Skip to content

FastMemory 0.1.1 [ALPHA-2026-08]

Latest

Choose a tag to compare

@andrestubbe andrestubbe released this 14 Aug 06:54
· 10 commits to main since this release

TITLE:
FastMemory: 4K Video Frame Buffer SIMD-Aligned Memory vs Java Heap 🚀

DESCRIPTION:
Demonstrating FastMemory, a SIMD-aligned off-heap memory allocator for Java. This performance comparison shows 4K video frame buffer processing (3840x2160 pixels, 31 MB) using 32-byte aligned native memory with page locking versus traditional Java heap arrays.

TECHNICAL HIGHLIGHTS:

  • 32-Byte SIMD Alignment: AVX2/AVX-512 vector operations without alignment penalties
  • Physical Page Locking: VirtualLock prevents OS swap for real-time video processing
  • Zero GC Overhead: Off-heap memory eliminates garbage collection pauses
  • FastPointer Integration: Direct 64-bit address arithmetic for pixel manipulation

RESULTS:

  • Write Speedup: 1.1x faster than Java heap arrays (21ms vs 23ms)
  • SIMD-Ready: 32-byte alignment optimal for AVX2 vector instructions
  • Page-Locked: Physical RAM residency prevents swap latency
  • Memory Type: Off-heap native vs heap allocation

FASTJAVA CORE REPOSITORIES:

  1. FastMemory: https://github.com/andrestubbe/FastMemory
  2. FastPointer: https://github.com/andrestubbe/FastPointer
  3. FastSIMD: https://github.com/andrestubbe/FastSIMD
  4. FastSharedMemory: https://github.com/andrestubbe/FastSharedMemory

Ecosystem Overview: https://github.com/andrestubbe/FastJava

Part of the FastJava Ecosystem — Making the JVM faster.
Made with ⚡ by Andre Stubbe

TAGS:
Java, FastJava, FastMemory, SIMD, Off-Heap, AVX2, Video Processing, Memory Alignment, VirtualLock, Page Locking