What Does Vivado Do? Synthesis to Bitstream, Step by Step
Vivado reads your Verilog and your constraints (an XDC file), synthesizes the RTL into a netlist of LUTs, flip-flops and other primitives, optimizes it, places each primitive on a site in the chip, routes the wires between them, reports timing and resource use, and writes a bitstream that configures the FPGA. You can drive it through the project GUI or a short Tcl script in batch mode; the script below built every Basys 3 and Arty A7 project on this site.
What you'll learn
- Each step from RTL to bitstream
- Project mode and non-project (Tcl) mode
- Which reports to read, and what WNS means
- How to rebuild a design from a script
The script
# Vivado non-project build of one design: synth, place, route, timing and
# utilization reports, and a bitstream (which also runs the DRCs).
# vivado -mode batch -source build.tcl -tclargs <top> <part> <xdc> <out_dir> <verilog files...>
set top [lindex $argv 0]
set part [lindex $argv 1]
set xdc [lindex $argv 2]
set out [lindex $argv 3]
file mkdir $out
foreach f [lrange $argv 4 end] { read_verilog $f }
read_xdc $xdc
synth_design -top $top -part $part
opt_design
place_design
route_design
report_timing_summary -file $out/timing.rpt
report_utilization -file $out/util.rpt
write_bitstream -force $out/$top.bit
Run it as vivado -mode batch -source build.tcl -tclargs top part xdc out_dir files. The project's build_all.sh calls it once per design and pulls the worst slack and the resource counts from the reports.
What each step does
| Step | Does | Watch for |
|---|---|---|
| synth_design | Maps RTL to LUTs, flip-flops, block RAM and DSP; infers memories and arithmetic | Latch warnings, removed logic, unconnected ports |
| opt_design | Removes redundant logic and merges constants | Logic you needed being removed because nothing reads it |
| place_design | Chooses a site for every primitive | Congestion on large designs |
| route_design | Connects every net through the routing | Unroutable nets, timing getting worse |
| report_timing_summary | Checks every path against the clock constraints | Negative WNS or TNS |
| write_bitstream | Runs design rule checks, then writes the .bit file | Unconstrained I/O, missing IOSTANDARD |
Results for the beginner projects
| Design | Board | WNS at 100 MHz | LUTs | Flip-flops |
|---|---|---|---|---|
| Blink | Basys 3 | 5.949 ns | 7 | 27 |
| Seven-segment counter | Basys 3 | 5.496 ns | 42 | 62 |
| UART echo | Basys 3 | 5.753 ns | 73 | 80 |
| VGA color bars | Basys 3 | 6.232 ns | 36 | 46 |
| Blink | Arty A7 | 6.056 ns | 7 | 27 |
| Counter on LEDs | Arty A7 | 5.202 ns | 23 | 39 |
| UART echo | Arty A7 | 6.058 ns | 73 | 76 |
Every design met timing with more than 5 ns to spare out of the 10 ns period. The blink design's 27 flip-flops are its 26-bit counter, which counts to 49,999,999, plus the LED output.
Project mode vs non-project mode
Project mode (the GUI) keeps runs, reports and the block design in a project directory and is easiest for IP and block designs such as a Zynq system. Non-project mode keeps everything in memory and in your script, so builds are reproducible and fit version control. Many teams use the GUI to explore and a script to build.
Common mistakes
- Ignoring synthesis warnings; most bugs show up there first.
- Reading the timing report after synthesis only. Routing adds delay, so check it after route_design.
- Leaving a port without a pin or IOSTANDARD; write_bitstream stops with a DRC error.
- Building only in the GUI, so nobody can reproduce the result.
Check your understanding
- Q1. Which step chooses where each LUT goes? Answer: place_design.
- Q2. What does a positive WNS mean? Answer: the slowest path still meets the clock with that much time to spare.
- Q3. Why does the blink design need 26 counter bits at 100 MHz? Answer: it counts to 49,999,999, and 2^26 = 67,108,864 is the smallest power of two above 50 million.
The FPGA series
Last verified
All figures are from Vivado 2025.2 runs of scripts/fpga/build_all.sh on 17 November 2026.