llama.cpp

21f3d1be - k-quants : fix build on armv7 (android only) (#2920)

k-quants : fix build on armv7 (android only) (#2920) * k-quants : fix build on armv7 * ggml : cleanup unused arm32 specific impl * k-quants : avoid some unused vzero / mzero define * ggml-alloc : use 4g for MEASURE_MAX_SIZE in 32-bit arm