How to speed up IRQ times at STM32F7xxx ?

[BSchm.0]

• CPU 210MHz
• If you run IRQ code from flash you have about 5 times more cpu cycles as needed (7WS)
• if you switch on ART accelerater and prefetch:

for loop gets slower

no influence on IRQ latency

• you could not run code from internal SRAM because of missing remap

STM32F4xxx can remap to SRAM at 0x00

there is no option to speed up IRQ times with this F7 family 👎

[Tesla DeLorean]

>>you could not run code from internal SRAM because of missing remap

You can still run from Internal SRAMs.

Use ITCMRAM at 0x00000000, or DTCMRAM at 0x20000000

[BSchm.0]

I can run code from internal SRAM too but I need IRQs and working vectortable, thats the problem !

I tyied already with DTCM RAM at 0x20000000 .. see https://community.st.com/s/question/0D50X0000AqqvO3SQI/stm32f767zi-could-not-remapped-to-internal-sram

[Tesla DeLorean]

And why wouldn't a vector table and code placed in ITCMRAM achieve this?

Both ITCM and DTCM RAM should permit instruction fetch.

I'd use ITCMRAM that's what it is explicitly designed and placed to achieve.