pub fn residual_add(
gpu: &dyn GpuBackend,
kernel: KernelHandle,
residual: DevicePtr,
src: DevicePtr,
num_elements: u32,
stream: u64,
) -> Result<()>Expand description
2026-09-25: In-place BF16 residual[i] += src[i] (kernel bf16_residual_add).