#milkymist IRC log for Monday, 2012-01-02

cladamwgood morning ! Happy new year 2012 !01:12
kristianpaulmorning, and thanks cladamw !01:17
wpwrakgrmbl. lists.milkymist.org is down11:15
wpwraksb0: seems that lists.milkymist.org had too much champagne. it's down :(11:48
sb0meh11:49
sb0dreamhost issue ... :(11:49
wpwraknightmarehost :)11:56
sb0yeah11:59
sb0there's hardly a month without an incident of some sort12:00
wpwrakwhy don't you just use qi-hw for this kind of infrastructure ?12:02
sb0they do fix them, though. but I'd indeed prefer they had a bit more focus on prevention...12:03
sb0well... then I have to move the web content (easy), database (semi-easy unless things go wrong), and mailman config and archives (messy)12:04
sb0and I hate sysadmin, it feels like cleaning a massive toilet12:05
wpwrakthat's why i think outsourcing the problem to wolfgang could be something you'd like to do :)12:05
wpwraksb0: btw, what's your plan for phasing in migen as the "standard" way of building the soc ? is this something you see happen soonish or will it be a while ?12:15
sb0it'll be in a while12:15
wpwraklots of infrastructure elements to convert ?12:15
sb0merely because I plan to do other things at the same time: redesign the memory infrastructure, support 32bpp graphics, and use migen flow for the tmu12:16
wpwrakyeah, i'm curious about the memory infrastructure. the speed improvements you predicted would be quite massive. but let's see how well practice and theory go together there :)12:18
sb0I do not plan to support the current SoC buses in Migen12:18
wpwrakyou mean wishbone and such ?12:18
sb0only the new, faster ones (except wishbone for lm32 compat)12:18
wpwrakah, i see. well, as long as there's some working bus where needed, who cares how it's done inside :)12:19
wpwrakon friday, i had a bit of fun trying to midi-ize the "digital flame". it was a disaster :) first, i hit a regression with audio levels. not sure yet where this comes from. audio from the internal mic on my rc3 is now very weak.12:24
wpwrakand then i realized that i still haven't killed the evil variable name typo "feature". i.e., that the compiler will just happily accept any typo and leave it to you to figure out what on earth is wrong this time.12:26
wpwraki thin what may work is to refuse any read of variables that haven't been assigned to. and to allow assignment (initialization) of user-defined variables in the initial section.12:27
sb0yes but be careful - there can be user variables, that need to be carried from frame to frame12:27
wpwrakyes, digital flame has some. nice little EWMA example.12:27
sb0and sometimes those variables are read and set in the same eqn12:27
sb0eg foo = foo*0.5 + bass*0.5 is perfectly valid12:28
wpwrakthere, we should probably require initialization12:29
sb0if you do not allow non-existing variable names, it'll complain about foo being incorrect when compiling the right hand side12:29
wpwrakyes12:29
wpwrakso foo = 0 into the initial section (or whatever value)12:29
sb0that's not how MD does it - so it should only be a warning12:29
wpwrakin the case of   foo = foo...., i could just "declare" foo before parsing the right-hand side, but that wouldn't work in more complex cases. e.g., if there's a foo/bar pair that affect each other12:30
wpwrakah, but we don't really have the concept of warnings :)12:31
sb0yes. MD (and my compiler) simply create a new variable when it reads an unknown identifier12:32
wpwrakbut what i can do is make it depend on the style detection. so if you write a new-style patch, you get the error. if your patch is ugly original style, the compiler will shut up and let you choose your own misery ;)12:32
wpwrakthe style is conveniently known by the time the first non-initial assignment happens. so the only divergence would be that old-style patches may accept user variables in the initial section, which they wouldn't do in MD. well, i could catch that when detecting new-style syntax.12:39
wpwrakbut then, it's probably not worth it. i'd rather convince people to limit their editing of old-style to a minimum and to spend those five minutes it takes to properly format a patch if they plan any larger changes12:40
wpwrakif we're good, maybe MD will "steal" our parser in a while ;-)12:41
wpwrakwolfspraul: btw, how is your midi shopping going ?12:53
wpwrakwolfspraul: yi is also looking for midi devices these days. i kinda suggested she get an icreativ. one of these days, i need to make a video comparing the controllers.12:54
wolfspraulwpwrak: I ordered 2 lv3 + 1 icreati13:23
wolfspraulicreativ13:23
wpwrakgood. you'll have fun :)13:26
--- Tue Jan 3 201200:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!