Caspar generates CUDA kernels from SymForce symbolic residuals for nonlinear solvers and reports 5-20x faster bundle adjustment than prior art on the BAL dataset with lower memory use.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.RO 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Caspar: CUDA Accelerator for Symbolic Programming with Adaptive Reordering
Caspar generates CUDA kernels from SymForce symbolic residuals for nonlinear solvers and reports 5-20x faster bundle adjustment than prior art on the BAL dataset with lower memory use.