Aim of this project is the development of a tightly-coupled cache coherence unit for a multicore processor based on CVA6. Like the ancient god 1, its responsibilities are to maintain order (and data consistency) among the memory accesses performed by the 2~4 CPUs which are part of the system.
git clone https://github.com/planvtech/culsans.git --recursive
make fpga
The top level file for FPGA synthesis is rtl/src/culsans_xilinx.sv
Make sure all dependencies specified in cva6-sdk
are fulfilled.
make sdk
Then follow the instruction in cva6-sdk
to copy the generated image to the SD.
Sanity check
make sanity-tests
Regression tests
make test
The CITS is a test platform that tests the integration of components of Culsans.
There are two layers to the tests.
There is c code self testing. This is achieved by c code within the test testing what it expects it has effected in the memory. If it fails, it exits early with a code indicative of the core and cacheline that incurred the problem.
The second (optional) layer is that where an external parser inspects the logs generated by the tests and compares them to an expected sequence. This layer is not further described here.
Tests are run in the tests/integration
directory. All commands below are executed in this directory.
To run a test in the CITS, type
make -C testlist/<test_name> all
To run a test in the CITS in a GUI, type
make GUI=1 -C testlist/<test_name> all
A CITS test consists of a batch of files within the folder testlist/<test_name>
. The folder should contain the following files:
File | Description |
---|---|
<test_name>.c/h | This file is the the actual test. See similar files in the CITS for an example of how to write one. |
main.c | This file co-ordinates the core execution and calls the test_name() function. See the main.c files in the CITS for an example on how to write one. |
Makefile | Symlink to ../../test_automation/Makefile |
sim.tcl | Symlink to ../../test_automation/sim.tcl |
The test function defined in <test_name>.c should be self-checking and call exit(arg)
with arg > 0 if the test fails, otherwise return 0
.