IEEE binary floating point is deterministic. There is no such thing as tolerated errors in compliant implementations.
Non-compliant implementations are widespread for performance reasons, but that is a completely different story.
Also I'm not sure about GPU, but FPU are typically compliant in HW, and it is typically the compilers that have faster approximate modes, and I think for mainstream ones only when you explicitly enable such optimizations, which are disabled by default.
Non-compliant implementations are widespread for performance reasons, but that is a completely different story.
Also I'm not sure about GPU, but FPU are typically compliant in HW, and it is typically the compilers that have faster approximate modes, and I think for mainstream ones only when you explicitly enable such optimizations, which are disabled by default.