June 22nd, 2024

Zlib-ng 2.2 Speeds Up Compression By ~12% On x86_64 CPUs

Zlib-ng 2.2 release candidate offers 12% faster compression on x86_64 CPUs with performance optimizations, revamped memory allocation, modern API, and CPU intrinsics support. Improvements focus on memory allocation, system calls reduction, and small buffer processing efficiency. Michael Larabel praises enhanced compression speed and memory handling.

Read original articleLink Icon
Zlib-ng 2.2 Speeds Up Compression By ~12% On x86_64 CPUs

Zlib-ng 2.2 has released its first candidate, offering a 12% speed improvement in compression on x86_64 CPUs compared to previous versions. The update includes performance optimizations and a revamped memory allocation system to enhance compression and decompression speeds. Zlib-ng maintains compatibility with Zlib, introduces a modernized native API, and supports various CPU intrinsics for different architectures. The latest release candidate focuses on optimizing memory allocation during initialization, reducing system calls, and improving the efficiency of processing small buffers. These changes aim to enhance overall compression performance and ensure smoother operation under memory constraints. Users can access the 2.2 release candidate announcement for more details and downloads to test the new optimizations. Michael Larabel, the principal author of Phoronix.com, highlighted the improvements brought by Zlib-ng 2.2, emphasizing the advancements in compression speed and memory handling for modern processors.

Link Icon 1 comments
By @hulitu - 4 months
Quiet thin article. 12% speedup on default (-6 ?) compression is not much. Will be nice to see what it brings for max compression.