MCPcopy Create free account
hub / github.com/MegEngine/MegEngine / fdim

Method fdim

dnn/include/megdnn/dtype/bfloat16.hpp:774–780  ·  view source on GitHub ↗

Source from the content-addressed store, hash-verified

772 /// \param y second operand
773 /// \return Positive difference stored in single-precision
774 MEGDNN_HOST MEGDNN_DEVICE static expr fdim(float x, float y) {
775#if defined(__CUDA_ARCH__)
776 return expr(fdimf(x, y));
777#else
778 return expr(std::fdim(x, y));
779#endif
780 }
781
782 /// Fused multiply-add implementation.
783 /// \param x first operand

Callers

nothing calls this directly

Calls 2

exprClass · 0.70
fdimFunction · 0.70

Tested by

no test coverage detected