[CI] Add job for CUDA testing on self-hosted runners (#14633)

I fixed the runners (updated drivers, fixed docker install) so we should
be able to use them now.

The total job runtime seems to be much better than the AWS job too:

AWS: [around](https://github.com/intel/llvm/actions/runs/9995414054) 46
min
Runner:
[around](https://github.com/intel/llvm/actions/runs/9995200242/job/27630350341?pr=14633)
30 min

Nightly manual run (runner 2):
https://github.com/intel/llvm/actions/runs/9996341629/job/27634460554

NVIDIA run from this PR (runner 2):
https://github.com/intel/llvm/actions/runs/9995200242/job/27630350341?pr=14633

Runs on runner 1 in another PR:

https://github.com/intel/llvm/actions/runs/9980056160/job/27623745128?pr=14114

https://github.com/intel/llvm/actions/runs/9980056160/job/27626193706?pr=14114

Runs on runner 2 in another PR:

https://github.com/intel/llvm/actions/runs/9980056160/job/27585488761?pr=14114

https://github.com/intel/llvm/actions/runs/9980056160/job/27620009329?pr=14114

https://github.com/intel/llvm/actions/runs/9980056160/job/27621993091?pr=14114

---------

Signed-off-by: Sarnie, Nick <[email protected]>

Author: sarnex

.github/workflows/sycl-linux-precommit.yml

@@ -76,6 +76,11 @@ jobs:
       fail-fast: false
       matrix:
         include:
+          - name: NVIDIA/CUDA
+            runner: '["Linux", "cuda"]'
+            image: ghcr.io/intel/llvm/ubuntu2204_build:latest
+            image_options: -u 1001 --gpus all --cap-add SYS_ADMIN
+            target_devices: ext_oneapi_cuda:gpu
           - name: AMD/HIP
             runner: '["Linux", "amdgpu"]'
             image: ghcr.io/intel/llvm/ubuntu2204_build:latest-0300ac924620a51f76c4929794637b82790f12ab

.github/workflows/sycl-nightly.yml

@@ -37,6 +37,13 @@ jobs:
             target_devices: ext_oneapi_hip:gpu
             tests_selector: e2e
 
+          - name: NVIDIA/CUDA
+            runner: '["Linux", "cuda"]'
+            image: ghcr.io/intel/llvm/ubuntu2204_build:latest
+            image_options: -u 1001 --gpus all --cap-add SYS_ADMIN
+            target_devices: ext_oneapi_cuda:gpu
+            tests_selector: e2e
+
           - name: Intel L0 GPU
             runner: '["Linux", "gen12"]'
             image: ghcr.io/intel/llvm/ubuntu2204_intel_drivers:latest