MatmulPlanPreferences#

class nvmath.linalg.advanced.MatmulPlanPreferences( reduction_scheme_mask: ReductionScheme | None = ReductionScheme.MASK, max_waves_count: float | None = 0.0, numerical_impl_mask: MatmulNumericalImplFlags | None = MatmulNumericalImplFlags.ALL, limit: int = 8, epilog: MatmulEpilogPreferences | None = None, )[source]#

A data class for providing options to the Matmul.plan() method and the wrapper function matmul().

reduction_scheme_mask#

Enumerators from nvmath.linalg.advanced.MatmulReductionScheme combined with bitwise operator |. The default is all reduction schemes.

Type:: nvmath.linalg.advanced.MatmulReductionScheme

max_waves_count#

The maximum wave count. Selecting a value greater than 0. will exclude algorithms with device utilization greater than specified. The default is 0.

Type:: float

numerical_impl_mask#

Enumerators from nvmath.nvmath.linalg.advanced.MatmulNumericalImplFlags combined with bitwise operator |. The default is all numerical implementation flag choices.

Type:: nvmath.linalg.advanced.MatmulNumericalImplFlags

limit#

The number of algorithms to consider. If not specified, a suitable default will be chosen.

Type:: int

epilog#

Epilog preferences (as an object of class MatmulEpilogPreferences or a dict).

Type:: nvmath.linalg.advanced.MatmulEpilogPreferences