STMH753 GPIO port pin read is slower compared to the STMF429

[TSosi.1]

Hi,

I am porting an application from the STMF429ZI to the STMH753ZI and have noticed that the STMH753 is slower when reading the GPIO inputs compared to the STMF429. In one function, multiple GPIOs are read to gather 16 input states. Measuring this function I see that the STMH753 with 360 MHz sysclock is twice as slow as the STMF429 with a sysclock of 180 MHz.

I see that the architectures are very different between the two processors. How much slower is the STMH753 when reading a port pin?

Note, I have verified the sysclock frequency by measuring it on MC02(360/15 = 24 MHz).

It also seems interrupt latency would be slower compared to the STMF429 correct?

Interesting previous topic: https://community.st.com/s/question/0D50X00009XkWN7SAN/stm32h7-gpio-togle-max-frequency

[waclawek.jan]

This has been discussed many times. The 'H7 is a number cruncher, and it performs relatively poorly as controller.

The GPIO is at a slow bus/matrix and the read has to travel from it to the processor through two matrices. There will be lots of mostly undocumented delays and latencies.

https://community.st.com/s/question/0D73W0000004bT4/detail

I am not sure about the interrupts (EXTI, I presume). It's in the same D3 domain as GPIO, but the links may be directly to NVIC. This is undocumented AFAIK. I wouldn't hold my breath, though; interrupts in ARMs generally are slow, given the big inherent ISR entry/exit latencies.

JW