How set NEMAGFX_MEM_POOL_SIZE in CubeMX

[nico23]

Hi,

I've a project with TouchGFX and FreeRTOS on an STM32H7RS. I was updating the library, regenerating the code from CubeMX but, I've noticed that, for some reason, the file \Appli\TouchGFX\target\generated\nema_hal.c has changed

// from
- #define NEMAGFX_MEM_POOL_SIZE          16128 /* NemaGFX byte pool size in byte */
// to
+ #define NEMAGFX_MEM_POOL_SIZE          24320 /* NemaGFX byte pool size in byte */

but I can't find in the .ioc inside the CubeMX where this value is set and why it has been changed

Thanks

[JohanAstrup · Accepted solution]

Hello @nico23.

You cannot directly control the entire NEMAGFX_MEM_POOL_SIZE from STM32CubeMX. The size is determined by the command list size, which you can configure in X-CUBE-TOUCHGFX in STM32CubeMX, and 8 KB required for other Nema allocations.

Based on your description, I assume that you have updated an older project to TouchGFX 4.25.0 or newer. In TouchGFX 4.25.0, the default command list size was increased from 8 KB to 16 KB to provide significant performance improvements in certain situations.
If you want to revert this change, you can set the command list size to 8 KB in STM32CubeMX.

Best regards,
Johan

[nico23]

Hi @JohanAstrup 

That's why I had issues with the .icf and the RAM used. Having doubled it, it was exciting my previous memory configuration in the .icf file.

Do you have an example of a situation in which the increased memory produced "significant performance improvements"?

Thanks

[JohanAstrup]

If the command list is not filled within one frame and does not reach an auto submit, you will likely not see any gain. However, if you have a complex screen that triggers many Nema commands per frame, a larger command list is beneficial.
An example of a situation where you will see performance improvements is if you are using vector fonts.
If you only use NeoChrom GPU for blitting, a larger command list does not provide a significant performance gain.

Best regards,
Johan