#milkymist IRC log for Sunday, 2011-05-22

azonenberglekernel: hi09:17
azonenbergI see you saw my latest test pattern, any thoughts?09:17
azonenberghttp://i.imgur.com/nsrQk.png is a lossless PNG version09:17
azonenbergtop right and top center of http://i.imgur.com/RQ1lO.png is the actual mask layout09:18
azonenbergat ~600 DPI09:18
azonenbergready to be printed and reduced 40x onto the wafer09:18
lekernelwell, I don't know.09:21
lekernelbut it does look good and I'm eager to see what it's going to do on silicon :-)09:21
azonenbergLol me too09:26
azonenbergPayday was thursday so i'm in the process of ordering blank wafers etc09:27
azonenbergwithin a month i should be starting fab runs09:27
azonenbergBut in the meantime i'm killing time by doing mask design09:28
azonenbergThe plan for this mask is, stick the bottom (red) mask in09:28
azonenbergLine up so that the horizontal axis is parallel to a <111> plane and roughly center it on the die09:28
azonenbergExpose and etch09:28
azonenbergThen do the second level on to pof that09:29
azonenbergAlign the central cross09:29
azonenbergExpose, develop, etch09:29
azonenbergAnd then use the vernier scales etc to see how close I got09:29
azonenbergto perfect alignment09:29
azonenbergIn addition, the squares made out of parallel lines below the cross are there to test how well double etches turn out09:30
azonenbergio, hole on top of hole and peak on top of peak09:31
azonenbergEach etch will probably be a micron or so deep09:31
azonenbergWhich is about a minute in 30% KOH at 80C09:31
azonenbergwell actually that'd be 1.25um but close enough for testing, i just need it deep enough to see edges clearly09:36
Fallenouwhich gcc are you using ATM ? 4.5.1 ? 4.5.2 ? 4.5.3 ?12:59
FallenouI am working a little bit on my lm32-rtems-gcc portfile for Mac OS13:00
Fallenoumy current version is 4.5.113:00
FallenouI am wondering if I should update13:00
kristianpaulgcc versiĆ³n 4.5.2 (GCC)13:04
kristianpaulI'm usign xiangfu's script13:05
Fallenoukristianpaul: is this the version lekernel is using ?13:20
Fallenoubecause rtems has a 4.5.3 version in its ftp13:20
lekernelyes, i'm using 4.5.213:20
Fallenoulekernel: why not 4.5.3 ?13:20
Fallenoujust not tested yet ?13:20
Fallenouor some reason13:20
lekernelthe rtems rpm's do not have the divider and sign extender enabled multilibs, and ralf is super anal about this, so i'm not using them13:21
lekernelbesides, 4.5.2 works fine13:21
FallenouI am not refering to their rpms13:22
Fallenoubut to their source tarballs13:22
lekernelsame ... it needs a little patch13:22
Fallenouhum ok13:22
FallenouI guess I will stick with 4.5.2 then13:22
FallenouOK will put this patch in addition to their patch13:23
lekernelI have fixed this in gcc svn, but now 4.6 is totally broken for lm32 ...13:23
lekernelit's a mess13:23
Fallenouhum ok :/13:24
Fallenouif I add the divide and sign extension, will it still be compatible with lm32_evr ?13:24
lekernelthe only reasonable outcomes I see are (a) switch to llvm or (b) maintain a gcc fork (e.g. 4.4 which has working c++)13:24
lekernelyes of course13:25
Fallenoumaitaining something has a big cost, especially if it's gcc ^^13:25
lekernelnot if you stick to the "if it's not broken do not touch it" rule13:25
lekernelthere's little innovation in gcc anyway13:26
Fallenouhum ok so you would not backport gcc features into your fork13:26
lekernelwhat features?13:26
lekernelthe only remotely interesting one is google go13:26
lekernelwrt lm32, the newer gcc releases have mostly brought regressions13:27
Fallenouoh really ?13:27
lekernelyeah, c++ doesn't work anymore for example13:28
lekernelin all 4.5.x13:28
FallenouI don't know if C++ is really a good choice for embedded applications13:28
lekernelit could definitely work on large embedded systems like mm13:29
Fallenoulekernel: you use newlib 1.19 or 1.18 ?13:59
lekernelI don't know ... try 1.19?14:10
xiangfuI am using 1.1914:39
xiangfuFallenou: what I can do/test on newlib-1.19?14:44
kristianpaulat least works :-)14:44
Fallenouno idea, really14:44
xiangfulekernel. so you think better fork gcc? if fix those problem will needs a lot of time in gcc 4.6 ?14:56
xiangfulekernel I know nothing about gcc code. never look into it. just asking.14:57
lekernelxiangfu: if you want to mess with the compiler, try to write a lm32 llvm backend14:58
lekernelbut this is low priority anyway14:58
lekernelI've tried to track down the current 4.6 problem, took an afternoon and gave up14:58
lekernelit's mess14:58
kristianpaulmay not a mess at all, just we born in different ages than gnu/gcc people? :-)15:07
xiangfumaybe I can try complain those bugs to gcc mailing list first : http://www.milkymist.org/wiki/index.php?title=GCC_bug_list :D15:16
lekerneloh, don't bother15:17
lekernelno one really cares about lm32 except us15:18
lekernelprobably you just won't get an answer15:18
lekernelbut we don't really need those bugs fixed anyway...15:18
xiangfuI just saw http://gcc.gnu.org/bugzilla/show_bug.cgi?id=46898#c15. same as AVR15:22
xiangfuhmm.. seems (a) switch to llvm  needs a lot of works: 1. write code 2. maintaining the code.15:38
kristianpaultwo days lost because a forgot to initialize a counter !!15:39
xiangfu(c) fix bug in gcc 4.6 needs: 1. fix bug 2. keep eyes on gcc15:39
xiangfu(b) maintain a gcc fork (e.g. 4.4 which has working c++). but we will far away from 'upstream' at someday, some new libs will not compiled someday :(15:41
Fallenoulekernel: do we need multilib support ?18:05
FallenouI think we just need one libgcc with our settings18:09
lekernelnot really, if you can only build the library with all the CPU features enabled it's fine for our purposes18:14
Fallenouyep ok18:17
FallenouI have some troubles compiling right now18:17
Fallenoumaybe you will understand18:17
Fallenouarg ok i'm missing ppl (cloog)18:19
CIA-51flickernoise: Sebastien Bourdeauducq master * re8509ad / src/keymap.c : No need to make curr_layout static - http://bit.ly/mgJ1Ak19:13
CIA-51flickernoise: Sebastien Bourdeauducq master * r7b6d091 / src/keymap.c : Remove irrelevant comment - http://bit.ly/j1sUlv19:13
CIA-51flickernoise: Sebastien Bourdeauducq master * r4629cc2 / src/flash.c : Cosmetic - http://bit.ly/lOpEMo19:41
CIA-51flickernoise: Sebastien Bourdeauducq master * r2452c7c / src/flash.c : Coding style - http://bit.ly/kSYPL119:41
CIA-51mtk: Sebastien Bourdeauducq master * r7ce0056 / lib/userstate.c : Use new key codes for repeat - http://bit.ly/jDArW420:09
--- Mon May 23 201100:00

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