FFmpeg just made one obscure filter 100x faster using ancient coding tricks most developers have abandoned
- The greatest acceleration of FFMPEG affects only one function that few people will have heard
- The handwritten assembly returns to a niche filter that most users will never even touch
- AVX512 gives FFMPEG an absurd 100x gain – but only if your CPU supports it
The FFMPEG project, known to supply some of the most used video editing and modification tools, makes headlines again.
The developers claim to have achieved what they call “the greatest acceleration so far”, offering a 100x performance gain in a recent update.
The capture? It only applies to a unique and obscure function, and the means of carrying out eyebrows – the handwritten assembly code, a technique widely considered to be overwhelmed by most of today’s developers.
Assembly coding stimulates nostalgia and skepticism
The assembly language, formerly essential to make the most of the limited material in the 80s and 1990s, became a niche practice.
However, FFMPEG developers continue to rely on this for extreme optimization, calling on “assembly evangelists”.
In their latest fix, they rewritten a filter called RangeDetect8_AvX512 using AVX512 instructions, part of a SIMD toolbox (single instructions, multiple data) which helps CPUs to do several tasks in parallel.
On AVX512 support systems, the AVX2 variant still offers an improvement of 65.63%.
As the team points out, “this is a unique function which is now 100 times faster, not the FFMPEG set.”
This news is following a similar boost reported in November 2024, where another patch brought some operations up to 94x faster.
In this case, part of the previous performance difference comes from the complexity of the inadequate filter: the generic version C used a convolution of 8 laps, while the SIMD version used a 6 -stroke approach.
Even the compilation of version C in liberation mode with a better compiler like Clang could fill more than 50% of the gap, which suggests that some of the speed gains claimed may have been exaggerated by comparing the worst case with the best cases of case.
“The recording allowance is zero on the compilers,” joked the developers on social networks, highlighting the ineffectiveness of the compiler.
Despite the warnings, this concentration renewed on low -level coding has sparked new conversations around performance optimization.
FFMPEG feeds everything, from the Player VLC Media to countless YouTube download tools, so even small improvements in isolated filters can wave through widely used software.
However, it should be noted that these results are often difficult to reproduce and apply in larger parts of the code base.
Although these types of deep optimizations are impressive, they may not reflect real world improvements for everyday users modifying images with video editing software.
Unless other basic functions receive similar treatment, the promise of a faster FFMPEG can remain limited to technical benchmarks.
Via Tomshardware
You might also love
