MERBIT achieves 1.27x single-precision and 1.25x double-precision average speedups over cuSPARSE COO on 50 large irregular datasets by merging merge-path partitioning with compact bit-field encoding for SpMV.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.DC 1years
2026 1verdicts
UNVERDICTED 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
MERBIT: A GPU-Based SpMV Method for Iterative Workloads
MERBIT achieves 1.27x single-precision and 1.25x double-precision average speedups over cuSPARSE COO on 50 large irregular datasets by merging merge-path partitioning with compact bit-field encoding for SpMV.