Current evidence covers selected one-element linear-static cases with replicated input and root-side assembly.
01 Model and scope
The recorded route launches selected linear-static cases with replicated input and root-side assembly across two ranks. It demonstrates that bounded execution path only; it does not show partitioned finite-element assembly or nonlinear distributed scaling.
Route assumptions and calculation checks
- It does not establish distributed assembly, nonlinear MPI, contact, dynamics, mixed distributed MPC, or arbitrary-mesh scaling.
- Historical structured-TET4 large-model workloads are environment-specific records, not portable performance promises.
- No general GPU support claim is made.
Useful foundations and prerequisites
02 Essential equations and symbols
This sheet records a scope or execution boundary. Its detailed steps explain the relevant conditions.
Definitions and conventions
- MPI ranks
- Processes participating in the recorded run.
- replicated input
- Each participating rank receives the model input.
- root-side assembly
- Assembly location in this bounded route.
- NOT_VALIDATED
- No general distributed capability claim.
03 Step by step derivation
Follow the assumptions, equations and discrete operations in order. The full explanation stays visible.
1 · Describe the evidenced execution layout
The recorded route launches selected one-element linear-static cases with replicated input and root-side assembly across the declared ranks. This describes where data and assembly live in that experiment; it is not a claim of distributed finite-element assembly.
2 · Separate execution from scaling evidence
An MPI job starting successfully proves that a bounded parallel execution path ran. It does not establish partitioning, communication efficiency, arbitrary-mesh scalability, nonlinear distributed mechanics, contact, or dynamic workflows. Large structured-TET4 records are separate named workloads.
3 · Bind every metric to environment and workload
Record mesh identity, active DOFs, rank count, backend, solver/preconditioner, hardware, timing boundary, memory, residual, and replay count. Historical records are observations for those exact configurations; no hardware-independent speedup or general GPU capability follows.
04 Algorithm and implementation
This sheet provides explanatory references rather than an inline source excerpt. Use the labeled sources to inspect the precise boundary.
05 Verification and limitations
OWNER-ACCEPTED BOUNDED EVIDENCE · two ranks
Each reference supports its own model, observable and analysis route. Read the original contract and limitations before transferring a conclusion.