MicroZed Chronicles: IOB Register Packing
Webinar 1/12/26: Accelerating FPGA & DSP Design with AMD Vitis™ Model Composer
Webinar 3/12/26: Designing a Low-Cost High-Performance FPGA Board
Last week, while running my FPGA development course, we had an interesting discussion about timing closure, constraints and how we ensure external devices such as memories meet their timing requirements.
This is important because we need to consider several elements, including the FPGA timing, PCB trace delays and the timing requirements of the external device. We cannot confidently say the interface will work until we have completed the system-level timing analysis.
We then started talking about techniques which can help us achieve timing closure. One of these techniques is register packing in the IOB, which I thought would make an interesting blog.
I mentioned memories, but timing closure is equally important when interfacing with ADCs, DACs and parallel data buses connected to processors.
When we drive a signal to an external device, the timing that matters is at the FPGA pin and not at the register. If the output register is located in a slice, each bit has its own route from the slice to the I/O buffer. This routing can add a nanosecond or more to the clock-to-output delay. Worse still, the routing delay can be different for each bit and may change when we rerun implementation.
Similarly, when receiving data at the FPGA, the timing that matters is when the data is captured by the register. The longer the route from the input pin to the register, the greater the impact on our timing closure.
One solution is simple, place the register as close as possible to the I/O pin. This minimizes the routing between the register and the I/O buffer for both input and output paths.
When working with 7 Series devices, each I/O has access to dedicated ILOGIC and OLOGIC resources located beside the I/O buffer. These provide input, output and three-state control registers.
Using these dedicated registers reduces the routing delay and also helps reduce timing differences between bits and variations caused by placement and routing changes between builds.
If we want to use the IOB registers, we can instruct Vivado to pack the registers into the I/O logic. By default, Vivado does not necessarily pack these registers automatically.
We can instruct Vivado to use the IOB registers by applying an RTL attribute to the register signal.
attribute IOB : string;
attribute IOB of ra : signal is "TRUE";Alternatively, we can define this in the XDC constraints file. We can apply the property to either the port or the register cell.
set_property IOB TRUE [get_ports {pad_a[*]}]
set_property IOB TRUE [get_cells {ra_reg[*]}]here may be cases where Vivado is unable to pack the register into the IOB. When this happens, Vivado can issue a critical warning, so make sure you review these messages and check the implemented design.
Typical reasons for IOB register packing to fail include additional combinational logic between the register and the output pin, the register output also being used within the FPGA fabric, or control signals which are incompatible with the IOB register resources.
One thing to remember is that once we have instructed Vivado to use the IOB register, the placer will not move it back into the fabric to help close an internal timing path. The IOB constraint takes precedence.
This is a good example of where the UltraFast design methodology can help. Registering module inputs and outputs and using additional register stages where appropriate can help with placement and timing closure. Of course, any additional register stages need to be considered in terms of latency and overall design requirements.
As always, it is good to demonstrate the benefits of a technique, so I created a simple design showing examples where IOB packing works and where the RTL structure can prevent it.
The sample design contains five examples.
First example: The register drives the output pad directly and nothing else, so it should be packed into the IOB register.
Second example: The output of the register passes through additional combinational logic before reaching the pad. This prevents the register from being packed into the IOB.
Third example: The register drives the output pad but is also used internally within the FPGA fabric. This can create conflicting placement requirements, depending on how Vivado implements the design.
Fourth example: This demonstrates how to correct the issue in the second example by moving the combinational logic before the register, allowing successful IOB packing.
Fifth example: This demonstrates how register duplication can be used to separate the external output path from the internal fabric path. Additional constraints are used to prevent synthesis from optimizing away the duplicated registers.
To see the impact of IOB packing on performance, the table below shows the clock-to-pad timing for each of the examples.
Bus | Coding style | Placement | Clock-to-pad | Spread |
pad_a | Register drives pad only | OLOGIC OUTFF | 3.991 ns | 0.0051 ns |
pad_b | Gate after the register | Slice FF + LUT | 6.334 ns | 0.726 ns |
pad_c | Register also used internally | OLOGIC | 4.005 ns | 0.032 ns |
pad_d | Gate moved before the register | OLOGIC OUTFF | 4.046 ns | 0.049 ns |
pad_e | Register duplicated | OLOGIC OUTFF | 4.055 ns | 0.039 ns |
Looking at the results, we can see the difference between the first and second examples.
With the register packed into the OLOGIC, the clock-to-pad delay is 3.991 ns. When the register is placed in the fabric with additional combinational logic, this increases to 6.334 ns, an increase of 2.343 ns.
The OLOGIC path example is shown below.


Failure to use IOB Register

Of course, the second example also includes an additional LUT, so not all of this difference is due to register placement alone.
What is also interesting is the difference in timing spread across the bus. The first example has a spread of just 5.1 ps, while the second has a spread of 726 ps.
This is important when working with parallel interfaces, where we need to ensure all the data bits meet the timing requirements of the receiving device.
Examples four and five show how relatively simple changes to the RTL can allow the registers to be packed into the OLOGIC and achieve timing similar to the first example.
The third example is also interesting because the timing is close to the successfully packed examples. We need to check the implemented placement to see whether Vivado packed or duplicated the register.
One thing to remember is that IOB packing does not remove the need for proper input and output timing constraints. We still need to consider the external device's setup and hold requirements, PCB delays and clock relationships. For input paths, reducing routing delay can also affect hold timing.
In this blog, everything has been done targeting a 7 Series device. However, the same approach can be used with UltraScale and UltraScale+ devices, although the I/O architectures differ slightly.
Hopefully, this demonstrates a little more about the benefits of using IOB register packing, along with some of the rules and considerations when using it.
The example source code, constraints and scripts are available on my GitHub.
FPGA Journal
Read about cutting edge FPGA developments, in the FPGA Horizons Journal or contribute an article.
Workshops and Webinars:
If you enjoyed the blog why not take a look at the free webinars, workshops and training courses we have created over the years. Highlights include:
Upcoming Webinars Timing, RTL Creation, FPGA Math and Mixed Signal
Professional PYNQ Learn how to use PYNQ in your developments
Introduction to Vivado learn how to use AMD Vivado
Ultra96, MiniZed & ZU1 three day course looking at HW, SW and PetaLinux
Arty Z7-20 Class looking at HW, SW and PetaLinux
Mastering MicroBlaze learn how to create MicroBlaze solutions
HLS Hero Workshop learn how to create High Level Synthesis based solutions
Perfecting Petalinux learn how to create and work with PetaLinux OS
Boards
Get an Adiuvo development board:
Adiuvo Embedded System Development board - Embedded System Development Board
Adiuvo Embedded System Tile - Low Risk way to add a FPGA to your design.
SpaceWire CODEC - SpaceWire CODEC, digital download, AXIS Interfaces
SpaceWire RMAP Initiator - SpaceWire RMAP Initiator, digital download, AXIS & AXI4 Interfaces
SpaceWire RMAP Target - SpaceWire Target, digital download, AXI4 and AXIS Interfaces
Embedded System Book
Do you want to know more about designing embedded systems from scratch? Check out our book on creating embedded systems. This book will walk you through all the stages of requirements, architecture, component selection, schematics, layout, and FPGA / software design. We designed and manufactured the board at the heart of the book! The schematics and layout are available in Altium here. Learn more about the board (see previous blogs on Bring up, DDR validation, USB, Sensors) and view the schematics here.
All words in this blog were written by a human.




