PackSELL packs delta-encoded indices and values into single words with tunable bit allocation, delivering up to 1.63x faster FP16 SpMV and FP32-accurate performance exceeding FP16 cuSPARSE while reducing memory traffic.
Title resolution pending
2 Pith papers cite this work, alongside 1 external citations. Polarity classification is still indexing.
2
Pith papers citing it
1
external citations · external index
citation-role summary
background 1
citation-polarity summary
fields
cs.DC 2years
2026 2verdicts
UNVERDICTED 2roles
background 1polarities
background 1representative citing papers
Block-level domain decomposition with WaveHoltz subdomain solver on NVIDIA A100 GPUs delivers 2x-25x speedup over MINRES for the Helmholtz equation, plus further 2x-10x gains in single precision.
citing papers explorer
-
PackSELL: A Sparse Matrix Format for Precision-Agnostic High-Performance SpMV
PackSELL packs delta-encoded indices and values into single words with tunable bit allocation, delivering up to 1.63x faster FP16 SpMV and FP32-accurate performance exceeding FP16 cuSPARSE while reducing memory traffic.
-
Efficient Domain Decomposition for the Helmholtz Equation on GPUs
Block-level domain decomposition with WaveHoltz subdomain solver on NVIDIA A100 GPUs delivers 2x-25x speedup over MINRES for the Helmholtz equation, plus further 2x-10x gains in single precision.