Popular repositories Loading

1. sparse-ternary-fma sparse-ternary-fma Public

   High-performance ternary arithmetic kernel with 2-bit encoding and AVX-512 SIMD acceleration for FHE and AI applications

   C 1 1

2. 2bit-ternary-bandwidth 2bit-ternary-bandwidth Public

   Surgical proof that 2-bit packed ternary encoding solves the memory bandwidth bottleneck in neural network inference.

   C 1

3. BitNet BitNet Public

   Forked from microsoft/BitNet

   Official inference framework for 1-bit LLMs

   C++ 1

4. fused-vs-unpacked-bench fused-vs-unpacked-bench Public

   Fused computation on packed ternary data is fundamentally more efficient than decode-then-compute approaches. This is not about a specific FHE implementation or BitNet optimization. This is a compu…

   C 1