2026 preprint: GPU mantle-convection benchmarks
Research context
development · proposed · As of 2026-10-10
preprint · 2026-09-18
- Study design
- Numerical convection benchmark and GPU strong/weak-scaling experiments on a specialized spherical-wedge discretization.
- Reported result
- The authors report portable large-scale computation and agreement with the selected convection benchmark suite.
- Limitations
- Computational verification is distinct from physical Earth validation; peer-reviewed acceptance was not established by this inspection.
The TERRA-NG v1.0 preprint reports a matrix-free GPU Stokes–energy solver specialized for spherical wedges and tests against a spherical-shell convection benchmark.
Key knowledge
The TERRA-NG v1.0 preprint reports a matrix-free GPU Stokes–energy solver specialized for spherical wedges and tests against a spherical-shell convection benchmark. Hardware scaling demonstrates computational performance for the stated discretization; it does not validate every Earth rheology or inferred mantle history.
Conditions and scope
- First submitted 18 September 2026; v3 dated 24 September. Preprint status at this inspection.
- Selected discretization and benchmark sections inspected; no independent code execution or timing reproduction.
Sources · 1
References
Sources retain their recorded scope. A source link alone does not establish verification of every statement.
Evidence and review
TERRA-NG v1.0: extreme-scale mantle convection: Sections 2 and 8: Stokes–energy discretization and verification benchmarks (selected source body inspected).
2026-10-10 · Bounded source check of the recorded statement and conditions. This does not certify a unique physical model, reproduce the cited study or replace independent branch review.
Entry details
Permanent link to this entry ↗- Record identity
- research-earth-terra-2026-preprint
- Revision
- research.1
- Structured reference record
- Postgraduate & research reference