Showcase Applications

RingKernel includes four comprehensive showcase applications demonstrating GPU-accelerated computing with the actor model. Each showcases different aspects of the framework.

WaveSim - 2D Acoustic Wave Simulation

Interactive wave propagation simulator implementing the Finite-Difference Time-Domain (FDTD) method.

Key Features

• Tile-based Actor Model: 16x16 cell tiles as actors with K2K messaging for halo exchange
• Educational Modes: Visualize computing paradigms from 1950s sequential to modern parallel
• Multi-backend: CPU (SoA + SIMD + Rayon) and CUDA

Performance

GPU vs CPU speedup: 3.1x at 256x256, 9.9x at 512x512

Run It

cargo run -p ringkernel-wavesim --release

Click anywhere on the canvas to inject wave impulses.

TxMon - Transaction Monitoring

Real-time transaction monitoring for banking/AML compliance scenarios with GPU-accelerated rule evaluation.

Key Features

• Transaction Factory: Configurable synthetic generation with realistic patterns
• Compliance Rules: Velocity breach, amount threshold, structuring detection
• Three GPU Approaches:
  • Batch Kernel: Maximum throughput (~93B elem/sec)
  • Ring Kernel: Persistent actors with HLC and K2K messaging
  • Stencil Kernel: Pattern detection in transaction networks

Performance

Run It

cargo run -p ringkernel-txmon --release

AccNet - Accounting Network Analytics

GPU-accelerated accounting network analysis transforming double-entry bookkeeping into graph analytics.

Key Features

• Network Visualization: Interactive graph showing account relationships and money flows
• Fraud Detection: Circular flows, threshold clustering, Benford’s Law violations
• GAAP Compliance: Automated detection of accounting rule violations
• Temporal Analysis: Seasonality, trends, behavioral anomalies
• Industry Chart of Accounts: Pre-built templates for different industries

Industry Templates

GPU Kernels

1. Suspense Detection: Identifies suspicious clearing accounts
2. GAAP Violation: Checks for improper account pairings
3. Benford Analysis: Statistical analysis of first-digit distribution
4. PageRank: Network centrality and influence analysis

Architecture

┌─────────────────┐     ┌──────────────────┐     ┌────────────────┐
│   Data Fabric   │────>│  GPU Kernels     │────>│  Visualization │
│ (Synthetic Gen) │     │ (CUDA)           │     │  (egui Canvas) │
└─────────────────┘     └──────────────────┘     └────────────────┘

Run It

# CPU backend
cargo run -p ringkernel-accnet --release

# With CUDA GPU acceleration
cargo run -p ringkernel-accnet --release --features cuda

ProcInt - Process Intelligence

GPU-accelerated process mining for discovering, analyzing, and monitoring business processes from event logs.

Key Features

• DFG Mining: Directly-Follows Graph construction from event streams with GPU acceleration
• Pattern Detection: Bottlenecks, loops, rework, and long-running activities
• Conformance Checking: Fitness and precision metrics against reference models
• Timeline View: Partial order traces with concurrent activity visualization
• Multi-Sector: Healthcare, Manufacturing, Finance, and IT process templates

GPU Kernels

1. DFG Construction: Builds directly-follows graph from event stream (batch kernel)
2. Pattern Detection: Identifies process patterns like bottlenecks and loops
3. Partial Order Derivation: Builds precedence matrix from interval events (stencil kernel)
4. Conformance Checking: Validates traces against reference models

Architecture

┌─────────────────┐     ┌──────────────────┐     ┌────────────────┐
│   Data Fabric   │────>│  GPU Kernels     │────>│  Visualization │
│ (Event Stream)  │     │ (DFG/Pattern)    │     │  (egui Canvas) │
└─────────────────┘     └──────────────────┘     └────────────────┘
        │                       │                       │
   Sector Templates      4 Kernel Types           Force-directed
   Anomaly Injection     CPU/CUDA fallback        DFG + Timeline

Performance

Run It

# CPU backend
cargo run -p ringkernel-procint --release

# Run benchmark
cargo run -p ringkernel-procint --bin procint-benchmark --release

Common Patterns Across Showcases

All four applications demonstrate RingKernel’s core capabilities:

Build All Showcases

# All showcases, CPU only
cargo build -p ringkernel-wavesim -p ringkernel-txmon -p ringkernel-accnet -p ringkernel-procint --release

# With CUDA support
cargo build -p ringkernel-wavesim -p ringkernel-txmon -p ringkernel-accnet -p ringkernel-procint --release --features cuda

Next: CUDA Code Generation