A game scripting language that runs speech-to-text 5x faster than llama.cpp on CPU.
Summary
This article介绍s dasllama, an inference engine built atop the daslang scripting language. It covers macro-based optimization, self-tuning kernels, and cross-CPU/GPU benchmarks for ASR, TTS, and LLM workloads, highlighting significant speedups on commodity hardware and ongoing GPU integration efforts.