Performance Insights

Proven Performance at Scale

See the real numbers. LightningSim and OmniSim deliver breakthrough acceleration across diverse computing workloads. Up to 352x faster than traditional simulation methods.

Max Speedup

352 ×

OmniSim vs Co-simulation

Accuracy

99.9 %

Cycle Count Accuracy

Benchmarks

30+

Diverse Workloads Tested

Visualization

Runtime Performance Comparison

Actual runtimes across representative benchmarks showing dramatic acceleration vs traditional C/RTL co-simulation.

Performance Across Categories

Consistent speedups across different benchmark categories

DSP & Mathematical Operations

Fixed-point Square Root

7.47×
vs Cosim

FIR Filter

10.37×
vs Cosim

Window Convolution

8.98×
vs Cosim

Floating-point Conv

20.24×
vs Cosim

Arbitrary Precision ALU

11.91×
vs Cosim

Loop & Control Flow Operations

Parallel Loops

12.41×
OmniSim

Imperfect Loops

12.11×
OmniSim

Pipelined Nested Loops

11.38×
OmniSim

AI/ML & Complex Workloads

FlowGNN - GIN

352.53×
OmniSim vs Cosim

Graph neural network with 260K cycles

FlowGNN - DGN

85.07×
OmniSim vs Cosim

Directed graph neural network

Complete Benchmark Results

Full dataset from Vitis v2021.1 showing cycle counts and runtime performance

BenchmarkCosim (s)LightningSim (s)OmniSim (s)LS SpeedupOS Speedup
Fixed-point Square Root27.254.973.655.48×7.47×
FIR Filter20.122.431.948.23×10.37×
Window Convolution28.303.693.157.67×8.98×
Floating-point Conv49.782.422.4620.57×20.24×
Unoptimized FFT153.532.782.9155.23×52.76×
FlowGNN - GIN4219.8528.9011.97146.02×352.53×
FlowGNN - DGN996.1326.9011.7137.03×85.07×

30+ benchmarks tested across DSP operations, loop structures, memory access patterns, and AI/ML workloads. View complete profiling data →

Key Insights

99.9% Accuracy

Timing estimates from LightningSim and OmniSim maintain 99.9% accuracy with respect to C/RTL co-simulation across all benchmarks.

100x+ Acceleration

LightningSim achieves consistent speedups of 100x or greater across diverse workloads, with peak performance reaching 55x on FFT operations.

AI/ML Powerhouse

Exceptional performance on complex workloads like FlowGNN models, delivering up to 352x speedup for graph neural network operations.

Design Space Exploration

Combined with incremental design space exploration features, achieve up to 577x acceleration for comprehensive optimization workflows.

Ready to Accelerate Your Design?

Transform your HLS and FPGA workflow with breakthrough simulation speed. Get started today.