Summary

Test rocroller (shard 4/5) on gfx94X-dcgpu failed in bump PR #5556 with a GPU hardware exception that simultaneously aborted both concurrent rocroller ctest processes. The rocroller test binary is unchanged; only rocm-systems was bumped (18 commits, b8c378e → 84ed18c).

Context

• Workflow: Multi-Arch CI
• Run: https://github.com/ROCm/TheRock/actions/runs/26766391968
• Affected arch: gfx94X-dcgpu (gfx942, MI300X)
• Runner group: linux-gfx942-1gpu-core42-ossci-rocm
• Platform: Windows

Runners

• linux-gfx942-1gpu-core42-ossci-rocm-jj2tj-runner-8kkt6
• linux-gfx942-1gpu-core42-ossci-rocm-jj2tj-runner-v7hlc
• linux-gfx942-1gpu-core42-ossci-rocm-jj2tj-runner-tjz6f
• linux-gfx942-1gpu-core42-ossci-rocm-jj2tj-runner-4htrn

Full logs:

• https://github.com/ROCm/TheRock/actions/runs/26766391968/job/79171606686?pr=5556
• https://github.com/ROCm/TheRock/actions/runs/26766391968/job/79171606683?pr=5556
• https://github.com/ROCm/TheRock/actions/runs/26791782958/job/78997621946?pr=5573
• https://github.com/ROCm/TheRock/actions/runs/26766391968/job/79171606683?pr=5556

Failure signature

4: HW Exception by GPU node-4 (Agent handle: 0x61372ad74ae0) reason :GPU Hang
8: HW Exception by GPU node-4 (Agent handle: 0x56387a41ea30) reason :GPU Hang
shared/rocroller/test/catch/GlobalLoadStoreTest.cpp:105: FAILED:
  {Unknown expression after the reported line}
due to a fatal error condition:
  SIGABRT - Abort (abnormal termination) signal

1/2 Test #4: rocroller-tests_full_suite .........Subprocess aborted***Exception: 113.49 sec
2/2 Test #8: rocroller-tests-catch_full_suite ...Subprocess aborted***Exception: 113.50 sec
0% tests passed, 2 tests failed out of 2