Instruction timing changing with memory location?

[Chris K]

Hi. I am writing very time sensitive code on the STM8 mostly using assembly. However, I have noticed that sometimes adding one cycle of delay at one point in the code changing the timing of the whole code. I will explain in more detail to clarify. I have an interrupt which fires based on the a GPIO. I start reading the inputs. My code works, data read is good. I respond, but the timing is slightly off. So I add in a couple of nop cycles to fix it. The read is now broken. Keep in mind this is all a single sequential file of assembly to handle the interrupt. The processor is always in the WFI state when the interrupt fires. But somehow adding a couple of nops somewhere in the code causes everything to break (in terms of timing). I sure there is some explanation for this, but I have yet to find anything in any of the documentation. Any insight would be greatly appreciated.

Thank you,

Chris

[Cristian Gyorgy]

Try posting your code (the relevant part of it). It would help us understand what you talk about.

[Chris K]

I can't post the exact code but I will try to create something representative which will illustrate the behaviour I am seeing. In the mean time though I just want to clarify if there are any alignment requirements for this device. Do misasligned data or function accesses cause stalls in the pipeline or elsewhere in the architecture? Because I have not been able to find anything in the document alluding that it may be the case. Thanks.

[Cristian Gyorgy · Accepted solution]

The following snip is from the STM8 programming manual, pg. 22, Doc ID 13590 Rev 3:

[Chris K]

That explains it. Thank you.