RedBoot can normally be configured to run in a number of startup modes (or just "modes" for short), determining its location of residence and execution:
In this mode, RedBoot both resides and executes from ROM memory (flash or EPROM). This mode is used when there are limited RAM resources. The flash commands cannot update the region of flash where the RedBoot image resides. In order to update the RedBoot image in flash, it is necessary to run a RAM mode instance of RedBoot.
In this mode, RedBoot resides in ROM memory (flash or EPROM), but is copied to RAM memory before it starts executing. The RAM footprint is larger than for ROM mode, but there are two advantages to make up for this: it normally runs faster (relevant only on slower boards) and it is able to update the flash region where the image resides.
In this mode, RedBoot both resides and executes from RAM memory. This is used for updating a primary ROM mode image in situ and sometimes as part of the RedBoot installation on the board when there's already an existing (non-RedBoot) boot monitor available.
You can only use ROM and ROMRAM mode images for booting a board - a RAM mode image cannot run unless loaded by another ROM monitor. There is no need for this startup mode if a RedBoot ROMRAM mode image is the primary boot monitor. When this startup mode is programmed into flash (as a convenience as it's fast to load from flash) it will generally be named as "RedBoot[RAM]" in the FIS directory.
The startup mode is controlled by the option CYG_HAL_STARTUP which resides in the platform HAL. Some platforms provide only some of the RAM, ROM, and ROMRAM modes, others provide additional modes.
To see mode of a currently executing RedBoot, issue the version command, which prints the RedBoot banner, including the startup mode (here ROM):
RedBoot>version RedBoot(tm) bootstrap and debug environment [ROM] Non-certified release, version UNKNOWN - built 13:31:57, May 17 2002