AM243x MCU+ SDK  08.04.00
Understanding the bootflow and bootloaders


Booting user defined applications on a SOC involves multiples steps as listed below,

  • Firstly, there are multiple steps involved to convert a user application, created using a compiler+linker toolchain, into a binary format that is suitable to be booted by the SOC
  • Next, we need to flash this binary to the EVM flash
  • Finally, when the SOC is powered on, the previously flashed binary is executed.
  • After powering on the EVM, the bootflow takes place mainly in two steps
    • ROM boot, in which the ROM bootloader boots a secondary bootloader or an SBL
    • SBL boot in which the secondary bootloader boots the application
  • Note, that a system application itself can consist of multiple CPU specific application binaries that all collaborate together to realize the overall system goal.

This section details these steps and gives an overview of the bootloaders to understand the process better.

This section also describes the steps to enable XIP (eXecute In Place) for your applications.

Additional References

See also these additional pages for more details and examples about the boot flow,

Preparing the application for boot

To see the exact sequence of steps in which applications and secondary bootloader (SBL) are converted from compiler generated .out files to boot images, see the makefile makefile_ccs_bootimage_gen that is included in every example and secondary bootloader (SBL) CCS project.
If you are using makefile based build, then see the file named makefile in the example folder.

Shown below are the different steps that are done to convert the compiler+linker generated application .out into a format suitable for flashing and booting

  • For each CPU, the compiler+linker toolchain is used to create the application .out "ELF" file which can be loaded and run via CCS
  • The below "post build" steps are then used to convert the application .out into a "flash" friendly format
    • For each CPU, out2rpc is used to convert the ELF .out to a binary file containing only the loadable sections. This is called a RPRC file.
    • For each CPU, xipGen is used to split this RPRC file into two RPRC files.
      • One RPRC, containing the section that during boot need to be loaded to RAM
      • Second RPRC, containing the section that during boot are not loaded to RAM but are instead "eXecuted In Place", i.e XIP
    • multiCoreGen is then used to combine all the non-XIP RPRC files per CPU into a single .appimage file which is a concatenation of the individual CPU specific RPRC files.
    • multiCoreGen is used again to combine all the XIP RPRC files per CPU into a single .appimage_xip file which is a concatenation of the individual CPU specific RPRC XIP files.
  • This .appimage and .appimage_xip is then flash to the EVM

Post build steps

Flashing the application for boot

  • Once the application images (.appimage and .appimage_xip) are created one needs to copy or flash these to a supported boot media so that the application can start executing once the SOC is powered ON
  • When flashing the application we also need to flash a bootloader or SBL image.
  • See Flashing Tools for detailed steps that are done to flash a user application

Booting the application

After a SBL and application image is flashed, shown below is the high level boot flow, after the SOC is powered on.


ROM Boot

  • As soon as the EVM is powered ON, the ROM bootloader or RBL starts running. The RBL is the primary bootloader.
  • Depending on which boot mode is selected on the EVM, the RBL will load the secondary bootloader or SBL from a boot media (OSPI flash, SD card or via UART).
  • Rest of the booting is done by the SBL.
  • The RBL expects the image it boots (SBL in our case) to always be signed. Refer Booting Tools for more information on signing scripts.

SBL Boot

  • The SBL is essentially an example application of the bootloader library.
  • We call it a secondary bootloader because it is booted by the RBL, which is the primary bootloader.
  • An SBL typically does a bunch of SOC specific initializations and proceeds to the application loading.
  • In case of AM243X EVM, the SBL loads the SYSFW to the Cortex M3 and sends the board cfg to the SYSFW once M3 is booted.
  • Depending on the type of SBL loaded, SBL looks for the multicore appimage (refer Booting Tools for more on multicore appimage) of the application binary at a specified location in a boot media.
  • If the appimage is found, the multicore appimage is parsed into multiple RPRCs. These are optimized binaries which are then loaded into individual CPUs.
  • Each RPRC image will have information regarding the core on which it is to be loaded, entry points and multiple sections of that application binary
  • The SBL uses this information to initialize each core which has a valid RPRC. It then loads the RPRC according to the sections specified, sets the entry points and releases the core from reset. Now the core will start running


Secondary Bootloaders

Depending on the boot media from which we load the application binary, we have multiple SBLs like sbl_ospi,sbl_uart etc. A bare minimum SBL called the sbl_null is also included which aids the users to load their applications via CCS.


  • The sbl_null is a secondary bootloader which doesn't load any application binary, but just does the SOC initialization and puts all the cores in WFI (Wait For Interrupt) mode.
  • This is referred to as the SOC initialization binary, refer Flash SOC Initialization Binary for more on this.


  • The sbl_sd is a secondary bootloader which reads the application image file from the SD card and then moves on to core initialization and other steps
  • To boot an application using the sbl_sd, the application image needs to be copied to the SD card as a file named "app". Make sure that the SD card is formatted to have a FAT partition. To know more about the SD card partitioning please refer SOC Initialization Using SD BOOT
  • Follow the steps in the above referred page to partition the SD card. For a complete boot from SD card, both the sbl_sd binary and the application image binary has to be present as files in the SD card. You have to rename the sbl_sd appimage as 'tiboot3.bin'.
      copy file to SD card => ${SDK_INSTALL_PATH}/tools/boot/sbl_prebuilt/am243x-evm/sbl_sd.release.tiimage
      rename in SD card as => tiboot3.bin
  • Similarly you can copy any appimage file to the SD card and rename in the SD card as "app" so that the SBL can pick it up.
  • Currently the sbl_sd reads the full appimage file into an MSRAM buffer and then parses the multicore appimage. Because of this reason appimages higher than ~380 KB in size can't be booted by sbl_sd as of now.


  • The sbl_ospi is a secondary bootloader which reads and parses the application image from a location in the OSPI flash and then moves on to core initialization and other steps
  • To boot an application using the sbl_ospi, the application image needs to be flashed at a particular location in the OSPI flash memory.
  • This location or offset is specified in the SysConfig of the sbl_ospi application. Currently this is 0x80000. In most cases you wouldn't need to change this.
  • To flash an application (or any file in fact) to a location in the OSPI flash memory, follow the steps mentioned in Basic steps to flash files


  • The sbl_uart is a secondary bootloader which receives the multicore appimage via UART, stores it in memory and then does the parsing, core initialization etc.
  • To boot an application using the sbl_uart, you can refer to UART Bootloader Python Script subsection. Detailed steps on the usage is mentioned in the same subsection.

Preparing the SBL for boot

The SBL is like any other application, created using the same compiler and linker toolchain. However the steps to convert the application .out into a bootable image are different for SBL as listed below

  • The SBL entry point needs to be different vs other applications. On AM243X after power-ON ROM boots the SBL and sets the entry point of SBL to both R5FSS0-0 as well as R5FSS0-1. However for SBL we need to detect the core and run SBL only on Core0 and keep Core1 in wfi loop. This is done by specifying a different entry point -e_vectors_sbl in the linker command file for the SBL application. In _vectors_sbl the very first thing it does is detect the core and continue execution for Core0, while if the core is Core1 then it enters wfi loop.
  • Other special factors for SBL application are listed below
    • After entering main(), make sure to call Bootloader_socLoadSysFw to load the SYSFW to DMSC M3 and setup a "board config"
    • The linker command file for SBL has to place vectors at address 0x70000000 and this is the entry point for the SBL.
    • Nothing should be placed in ATCM or BTCM
    • Only the region 0x70000000 to 0x70080000 should be used by SBL code, data, stack etc
  • After building, the SBL application .out file is first converted to a binary format .bin using the GCC objcopy tool.
    • This copies the loadable sections from the .out into a binary image stripping all symbol and section information.
    • If there are two loadable sections in the image which are not contiguous then objcopy fills the gaps with 0xFF.
    • It is highly recommended to keep all loadable sections together within a SBL application.
  • This .bin file is then signed using the Signing Scripts to create the final .tiimage bootable image.
    • A default key is used for this.
    • This is a ROM bootloader requirement and is needed even on a non-secure device.
  • This .tiimage file can then be flashed or copied to a boot image using the Flashing Tools