#avr Logs

Jul 14 2017

01:42 AM Emil: Jartza: what's that for?
01:49 AM Emil: Jartza: umm
01:49 AM Emil: Jartza: why do you design it like that?
01:50 AM Emil: Jartza: why not connecto directly to the ground plane instead of going under the chip?
02:50 AM day_ is now known as daey
03:34 AM _abc_: Hello. avr-gcc is not smart enough to deduce that volatile global vars used in an isr are ... not volatile in that context? (normal vars, not SFRs)?
03:35 AM _abc_: Iow it makes sense to cache a volatile global in an uninterruptible isr because otherwise the compiler will generate volatile access to the var in every use instance?
03:35 AM _abc_: The .lst I am looking at seems to show exactly this.
03:35 AM _abc_: Are there any pragmas in avr-gcc to control this behavior? I assume not?
03:58 AM Emil: WAT?
03:58 AM Emil: Wat?*
04:00 AM Emil: _abc_: if you need a volatile global but want to cache it in an isr then you must do type name = global_var at the beginning and use that instead
04:05 AM _abc_: that is caching locally Emil.
04:05 AM _abc_: exactly what I said.
04:09 AM Emil: ...
04:14 AM _abc_: fyi optimizing compilers DO have such features, deducing that using the same volatile var 3 times in the same statement block in an isr context marked uninterruptible means it can and should be cached locally
04:16 AM _abc_: https://stackoverflow.com/questions/27816535/when-can-a-volatile-variable-be-optimized-away-completely example of unwanted volatile optimization. clang/llvm does the wrong thing, deleting the f assignment. Typically in embedded such an assignment would ...
04:16 AM _abc_: ... set up a sfr. That optimization is a bad idea.
04:25 AM * _abc_ has trouble timing an external signal with an avr328 in mysterious ways. Strange things which used to work no longer do. Bit rot!
04:57 AM Jartza: Emil: because both vcc and ground need to go through bypass cap
05:06 AM Jartza: it would be different case if I used 4-side board with ground and power planes
06:05 AM Emil: Jartza: wat
06:05 AM Emil: Jartza: your explanation makes no sense
06:05 AM Jartza: it's recommended that return path goes via capacitor too
06:06 AM Emil: Jartza: also, where is the pinout for measurements?
06:06 AM Jartza: nowhere yet
06:06 AM Jartza: it's a work in progress
06:07 AM Emil: Wtf :D why route things like that
06:10 AM Haohmaru: abstract arts
06:17 AM Jartza: like what?
06:33 AM Jartza: http://gerblook.org/pcb/tFY94vQnD6iU2TJhjooZ7j
06:33 AM Jartza: that's the routing
06:34 AM Jartza: but yea. still misses half of the parts
06:49 AM Emil: But
06:50 AM Emil: Why would you not design the schematic first
07:24 AM Emil_: Well, if it works for oyu
08:32 AM day_ is now known as daey
01:50 PM Jartza: Emil: of course I design schematic first, but I started from the thermocouples
01:50 PM Jartza: there are lot of chips outside the picture :)