Stencil kernels on Tenstorrent Wormhole match CPU speed in isolation but lag overall due to PCIe and initialization overheads, with lower energy use for large inputs via the Axpy mapping.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Stencil Computations on Tenstorrent Wormhole
Stencil kernels on Tenstorrent Wormhole match CPU speed in isolation but lag overall due to PCIe and initialization overheads, with lower energy use for large inputs via the Axpy mapping.