/tmp -s

./GenCrossTools.py -a powerpc   -p /opt/H4.0/ -w /opt/H4.0/tmp
./GenCrossTools.py -a powerpc   -p /opt/H4.0/ -w /opt/H4.0/tmp -s

./GenCrossTools.py -a cxram-linux   -p /opt/H4.0/ -w /opt/H4.0/tmp
./GenCrossTools.py -a cxram-linux   -p /opt/H4.0/ -w /opt/H4.0/tmp -s

This step could take time. Count between 5mn and 20mn for each architecture depending on your computing power and bandwith.

Build HybroGen

HybroGen is mainly written in with python but need some build

  • Run make DbPopulate to populate the SQL database with instructions description
  • Congratulation, HybroGen is ready to work !

If you want to play with grammar / lexer / parser, you'll need some more steps:

  • Install
    • sudo apt install antlr4
    • make buildGrammar to build the ANTLR lexer / parser

For Computing in memory platform aka CXRAM

For this platform we need a qemu plugin which emulate the C-SRAM accelerator and give statistics about executed instructions.

Follow instructions on this repository : https://github.com/CEA-LIST/csram-qemu-plugin

Run some examples / démonstration

  • Some code examples are located in the this sub directory : CodeExamples

For example to run an demonstration for the power architecture here is the command. Adapt for other architectures / demonstrations.

  • cd CodeExamples/

  • source /opt/H4.0/powerpc/.cshrc

  • ./RunDemo.py -a power -i Array-Mult-Specialization

  • Regression can be run in the same directory :

    • ./Regression.py power

Execution dependencies

GenCrossTools used to generate compiler / debugger and qemu has it's own documentation :