Intel® Simics® Simulator for Intel® FPGAs: User Guide

ID 784383
Date 4/01/2024

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents Branch Synchronization

The execution of the code inside of a branch is suspended once it reaches wait-for command and continues until the condition defined in the command is satisfied. This allows synchronizing the branch execution with an event that occurs in the simulation. The condition can be a specific value written or read in a register, a breakpoint trigger, elapse a certain amount of time, execute a certain number of steps or cycles, or also a specific string printed on the Intel® Simics® simulator CLI.

The create-script-barrier command creates a barrier, which is a mechanism that allows the synchronization in several branches in the target script. This command receives as a parameter a number that defines the number of branches that need to reach the barrier location. The location in the branches is defined using the wait-for-script-barrier command that receives as parameter the return value from the create-script-barrier.

The following figure describes the functionality of this synchronization mechanism. Here, initially the barrier is created with the create-script-barrier command, indicating the number of branches that requires to be synchronized. In each one of the branches to synchronize, the wait-for-script-barrier command is used, and the branch execution is suspended when they get to this command and resumed until the number of branches defined reaches to this barrier. After this, the branches execution resumes.

Figure 11. Description of Branch Synchronization

The number of the branches that need to reach the barrier can be updated using the script-barrier-limit command.

The following example shows an use of the description synchronization. This example starts creating the barrier in which the number of branches to synchronize is 2. The ID of this barrier is the $barrier_id variable. Later, two branches are defined in which some cycles are executed before setting the barrier. The number of cycles is different in both branches (500 and 200) and the objective is that after the branch sync, both branches get synchronized resuming the execution at the same time (after 500 cycles). This is achieved as this can be seen from the output log after printing some messages during the execution of the script.
# CLI script to demonstrate branch synchronization
$barrier_id = (create-script-barrier 2)
$time_org = (ptime -c)
echo "Setting barrier at time " + $time_org + " ID: " + $barrier_id
script-branch {
     $timeB1Pre = (ptime -c)
     echo "Branch1: @" + $timeB1Pre + " will get to barrier in 500 cycles"
     bp.cycle.wait-for 500
     $timeB1Pre = (ptime -c)
     echo "Wait in branch 1 at: " + $timeB1Pre
     wait-for-script-barrier $barrier_id 
     $timeB1Post = (ptime -c)
     echo "Sync in branch 1 at: " + $timeB1Post

script-branch {
     $timeB2Pre = (ptime -c)
     echo "Branch2: @" + $timeB2Pre + " will get to barrier in 200 cycles"
     bp.cycle.wait-for 200
     $timeB2Pre = (ptime -c)
     echo "Wait in branch 2 at: " + $timeB2Pre
     wait-for-script-barrier $barrier_id 
     $timeB2Post = (ptime -c)
     echo "Sync in branch 2 at: " + $timeB2Post

# Output log from the script execution
Setting barrier at time 0 ID: 1
Branch1: @0 will get to barrier in 500 cycles
Branch2: @0 will get to barrier in 200 cycles
Wait in branch 2 at: 200
Wait in branch 1 at: 500
Sync in branch 1 at: 500
Sync in branch 2 at: 500