Execution delay different for NUCLEO-F207 and NUCLEO-F767 boards for the same set of lines

[Shubhamkeshari]

Hi, I wrote a standard (DFT)Discreate fourier transform code (code is not the issue), in NUCLEO-F207 board using zephyr RTOS. I build the same program for NUCLEO-F767 board and flashed.

The time taken to execute DFT code in NUCLEO-F207 was 8700 micro seconds.

The time taken to execute the same code in NUCLEO-F767 is 65000 micro seconds.

FPU(Floating point unit) is enabled for NUCELO-F767 board 

Hardware and softwares are completely same for both the boards

How is it happening what is the potential cause ?

[Andrew Neil]

---

@Shubhamkeshari wrote:

Hardware and softwares are completely same for both the boards

---

Hardly: they are different MCUs Cortex-M3 and -M7!

Are you sure that the optimisation level is the same on both?

What about clocking?

---

@Shubhamkeshari wrote:

FPU(Floating point unit) is enabled for NUCELO-F767 board 

---

So what happens if you disable it?

[mƎALLEm]

Hello,

As you are measuring the exec duration based on the time, need to check the System frequency first. 

Did you enable the cache/ART accelerator on the F7 product?

[Shubhamkeshari]

Are you sure that the optimisation level is the same on both?

- Yes same optimization level is maintained

What about clocking?

- NUCLEO-F767

&clk_hse {

hse-bypass;

clock-frequency = <DT_FREQ_M(8)>; /* STLink 8MHz clock */

status = "okay";

};

&pll {

div-m = <4>;

mul-n = <216>;

div-p = <2>;

div-q = <9>;

clocks = <&clk_hse>;

status = "okay";

};

&rcc {

clocks = <&pll>;

clock-frequency = <DT_FREQ_M(216)>;

ahb-prescaler = <1>;

apb1-prescaler = <4>;

apb2-prescaler = <2>;

};

- NUCLEO-F207

&clk_hse {
hse-bypass;
clock-frequency = <DT_FREQ_M(8)>; /* STLink 8MHz clock */
status = "okay";
};

&pll {
div-m = <8>;
mul-n = <240>;
div-p = <2>;
div-q = <5>;
clocks = <&clk_hse>;
status = "okay";
};

&rcc {
clocks = <&pll>;
clock-frequency = <DT_FREQ_M(120)>;
ahb-prescaler = <1>;
apb1-prescaler = <4>;
apb2-prescaler = <2>;
};

If I disable the floating point unit in NUCLEO-F767 the delay increases

[Shubhamkeshari]

No i have not enabled cache/ART accelerator , how to enable that ?

[Andrew Neil]

Please see: How to insert source code.

[mƎALLEm]

What is the environment are you using? Seems it's not a ST environment!

[Shubhamkeshari]

I am using Zephyr RTOS

[Shubhamkeshari]

/*NUCLEO-F207*/
&clk_hse {
	hse-bypass;
	clock-frequency = <DT_FREQ_M(8)>; /* STLink 8MHz clock */
	status = "okay";
};

&pll {
	div-m = <8>;
	mul-n = <240>;
	div-p = <2>;
	div-q = <5>;
	clocks = <&clk_hse>;
	status = "okay";
};

&rcc {
	clocks = <&pll>;
	clock-frequency = <DT_FREQ_M(120)>;
	ahb-prescaler = <1>;
	apb1-prescaler = <4>;
	apb2-prescaler = <2>;
};

/*NUCLEO-F767*/
&clk_hse {
	hse-bypass;
	clock-frequency = <DT_FREQ_M(8)>; /* STLink 8MHz clock */
	status = "okay";
};

&pll {
	div-m = <4>;
	mul-n = <216>;
	div-p = <2>;
	div-q = <9>;
	clocks = <&clk_hse>;
	status = "okay";
};

&rcc {
	clocks = <&pll>;
	clock-frequency = <DT_FREQ_M(216)>;
	ahb-prescaler = <1>;
	apb1-prescaler = <4>;
	apb2-prescaler = <2>;
};

[mƎALLEm]

Hello,

I don't know about that Zephyr environment. 

But you need to refer to our documentation: Especially the AN4667 "STM32F7 Series system architecture and performance".

In the HAL examples, to enable the data and the instruction cache:

static void CPU_CACHE_Enable(void)
{
  /* Enable I-Cache */
  SCB_EnableICache();

  /* Enable D-Cache */
  SCB_EnableDCache();
}

 For ART it is enabled in HAL_Init():

HAL_StatusTypeDef HAL_Init(void)
{
  /* Configure Instruction cache through ART accelerator */ 
#if (ART_ACCELERATOR_ENABLE != 0)
  __HAL_FLASH_ART_ENABLE();
#endif /* ART_ACCELERATOR_ENABLE */

  So you need to check in the Zephyr environment where to enable all these features.