[09:59 Sat,9.November 2024 by Thomas Richter] |
Owners of CPUs with the AVX-512 instruction set extension have reason to rejoice: the developers of the open-source tool FFmpeg have written manually optimized assembly code specifically for this purpose, leveraging these special instructions to accelerate certain functions by a factor of 3 to 94. On newer CPUs, the code is expected to run even faster. AVX-512 enables large amounts of data to be processed in parallel, as the 512-bit registers can execute a multitude of operations simultaneously—ideal for FFmpeg, which is often used for compute-intensive tasks in video editing that can be easily parallelized. ;)
By manually adjusting the code at the assembly level, the developers can tap into the potential of AVX-512 and achieve far more performance than would be possible with traditional high-level programming languages. Benchmark tests show impressive results compared to standard implementations and previous SIMD instruction sets like AVX2 and SSE3. FFmpeg? It’s Everywhere!If you don’t know FFmpeg, you probably don’t realize you’re already using it, as the various functions of this "Swiss Army Knife"—especially encoding, decoding, and converting different audio and video formats—are integrated into numerous audio and video programs due to its open nature. For example, the popular VLC player, Blackmagic’s video editing software DaVinci Resolve, Adobe Premiere, the free media center Kodi, and the open-source audio editing program Audacity all use the FFmpeg library for various tasks. ;) An acceleration of FFmpeg, therefore, impacts numerous tools once the new code is integrated into various programs, speeding up many important functions that deal with video formats. Once this experimental code is officially integrated into FFmpeg, it will be interesting to see which functions and which programs specifically benefit from the speedup. Who Benefits from the AVX-512 Optimizations?However, not all hardware benefits from these improvements. While processors like AMD’s Ryzen 9000 series have full access to AVX-512, Intel has disabled support for AVX-512 on its 12th, 13th, and 14th generation Core processors. Owners of these CPUs cannot take advantage of the performance boost, while users with compatible hardware can fully exploit the acceleration.  Here is an overview of all CPUs with AVX-512.  more infos at bei x.comdeutsche Version dieser Seite: FFmpeg bis zu 94x schneller durch handoptimierten AVX-512 Assembler-Code |
 |
|
|