For the complete documentation index, see llms.txt. Markdown versions of all pages are available by appending .md to any URL (e.g. /get-started.md).

Mojo function

cast_fp_to_fp4e2m1

def cast_fp_to_fp4e2m1[dtype: DType, width: SIMDLength, //](x: SIMD[dtype, width]) -> SIMD[dtype, width]

Rounds each floating-point element to the nearest FP4 E2M1 representable value.

Implements nearest-even rounding into the 16 representable FP4 E2M1 values: {±0, ±0.5, ±1, ±1.5, ±2, ±3, ±4, ±6}. Values outside [−6, 6] saturate to ±6. The result is returned in the original dtype, not packed as nibbles.

Parameters:

• ​dtype (DType): Floating-point element type (float32, bfloat16, or float16).
• ​width (SIMDLength): SIMD lane count.

Args:

• ​x (SIMD[dtype, width]): Input floating-point vector.

Returns:

SIMD[dtype, width]: Vector of the same dtype with each element rounded to the nearest FP4 E2M1 value.