Skip to content
Telecommunications & Carrier Systems·Tier-1 Telecom Infrastructure Operator

EVS Codec SIMD Acceleration for Telecom Networks

Optimizing 3GPP Enhanced Voice Services (EVS) transcoding with AVX-512 and ARM Neon assembly, reducing CPU footprint by 78% on high-density media gateways.
Tech Stack:C99RustAVX-512ARM NeonSIP / RTPLinux Kernel Bypass

Engineering Deliverables

  • ›Bespoke SIMD audio resampler in C/Rust
  • ›Deterministic latency under 12ms per frame
  • ›Zero packet memory allocation in hot path

Measured Production Impact

  • Quadrupled concurrent voice call capacity on existing rack hardware
  • Saved over €340,000 in planned server expansions
  • 100% 3GPP conformance test suite validation

Engineering Architecture & Implementation

Telecommunication gateways handle millions of audio frames every minute. Traditional software implementations of the 3GPP EVS codec saturated rack servers with heavy floating-point operations. Vext re-engineered the mathematical DSP pipeline directly into vectorized SIMD instructions. The resulting engine processes 4,800 simultaneous HD audio streams on a single dual-socket server while strictly adhering to ITU-T and 3GPP bit-exact conformance specs.

Have a similar engineering challenge?

Speak directly with the lead architect behind this project.