Collected sources and patterns will appear here. Add from search or the patterns library.
EnvConfig -> HardwarePrecisionConfig
Force the runtime engine to compute matrix multiplications at full float32 precision rather than defaulting to hardware-accelerated mixed precision.
Problem it solves
Implicit low-precision approximations (such as TF32) introduce chaotic numerical divergence in parallel physics rollouts.
Consumes
Emits
Establishes
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.