STM32H7 with DTCM of 192K getting elf load problem with debugger

[JBias]

I'm using the NUCLEO-H7S3L8 and have it setup 192K of DTCM memory. However, when I start to allocate DTCM memory above 64K the .elf file fails to load in the debugger. If I stay below 64K not a problem.

 

volatile uint8_t __attribute__((section (".dtcm_data"))) __attribute__((aligned(32))) gShadowPQ_WFB[65536] = {1, 2, 3, 4, 5, 6, 7, 8};
volatile uint8_t __attribute__((section (".dtcm_data"))) __attribute__((aligned(32))) gShadowS1_WFB[8192]; //<-- gives problems with loading of STS_Bias.elf

 

 

[TDK]

Did you set DTCM_AXI_SHARED option byte to allocate the max 192 kB to DTCM? Show in CubeProgrammer.

 

[JBias]

@TDK I set these bits via the MX tool:

and they appear to increase the memory size. However, specifying more than 64K in my code results in the .elf file not being able to be loaded in the debugger

[Pavel A.]

.dtcm_data :
{
 . = ALIGN(4);
 *(.dtcm_data) /* Include data */
 . = ALIGN(4);
} AT> DTCM

In this part of the .ld file, remove "AT" .  If you want initialized data in DTCM or ITCM, please check examples.

For simplicity and robustness,  initialize variables in non-default data segments in runtime. Compile-time initialization requires extra code and is error-prone.

 

 

 

[JBias]

Thanks this is good to know. I do not need to have them initialized at all, I just did so for debug purposes. However, I still have the same issue with the .elf file not being able to be loaded if I specify more than 64K in the DTCM area. I want to place some large arrays that are at max 64K and just can not seem to be able to do so without the elf file not wanting to load.

[Pavel A.]

Then there should be other problems in the ld file or in the code. 

Try to obtain more detailed log of the debugger.