This repository contains a set of test, implemented in C, that are used as regression tests in our CI flow. Indeed, these tests are performed any time a push is done in the pulp-open's branches and are intended to verify that the basic functionalities of Pulp have not been compromised by the changes. Through time, this repo will be ppulated by more and more specific tests, granted that they work with the current pulp implementation.
You can run tests either locally in a manual fashion or use your GitLab CI instance.
Call the ./update-regression-tests
script in the pulp directory. This will populate this directory with all software tests.
You have two options to run these tests, the first one is to test without the peripheral tests and the second one is to use them. The main difference is just the time the simulation will need: withouth the peripheral tests, there is no need to instantiate the Verification IPs, which are heavy and with tiny time steps (1ps).
If you chose the first option, please do the following, from the pulp repo:
source setup/vsim.sh
./update-ips
./generate-scripts
cd sim
make all
cd ..
source pulp-runtime/configs/pulp.sh
export PULP_RISCV_GCC_TOOLCHAIN= *path to riscv gcc toolchain*
make test-fast-regressions
If you chose the second option, you will have to download the VIPs in the pulp/rtl/vip
folder, following the instruction in the relative READMEs. (You will not need the psram). And the you will have to do the following:
source setup/vsim.sh
./update-ips
./generate-scripts --rt-dpi --i2c-vip --flash-vip --i2s-vip --hyper-vip --use-vip --verbose
cd sim
make all
cd ..
source pulp-runtime/configs/pulp.sh
export PULP_RISCV_GCC_TOOLCHAIN= *path to riscv gcc toolchain*
make test-local-regressions
You can add your own tests by putting them in a repository and adding them to
update-tests
and update-tests-gitlab
scripts in the top-level folder. Make
sure you define an appropriate testset.cfg
that lists all the tests in your
repository and for each test provide another testset.cfg
with instructions on
how to run it. Have a look at the already existing tests for an idea how such
files should look like.