You're reading an pre-release version of this documentation.
For the latest stable release version, please have a look at master.

Boot a simulation


Below is an example hardware definition + testbench:

import spinal.core._

// Identity takes n bits in a and gives them back in z
class Identity(n: Int) extends Component {
  val io = new Bundle {
    val a = in Bits(n bits)
    val z = out Bits(n bits)

  io.z := io.a
import spinal.core.sim._

object TestIdentity extends App {
  // Use the component with n = 3 bits as "dut" (device under test)
  SimConfig.withWave.compile(new Identity(3)).doSim{ dut =>
    // For each number from 3'b000 to 3'b111 included
    for (a <- 0 to 7) {
      // Apply input #= a
      // Wait for a simulation time unit
      // Read output
      val z =
      // Check result
      assert(z == a, s"Got $z, expected $a")


SimConfig will return a default simulation configuration instance on which you can call multiple functions to configure your simulation:




Enable simulation wave capture (default format)


Enable simulation wave capture (VCD text format)


Enable simulation wave capture (FST binary format)


Specify the SpinalConfig that should be use to generate the hardware


Enable all the RTL compilation optimizations to reduce simulation time (will increase compilation time)


Change the folder where the sim files are generated


Use Verilator as simulation backend (default)


Use GHDL as simulation backend


Use Icarus Verilog as simulation backend


Use Synopsys VCS as simulation backend

Then you can call the compile(rtl) function to compile the hardware and warm up the simulator. This function will return a SimCompiled instance.

On this SimCompiled instance you can run your simulation with the following functions:

doSim[(simName[, seed])]{dut => /* main stimulus code */}

Run the simulation until the main thread runs to completion and exits/returns. It will detect and report an error if the simulation gets fully stuck. As long as e.g. a clock is running the simulation can continue forever, it is therefore recommended to use SimTimeout(cycles) to limit the possible runtime.

doSimUntilVoid[(simName[, seed])]{dut => ...}

Run the simulation until it is ended by calling either simSuccess() or simFailure(). The main stimulus thread can continue or exit early. As long as there are events to process, the simulation will continue. The simulation will report an error if it gets fully stuck.

The following testbench template will use the following toplevel :

class TopLevel extends Component {
   val counter = out(Reg(UInt(8 bits)) init (0))
   counter := counter + 1

Here is a template with many simulation configurations:

val spinalConfig = SpinalConfig(defaultClockDomainFrequency = FixedFrequency(10 MHz))

  .compile(new TopLevel)
  .doSim { dut =>
    // Simulation code here

Here is a template where the simulation ends by completing the simulation main thread execution:

SimConfig.compile(new TopLevel).doSim { dut =>
  dut.clockDomain.waitSamplingWhere(dut.counter.toInt == 20)

Here is a template where the simulation ends by explicitly calling simSuccess():

SimConfig.compile(new TopLevel).doSimUntilVoid{ dut =>
  fork {
    dut.clockDomain.waitSamplingWhere(dut.counter.toInt == 20)

Note is it equivalent to:

SimConfig.compile(new TopLevel).doSim{ dut =>
  fork {
    dut.clockDomain.waitSamplingWhere(dut.counter.toInt == 20)
  simThread.suspend() // Avoid the "doSim" completion

Note that by default, the simulation files will be placed into the simWorkspace/xxx folders. You can override the simWorkspace location by setting the SPINALSIM_WORKSPACE environment variable.

Running multiple tests on the same hardware

val compiled = SimConfig.withWave.compile(new Dut)

compiled.doSim("testA") { dut =>
   // Simulation code here

compiled.doSim("testB") { dut =>
   // Simulation code here

Throw Success or Failure of the simulation from a thread

At any moment during a simulation you can call simSuccess or simFailure to end it.

It is possible to make a simulation fail when it is too long, for instance because the test-bench is waiting for a condition which never occurs. To do so, call SimTimeout(maxDuration) where maxDuration is the time (in simulation units of time) after the which the simulation should be considered to have failed.

For instance, to make the simulation fail after 1000 times the duration of a clock cycle:

val period = 10
SimTimeout(1000 * period)

Capturing wave for a given window before failure

In the case you have a very long simulation, and you don’t want to capture the wave on all of it (too bug, too slow), you have mostly 2 ways to do it.

Either you know already at which simTime the simulation failed, in which case you can do the following in your testbench :


Or you can run a dual lock-step simulation, with one running a bit delayed from the the other one, and which will start recording the wave once the leading simulation had a failure.

To do this, you can use the DualSimTracer utility, with parameters for the compiled hardware, the window of time you want to capture before failure, and a seed.

Here is an example :

package spinaldoc.libraries.sim

import spinal.core._
import spinal.core.sim._
import spinal.lib.misc.test.DualSimTracer

class Toplevel extends Component{
  val counter = out(Reg(UInt(16 bits))) init(0)
  counter := counter + 1

object Example extends App {
  val compiled = SimConfig.withFstWave.compile(new Toplevel())

  DualSimTracer(compiled, window = 10000, seed = 42){dut=>
      val value = dut.counter.toInt

      if(value % 0x1000 == 0){
        println(f"Value=0x$value%x at ${simTime()}")

      // Throw a simulation failure after 64K cycles
      if(value == 0xFFFF){

This will generate the following file structure :

  • simWorkspace/Toplevel/explorer/stdout.log : stdout of the simulation which is ahead

  • simWorkspace/Toplevel/tracer/stdout.log : stdout of the simulation doing the wave tracing

  • simWorkspace/Toplevel/tracer.fst : Waveform of the failure

The scala terminal will show the explorer simulation stdout.