The tension between fmax and Tco in a specialized processing block of a programmable integrated circuit device can be reduced by providing variable delays on the clock inputs of the pipeline registers within the specialized processing block. This allows the introduction of beneficial skew that allows slower functions to be performed within the specialized processing block rather than outside the block, thereby reducing Tco, without slowing down the clock—i.e., without reducing fmax. This technique may also apply to other specialized blocks such as memory.