[SWDEV-357472] Add evicted_ms metric (#620)

- **Added evicted_time metric for kfd processes**.  
  - Time that queues are evicted on a GPU in milliseconds
  - Added to CLI in `amd-smi monitor -q` and `amd-smi process`
  - Added to C API and Python API:
    - amdsmi_get_gpu_process_list()
    - amdsmi_get_gpu_compute_process_info()
    - amdsmi_get_gpu_compute_process_info_by_pid()

---------

Signed-off-by: Pryor, Adam <Adam.Pryor@amd.com>
Этот коммит содержится в:
Pryor, Adam
2025-10-28 14:49:03 -05:00
коммит произвёл GitHub
родитель 6f222c11a6
Коммит 2144cfbba4
11 изменённых файлов: 95 добавлений и 31 удалений
+4 -1
Просмотреть файл
@@ -2953,7 +2953,8 @@ def amdsmi_get_gpu_process_list(
"cpu_mem": process_list[index].memory_usage.cpu_mem,
"vram_mem": process_list[index].memory_usage.vram_mem,
},
"cu_occupancy": _validate_if_max_uint(process_list[index].cu_occupancy, MaxUIntegerTypes.UINT32_T)
"cu_occupancy": _validate_if_max_uint(process_list[index].cu_occupancy, MaxUIntegerTypes.UINT32_T),
"evicted_time": _validate_if_max_uint(process_list[index].evicted_time, MaxUIntegerTypes.UINT32_T)
})
return result
@@ -5303,6 +5304,7 @@ def amdsmi_get_gpu_compute_process_info() -> List[Dict[str, int]]:
"vram_usage": proc.vram_usage,
"sdma_usage": proc.sdma_usage,
"cu_occupancy": proc.cu_occupancy,
"evicted_time": proc.evicted_time,
}
for proc in procs
]
@@ -5324,6 +5326,7 @@ def amdsmi_get_gpu_compute_process_info_by_pid(pid: int) -> Dict[str, int]:
"vram_usage": proc.vram_usage,
"sdma_usage": proc.sdma_usage,
"cu_occupancy": proc.cu_occupancy,
"evicted_time": proc.evicted_time,
}