Space Pulp Tutorial

Install pulp-riscv-gnu-toolchain

• Clone the repositories and submodules:

git clone --recursive https://github.com/pulp-platform/pulp-riscv-gnu-toolchain.git

• Install dependencies (Ubuntu 20.04):

sudo apt-get install autoconf automake autotools-dev curl libmpc-dev libmpfr-dev libgmp-dev gawk
build-essential bison flex texinfo gperf libtool patchutils bc zlib1g-dev

• Add install path to PATH:

export PATH=$PATH:/opt/riscv/bin
./configure --prefix=/opt/riscv --with-arch=rv32imc --with-cmodel=medlow --enable-multilib

• Compile and install to --prefix:

sudo make

• Install pyelftools:

sudo apt install python3-pip
sudo pip3 install pyelftools

Michael’s Space Pulp instructions

• Clone my fork of the pulp repository and check out the space_pulp_tmp branch.

• There’s also the cevero fork, just clone the repository and check out the cevero-space-pulp-tmp branch. This was created by:

  • Adding Michael’s fork as a remote:

    git remote add micprog https://github.com/micprog/pulp.git
    

  • Creating a local branch named cevero-space-pulp-tmp and switching to it.
  • Pulling micprog/space_pulp_tmp into cevero-space-pulp-tmp with git pull micprog space_pulp_tmp
• Here, run the following make targets to get the corresponding software with changes:

  make pulp-runtime test-checkout-gitlab
  

• As I did all my development using bender for dependency management, before collecting dependencies first run export BENDER=1
• Make sure your git is configured with github ssh keys;
• To collect all dependencies and generate the corresponding scripts, run make checkout
• To build, first set the correct paths by running source setup/vsim.sh and then run make build
• To run a simulation, first source the correct runtime config by running

  source pulp-runtime/configs/tcls_pulp_ibex.sh
  

• I currently set up one SW test to enable TCLS, to test this run:

  make -C regression_tests/parallel_bare_tests/parMatrixMul/ clean all run
  

• If you would like to inject faults into the simulation, the rather crude script can be enabled by running export INJECT_FAULT=1 and disabled by running unset INJECT_FAULT

back