Compilation error when enabling NRF5340 FPU

[kpi_xie]

Dear Nordic

hardware:NRF5340, sdk:2.6.1

Our project needs to use floating-point calculations. Attempting to add CONFIG_FPU=y and CONFIG_FPU_SHARING=y to prj.conf will result in an error during compilation
as follows

-- Cache files will be written to: E:/nordic-sdk/v2.6.1/zephyr/.cache
CMake Error at cmake/utils.cmake:67 (message):
INVALID CONFIG: ( CONFIG_TFM_ENABLE_FP OR CONFIG_TFM_ENABLE_MVE OR
CONFIG_TFM_ENABLE_MVE_FP ) AND NOT CONFIG_TFM_ENABLE_CP10CP11
Call Stack (most recent call first):
config/cp_check.cmake:35 (tfm_invalid_config)
config/check_config.cmake:85 (include)
CMakeLists.txt:79 (include)

Our project is developed based on the Zephyr example Blinky, and the board uses nrf5340dk_crf5340_cpuapp_ns

[Hardware CI]

I'm having a similar error with TFM and FPU configured together. To be exact:

CONFIG_FPU=y
CONFIG_TFM_PROFILE_TYPE_SMALL=y

for target nrf5340dk_nrf5340_cpuapp_ns it returns the error:

CMake Error at cmake/utils.cmake:67 (message):
INVALID CONFIG: TFM_PROFILE = profile_small AND CONFIG_TFM_SPM_BACKEND_IPC
Call Stack (most recent call first):
config/check_config.cmake:77 (tfm_invalid_config)
CMakeLists.txt:79 (include)

which doesn't make much sense to me.

SDK ver is 2.6.1

[Elfving]

Hi,

I am seeing the same thing on my side. It seems that enabling FPU also enables it in the TFM image, which is not compatible with using the smaller TFM type.

Though we technically only support the MINIMAL and NOT_SET option for TFM_IMAGE either way, so I would recommend you to rather use eg. NOT_SET instead.

Regards,

Elfving