Class EnableCHWN4Pass

Inheritance Relationships

Base Type

Class Documentation

class mgb::gopt::EnableCHWN4Pass : public mgb::gopt::TensorReformatPass

enable using chwn4 tensor format on Nvidia Platform with compute capability 6.1 or later

Public Functions

const char *name() const override

Public Static Functions

std::unique_ptr<EnableCHWN4Pass> make_chwn4_converter()

make nchw4 -> chwn4 converter opt pass