method · first seen 27 Oct 2026
GPU kernel for sparse matrix-vector multiplication
Also: sparse GEMV
Specialized GPU kernels that compute matrix-vector products on compressed weights to speed up LLM decoding.
Mentions per week
PapersCodeDiscussion
Earliest sources
- metalSPARSE: A 2:4 Semi-Structured Sparse GEMV Kernel for Apple Silicon GPUsPapers · 27 Oct 2026 · Tek Raj Awasthi, Babita Awasthi
Latest sources
- metalSPARSE: A 2:4 Semi-Structured Sparse GEMV Kernel for Apple Silicon GPUsPapers · 27 Oct 2026 · Tek Raj Awasthi, Babita Awasthi