NVIDIA GPU
Public
NVIDIA GPU loadable plugin
Source
xxxxxxxxxx1
1
package nvml2
2
3
3
// Runner defines the interface for an NVML runner.4
4
//5
5
//nolint:interfacebloat6
6
type Runner interface {7
-
InitNVML() error7
+
// InitNVML initializes the NVML library using the older NVML interface.8
+
Init() error9
+
// InitNVMLv2 initializes the NVML library using the NVML v2 interface.10
+
InitV2() error8
11
9
-
InitNVMLv2() error10
-
// Initialize NVML and any necessary resources12
+
// GetDeviceCount retrieves the number of NVIDIA devices using the standard NVML interface.11
13
GetDeviceCount() (uint, error)12
14
13
-
// Initialize NVML and any necessary resources15
+
// GetDeviceCountV2 retrieves the number of NVIDIA devices using the NVML v2 interface.14
16
GetDeviceCountV2() (uint, error)15
17
16
-
// Get a device by index18
+
// GetDeviceByIndexV2 retrieves a handle to an NVIDIA device by its index using the NVML v2 interface.17
19
GetDeviceByIndexV2(index uint) (*NVMLDevice, error)18
20
21
+
// GetDeviceByUUID retrieves a handle to an NVIDIA device by its UUID.19
22
GetDeviceByUUID(uuid string) (*NVMLDevice, error)20
23
21
-
// Get NVML version24
+
// GetNVMLVersion retrieves the version of the NVML library currently in use.22
25
GetNVMLVersion() (string, error)23
26
24
-
// Get Driver version27
+
// GetDriverVersion retrieves the version of the NVIDIA driver currently in use.25
28
GetDriverVersion() (string, error)26
29
27
30
// Shutdown NVML and clean up resources28
31
ShutdownNVML() error29
32
33
+
// Close releases the resources associated with the loaded library in the Runner.30
34
Close() error31
35
}32
36
33
37
// Device defines the methods for interacting with a GPU device.34
38
//35
39
//nolint:interfacebloat36
40
type Device interface {37
-
// Get the temperature of the device.41
+
// GetTemperature retrieves the temperature of the NVIDIA device using the default sensor.38
42
GetTemperature() (int, error)39
43
40
-
// Get the memory information for the device.44
+
// GetMemoryInfo retrieves memory information for the NVIDIA device.41
45
GetMemoryInfo() (*MemoryInfo, error)42
46
47
+
// GetBAR1MemoryInfo retrieves BAR1 memory information for the NVIDIA device.43
48
GetBAR1MemoryInfo() (*MemoryInfo, error)44
49
45
-
// Get the memory information for the device.50
+
// GetMemoryInfoV2 retrieves detailed memory information for the NVIDIA device using the NVML v2 interface.46
51
GetMemoryInfoV2() (*MemoryInfoV2, error)47
52
48
-
// Get the fan speed of the device.53
+
// GetFanSpeed retrieves the current fan speed of the NVIDIA device as a percentage of its maximum speed.49
54
GetFanSpeed() (uint, error)50
55
51
-
// Get PCIe throughput (TX or RX).52
-
GetPcieThroughput(metricType PcieMetricType) (uint, error)53
-
54
-
// GetUtilizationRates returns the GPU and memory utilization in that order (GPU, Memory).56
+
// GetPCIeThroughput retrieves the PCIe throughput for the NVIDIA device, based on the specified metric type.57
+
GetPCIeThroughput(metricType PcieMetricType) (uint, error)58
+
59
+
// GetUtilizationRates retrieves the GPU and memory utilization rates for the device.60
+
// The GPU utilization represents the percentage of time over the past sampling period61
+
// that the GPU was actively processing, while the memory utilization indicates the62
+
// percentage of time the memory was being accessed.63
+
//64
+
// Returns:65
+
// - gpuUtilization (uint): The GPU utilization rate as a percentage (0-100).66
+
// - memoryUtilization (uint): The memory utilization rate as a percentage (0-100).67
+
// - error: An error object if there is a failure in verifying the NVML symbol existence68
+
// or in retrieving the utilization rates from NVML.55
69
GetUtilizationRates() (uint, uint, error)56
70
57
-
// Get the UUID of the device.71
+
// GetUUID retrieves the UUID of the NVIDIA device.58
72
GetUUID() (string, error)59
73
60
-
// Get the name of the device.74
+
// GetName retrieves the name of the NVIDIA device.61
75
GetName() (string, error)62
76
63
-
// Get the serial number of the device.77
+
// GetSerial retrieves the serial number of the NVIDIA device.64
78
GetSerial() (string, error)65
79
80
+
// GetPowerUsage retrieves the power usage of the NVIDIA device in milliwatts.66
81
GetPowerUsage() (uint, error)67
82
83
+
// GetPerformanceState retrieves the performance state (P-state) of the NVIDIA device.68
84
GetPerformanceState() (uint, error)69
85
86
+
// GetClockInfo retrieves the clock rate for the specified clock type of the NVIDIA device.70
87
GetClockInfo(clockType ClockType) (uint, error)71
88
89
+
// GetPowerManagementLimit retrieves the power management limit of the NVIDIA device in milliwatts.72
90
GetPowerManagementLimit() (uint, error)73
91
74
92
GetTotalEnergyConsumption() (uint64, error)75
93
94
+
// GetEncoderStats retrieves statistics related to the encoder activity on the device.95
+
// It returns the following statistics:96
+
// - sessionCount: the number of active encoder sessions.97
+
// - averageFps: the average frames per second across all active encoder sessions.98
+
// - averageLatency: the average latency (in milliseconds) across all active encoder sessions.76
99
GetEncoderStats() (uint, uint, uint, error)77
100
101
+
// GetEncoderUtilization retrieves the encoder utilization statistics for the device.102
+
// It returns the following values:103
+
// - utilization: the percentage of time over the past sampling period during which the encoder was active.104
+
// - samplingPeriodUs: the sampling period duration in microseconds,105
+
// indicating how long the utilization metric was measured.78
106
GetEncoderUtilization() (uint, uint, error)79
107
108
+
// GetDecoderUtilization retrieves the decoder utilization statistics for the device.109
+
// It returns the following values:110
+
// - utilization: the percentage of time over the past sampling period during which the decoder was active.111
+
// - samplingPeriodUs: the sampling period duration in microseconds, indicating how long the utilization112
+
// metric was measured.80
113
GetDecoderUtilization() (uint, uint, error)81
114
115
+
// GetMemoryErrorCounter retrieves the ECC memory error count for the specified error type,116
+
// memory location, and counter type.82
117
GetMemoryErrorCounter(83
118
errorType MemoryErrorType,84
119
memoryLocation MemoryLocation,85
120
counterType EccCounterType,86
121
) (uint64, error)87
122
123
+
// GetEccMode retrieves the current and pending ECC (Error Correction Code) modes for the device.124
+
// ECC mode indicates whether error correction is enabled or disabled on the device.125
+
//126
+
// Returns:127
+
// - `currentEnabled` (bool): `true` if ECC is currently enabled, `false` if disabled.128
+
// - `pendingEnabled` (bool): `true` if ECC will be enabled on the next reboot, `false` if it will be disabled.129
+
// - `error` (error): An error if the function fails to retrieve the ECC mode, otherwise `nil`.88
130
GetEccMode() (bool, bool, error)89
131
}90
132
91
133
// MemoryInfoV2 represents the memory information of the device (in bytes).92
134
type MemoryInfoV2 struct {93
135
Total uint64 `json:"total_memory_bytes"` // Total memory available94
136
Reserved uint64 `json:"reserved_memory_bytes"` // Memory reserved by the system95
137
Free uint64 `json:"free_memory_bytes"` // Free memory available96
138
Used uint64 `json:"used_memory_bytes"` // Memory currently being used + reserved97
139
}