Function dispatch_cast_bf16_to_f32_with_encoder 

pub fn dispatch_cast_bf16_to_f32_with_encoder(
    encoder: &mut CommandEncoder,
    registry: &mut KernelRegistry,
    device: &DeviceRef,
    input: &MlxBuffer,
    output: &MlxBuffer,
    n_elements: u32,
) -> Result<()>

Cast bf16 to f32 using an externally-provided encoder (no commit).

Encodes the cast_bf16_to_f32 kernel into the given encoder without committing or waiting. Use this to chain the cast into a mega-encoder alongside other GPU work, avoiding CPU round-trips.

Arguments

• encoder - Command encoder to record the dispatch into.
• registry - Kernel registry (mutable for lazy pipeline compilation).
• device - Metal device for pipeline compilation.
• input - Input buffer (bf16).
• output - Output buffer (f32, pre-allocated with n_elements * 4 bytes).
• n_elements - Number of elements to cast.

Errors

Returns MlxError::InvalidArgument if n_elements is zero or buffers are too small.