A GEMM accelerator exploits subset-pattern reuse in bit-sliced weights to cut multiplications, reporting 7.46x and 3.97x speedups over Olive and BitVert at comparable LLaMA perplexity.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Transitive Array: An Efficient GEMM Accelerator with Result Reuse
A GEMM accelerator exploits subset-pattern reuse in bit-sliced weights to cut multiplications, reporting 7.46x and 3.97x speedups over Olive and BitVert at comparable LLaMA perplexity.