Is this page helpful?

Fabric definitions#

Macros#

NVML_GPU_FABRIC_CLIQUE_MAX: Maximum number of fabric clique entries.
NVML_GPU_FABRIC_CLIQUE_TYPE_MULTICAST_LOGICAL_ENDPOINT: Multicast logical-endpoint-based access.
NVML_GPU_FABRIC_CLIQUE_TYPE_MULTICAST_POINTER: Multicast pointer-based access.
NVML_GPU_FABRIC_CLIQUE_TYPE_UNICAST_LOGICAL_ENDPOINT: Unicast logical-endpoint-based access.
NVML_GPU_FABRIC_CLIQUE_TYPE_UNICAST_POINTER: Unicast pointer-based access.
NVML_GPU_FABRIC_HEALTH_GET: GPU Fabric Health Status Mask for various fields can be obtained using the below macro.
NVML_GPU_FABRIC_HEALTH_MASK_ACCESS_TIMEOUT_RECOVERY_FALSE: Fabric Health Mask: Access Timeout Recovery not in progress.
NVML_GPU_FABRIC_HEALTH_MASK_ACCESS_TIMEOUT_RECOVERY_NOT_SUPPORTED: Fabric Access Timeout Recovery.
NVML_GPU_FABRIC_HEALTH_MASK_ACCESS_TIMEOUT_RECOVERY_TRUE: Fabric Health Mask: Access Timeout Recovery in progress.
NVML_GPU_FABRIC_HEALTH_MASK_DEGRADED_BW_FALSE: Fabric Health Mask: Bandwidth not degraded.
NVML_GPU_FABRIC_HEALTH_MASK_DEGRADED_BW_NOT_SUPPORTED: Fabric Degraded BW.
NVML_GPU_FABRIC_HEALTH_MASK_DEGRADED_BW_TRUE: Fabric Health Mask: Bandwidth degraded.
NVML_GPU_FABRIC_HEALTH_MASK_GFM_STATE_CONNECTED: Fabric Health Mask: Global Fabric Manager State is Connected.
NVML_GPU_FABRIC_HEALTH_MASK_GFM_STATE_DISCONNECTED: Fabric Health Mask: Global Fabric Manager State is Disconnected.
NVML_GPU_FABRIC_HEALTH_MASK_GFM_STATE_NOT_SUPPORTED: Global Fabric Manager State.
NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_GPU_STATE_INVALID: Fabric Health Mask: Incorrect Configuration - GPU State Invalid.
NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_INCOMPATIBLE_GPU_FW: Fabric Health Mask: Incorrect Configuration - Incompatible GPU Firmware.
NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_INCORRECT_CHASSIS_SN: Fabric Health Mask: Incorrect Configuration - Chassis Serial Number.
NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_INCORRECT_SYSGUID: Fabric Health Mask: Incorrect Configuration - SysGUID.
NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_INSUFFICIENT_NVLINKS: Fabric Health Mask: Incorrect Configuration - Insufficient Nvlinks.
NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_INVALID_LOCATION: Fabric Health Mask: Incorrect Configuration - Invalid Location.
NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_NONE: Fabric Health Mask: Correct Configuration.
NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_NOT_SUPPORTED: Fabric Incorrect Configuration.
NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_NO_PARTITION: Fabric Health Mask: Incorrect Configuration - No Partition.
NVML_GPU_FABRIC_HEALTH_MASK_PARTITION_ASSIGNED_FALSE: Fabric Health Mask: Partition is not Assigned.
NVML_GPU_FABRIC_HEALTH_MASK_PARTITION_ASSIGNED_NOT_SUPPORTED: Fabric Partition Assigned.
NVML_GPU_FABRIC_HEALTH_MASK_PARTITION_ASSIGNED_TRUE: Fabric Health Mask: Partition is Assigned.
NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_RECOVERY_FALSE: Fabric Health Mask: Route Recovery not in progress.
NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_RECOVERY_NOT_SUPPORTED: Fabric Route Recovery.
NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_RECOVERY_TRUE: Fabric Health Mask: Route Recovery in progress.
NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_UNHEALTHY_FALSE: Fabric Health Mask: Route is healthy.
NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_UNHEALTHY_NOT_SUPPORTED: Nvlink Fabric Route Unhealthy.
NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_UNHEALTHY_TRUE: Fabric Health Mask: Route is unhealthy.
NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_ACCESS_TIMEOUT_RECOVERY: Fabric Health Mask Bit Shift for Access Timeout Recovery.
NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_DEGRADED_BW: Fabric Health Mask Bit Shift for Degraded Bandwidth.
NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_GFM_STATE: Fabric Health Mask Bit Shift for Global Fabric Manager State.
NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_INCORRECT_CONFIGURATION: Fabric Health Mask Bit Shift for Incorrect Configuration.
NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_PARTITION_ASSIGNED: Fabric Health Mask Bit Shift for Partition Assigned.
NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_ROUTE_RECOVERY: Fabric Health Mask Bit Shift for Route Recovery.
NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_ROUTE_UNHEALTHY: Fabric Health Mask Bit Shift for Route Unhealthy.
NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_ACCESS_TIMEOUT_RECOVERY: Fabric Health Mask Width for Access Timeout Recovery.
NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_DEGRADED_BW: Fabric Health Mask Width for Degraded Bandwidth.
NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_GFM_STATE: Fabric Health Mask Width for Global Fabric Manager State.
NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_INCORRECT_CONFIGURATION: Fabric Health Mask Width for Incorrect Configuration.
NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_PARTITION_ASSIGNED: Fabric Health Mask Width for Partition Assigned.
NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_ROUTE_RECOVERY: Fabric Health Mask Width for Route Recovery.
NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_ROUTE_UNHEALTHY: Fabric Health Mask Width for Route Unhealthy.
NVML_GPU_FABRIC_HEALTH_SUMMARY_HEALTHY: Fabric Health Summary: Healthy.
NVML_GPU_FABRIC_HEALTH_SUMMARY_LIMITED_CAPACITY: Fabric Health Summary: Limited Capacity.
NVML_GPU_FABRIC_HEALTH_SUMMARY_NOT_SUPPORTED: Fabric Health.
NVML_GPU_FABRIC_HEALTH_SUMMARY_UNHEALTHY: Fabric Health Summary: Unhealthy.
NVML_GPU_FABRIC_HEALTH_TEST: GPU Fabric Health Status Mask for various fields can be tested using the below macro.
NVML_GPU_FABRIC_STATE_COMPLETED: Fabric Probe State completed.
NVML_GPU_FABRIC_STATE_IN_PROGRESS: Fabric Probe in progress.
NVML_GPU_FABRIC_STATE_NOT_STARTED: Fabric Probe has not started.
NVML_GPU_FABRIC_STATE_NOT_SUPPORTED: Fabric Probe States.
NVML_GPU_FABRIC_UUID_LEN: Length of Fabric UUID.
nvmlGpuFabricInfo_v2: Version identifier value for nvmlGpuFabricInfo_v2_t::version .
nvmlGpuFabricInfo_v3: Version identifier value for nvmlGpuFabricInfo_v3_t::version .

Structs#

nvmlGpuFabricClique_v1_t: Fabric clique entry.
nvmlGpuFabricInfo_t: Contains the device fabric information.
nvmlGpuFabricInfo_v2_t: GPU Fabric information (v2).
nvmlGpuFabricInfo_v3_t: GPU Fabric information (v3).
nvmlGpuFabricInfo_v4_t: GPU Fabric information (v4).

Typedefs#

nvmlGpuFabricInfoV_t
nvmlGpuFabricState_t: Probe State of GPU registration process.

Macros#

NVML_GPU_FABRIC_CLIQUE_MAX#: Maximum number of fabric clique entries.

NVML_GPU_FABRIC_CLIQUE_TYPE_MULTICAST_LOGICAL_ENDPOINT#: Multicast logical-endpoint-based access.

NVML_GPU_FABRIC_CLIQUE_TYPE_MULTICAST_POINTER#: Multicast pointer-based access.

NVML_GPU_FABRIC_CLIQUE_TYPE_UNICAST_LOGICAL_ENDPOINT#: Unicast logical-endpoint-based access.

NVML_GPU_FABRIC_CLIQUE_TYPE_UNICAST_POINTER#: Unicast pointer-based access.

NVML_GPU_FABRIC_HEALTH_GET(var, type)#

GPU Fabric Health Status Mask for various fields can be obtained using the below macro.

Ex - NVML_GPU_FABRIC_HEALTH_GET(var, _DEGRADED_BW) Macro to get GPU fabric health status.

NVML_GPU_FABRIC_HEALTH_MASK_ACCESS_TIMEOUT_RECOVERY_FALSE#: Fabric Health Mask: Access Timeout Recovery not in progress.

NVML_GPU_FABRIC_HEALTH_MASK_ACCESS_TIMEOUT_RECOVERY_NOT_SUPPORTED#

Fabric Access Timeout Recovery.

Fabric Health Mask: Access Timeout Recovery not supported

NVML_GPU_FABRIC_HEALTH_MASK_ACCESS_TIMEOUT_RECOVERY_TRUE#: Fabric Health Mask: Access Timeout Recovery in progress.

NVML_GPU_FABRIC_HEALTH_MASK_DEGRADED_BW_FALSE#: Fabric Health Mask: Bandwidth not degraded.

NVML_GPU_FABRIC_HEALTH_MASK_DEGRADED_BW_NOT_SUPPORTED#

Fabric Degraded BW.

Fabric Health Mask: Degraded Bandwidth not supported

NVML_GPU_FABRIC_HEALTH_MASK_DEGRADED_BW_TRUE#: Fabric Health Mask: Bandwidth degraded.

NVML_GPU_FABRIC_HEALTH_MASK_GFM_STATE_CONNECTED#: Fabric Health Mask: Global Fabric Manager State is Connected.

NVML_GPU_FABRIC_HEALTH_MASK_GFM_STATE_DISCONNECTED#: Fabric Health Mask: Global Fabric Manager State is Disconnected.

NVML_GPU_FABRIC_HEALTH_MASK_GFM_STATE_NOT_SUPPORTED#

Global Fabric Manager State.

Fabric Health Mask: Global Fabric Manager State not supported

NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_GPU_STATE_INVALID#: Fabric Health Mask: Incorrect Configuration - GPU State Invalid.

NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_INCOMPATIBLE_GPU_FW#: Fabric Health Mask: Incorrect Configuration - Incompatible GPU Firmware.

NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_INCORRECT_CHASSIS_SN#: Fabric Health Mask: Incorrect Configuration - Chassis Serial Number.

NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_INCORRECT_SYSGUID#: Fabric Health Mask: Incorrect Configuration - SysGUID.

NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_INSUFFICIENT_NVLINKS#: Fabric Health Mask: Incorrect Configuration - Insufficient Nvlinks.

NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_INVALID_LOCATION#: Fabric Health Mask: Incorrect Configuration - Invalid Location.

NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_NONE#: Fabric Health Mask: Correct Configuration.

NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_NOT_SUPPORTED#

Fabric Incorrect Configuration.

Fabric Health Mask: Incorrect Configuration not supported

NVML_GPU_FABRIC_HEALTH_MASK_INCORRECT_CONFIGURATION_NO_PARTITION#: Fabric Health Mask: Incorrect Configuration - No Partition.

NVML_GPU_FABRIC_HEALTH_MASK_PARTITION_ASSIGNED_FALSE#: Fabric Health Mask: Partition is not Assigned.

NVML_GPU_FABRIC_HEALTH_MASK_PARTITION_ASSIGNED_NOT_SUPPORTED#

Fabric Partition Assigned.

Fabric Health Mask: Partition Assigned not supported

NVML_GPU_FABRIC_HEALTH_MASK_PARTITION_ASSIGNED_TRUE#: Fabric Health Mask: Partition is Assigned.

NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_RECOVERY_FALSE#: Fabric Health Mask: Route Recovery not in progress.

NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_RECOVERY_NOT_SUPPORTED#

Fabric Route Recovery.

Fabric Health Mask: Route Recovery not supported

NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_RECOVERY_TRUE#: Fabric Health Mask: Route Recovery in progress.

NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_UNHEALTHY_FALSE#: Fabric Health Mask: Route is healthy.

NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_UNHEALTHY_NOT_SUPPORTED#

Nvlink Fabric Route Unhealthy.

Fabric Health Mask: Route Unhealthy not supported

NVML_GPU_FABRIC_HEALTH_MASK_ROUTE_UNHEALTHY_TRUE#: Fabric Health Mask: Route is unhealthy.

NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_ACCESS_TIMEOUT_RECOVERY#: Fabric Health Mask Bit Shift for Access Timeout Recovery.

NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_DEGRADED_BW#: Fabric Health Mask Bit Shift for Degraded Bandwidth.

NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_GFM_STATE#: Fabric Health Mask Bit Shift for Global Fabric Manager State.

NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_INCORRECT_CONFIGURATION#: Fabric Health Mask Bit Shift for Incorrect Configuration.

NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_PARTITION_ASSIGNED#: Fabric Health Mask Bit Shift for Partition Assigned.

NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_ROUTE_RECOVERY#: Fabric Health Mask Bit Shift for Route Recovery.

NVML_GPU_FABRIC_HEALTH_MASK_SHIFT_ROUTE_UNHEALTHY#: Fabric Health Mask Bit Shift for Route Unhealthy.

NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_ACCESS_TIMEOUT_RECOVERY#: Fabric Health Mask Width for Access Timeout Recovery.

NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_DEGRADED_BW#: Fabric Health Mask Width for Degraded Bandwidth.

NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_GFM_STATE#: Fabric Health Mask Width for Global Fabric Manager State.

NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_INCORRECT_CONFIGURATION#: Fabric Health Mask Width for Incorrect Configuration.

NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_PARTITION_ASSIGNED#: Fabric Health Mask Width for Partition Assigned.

NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_ROUTE_RECOVERY#: Fabric Health Mask Width for Route Recovery.

NVML_GPU_FABRIC_HEALTH_MASK_WIDTH_ROUTE_UNHEALTHY#: Fabric Health Mask Width for Route Unhealthy.

NVML_GPU_FABRIC_HEALTH_SUMMARY_HEALTHY#: Fabric Health Summary: Healthy.

NVML_GPU_FABRIC_HEALTH_SUMMARY_LIMITED_CAPACITY#: Fabric Health Summary: Limited Capacity.

NVML_GPU_FABRIC_HEALTH_SUMMARY_NOT_SUPPORTED#

Fabric Health.

Fabric Health Summary: Not supported

NVML_GPU_FABRIC_HEALTH_SUMMARY_UNHEALTHY#: Fabric Health Summary: Unhealthy.

NVML_GPU_FABRIC_HEALTH_TEST(var, type, val)#

GPU Fabric Health Status Mask for various fields can be tested using the below macro.

Ex - NVML_GPU_FABRIC_HEALTH_TEST(var, _DEGRADED_BW, _TRUE) Macro to test GPU fabric health status.

NVML_GPU_FABRIC_STATE_COMPLETED#: Fabric Probe State completed.

NVML_GPU_FABRIC_STATE_IN_PROGRESS#: Fabric Probe in progress.

NVML_GPU_FABRIC_STATE_NOT_STARTED#: Fabric Probe has not started.

NVML_GPU_FABRIC_STATE_NOT_SUPPORTED#

Fabric Probe States.

Fabric Probe State not supported

NVML_GPU_FABRIC_UUID_LEN#: Length of Fabric UUID.

nvmlGpuFabricInfo_v2#

Version identifier value for nvmlGpuFabricInfo_v2_t::version.

Version macro for nvmlGpuFabricInfo_v2_t

nvmlGpuFabricInfo_v3#

Version identifier value for nvmlGpuFabricInfo_v3_t::version.

Version macro for nvmlGpuFabricInfo_v3_t

Typedefs#

typedef nvmlGpuFabricInfo_v3_t nvmlGpuFabricInfoV_t#

typedef unsigned char nvmlGpuFabricState_t#: Probe State of GPU registration process.