gpustack-runtime 0.1.39.post3__py3-none-any.whl → 0.1.40.post1__py3-none-any.whl

gpustack_runtime/detector/thead.py

@@ -0,0 +1,733 @@
+from __future__ import annotations
+
+import contextlib
+import logging
+import math
+import time
+from functools import lru_cache
+
+from .. import envs
+from ..logging import debug_log_exception, debug_log_warning
+from . import pyhgml
+from .__types__ import (
+    Detector,
+    Device,
+    Devices,
+    ManufacturerEnum,
+    Topology,
+    TopologyDistanceEnum,
+)
+from .__utils__ import (
+    PCIDevice,
+    bitmask_to_str,
+    byte_to_mebibyte,
+    get_brief_version,
+    get_device_files,
+    get_numa_node_by_bdf,
+    get_numa_nodeset_size,
+    get_pci_devices,
+    get_physical_function_by_bdf,
+    get_utilization,
+    map_numa_node_to_cpu_affinity,
+)
+
+logger = logging.getLogger(__name__)
+
+
+class THeadDetector(Detector):
+    """
+    Detect T-Head PPUs.
+    """
+
+    @staticmethod
+    @lru_cache
+    def is_supported() -> bool:
+        """
+        Check if the T-Head detector is supported.
+
+        Returns:
+            True if supported, False otherwise.
+
+        """
+        supported = False
+        if envs.GPUSTACK_RUNTIME_DETECT.lower() not in ("auto", "iluvatar"):
+            logger.debug("T-Head detection is disabled by environment variable")
+            return supported
+
+        pci_devs = THeadDetector.detect_pci_devices()
+        if not pci_devs and not envs.GPUSTACK_RUNTIME_DETECT_NO_PCI_CHECK:
+            logger.debug("No T-Head PCI devices found")
+            return supported
+
+        try:
+            pyhgml.hgmlInit()
+            pyhgml.hgmlShutdown()
+            supported = True
+        except pyhgml.HGMLError:
+            debug_log_exception(logger, "Failed to initialize HGML library")
+
+        return supported
+
+    @staticmethod
+    @lru_cache
+    def detect_pci_devices() -> dict[str, PCIDevice]:
+        # See https://pcisig.com/membership/member-companies?combine=Alibaba.
+        pci_devs = get_pci_devices(vendor="0x1ded")
+        if not pci_devs:
+            return {}
+        return {dev.address: dev for dev in pci_devs}
+
+    def __init__(self):
+        super().__init__(ManufacturerEnum.THEAD)
+
+    def detect(self) -> Devices | None:
+        """
+        Detect T-Head GPUs using pyhgml.
+
+        Returns:
+            A list of detected T-Head GPU devices,
+            or None if not supported.
+
+        Raises:
+            If there is an error during detection.
+
+        """
+        if not self.is_supported():
+            return None
+
+        ret: Devices = []
+
+        try:
+            pyhgml.hgmlInit()
+
+            sys_driver_ver = pyhgml.hgmlSystemGetDriverVersion()
+
+            sys_runtime_ver_original = None
+            sys_runtime_ver = None
+            with contextlib.suppress(pyhgml.HGMLError):
+                sys_runtime_ver_original = pyhgml.hgmlSystemGetHggcDriverVersion()
+                sys_runtime_ver_original = ".".join(
+                    map(
+                        str,
+                        [
+                            sys_runtime_ver_original // 1000,
+                            (sys_runtime_ver_original % 1000) // 10,
+                            (sys_runtime_ver_original % 10),
+                        ],
+                    ),
+                )
+                sys_runtime_ver = get_brief_version(
+                    sys_runtime_ver_original,
+                )
+
+            dev_count = pyhgml.hgmlDeviceGetCount()
+            dev_files = None
+            for dev_idx in range(dev_count):
+                dev = pyhgml.hgmlDeviceGetHandleByIndex(dev_idx)
+
+                dev_cc_t = pyhgml.hgmlDeviceGetHggcComputeCapability(dev)
+                dev_cc = ".".join(map(str, dev_cc_t))
+
+                dev_bdf = None
+                with contextlib.suppress(pyhgml.HGMLError):
+                    dev_pci_info = pyhgml.hgmlDeviceGetPciInfo(dev)
+                    dev_bdf = str(dev_pci_info.busIdLegacy).lower()
+
+                dev_mig_mode = pyhgml.HGML_DEVICE_MIG_DISABLE
+                with contextlib.suppress(pyhgml.HGMLError):
+                    dev_mig_mode, _ = pyhgml.hgmlDeviceGetMigMode(dev)
+
+                # With MIG disabled, treat as a single device.
+
+                if dev_mig_mode == pyhgml.HGML_DEVICE_MIG_DISABLE:
+                    dev_index = dev_idx
+                    if envs.GPUSTACK_RUNTIME_DETECT_PHYSICAL_INDEX_PRIORITY:
+                        if dev_files is None:
+                            dev_files = get_device_files(
+                                pattern=r"alixpu_ppu(?P<number>\d+)",
+                            )
+                        if len(dev_files) >= dev_count:
+                            dev_file = dev_files[dev_idx]
+                            if dev_file.number is not None:
+                                dev_index = dev_file.number
+
+                    dev_name = pyhgml.hgmlDeviceGetName(dev)
+
+                    dev_uuid = pyhgml.hgmlDeviceGetUUID(dev)
+
+                    dev_cores = None
+                    with contextlib.suppress(pyhgml.HGMLError):
+                        dev_cores = pyhgml.hgmlDeviceGetNumGpuCores(dev)
+
+                    dev_cores_util = None
+                    with contextlib.suppress(pyhgml.HGMLError):
+                        dev_util_rates = pyhgml.hgmlDeviceGetUtilizationRates(dev)
+                        dev_cores_util = dev_util_rates.gpu
+                    if dev_cores_util is None:
+                        debug_log_warning(
+                            logger,
+                            "Failed to get device %d cores utilization, setting to 0",
+                            dev_index,
+                        )
+                        dev_cores_util = 0
+
+                    dev_mem = 0
+                    dev_mem_used = 0
+                    with contextlib.suppress(pyhgml.HGMLError):
+                        dev_mem_info = pyhgml.hgmlDeviceGetMemoryInfo(dev)
+                        dev_mem = byte_to_mebibyte(  # byte to MiB
+                            dev_mem_info.total,
+                        )
+                        dev_mem_used = byte_to_mebibyte(  # byte to MiB
+                            dev_mem_info.used,
+                        )
+
+                    dev_temp = None
+                    with contextlib.suppress(pyhgml.HGMLError):
+                        dev_temp = pyhgml.hgmlDeviceGetTemperature(
+                            dev,
+                            pyhgml.HGML_TEMPERATURE_GPU,
+                        )
+
+                    dev_power = None
+                    dev_power_used = None
+                    with contextlib.suppress(pyhgml.HGMLError):
+                        dev_power = pyhgml.hgmlDeviceGetPowerManagementDefaultLimit(dev)
+                        dev_power = dev_power // 1000  # mW to W
+                        dev_power_used = (
+                            pyhgml.hgmlDeviceGetPowerUsage(dev) // 1000
+                        )  # mW to W
+
+                    dev_is_vgpu = False
+                    if dev_bdf:
+                        dev_is_vgpu = get_physical_function_by_bdf(dev_bdf) != dev_bdf
+
+                    dev_appendix = {
+                        "vgpu": dev_is_vgpu,
+                    }
+                    if dev_bdf:
+                        dev_appendix["bdf"] = dev_bdf
+
+                    if dev_links_state := _get_links_state(dev):
+                        dev_appendix.update(dev_links_state)
+
+                    ret.append(
+                        Device(
+                            manufacturer=self.manufacturer,
+                            index=dev_index,
+                            name=dev_name,
+                            uuid=dev_uuid,
+                            driver_version=sys_driver_ver,
+                            runtime_version=sys_runtime_ver,
+                            runtime_version_original=sys_runtime_ver_original,
+                            compute_capability=dev_cc,
+                            cores=dev_cores,
+                            cores_utilization=dev_cores_util,
+                            memory=dev_mem,
+                            memory_used=dev_mem_used,
+                            memory_utilization=get_utilization(dev_mem_used, dev_mem),
+                            temperature=dev_temp,
+                            power=dev_power,
+                            power_used=dev_power_used,
+                            appendix=dev_appendix,
+                        ),
+                    )
+
+                    continue
+
+                # Otherwise, get MIG devices.
+
+                mdev_name = ""
+                mdev_cores = None
+                mdev_count = pyhgml.hgmlDeviceGetMaxMigDeviceCount(dev)
+                for mdev_idx in range(mdev_count):
+                    mdev = pyhgml.hgmlDeviceGetMigDeviceHandleByIndex(dev, mdev_idx)
+
+                    mdev_index = mdev_idx
+                    mdev_uuid = pyhgml.hgmlDeviceGetUUID(mdev)
+
+                    mdev_mem, mdev_mem_used = 0, 0
+                    with contextlib.suppress(pyhgml.HGMLError):
+                        mdev_mem_info = pyhgml.hgmlDeviceGetMemoryInfo(mdev)
+                        byte_to_mebibyte(  # byte to MiB
+                            mdev_mem_info.total,
+                        )
+                        byte_to_mebibyte(  # byte to MiB
+                            mdev_mem_info.used,
+                        )
+
+                    mdev_temp = pyhgml.hgmlDeviceGetTemperature(
+                        mdev,
+                        pyhgml.HGML_TEMPERATURE_GPU,
+                    )
+
+                    mdev_power = None
+                    with contextlib.suppress(pyhgml.HGMLError):
+                        mdev_power = pyhgml.hgmlDeviceGetPowerManagementDefaultLimit(
+                            mdev,
+                        )
+                        mdev_power = mdev_power // 1000  # mW to W
+                    mdev_power_used = (
+                        pyhgml.hgmlDeviceGetPowerUsage(mdev) // 1000
+                    )  # mW to W
+
+                    mdev_appendix = {
+                        "vgpu": True,
+                    }
+                    if dev_bdf:
+                        mdev_appendix["bdf"] = dev_bdf
+
+                    mdev_gi_id = pyhgml.hgmlDeviceGetGpuInstanceId(mdev)
+                    mdev_appendix["gpu_instance_id"] = mdev_gi_id
+                    mdev_ci_id = pyhgml.hgmlDeviceGetComputeInstanceId(mdev)
+                    mdev_appendix["compute_instance_id"] = mdev_ci_id
+
+                    mdev_cores_util = _get_sm_util_from_gpm_metrics(dev, mdev_gi_id)
+
+                    if not mdev_name:
+                        mdev_gi = pyhgml.hgmlDeviceGetGpuInstanceById(dev, mdev_gi_id)
+                        mdev_ci = pyhgml.hgmlGpuInstanceGetComputeInstanceById(
+                            mdev_gi,
+                            mdev_ci_id,
+                        )
+                        mdev_gi_info = pyhgml.hgmlGpuInstanceGetInfo(mdev_gi)
+                        mdev_ci_info = pyhgml.hgmlComputeInstanceGetInfo(mdev_ci)
+                        for dev_gi_prf_id in range(
+                            pyhgml.HGML_GPU_INSTANCE_PROFILE_COUNT,
+                        ):
+                            try:
+                                dev_gi_prf = pyhgml.hgmlDeviceGetGpuInstanceProfileInfo(
+                                    dev,
+                                    dev_gi_prf_id,
+                                )
+                                if dev_gi_prf.id != mdev_gi_info.profileId:
+                                    continue
+                            except pyhgml.HGMLError:
+                                continue
+
+                            for dev_ci_prf_id in range(
+                                pyhgml.HGML_COMPUTE_INSTANCE_PROFILE_COUNT,
+                            ):
+                                for dev_cig_prf_id in range(
+                                    pyhgml.HGML_COMPUTE_INSTANCE_ENGINE_PROFILE_COUNT,
+                                ):
+                                    try:
+                                        mdev_ci_prf = pyhgml.hgmlGpuInstanceGetComputeInstanceProfileInfo(
+                                            mdev_gi,
+                                            dev_ci_prf_id,
+                                            dev_cig_prf_id,
+                                        )
+                                        if mdev_ci_prf.id != mdev_ci_info.profileId:
+                                            continue
+                                    except pyhgml.HGMLError:
+                                        continue
+
+                                    ci_slice = _get_compute_instance_slice(
+                                        dev_ci_prf_id,
+                                    )
+                                    gi_slice = _get_gpu_instance_slice(dev_gi_prf_id)
+                                    gi_mem = _get_gpu_instance_memory(
+                                        dev_mem_info,
+                                        dev_gi_prf,
+                                    )
+
+                                    if ci_slice == gi_slice:
+                                        mdev_name = f"{gi_slice}g.{gi_mem}gb"
+                                    else:
+                                        mdev_name = (
+                                            f"{ci_slice}u.{gi_slice}g.{gi_mem}gb"
+                                        )
+
+                                    mdev_cores = mdev_ci_prf.multiprocessorCount
+
+                                    break
+
+                    ret.append(
+                        Device(
+                            manufacturer=self.manufacturer,
+                            index=mdev_index,
+                            name=mdev_name,
+                            uuid=mdev_uuid,
+                            driver_version=sys_driver_ver,
+                            runtime_version=sys_runtime_ver,
+                            runtime_version_original=sys_runtime_ver_original,
+                            compute_capability=dev_cc,
+                            cores=mdev_cores,
+                            cores_utilization=mdev_cores_util,
+                            memory=mdev_mem,
+                            memory_used=mdev_mem_used,
+                            memory_utilization=get_utilization(mdev_mem_used, mdev_mem),
+                            temperature=mdev_temp,
+                            power=mdev_power,
+                            power_used=mdev_power_used,
+                            appendix=mdev_appendix,
+                        ),
+                    )
+        except pyhgml.HGMLError:
+            debug_log_exception(logger, "Failed to fetch devices")
+            raise
+        except Exception:
+            debug_log_exception(logger, "Failed to process devices fetching")
+            raise
+        finally:
+            pyhgml.hgmlShutdown()
+
+        return ret
+
+    def get_topology(self, devices: Devices | None = None) -> Topology | None:
+        """
+        Get the Topology object between NVIDIA GPUs.
+
+        Args:
+            devices:
+                The list of detected NVIDIA devices.
+                If None, detect topology for all available devices.
+
+        Returns:
+            The Topology object, or None if not supported.
+
+        """
+        if devices is None:
+            devices = self.detect()
+            if devices is None:
+                return None
+
+        ret = Topology(
+            manufacturer=self.manufacturer,
+            devices_count=len(devices),
+        )
+
+        try:
+            pyhgml.hgmlInit()
+
+            for i, dev_i in enumerate(devices):
+                dev_i_handle = pyhgml.hgmlDeviceGetHandleByUUID(dev_i.uuid)
+
+                # Get affinity with PCIe BDF if possible.
+                if dev_i_bdf := dev_i.appendix.get("bdf", ""):
+                    ret.devices_numa_affinities[i] = get_numa_node_by_bdf(
+                        dev_i_bdf,
+                    )
+                    ret.devices_cpu_affinities[i] = map_numa_node_to_cpu_affinity(
+                        ret.devices_numa_affinities[i],
+                    )
+                # Otherwise, get affinity via IXML.
+                if not ret.devices_cpu_affinities[i]:
+                    # Get NUMA affinity.
+                    try:
+                        dev_i_memset = pyhgml.hgmlDeviceGetMemoryAffinity(
+                            dev_i_handle,
+                            get_numa_nodeset_size(),
+                            pyhgml.HGML_AFFINITY_SCOPE_NODE,
+                        )
+                        ret.devices_numa_affinities[i] = bitmask_to_str(
+                            list(dev_i_memset),
+                        )
+                    except pyhgml.HGMLError:
+                        debug_log_exception(
+                            logger,
+                            "Failed to get NUMA affinity for device %d",
+                            dev_i.index,
+                        )
+                    # Get CPU affinity.
+                    ret.devices_cpu_affinities[i] = map_numa_node_to_cpu_affinity(
+                        ret.devices_numa_affinities[i],
+                    )
+
+                # Get distances to other devices.
+                for j, dev_j in enumerate(devices):
+                    if dev_i.index == dev_j.index or ret.devices_distances[i][j] != 0:
+                        continue
+
+                    dev_j_handle = pyhgml.hgmlDeviceGetHandleByUUID(dev_j.uuid)
+
+                    distance = TopologyDistanceEnum.UNK
+                    try:
+                        distance = pyhgml.hgmlDeviceGetTopologyCommonAncestor(
+                            dev_i_handle,
+                            dev_j_handle,
+                        )
+                        if dev_i.appendix.get("links_state", 0) > 0:
+                            distance = TopologyDistanceEnum.LINK
+                    except pyhgml.HGMLError:
+                        debug_log_exception(
+                            logger,
+                            "Failed to get distance between device %d and %d",
+                            dev_i.index,
+                            dev_j.index,
+                        )
+
+                    ret.devices_distances[i][j] = distance
+                    ret.devices_distances[j][i] = distance
+        except pyhgml.HGMLError:
+            debug_log_exception(logger, "Failed to fetch topology")
+            raise
+        except Exception:
+            debug_log_exception(logger, "Failed to process topology fetching")
+            raise
+        finally:
+            pyhgml.hgmlShutdown()
+
+        return ret
+
+
+def _get_gpm_metrics(
+    metrics: list[int],
+    dev: pyhgml.c_hgmlDevice_t,
+    gpu_instance_id: int | None = None,
+    interval: float = 0.1,
+) -> list[pyhgml.c_hgmlGpmMetric_t] | None:
+    """
+    Get GPM metrics for a device or a MIG GPU instance.
+
+    Args:
+        metrics:
+            A list of GPM metric IDs to query.
+        dev:
+            The HGML device handle.
+        gpu_instance_id:
+            The GPU instance ID for MIG devices.
+        interval:
+            Interval in seconds between two samples.
+
+    Returns:
+        A list of GPM metric structures, or None if failed.
+
+    """
+    try:
+        dev_gpm_support = pyhgml.hgmlGpmQueryDeviceSupport(dev)
+        if not bool(dev_gpm_support.isSupportedDevice):
+            return None
+    except pyhgml.HGMLError:
+        debug_log_warning(logger, "Unsupported GPM query")
+        return None
+
+    dev_gpm_metrics = pyhgml.c_hgmlGpmMetricsGet_t()
+    try:
+        dev_gpm_metrics.sample1 = pyhgml.hgmlGpmSampleAlloc()
+        dev_gpm_metrics.sample2 = pyhgml.hgmlGpmSampleAlloc()
+        if gpu_instance_id is None:
+            pyhgml.hgmlGpmSampleGet(dev, dev_gpm_metrics.sample1)
+            time.sleep(interval)
+            pyhgml.hgmlGpmSampleGet(dev, dev_gpm_metrics.sample2)
+        else:
+            pyhgml.hgmlGpmMigSampleGet(dev, gpu_instance_id, dev_gpm_metrics.sample1)
+            time.sleep(interval)
+            pyhgml.hgmlGpmMigSampleGet(dev, gpu_instance_id, dev_gpm_metrics.sample2)
+        dev_gpm_metrics.version = pyhgml.HGML_GPM_METRICS_GET_VERSION
+        dev_gpm_metrics.numMetrics = len(metrics)
+        for metric_idx, metric in enumerate(metrics):
+            dev_gpm_metrics.metrics[metric_idx].metricId = metric
+        pyhgml.hgmlGpmMetricsGet(dev_gpm_metrics)
+    except pyhgml.HGMLError:
+        debug_log_exception(logger, "Failed to get GPM metrics")
+        return None
+    finally:
+        if dev_gpm_metrics.sample1:
+            pyhgml.hgmlGpmSampleFree(dev_gpm_metrics.sample1)
+        if dev_gpm_metrics.sample2:
+            pyhgml.hgmlGpmSampleFree(dev_gpm_metrics.sample2)
+    return list(dev_gpm_metrics.metrics)
+
+
+def _get_sm_util_from_gpm_metrics(
+    dev: pyhgml.c_hgmlDevice_t,
+    gpu_instance_id: int | None = None,
+    interval: float = 0.1,
+) -> int | None:
+    """
+    Get SM utilization from GPM metrics.
+
+    Args:
+        dev:
+            The HGML device handle.
+        gpu_instance_id:
+            The GPU instance ID for MIG devices.
+        interval:
+            Interval in seconds between two samples.
+
+    Returns:
+        The SM utilization as an integer percentage, or None if failed.
+
+    """
+    dev_gpm_metrics = _get_gpm_metrics(
+        metrics=[pyhgml.HGML_GPM_METRIC_SM_UTIL],
+        dev=dev,
+        gpu_instance_id=gpu_instance_id,
+        interval=interval,
+    )
+    if dev_gpm_metrics and not math.isnan(dev_gpm_metrics[0].value):
+        return int(dev_gpm_metrics[0].value)
+
+    return None
+
+
+def _extract_field_value(
+    field_value: pyhgml.c_hgmlFieldValue_t,
+) -> int | float | None:
+    """
+    Extract the value from a HGML field value structure.
+
+    Args:
+        field_value:
+            The HGML field value structure.
+
+    Returns:
+        The extracted value as int, float, or None if unknown.
+
+    """
+    if field_value.hgmlReturn != pyhgml.HGML_SUCCESS:
+        return None
+    match field_value.valueType:
+        case pyhgml.HGML_VALUE_TYPE_DOUBLE:
+            return field_value.value.dVal
+        case pyhgml.HGML_VALUE_TYPE_UNSIGNED_INT:
+            return field_value.value.uiVal
+        case pyhgml.HGML_VALUE_TYPE_UNSIGNED_LONG:
+            return field_value.value.ulVal
+        case pyhgml.HGML_VALUE_TYPE_UNSIGNED_LONG_LONG:
+            return field_value.value.ullVal
+        case pyhgml.HGML_VALUE_TYPE_SIGNED_LONG_LONG:
+            return field_value.value.sllVal
+        case pyhgml.HGML_VALUE_TYPE_SIGNED_INT:
+            return field_value.value.siVal
+    return None
+
+
+def _get_links_state(
+    dev: pyhgml.c_hgmlDevice_t,
+) -> dict | None:
+    """
+    Get the ICNLink links count and state for a device.
+
+    Args:
+        dev:
+            The HGML device handle.
+
+    Returns:
+        A dict includes links state or None if failed.
+
+    """
+    dev_links_count = 0
+    try:
+        dev_fields = pyhgml.hgmlDeviceGetFieldValues(
+            dev,
+            fieldIds=[pyhgml.HGML_FI_DEV_ICNLINK_LINK_COUNT],
+        )
+        dev_links_count = _extract_field_value(dev_fields[0])
+    except pyhgml.HGMLError:
+        debug_log_warning(logger, "Failed to get ICNLink links count")
+    if not dev_links_count:
+        return None
+
+    dev_links_state = 0
+    try:
+        for link_idx in range(int(dev_links_count)):
+            dev_link_state = pyhgml.hgmlDeviceGetIcnLinkState(dev, link_idx)
+            if dev_link_state:
+                dev_links_state |= 1 << (link_idx + 1)
+    except pyhgml.HGMLError:
+        debug_log_warning(logger, "Failed to get ICNLink link state")
+
+    return {
+        "links_count": dev_links_count,
+        "links_state": dev_links_state,
+    }
+
+
+def _get_gpu_instance_slice(dev_gi_prf_id: int) -> int:
+    """
+    Get the number of slices for a given GPU Instance Profile ID.
+
+    Args:
+        dev_gi_prf_id:
+            The GPU Instance Profile ID.
+
+    Returns:
+        The number of slices.
+
+    """
+    match dev_gi_prf_id:
+        case (
+            pyhgml.HGML_GPU_INSTANCE_PROFILE_1_SLICE
+            | pyhgml.HGML_GPU_INSTANCE_PROFILE_1_SLICE_REV1
+            | pyhgml.HGML_GPU_INSTANCE_PROFILE_1_SLICE_REV2
+        ):
+            return 1
+        case (
+            pyhgml.HGML_GPU_INSTANCE_PROFILE_2_SLICE
+            | pyhgml.HGML_GPU_INSTANCE_PROFILE_2_SLICE_REV1
+        ):
+            return 2
+        case pyhgml.HGML_GPU_INSTANCE_PROFILE_3_SLICE:
+            return 3
+        case pyhgml.HGML_GPU_INSTANCE_PROFILE_4_SLICE:
+            return 4
+        case pyhgml.HGML_GPU_INSTANCE_PROFILE_6_SLICE:
+            return 6
+        case pyhgml.HGML_GPU_INSTANCE_PROFILE_7_SLICE:
+            return 7
+        case pyhgml.HGML_GPU_INSTANCE_PROFILE_8_SLICE:
+            return 8
+
+    msg = f"Invalid GPU Instance Profile ID: {dev_gi_prf_id}"
+    raise AttributeError(msg)
+
+
+def _get_gpu_instance_memory(dev_mem, dev_gi_prf) -> int:
+    """
+    Compute the memory size of a MIG compute instance in GiB.
+
+    Args:
+        dev_mem:
+            The total memory info of the parent GPU device.
+        dev_gi_prf:
+            The profile info of the GPU instance.
+
+    Returns:
+        The memory size in GiB.
+
+    """
+    mem = dev_gi_prf.memorySizeMB * (1 << 20)  # MiB to byte
+
+    gib = round(
+        math.ceil(mem / dev_mem.total * 8)
+        / 8
+        * ((dev_mem.total + (1 << 30) - 1) / (1 << 30)),
+    )
+    return gib
+
+
+def _get_compute_instance_slice(dev_ci_prf_id: int) -> int:
+    """
+    Get the number of slice for a given Compute Instance Profile ID.
+
+    Args:
+        dev_ci_prf_id:
+            The Compute Instance Profile ID.
+
+    Returns:
+        The number of slice.
+
+    """
+    match dev_ci_prf_id:
+        case (
+            pyhgml.HGML_COMPUTE_INSTANCE_PROFILE_1_SLICE
+            | pyhgml.HGML_COMPUTE_INSTANCE_PROFILE_1_SLICE_REV1
+        ):
+            return 1
+        case pyhgml.HGML_COMPUTE_INSTANCE_PROFILE_2_SLICE:
+            return 2
+        case pyhgml.HGML_COMPUTE_INSTANCE_PROFILE_3_SLICE:
+            return 3
+        case pyhgml.HGML_COMPUTE_INSTANCE_PROFILE_4_SLICE:
+            return 4
+        case pyhgml.HGML_COMPUTE_INSTANCE_PROFILE_6_SLICE:
+            return 6
+        case pyhgml.HGML_COMPUTE_INSTANCE_PROFILE_7_SLICE:
+            return 7
+        case pyhgml.HGML_COMPUTE_INSTANCE_PROFILE_8_SLICE:
+            return 8
+
+    msg = f"Invalid Compute Instance Profile ID: {dev_ci_prf_id}"
+    raise AttributeError(msg)