[FEA]: Build MatX as part of our third-party CI testing

[alliepiper]

Is this a duplicate?

• I confirmed there appear to be no duplicate issues for this request and that I agree to the Code of Conduct

Area

Infrastructure

Is your feature request related to a problem? Please describe.

@cliffburdick requested that we build MatX in our CI, similar to what we do for RAPIDS.

Describe the solution you'd like

Build-only checks are sufficient. This can be done via:

git clone [email protected]:NVIDIA/MatX.git
cd MatX
mkdir build
cd build
cmake -G Ninja .. -DMATX_BUILD_TESTS=ON -DMATX_BUILD_EXAMPLES=ON -DMATX_BUILD_BENCHMARKS=ON -DMATX_EN_CUTENSOR=ON
cmake --build

Describe alternatives you've considered

No response

Additional context

MatX depends on features that we do not provide in our default devcontainers. We'll need to make a dedicated devcontainer for MatX.

Our devcontainer infra repo's yaml file disables installCTKLibraries and installcuTensor for CCCL. We should be able to just define a new config with those enabled.

https://github.com/rapidsai/devcontainers/blob/1d4d433352d7ab4277a7d2962db72deca39fc94f/matrix.yml#L49-L57

Tasks

• Add deps to cuda_ext containers
• Patch cccl version to use PR SHA (https://github.com/NVIDIA/MatX/blob/main/cmake/versions.json)