In AMD 7 series FPGAs, there are no local inverters on control signals (resets or clock enables). The following description uses a reset as the example, but the same applies for clock enables.
If a design uses an active-Low reset, a LUT must be used to invert the signal. In non-DFX designs that use all active-Low resets multiple LUTs are inferred but can be combined into a single LUT and pushed into the I/O elements (the LUT goes away). In non-DFX designs that use a mix of High and Low, the LUT inverters can be combined into one LUT that remains in the design, but that has minimal effect on routing and the timing of the reset net (output of LUT can still be put on global resources). However, for a design that uses active-Low resets on a partition, it is possible to have inverters inferred inside the partition that cannot be pulled out and combined. This makes it impossible to put the reset on global resources, and can lead to poor reset timing and to routing issues if the design is already congested.
The best way to avoid this is to avoid using active-Low control signals. However, there are cases where this is not possible (for example, when using an IP core with an Advanced eXtensible Interface (AXI) interface). In these cases the design should assign the active-Low reset to a signal at the top level, and use that new signal everywhere in the design.
As an example:
reset_n <= !reset;
reset_n signal for all cases, and do not use the
!reset assignments on signals or ports.
This ensures that a LUT is inferred only for the reset net for the whole design and has a minimal effect on design performance.