GGML : X2 SPEED FOR WASM BY OPTIMIZING SIMD

Summary from the TLDR Web Dev newsletter

Code generated by DeepSeek resulted in significant speed improvements in WebAssembly code for llama.cpp.