Skip to main content

Module monitor

trueno_gpu

Module monitor 

Source
Expand description

TUI Compute Mode Flow: CPU, GPU, and Memory Monitoring

TRUENO-SPEC-020: Unified compute device abstraction for real-time monitoring.

§Architecture

┌─────────────┐   ┌─────────────┐   ┌─────────────┐
│   CPU       │   │ NVIDIA GPU  │   │  AMD GPU    │
│  Backend    │   │   Backend   │   │   Backend   │
└──────┬──────┘   └──────┬──────┘   └──────┬──────┘
       └────────────┬────┴────────────┬────┘
             ┌──────▼─────────────────▼──────┐
             │   Unified Telemetry Collector │
             └───────────────┬───────────────┘
                             │
             ┌───────────────▼───────────────┐
             │    TUI Renderer (presentar)    │
             └───────────────────────────────┘

§References

  • [Volkov2008] Tile size optimization, memory bandwidth modeling
  • [Liker2004] Toyota Way principles (Genchi Genbutsu, Jidoka)
  • [LAMBDA-0002] Memory pressure levels specification

Structs§

ColorScheme
Colorblind-safe color scheme based on Viridis
ComputeMetrics
Compute pipeline metrics
CpuDevice
CPU compute device using sysinfo
CudaDeviceInfo
CUDA device information from native driver API
CudaMemoryInfo
Real-time CUDA memory information from cuMemGetInfo
DataFlowMetrics
Data flow and transfer metrics
DataFlowRenderState
Data flow render state
DeviceComputeMetrics
Per-device compute metrics
DeviceId
Unique identifier for a compute device
DeviceRenderState
Device render state
DeviceSnapshot
Point-in-time snapshot of device metrics
GaugeWidget
Gauge widget for showing percentages
GpuVramMetrics
GPU VRAM metrics for a single device
KernelExecution
Active kernel execution
KernelRenderState
Kernel render state
MemoryMetrics
Comprehensive memory metrics for system and GPU
MemoryRenderState
Memory render state
PressureAnalysis
Detailed memory pressure analysis
ProgressBarWidget
Progress bar widget
RgbColor
RGB color
Section
TUI section definition
SparklineWidget
Sparkline widget for showing history
StressMetrics
Stress test metrics collected during execution
StressTestConfig
Stress test configuration
StressTestReport
Stress test report
TableWidget
Table widget
TextWidget
Text widget
Transfer
Memory transfer between host and device
TransferId
Unique transfer identifier
TuiLayout
TUI layout configuration
TuiRenderState
Complete TUI render state

Enums§

ChaosPreset
Chaos engineering preset
DeviceType
Type of compute device
GaugeColor
Gauge color
KernelStatus
Kernel execution status
KeyAction
Keyboard action
MemoryLocation
Memory location
PressureLevel
Memory pressure level based on available memory percentage
SizeCheck
Terminal size check result
StressTarget
Stress target
StressTestState
Stress test runner state
StressTestVerdict
Stress test verdict
TextStyle
Text style
ThrottleReason
Reason for compute throttling
TransferDirection
Transfer direction
TransferStatus
Transfer status
Widget
TUI widget

Traits§

ComputeDevice
Unified compute device abstraction

Functions§

achieved_gflops
Calculate achieved GFLOPS from operation count and time
compute_efficiency
Calculate compute efficiency percentage
cuda_device_count
Get the number of CUDA devices
cuda_monitoring_available
Check if CUDA monitoring is available
gemm_flops
Calculate theoretical FLOPS for a GEMM operation