Step 1: Prepare Your DSL DAG Projec

Make sure your DSL DAG project has been properly generated. The following files are required for integration with the L1 simulator:

• The compiled DAG binary image: <project_path>/dsl/bin/<dag_name>.bin
• The DAG metadata file: <project_path>/dsl/final_output/<dag_name>.json
• The individual task binary images: <project_path>/dsl/venus_test/ir/<task_name>.hex

These files will be used by the L1 simulator for joint debugging and integration.

Step 2: Write the L1 Runtime Code

2.1 Acquire the input and output sequence of the DAG by running dagInfoPaser script:

cd <project_path>/l1_main && python ./dagInfoPaser.py <dag_name>

2.2 Retrieve the order of input and output variables from the file:

<project_path>/l1_main/daginfo/<dag_name>.txt

2.3 Update <dag_name>/source/l1.cpp to declare the input and output variables for your DAG. Ensure that the variable types and lengths strictly match those defined in the associated .bas file.

Then, establish the DAG-to-DAG dependencies to define execution order and data flow.

For example, suppose the following input and output variables are declared in your .bas file:

' input parameter
  dfedata char dfe_input[4388]
  dag_input short nCellid[1]

' output parameter
  return_value short subFrameNum[1]

Suppose your DAG’s input and output variable names are defined as follows:

# Input Names
nCellid
dfe_input
# Return Output Names
subFrameNum

You can then invoke the DAG in l1.cpp as follows:

• fire_dag API Parameters:

You can use the fire_dag function to instruct L1 to deploy and run the specified DAG for computation.

#include "l1_api.h"
char dfe_input[4388] = {...};
short nCellid[1] = {2};
short subFrameNum[1];

void l1_process() {
  fire_dag(dag_name, 2, 1, &dfe_input, &nCellid, &subFrameNum);

  // You can connect the output of this DAG as the input
  // to the next DAG to represent data flow and dependencies between DAGs.
}