MKV56 (Cortex M7) is slower than MKV46 (Cortex M4)

[anup_gandra]

We are using MKV46 and MKV56 as alternate in one of our projects with below configuration of clock.

1. System clock - 168 Mhz for both 

What i observed is that execution time of one algorithm is taking more time in MKV56 (64uS) than MKV46 (49uS).  What can be the reason for this?

Also for MKV46 i am running clock at 168 Mhz in Normal Run mode but reference manual says it should be 100Mhz or less in normal run mode. What can be the issues/problems that can come with 168 Mhz in normal run mode?

Thanks in advance.

[_Leo_]

Hi Anup Kumar Gandra,

Thank you so much for your interest in our products and for using our community.

Regarding your first question, surely the MKV56 is doing more things. It would be a matter of checking in more detail the algorithm with which this test is being carried out.

On the other hand, unfortunately, we do not have tests for MKV46 running at 168 Mhz in Normal Run mode, so we cannot guarantee functional operation with ratings that are out of spec. However for that frequency, you can use the HSRUN mode.

Sorry for the inconvenience. And please let me know if you have more questions about it.

Have a nice day!

[anup_gandra]

Regarding your first question, surely the MKV56 is doing more things. It would be a matter of checking in more detail the algorithm with which this test is being carried out. 

Regarding this. Its not specific to algorithm but any c code like function call or PWM register access also taking more time in MKV56 than MKV46.

Even this below code snippet also takes higher time in MKV56 with same clock configuration in both.

if (((PWM_BASE_ADDR->SM[0].OCTRL & PWM_OCTRL_PWMA_IN_MASK) && (PWM_BASE_ADDR->SM[0].OCTRL & PWM_OCTRL_PWMB_IN_MASK)) ||

((PWM_BASE_ADDR->SM[1].OCTRL & PWM_OCTRL_PWMA_IN_MASK) && (PWM_BASE_ADDR->SM[1].OCTRL & PWM_OCTRL_PWMB_IN_MASK)) ||

((PWM_BASE_ADDR->SM[2].OCTRL & PWM_OCTRL_PWMA_IN_MASK) && (PWM_BASE_ADDR->SM[2].OCTRL & PWM_OCTRL_PWMB_IN_MASK)))

Regards,

Anup

Contact No: 9505829851

[_Leo_]

Hi again Anup Kumar Gandra,

It is possible that PWM register access takes more time in MKV56 than MKV46. It depends on design factors. But to make a fair performance comparison between different ARM architectures it is better to refer to CoreMark, which is a benchmark that measures the performance of MCUs.

I hope this will help you. Please let me know if you have more questions about it.

Have a nice day!

[anup_gandra]

But coremark is better for cortex M7(5.01) than cortex M4(3.42) right?

[_Leo_]

That's right, the one from the cortex M7 is better.

[anup_gandra]

Hi,

Thanks. 

Also i just gave PWM register access as an example but normal C specific code without peripheral register access is also taking more time.

Where can i check the specifications of CoreMark.

Regards,

Anup

[_Leo_]

In the following links in the characteristics section you can find their respective performance efficiency for the Cortex M7 and M4:

Cortex-M7

Cortex-M4

[anup_gandra]

Hi,

I have gone through this differences but from this i feel cortex M7 should give higher performance than cortex M4.

Can you perform some tests between MKV46 and MKV56 to see what are the observations?

Thanks and Regards,

Anup

[_Leo_]

The performance comparison between the MKV5x and the MKV4x can be seen in the Kinetis V Series Fact Sheet.

Regarding the CoreMark of the Cortex M7 and M4 cores, they are presented by ARM in the links mentioned above.

Any other comparison test is out of our scope. An apology for any inconvenience this may cause you.

Hope it helps you!

Have a nice weekend!

[anup_gandra]

HI,

I have understood your points but i have a request as follows.

Can you share the execution time of same code (with same clock settings) between MKV46 and MKV56?

Regards.

Anup

[anup_gandra]

Thanks for the help.

Can you compare results between MKV46 and MKV56 and see what's it like?

[_Leo_]

As I mentioned before this comparison test is out of our scope. An apology for any inconvenience this may cause you.