Search papers, labs, and topics across Lattice.
This paper introduces Baseline-Mechanism-Outcome Attribution (BMOA), a framework designed to dissect and analyze compiler-induced numerical deviations in floating-point computations. By separating the comparison relation, compiler mechanisms, and accuracy outcomes, BMOA enables a more nuanced understanding of numerical correctness beyond simple pass/fail metrics. Evaluations on various scientific-computing kernels reveal that the choice of baseline significantly influences diagnostic outcomes, and that not all compiler-induced deviations result in accuracy loss, highlighting the framework's potential for improving numerical correctness assessments in compiler design.
Baseline choice can radically alter the diagnosis of numerical deviations, revealing that compiler-induced discrepancies don't always equate to accuracy loss.
Formalizing compiler-aware numerical correctness requires distinguishing what an observed floating-point difference means, what compiler behavior the evidence supports, and what numerical consequence follows. Existing testing workflows often collapse these questions into a pass/fail mismatch. We introduce Baseline--Mechanism--Outcome Attribution (BMOA), a diagnostic framework that separates the comparison relation and system boundary, the evidence-supported compiler mechanism, and the reference-qualified accuracy outcome. BMOA combines operational strict floating-point, transformation-local, reproducibility, cross-compiler, and higher-precision comparisons, while preserving mixed, ambiguous, and unknown attributions when evidence is insufficient. Each record retains inputs, configurations, numerical metrics, and supporting artifacts for audit. We evaluate BMOA on six scientific-computing kernels, deterministic stress-input families, and controlled Clang configurations on ARM64. A 1,276-record attribution corpus and a 162-instance controlled mechanism matrix show that baseline choice changes diagnoses, compiler-induced deviation does not imply accuracy loss, and cancellation and large dynamic range expose the strongest effects within the targeted matrix. BMOA converts raw mismatches into explicit, auditable, evidence-bounded records. Although it is not itself a proof system, these records provide an empirical foundation for future formal specifications and proof obligations for compiler-aware numerical correctness.
