#avr Logs

Sep 16 2022

08:40 AM saxicola is now known as saxicola-afk
08:50 AM saxicola-afk is now known as saxicola
08:51 AM cehteh: making the context save minimal needs some more tricks and looking at the generated asm, usually you dont want to call any other functions from an ISR or inline everything there, then the compiler knows what to save
08:53 AM cehteh: (and c++ member methods as isr .. glhf)
09:06 AM exp: cehteh: i'm pretty sure they (vector table entries) are just symbols
09:48 AM saxicola is now known as saxicola-afk
11:41 AM Phantom: you can also inline the function
01:28 PM exp: Phantom: oh?
01:29 PM exp: skempf: can you give more details about your requirements?
03:10 PM skempf: Sorry, was out.
03:12 PM skempf: I was thinking that the C inline was a hint, not guaranteed. But I was reading up on c++ and it looks like inline might be guaranteed.
03:12 PM exp: what's actually driving this though
03:12 PM exp: why do you need to avoid the potential function preamble?
03:13 PM skempf: I have multiple rs-232 ports talking and I want to implement my framing protocol just once and have each of the ports use the same code, just with different internal pointers.
03:13 PM skempf: ^ pointers not correct word
03:13 PM skempf: different internal variables
03:14 PM exp: you will probably need to utilise inlining then AFAIK
03:14 PM skempf: I know not to call functions inside of an ISR, because the context has to be saved. Too much overhead.
03:14 PM exp: as using the exact same address for each vector would not give you variables as you desire
03:14 PM exp: really it depends on what you're doing
03:14 PM exp: as long as you understand interrupts are disabled while in the ISR etc
03:15 PM exp: then you can do whatever you like
03:15 PM skempf: I was originally trying to come up with a way to register an class instance's method as the ISR (maybe put the pointer to the method in the isr jump table)
03:16 PM skempf: I think a simple function that is inlined might be the way to go. I can at least try it and see how much the code is inflated.
03:16 PM exp: i don't know much about the c++ execution model so i'm lost there really
03:16 PM exp: i think that's a good approach
03:30 PM Cracki: >inflation