Neon CpuF16 and CpuBF16 implementations#3110

Draft

ivarflakstad wants to merge 11 commits intomainfrom

neon-cpu-impl-f16-bf16

Commits on Sep 25, 2025

Add direct copy for floats in quantization
ivarflakstad
committed
Calling q matmul directly gives slightly different results, but within ggml error leniency
ivarflakstad
committed

Commits on Oct 1, 2025

Merge branch 'main' into quant-direct-copy-floats
ivarflakstad
committed

Commits on Oct 3, 2025