← Documentation index

Adding a new built-in target

This guide describes how to manually add support for a new target and/or board to pyOCD. In most cases you do not need to add a builtin target anymore, and can use pyOCD’s support for CMSIS Device Family Packs.

For background information, review the architecture overview document first. The CMSIS Pack documentation may also be helpful.

Device Family Pack intro

The instructions below assume you have a CMSIS Device Family Pack (DFP) available for your target. See the list of all publicly available Packs to find and download the DFP for your target.

A DFP is simply a zip file with a .pack extension. To extract the contents you can change the extension to .zip and extract with your favourite archive utility.

For this context, the most important thing inside the DFP are .FLM files that contain the flash programming algorithms used in step 5 below. The .pdsc file can also be useful. It is an XML file that contains details such as the memory map for the target devices described by the DFP.

Steps to add a new target

  1. Create a new CoreSightTarget subclass in a file under pyocd/target/builtin/. You can copy one of the existing target files like pyocd/target/builtin/target_ncs36510.py and rename the class.

    The target source file name must follow the pattern “target_<device>.py”, where “<device>” is the device’s Dname or Dvariant part number value from the appropriate CMSIS Device Family Pack (DFP). For instance, target_LPC54608J512ET180.py. You may substitute an “x” for certain fields in the part number, such as a package or pin count code, temperature code, or memory size (if multiple memory sizes are supported via classes within the one source file). For instance, target_STM32F412xx.py. If the device doesn’t have a DFP, then use a similar, complete part number.

  2. Set the VENDOR class attribute on the CoreSightTarget subclass. The vendor name must be one of the standard values defined by the DeviceVendorEnum type for CMSIS Packs. (If the vendor is not listed, please contact Arm.)

  3. You may optionally add PART_NUMBER and/or PART_FAMILIES class attributes to your target class. T

  4. Create the target’s memory map from information in the device’s reference manual. The memory map should be a MEMORY_MAP class attribute of the target class. Modifying an existing memory map is easiest, and there are many examples in the other targets.

  5. To create the flash algo, the recommended method is to use the scripts/generate_flash_algo_.py script included in the pyocd repo. This script will generate an output file in the form required for pyocd from an .FLM file that is included as part of a CMSIS DFP.

    1. Locate the correct .FLM file from the DFP for your target.

    2. Run scripts/generate_flash_algo_.py \<path to FLM>. It will write the output files to the working directory from where you called the script.

    3. The pyocd_blob.py output file contains the Python code for the flash algo. Copy the FLASH_ALGO_* dictionary (the name will have a suffix based on the FLM name) into the target source file.

    4. Review the addresses in the flash_algo dictionary to make sure they are valid. The memory layout should look like:

      |----------------|-------------|------------|-----|-----------------|-----------------|
|  load_address  | static_base | << (stack) | ... | page_buffers[0] | page_buffers[1] |
|----------------|-------------|------------|-----|-----------------|-----------------|
^                                           ^     ^
RAM start            begin_stack (grows down)     also begin_data

      Each of the addresses in the page_buffers list points to a buffer of the maximum page size of any flash region. If there is not enough RAM to hold two page buffers, then remove one of the addresses from the list. This will disable double buffered flash programming.

    5. To enable efficient scanning for modified pages via CRC checking, you can set the analyzer_supported key to True and the analyzer_address to the start address for an unused range of (1224 + 4 * number-of-flash-pages) bytes of RAM.

    6. Pass the FLASH_ALGO_* dict for the algo parameter to the FlashRegion constructor in your memory map. This binds the flash algo to that flash memory region.

  6. If the target has multiple flash algos for different flash types, repeat step 5 as necessary.

  7. Edit pyocd/target/builtin/__init__.py to import your target source file and add your new target to the BUILTIN_TARGETS dict.

  8. You or your employer own the copyright on the new code, so make sure you set the copyright on the new target file. You should also add a copyright to any existing files you modified.

Now your new target is available for use via the --target command line option! You can test its availability by running pyocd list --targets --name <target-name>.

Steps to add a new board

This section only applies if your board has an on-board debug probe that either:

  • Uses the Arm DAPLink firmware. DAPLink presents the board ID as the first 4 characters of the USB serial number.
  • Uses the STLinkV2-1 or STLinkV3 firmware and the board is Mbed-enabled. STLink presents the board ID as the first 4 characters of the code in the HTML file on the USB mass storage volume.

If neither applies, then pyOCD will be unable to automatically detect the board type. However, you can still use the target by passing the --target argument to pyOCD.

Follow these steps:

  1. Identify the 4-character board ID.

  2. Insert a row into the BOARD_ID_TO_INFO table in pyocd/board/board_ids.py with the board ID, board name, target type, and test binary file name.

    The new row should look similar to this:

     "0205": BoardInfo(  "FRDM-KL28Z",           "kl28z",            "l1_kl28z.bin",         ),

    Be sure to insert the row in sorted order by board ID, and please align columns.

  3. Place a test firmware binary file listed in the board info into the test/data/binaries/ directory. The test firmware can be nothing more than a tiny LED blinky demo. It must not require any user input, and should provide immediate visual feedback that the code is successfully running, assuming there are LEDs on the board.