Mojo struct
FastMathFlag
@register_passable
struct FastMathFlag
Flags for controlling fast-math optimizations in floating-point operations.
FastMathFlag provides compile-time controls for various floating-point math optimization modes that trade strict IEEE 754 compliance for performance.
Available flags:
NONE: No fast-math optimizations.NNAN: Assume operands and results are not NaN.NINF: Assume operands and results are not +/- infinity.NSZ: Treat the sign of a zero as insignificant.ARCP: Allow reciprocal of values.CONTRACT: Allow floating-point contraction (e.g., fused multiply-add).AFN: Allow algebraic function approximations.REASSOC: Allow reassociation of floating-point operations.FAST: Enable all fast-math optimizations.
Examples:
# Use contract flag for fused multiply-add
var result = value.fma[FastMathFlag.CONTRACT](multiplier, accumulator)
# Use fast flag for maximum optimization
var fast_result = value.fma[FastMathFlag.FAST](multiplier, accumulator)Implemented traits
AnyType,
Copyable,
ImplicitlyCopyable,
Movable,
UnknownDestructibility
Aliases
__copyinit__is_trivial
alias __copyinit__is_trivial = True
__del__is_trivial
alias __del__is_trivial = True
__moveinit__is_trivial
alias __moveinit__is_trivial = True
AFN
alias AFN = FastMathFlag(6)
ARCP
alias ARCP = FastMathFlag(4)
CONTRACT
alias CONTRACT = FastMathFlag(5)
FAST
alias FAST = FastMathFlag(8)
NINF
alias NINF = FastMathFlag(2)
NNAN
alias NNAN = FastMathFlag(1)
NONE
alias NONE = FastMathFlag(0)
NSZ
alias NSZ = FastMathFlag(3)
REASSOC
alias REASSOC = FastMathFlag(7)
Methods
__is__
__is__(self, other: Self) -> Bool
Compares two FastMathFlag values for identity.
Args:
- other (
Self): The FastMathFlag to compare against.
Returns:
Bool: True if both flags have the same value, False otherwise.
Was this page helpful?
Thank you! We'll create more content like this.
Thank you for helping us improve!