veritatisquaesitoressumus
veritatisquaesitoressumus
/ ggml_turboquant.c
Created
March 25, 2026 14:38
TurboQuant KV Cache Compression for llama.cpp (Zandieh et al., ICLR 2026) — 3-bit, 4.9x compression, 18/18 tests passing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| /* | |
| * TurboQuant: CPU Reference Implementation | |
| * ========================================== | |
| * Implements Algorithm 1 (TurboQuant_mse) from Zandieh et al., ICLR 2026. | |
| * | |
| * This is the portable C implementation that runs on CPU. | |
| * The CUDA implementation (ggml_turboquant.cu) mirrors this logic | |
| * with GPU-optimized kernels. | |
| * | |
| * Authors: Jim Sullivan / Claude collaboration |