Struct GpuMemoryDashboard 

pub struct GpuMemoryDashboard {
    allocations: RwLock<HashMap<u64, GpuMemoryAllocation>>,
    device_stats: RwLock<HashMap<u32, GpuDeviceMemoryStats>>,
    thresholds: GpuMemoryThresholds,
    allocation_counter: AtomicU64,
    total_allocated: AtomicU64,
    peak_allocated: AtomicU64,
}

GPU Memory Dashboard for monitoring and visualization.

Provides real-time GPU memory tracking with allocation history, per-kernel usage, and memory pressure alerts.

Example

use ringkernel_core::observability::GpuMemoryDashboard;

let dashboard = GpuMemoryDashboard::new();

// Track an allocation
dashboard.track_allocation(
    1,
    "input_queue",
    65536,
    GpuMemoryType::QueueBuffer,
    0,
    Some("processor_kernel"),
);

// Get current stats
let stats = dashboard.get_device_stats(0);
println!("GPU 0 utilization: {:.1}%", stats.utilization());

// Generate Grafana panel JSON
let panel = dashboard.grafana_panel();

Fields

allocations: RwLock<HashMap<u64, GpuMemoryAllocation>>

Active allocations.

device_stats: RwLock<HashMap<u32, GpuDeviceMemoryStats>>

Per-device statistics.

thresholds: GpuMemoryThresholds

Memory pressure thresholds.

allocation_counter: AtomicU64

Allocation counter for unique IDs.

total_allocated: AtomicU64

Total bytes allocated.

peak_allocated: AtomicU64

Peak bytes allocated.

Implementations

impl GpuMemoryDashboard

pub fn new() -> Arc<Self>

Create a new GPU memory dashboard.

pub fn with_thresholds(thresholds: GpuMemoryThresholds) -> Arc<Self>

Create with custom thresholds.

pub fn track_allocation( &self, id: u64, name: impl Into<String>, size: usize, memory_type: GpuMemoryType, device_index: u32, kernel_id: Option<&str>, )

Track a new GPU memory allocation.

pub fn next_allocation_id(&self) -> u64

Generate a new unique allocation ID.

pub fn track_deallocation(&self, id: u64)

Track deallocation.

pub fn mark_unused(&self, id: u64)

Mark an allocation as no longer in use (but not freed).

pub fn register_device( &self, device_index: u32, name: impl Into<String>, total_memory: u64, )

Register a GPU device.

pub fn update_device_stats( &self, device_index: u32, free_memory: u64, ringkernel_used: u64, )

Update device memory statistics.

pub fn get_device_stats( &self, device_index: u32, ) -> Option<GpuDeviceMemoryStats>

Get device statistics.

pub fn get_all_device_stats(&self) -> Vec<GpuDeviceMemoryStats>

Get all device statistics.

pub fn get_allocations(&self) -> Vec<GpuMemoryAllocation>

Get all active allocations.

pub fn get_kernel_allocations( &self, kernel_id: &str, ) -> Vec<GpuMemoryAllocation>

Get allocations for a specific kernel.

pub fn total_allocated(&self) -> u64

Get total allocated memory.

pub fn peak_allocated(&self) -> u64

Get peak allocated memory.

pub fn allocation_count(&self) -> usize

Get allocation count.

pub fn check_pressure(&self, device_index: u32) -> MemoryPressureLevel

Check memory pressure level for a device.

pub fn grafana_panel(&self) -> GrafanaPanel

Generate Grafana dashboard panel for GPU memory.

pub fn prometheus_metrics(&self) -> String

Generate Prometheus metrics for GPU memory.

pub fn summary_report(&self) -> String

Generate a memory summary report.

Trait Implementations

impl Default for GpuMemoryDashboard

fn default() -> Self

Returns the “default value” for a type.