1
High-Rate Quantized Matrix Multiplication II
水填充理论指导LLM权重量化中的比特分配，揭示经典源编码方案的实际改进效果。
arXiv:2605.13768v2 Announce Type: replace Abstract: This is the second part of the work investigating quantized matrix multiplication (MatMul). In par…
2
MCemm a GEMM (General Matrix Multiply) Kernel Generato
专为Apple Metal优化的GEMM内核生成器开源发布，支持自定义tile、warp、数据类型与激活函数，加速Metal计算性能。
Released mCemm a GEMM (General Matrix Multiply) kernel generator for Apple Metal which generates optimized Metal shaders with configurable tile sizes,…
3
Hawkeye: Reproducing GPU-Level Non-Determinism
揭秘GPU计算的黑箱，在CPU上精确复现NVIDIA矩阵乘法，解决机器学习非确定性难题。
arXiv:2603.20421v2 Announce Type: replace-cross Abstract: We present Hawkeye, a system for analyzing and reproducing GPU-level arithmetic operations. …