[UR][L0v2] Introduce urGraphGetIdExp to provide unique graph identifiers

[mmichel11]

• Implements urGraphGetIdExp which internally invokes zeGraphGetIdExt to align SYCL native graph IDs with L0
• The identifier is process-unique and persists beyond graph destruction

[mmichel11]

I first observed these non-graph test failures in: https://github.com/intel/llvm/actions/runs/28179054588 where the driver does not yet support zeGraphGetIdExp. I reproduced this issue locally and confirmed the affected tests pass once I use a driver that supports zeGraphGetIdExt and skip the zeDriverGetExtensionFunctionAddress call for older drivers.

There's an issue to be filed to L0 which seems to be the source of this memory corruption, but this check provides us a workaround for now.

[mmichel11]

The added UR test is skipped in CI as the driver does not yet contain the new L0 graph function. Here are the results from a local BMG run with a new driver build:

urGraphGetIdExp L0v2 BMG

$ ./test/conformance/exp_graph/exp_graph-test --gtest_filter="*urGraphGetIdExp*Level_Zero_V2*"
Note: Google Test filter = *urGraphGetIdExp*Level_Zero_V2*
[==========] Running 8 tests from 1 test suite.
[----------] Global test environment set-up.
[----------] 8 tests from urGraphGetIdExpTest
[ RUN      ] urGraphGetIdExpTest.SuccessMonotonicallyIncreasing/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________default_queue_submission_mode
[       OK ] urGraphGetIdExpTest.SuccessMonotonicallyIncreasing/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________default_queue_submission_mode (6 ms)
[ RUN      ] urGraphGetIdExpTest.SuccessMonotonicallyIncreasing/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________UR_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE
[       OK ] urGraphGetIdExpTest.SuccessMonotonicallyIncreasing/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________UR_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE (3 ms)
[ RUN      ] urGraphGetIdExpTest.SuccessUniqueAcrossContexts/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________default_queue_submission_mode
[       OK ] urGraphGetIdExpTest.SuccessUniqueAcrossContexts/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________default_queue_submission_mode (2 ms)
[ RUN      ] urGraphGetIdExpTest.SuccessUniqueAcrossContexts/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________UR_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE
[       OK ] urGraphGetIdExpTest.SuccessUniqueAcrossContexts/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________UR_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE (2 ms)
[ RUN      ] urGraphGetIdExpTest.InvalidNullHandleGraph/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________default_queue_submission_mode
[       OK ] urGraphGetIdExpTest.InvalidNullHandleGraph/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________default_queue_submission_mode (1 ms)
[ RUN      ] urGraphGetIdExpTest.InvalidNullHandleGraph/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________UR_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE
[       OK ] urGraphGetIdExpTest.InvalidNullHandleGraph/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________UR_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE (1 ms)
[ RUN      ] urGraphGetIdExpTest.InvalidNullPtrGraphId/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________default_queue_submission_mode
[       OK ] urGraphGetIdExpTest.InvalidNullPtrGraphId/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________default_queue_submission_mode (1 ms)
[ RUN      ] urGraphGetIdExpTest.InvalidNullPtrGraphId/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________UR_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE
[       OK ] urGraphGetIdExpTest.InvalidNullPtrGraphId/UR_BACKEND_LEVEL_ZERO__Intel_R__oneAPI_Unified_Runtime_over_Level_Zero_V2__Intel_R__Arc_TM__B580_Graphics_ID0ID______________________UR_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE (1 ms)
[----------] 8 tests from urGraphGetIdExpTest (20 ms total)

[----------] Global test environment tear-down
[==========] 8 tests from 1 test suite ran. (20 ms total)
[  PASSED  ] 8 tests.

The MacOS build and Adapters/level_zero/batch_event_status.cpp failure appear unrelated and are showing up in other PR jobs.

[bratpiorka]

CUDA/HIP/OpenCL/native cpu LGTM