Dense-layer-benchmark

This repository benchmarks the performance of various dense (fully connected) layer implementations in Java using JMH (Java Microbenchmark Harness).

📌 Purpose

This project aims to evaluate and compare the performance of:

✅ Scalar (naive) implementation

✅ EJML matrix multiplication

✅ DeepLearning4j (DL4J)

✅ Java Vector API

✅ Java Vector API with Fused Multiply Add (FMA)

across varying input and output sizes, and multi-threaded environments.

🚀 Benchmarked Implementations

Implementation	Description
Scalar	Plain nested loop multiplication with bias addition
EJML	Matrix multiplication using EJML's FMatrixRMaj and CommonOps_FDRM
DL4J	Dense layer using DL4J with identity activation
Vector	SIMD-based computation using Java Vector API
FMA	SIMD-based computation using Java Vector API with Fused Multiply Add

📊 Benchmark Modes

Two benchmark modes are evaluated:

AverageTime – time per operation (μs)

Throughput – operations per second (ops/µs)

✅ Sanity Test

A sanity test is included to ensure correctness across all implementations. It verifies that outputs are numerically consistent (within float tolerance).

🧪 How to Run

Build the Project

  mvn clean install  

Run the benchmark

 
java --enable-preview --add-modules jdk.incubator.vector -jar target/benchmark.jar \
-t 8 -rf csv -rff results_threads8.csv
 

You can also customize threads, benchmark mode and output format.