Warren Deng

Taming Bitwise Behavior in GPU Kernels with Tensor Core: Black-Box Reconstruction, Compiler Enforcement, and Static Verification

Determinism and numerical reproducibility are increasingly required of GPU kernels in machine learning systems, yet deterministic implementations of the same kernel can still differ bit for bit. Floating-point reduction order is the primary cause, alongside partial-sum precision, fused multiply-add operations, and roun...