performance improvement to run in PSPR

[User19977]

Hi there,

the cpu load is a little bit high. I am thinking to run some tasks out of PSPR. What performance improvement should I expect comparing running from PFLASH?

Thanks,
Ke

[User19977]

anybody knows

[Andrew_Q]

if you have caches enabled, then in my personal opinion you will not get any performance gain.

[Epistemon]

You could follow the advice in this document:

https://www.infineon.com/dgdl/ap3216810_TriCore_Performace_Optimization.pdf?fileId=db3a304336797ff90...

[User19977]

what is the use case for PSPR then?

[Epistemon]

As per Table 2 in the document link, the use case for PSPR: 

"Used for deterministic and performance critical code (e.g. DSP Algorithms, OS)"

This is on the basis that it doesn't have any wait-states compared to running from flash. It seems reasonable to try it out given the relative ease of moving the code to PSPR using the linker script.

@Andrew_Q is expressing a personal view which seems to counter the advice in the document but that's his view.

It seems no one can give you any outright performance improvement figures so you should probably measure the improvement yourself using the tips in Section 2.2 of the document.

[Andrew_Q]

@Epistemon is absolutely right! You need to check each individual case.  In my case, I tested one core that works with flash memory. If you are using a lot of cores, then porting the code to PRPS will have an effect. Or you are using large computational algorithms that force the cache to be flushed. But how much?