This is a low-level debugging tool for Tenstorrent hardware. It enables access to and communication with the device. It supports Wormhole and Blackhole devices.
The wheel can be installed directly from GitHub with:
pip install git+https://github.com/tenstorrent/tt-exalens.git
The CLI application can be run by invoking tt-exalens command after installing the wheel.
After cloning the repository, run:
git submodule update --init --recursiveto ensure all submodules are properly initialized.
Project has been tested on Ubuntu 22.04.
To build it, you need the following dependencies:
- software-properties-common,
- build-essential,
- python3.X-venv,
- libyaml-cpp-dev,
- libhwloc-dev,
- ninja-build
Install them with:
sudo apt install software-properties-common build-essential libyaml-cpp-dev libhwloc-dev libgtest-dev libgmock-dev xxd ninja-buildPython 3.10 is the only supported version. Install it with:
sudo apt install python3.10-venvAdditional Python dependencies are listed in ttexalens/requirements.txt and can be installed with:
pip install -r ttexalens/requirements.txtOnce the dependencies are installed, building the project is straightforward and can be done by running:
make build
from the project's root directory.
To verify that the build was successful, try running:
python tt-exalens.pyor
./tt-exalens.pyfrom the root directory.
The wheel can be installed from the GitHub release, built from source, or installed directly from GitHub with
pip install git+https://github.com/tenstorrent/tt-exalens.git
Another option is to build the wheel from local source, run:
make wheel
in the root directory. Then install it using:
pip install build/ttexalens_wheel/<ttexalens_wheel>.whl
where <ttexalens_wheel> is an automatically generated filename for the build.
The CLI application can be run via the tt-exalens.py script or by invoking tt-exalens command after installing the wheel.
It currently operates in Limited mode by default, with plans to support additional modes in the future.
Limited mode allows basic communication with the device, such as reading/writing registers and memory, and running .elf files on RISC cores.
The CLI application can target local or remote devices. For remote runs, a server instance is required and can be started using the same application.
It’s also possible to cache session results and replay commands on a machine without Tenstorrent hardware.
A GDB server can be launched, enabling stepping, breakpoints, and other debugging features via a GDB client.
For more usage information, refer to the tutorial or the documentation.
The application functionalities can also be accessed through the ttexalens Python library to create scripts that interact with Tenstorrent hardware.
For a quick start with the library, check out the tutorial. Full documentation is also available here.
Apart from the base dependencies, running tests requires additional Python packages:
pytestcoverageparameterized
and system libraries:
libgtest-devlibgmock-dev
Install the Python packages with:
pip install -r test/test_requirements.txtIt is currently possible to run tests locally, by running
make test
from the project root directory.
In order to update documentation virtual environment has to be created first by running the following command:
./scripts/create-venv.shTo activate created environment use
.venv/bin/activateLibrary documentation is automatically generated from source code docstrings. To update the library docs, you need to run (while in created environment):
make docs
in the project root.
For more advanced use cases, refer to the source code of the documentation generation scripts, located in docs/bin.
We have defined various pre-commit hooks that check the code for formatting, licensing issues, etc.
To install pre-commit , run the following command:
pip install pre-commitAfter installing pre-commit, you can install the hooks by running:
pre-commit installNow, each time you run git commit the pre-commit hooks (checks) will be executed.
If you have already committed before installing the pre-commit hooks, you can run on all files to "catch up":
pre-commit run --all-filesFor more information visit pre-commit
To update docker images used on CI and in development, you need to update Dockerfiles in .github directory and then run release-docker-images workflow.
This can be done through GitHub CLI (see here how to install it) by using command
gh workflow run 116548537 --ref <your-branch-name>This will automatically generate new image releases and tag them as latest.
test/wheel-test/wheel-test.sh: line 11: python: command not found
make: *** [Makefile:182: test] Error 127
Fix:
alias python=python3
This error might occur when trying to run unit tests:
Failed to import test module: test.ttexalens.pybind.test_bindings
Traceback (most recent call last):
File "/usr/lib/python3.10/unittest/loader.py", line 436, in _find_test_path
module = self._get_module_from_name(name)
File "/usr/lib/python3.10/unittest/loader.py", line 377, in _get_module_from_name
__import__(name)
File "./test/ttexalens/pybind/test_bindings.py", line 17, in <module>
from ttexalens_pybind_unit_tests import set_ttexalens_test_implementation
ModuleNotFoundError: No module named 'ttexalens_pybind_unit_tests'
To fix, just build the neccessary test library:
make build
