IRC logs for #openrisc Sunday, 2013-02-17

stekernjuliusb: the problem that the patch solves is that the timer exception wasn't turned off when the test was done, so the exception had time to hit once more (increasing the test counter) before the final test code was run02:50
stekern"uncommon that you have much between your code and the processor" <- what does that mean?02:51
stekernlittle abstraction in your application?02:52
olofkFirst patches since christmas applied for orpsocv3. Can't take credit for any of the work though03:04
juliusbohhh ok, I get it12:17
juliusb(regarding timer test)12:18
juliusbum,by the "uncommon" comment, I meant it's usually not the case that you have a lot of stuff on the instruction fetch bus of the processor12:19
juliusbyou couple it as closely as possible to where the instructions are coming from, or you put in a nice cache12:19
juliusbbut in deeply embedded stuff you've got a ROM or a biggish RAM, so just put it as closely as possible to that12:19
juliusbanyone know how to get the name of the recipe/target being run in Make as a variable?14:42
juliusblike, to alter a variable if we're running a particular target?14:43
juliusbso the name of thing you pass to Make, ie. "make foo" I want to know how to test if one of the recipes we're going to run is foo14:44
juliusbMAKECMDGOALS is it i believe14:48
blueCmd_franck_: great!15:40
stekernjuliusb: ah, ok. Yeah, that makes sense, and is kind of expected, I mean, why waste logic on some biu and fetch stall logic if you're going to run from a fast easily determinable ROM16:58
juliusbOk I think I have carry and overflow logic working in espresso and prontoespresso17:21
juliusbI think I'll push that for now17:22
stekernI assume you have some tests for it? I can take a look at adding it to cappuccino17:24
juliusbno just coded it up, testing? why bother testing? ;)17:24
juliusbhaha yes17:24
juliusba overflow and carry test, plus a test written in C which will generate random numbers, test if they should generate overflow or carry and then do them in hardware17:25
juliusband check the results17:25
juliusbi pushed that last night to mor1kx-dev-env17:25
juliusbwel, the fancy test anyone, but I have more to push today17:25
juliusbnoticed a discrepency between my implementation in espresso and pronto for the overflow exception stuff - on one the result was written to  the RF, but in the other the result was not. I think it's sensible to write the result, so I've done that17:27
juliusbbut that made me think more about the software doing the resting17:27
juliusbwhich is good, so anyone, I'm reasonably sure it's robust17:27
stekernI'm making slow progress with everything atm now though, busy with laying in the sun and bathing ;)17:27
stekernso only hack a bit to the morning coffee and the night cap beer 17:28
juliusbit only does it for add, though, my brain broke when I tried to figure out how l.sub should work17:28
juliusboh where are you?17:28
juliusbsomewhre sunny I hope17:28
juliusband not finland17:28
juliusboh nice! :)17:28
juliusbi see it's a bit of a night-hack for you at the moment then17:28
stekernwe're making the escape from snow-hell a yearly event ;)17:28
stekernyeah, half past ten here17:29
juliusbgood call. I've had to do it every year I've been here in norther europe17:30
juliusbbrings you back to life :)17:30
stekern(written to rf) I'm not sure what you meant there, but I've just made some changes to cappuccino that cuts all wb off on exception17:33
stekerni.e. it also flushes the tail of the pipeline17:33
juliusbwell, for range i think it's obvious that you want to keep the result, you just also want to know if it caused a range exception too17:34
stekernI think taking all exceptions in the "ctrl" stage was a good call (you made that in fourstage already)17:34
juliusbsure, the result overflowed but you probably want both the result and the record of it overflowing, not one or the other17:35
juliusbha yeah spreading out the exception handling logic everywhere would make it a nightmare17:36
stekernhmm, yes, that sounds right17:36
juliusband allowing them to flow through makes things a little easier I think17:36
juliusbflow through from the earlier stages, that is17:36
stekernyes, that's exactly my point17:36
stekernjust push in noops if they are detected early17:37
juliusbya, too easy right?17:37
juliusbsomething that's less easy - tiger or singha?17:37
juliusb(don't say carslberg!)17:38
stekernyup, tlb miss exceptions coupled with ibus errors etc have not caused me any head aches at all ;)17:38
stekernnah, I didn't even drink any carlsberg when I was in copenhagen last month17:38
stekernI think I prefer Chang the most, but it's pretty watery... but what's not here17:39
juliusbah yes, chang17:40

Generated by 2.15.2 by Marius Gedminas - find it at!