#milkymist IRC log for Thursday, 2011-12-08

wpwrakkristianpaul: i don't she's here a lot. better try mail01:07
kristianpaulsent01:16
kristianpaulhum topic changed, llhdl and antares not part of the free technology is beautifu slogan? :-)01:20
wpwrakantares, qu'est-ce que c'est ?01:22
kristianpauli think was P&R for s6 if no the bitstream generator too01:25
kristianpauls/was/is01:25
wpwrakoh. from xilinx ? didn't think they used "nice" names01:26
kristianpaulno no, free implementatio from lekernel research01:28
kristianpaullol01:28
kristianpaulwpwrak: https://github.com/sbourdeauducq/antares/01:28
wpwrakaah ! it did sound a little french :)01:29
kristianpaulxD01:29
Thihiwolfspraul, So how does one update the milkymist?01:56
Thihi Do you distribute updates as files manually downloadable somewhere and then transferred to it, or is it in some way automatic or semiautomatic?01:56
wolfspraulah yes, hi Thihi! :-)01:57
Thihi:)01:57
ThihiHi.01:57
wolfspraulThihi is our finnish independent journalist friend who got one of our prized European review units for a few months, to hopefully fall in love with :-)01:57
wolfsprauland he just got it!01:57
wolfspraulgreat01:57
ThihiI'm getting it on friday, to be precise. It's still being photographed. :/01:57
wolfspraulabout updating, there is something called 'web update'01:58
ThihiThey won't let me make fingerprints.01:58
ThihiBefore it's been in front of a really expensive camera ;P01:58
wolfspraulouch01:58
Thihi:)01:58
wolfspraulthere might be fingerprints on it from where it was before (MSL), but ok01:58
ThihiThat's ok.01:58
ThihiThey just want it as pristine as possible.01:58
wolfspraulI suggest you first try it the way it is, without update01:58
Thihi:)01:58
wolfspraulif that works well, we can try web update01:58
ThihiYeah, ok cool.01:59
wolfspraulweb update means that you plug in the Ethernet cable going to your dhcp-capable router, then turn on the m101:59
wolfspraul(has to be like that because the older software is quite fragile on dhcp, will only try once at boot)01:59
ThihiOk.01:59
wolfspraulif the dhcp works, you should be able to just click on an 'update' button in the gui, which will download and reflash stuff, and eventually reboot01:59
ThihiEasy enough then.01:59
wolfspraulyeah if it works :-)02:00
ThihiIf all goes as planned.02:00
ThihiYeah.02:00
Thihi:)02:00
wolfspraulyes02:00
wolfsprauland that will not update the very first bitstream (called 'standby bitstream')02:00
wolfspraulwe also made a small change in the standby bitstream by removing standby mode02:00
wolfspraulthat is you don't have to press the middle button anymore to boot, it will boot automatically after power-on02:00
wolfspraulif you want to get that, you need to update the standby bitstream with either ftp or jtag, not exactly sure02:01
wolfspraulbut let's update the rest first, that's far more important than the 1 fix in the standby bitstream02:01
wolfspraulthat's about all02:01
ThihiOk.02:01
ThihiSounds cool. I'll try it on friday.02:01
wolfspraulit would be great if you would be live in the channel here when you do the update02:01
ThihiSure.02:01
wolfspraulthat's helpful for you and everybody else here to smooth out things, misunderstandings etc.02:02
wolfspraulfirst impressions are priceless!02:02
wolfspraulplease share them02:02
wolfspraulall the little things you run into that confuse you...02:02
wolfspraulxiangfu: hi good morning03:06
xiangfuHi03:07
wolfspraulThihi got a review M1 with the original 07-13 software. can he just run web update and it will all get updated?03:07
wolfspraulincluding latest patches and pictures and all?03:07
xiangfuwolfspraul, yes. I can't remember if there is 'webupdate' button in 07-13. if there is the button. yes.03:10
xiangfuwhen click 'webupdate' the m1 will get updated include all patches not picture.03:11
xiangfuafter update to latest version. click webupdate again. will download the pictures.03:11
xiangfuwolfspraul, 'webupdate' will not touch 'standby.fpg'03:12
wolfspraulthere is an update button in 07-1303:12
wolfspraulbut dhcp will only happen at boot time, I think03:12
wolfspraulcan Thihi update the standby with ftp? or only jtag?03:12
xiangfuok. then 1. webupdate under 07-13 will update soc, flickernoise, patches only .fnp files03:13
xiangfu(dhcp) there is DHCP button under 'setting' so. if failed click the button again. or manually set the ip address03:14
xiangfustandby only in jtag.03:14
wolfspraulok03:14
wolfspraulthanks!03:14
n0carri3revening guys :)03:32
wolfspraulgood evening!03:35
wolfsprauln0carri3r: hey, sorry, but is there a chance still for me to get a high-res download link for the cheap dinosaurs video?03:35
n0carri3rahh, emily never replied?03:36
wolfspraulnot to me03:36
n0carri3rOK i'll remind her :)03:36
wolfspraulyou think she is ok with it being freely licensed?03:36
n0carri3ri'm sure03:36
wolfspraulI really hate to bug people on this licensing mess, it's such a drag03:36
wolfspraulok, done03:36
wolfspraulI will scrape from youtube and use that03:36
wolfspraullet's focus on the good stuff03:36
n0carri3rthats good enough?03:36
wolfspraulthat's the best considering that anything else bugs people and wastes their time :-)03:37
wolfspraulhow about your patches? can you find time to upload some? (we need to make that easier as well, asap)03:38
wolfspraulwhich controllers do you typically use?03:39
wolfspraulwe are quite serious on better controller integration now, maybe we can focus on the one you have if it's easy to get/cheap...03:39
n0carri3ri'm hooking up the MM now03:39
n0carri3ri'll clean up and copy patches to upload :)03:39
wolfspraulexcellent03:39
n0carri3ryou mean MIDI controllers?03:40
wolfspraulyes03:40
wolfspraulor OSC03:40
n0carri3rhavent had time to experiment with either03:41
n0carri3rhowever, i have a few MIDI controllers around here03:41
n0carri3rso far i've been using keyboard and mouse, but i really want to use USBMIDI or OSC soon!03:41
wolfspraulwhich controllers do you have?03:45
wolfspraulcan you give us a quick list?03:45
n0carri3ryeah one sec03:45
wolfspraulwpwrak: should I make a quick more realistic invoice for your package? I can send you the PDF if that helps you03:48
n0carri3rwell, i have a few around here, mostly my roommates03:49
n0carri3rbut i would like to use the nanokontrol03:49
n0carri3ras it has buttons, faders, and dials on it03:49
n0carri3rand its only about US$4003:50
wolfspraulok03:54
wolfspraulwerner has a nanokontrol203:54
n0carri3rbacking up my SSD now03:55
n0carri3rthen gonna update - finally :)03:55
xiangfuwolfspraul, http://www.openmobilefree.net/?p=121503:55
xiangfuI order 5 DMX cables. for connect those DMX device together.03:58
wolfsprauln0carri3r: any other controllers you like/would like to use?03:59
n0carri3rthat one is very portable, so i would probably start with that04:00
n0carri3rby the way, how is OSC currently implemented?04:01
n0carri3rdo i have to have it hooked up to a computer for network sharing?04:01
wolfspraulprobably the Ethernet has to be plugged into a router or so that can answer DHCP and make the m1 visible on the network04:02
wolfspraulxiangfu: is that correct to get OSC working?04:02
n0carri3ri see04:02
n0carri3ri am updating data, with the ethernet cable into my laptop04:02
n0carri3rwith internet sharing turned on04:02
n0carri3rwould that work for OSC, too?04:02
xiangfun0carri3r, yes. that is also ok.04:02
n0carri3rOK, great.04:03
xiangfun0carri3r, yes. sure.04:03
n0carri3ri'll try it with touch OSC04:03
xiangfun0carri3r, as long as network works. OSC works. :)04:03
n0carri3r(for iPod / iPad)04:03
xiangfun0carri3r, hmm... touch OSC is different.04:03
n0carri3roh?04:03
xiangfun0carri3r, the 'address' of touch OSC is hardcode04:04
xiangfun0carri3r, and not meet with m1's osc.04:04
n0carri3rahh i see04:04
wolfspraulxiangfu: why not? how can we make this work?04:05
xiangfuwolfspraul, hardcode the address to flickernosie. make it meet with touch OSC04:06
xiangfuwolfspraul, but I think every OSC application is different and all hardcode inside.(I tried three different OSC apps under Android)04:07
xiangfuall of them is different.04:07
xiangfuwolfspraul, the best is we could configure those under flickernoise.04:07
xiangfuthat is needs some work.04:07
xiangfuhttp://www.milkymist.org/wiki/index.php?title=Flickernoise_user_manual#OSC04:08
wolfspraulyou mean ports?04:09
wolfspraulcan't we just listen on multiple ports?04:09
xiangfunot ports.04:09
xiangfunow we have four 'addresses' : /midi, /patch, /variable, /osd04:09
xiangfuthe touchOSC output would like: /1/fader504:12
xiangfun0carri3r, oscdump and oscsend will help a lot at begin.04:13
n0carri3rthe update worked! the star ballet patches look good :)04:14
n0carri3rgonna try some of my own images now!04:14
xiangfuOSC, have hostname, port, address and types of address.04:14
xiangfun0carri3r, great. :)04:15
n0carri3rhmm when i FTP in, something is different now04:16
n0carri3ri can't seem to access the directories04:16
n0carri3rbefore i saw ramdisk, SSD, and dev04:17
n0carri3rnow i see dev and SSD, but i can't seem to view the contents of SSD04:17
wolfsprauln0carri3r: did you backup your patches somewhere?04:20
wolfspraulI hope you don't loose them, it already happened a few times04:20
n0carri3ri did :)04:20
n0carri3rjust before i updated04:20
wolfspraulphew, great04:20
n0carri3rbut they are still on the MM04:20
wolfspraulok good04:20
n0carri3ri just cant seem to FTP in correctly for some reason04:20
n0carri3rhmm this is strange04:22
wolfspraulsorry have to run, bbiab04:23
n0carri3rxiangfu: any idea why i can't access the SSD after the update?04:26
n0carri3ri'm FTP'ing in to the MM, but i can't access the directory04:27
wolfspraulI think we changed the root04:29
wolfspraulso that people don't accidentally write to an area that will get lost on a power cycle04:29
wolfspraulso maybe the former /ssd is now the root? I am not sure...04:29
n0carri3ryeah i cant seem to access it at all now04:30
n0carri3rwhich isnt good!04:30
wolfspraulwhat do you want to access?04:31
wolfspraulsorry have to run, hopefully xiangfu is back soon04:31
wolfsprauljust write what you want here, we'll do it04:31
n0carri3ri just want to FTP in, so i can back up new patches i write04:32
n0carri3rand right now, i want to FTP in so i can upload new images to the SSD04:32
n0carri3rto use in patches04:32
n0carri3rbut i cant access the SSD folder now04:32
wolfspraulbut you maybe in the /ssd folder already04:32
wolfspraulit's the root now04:32
n0carri3ri see the dev and ssd folders when i log in there04:32
n0carri3rso i dont think i'm there04:32
wolfspraulhmm04:32
n0carri3rand if i am, i dont see any of my patches04:32
n0carri3rthey are all there, as i can see them on my MM if i browse files04:33
wolfspraulsounds like a bug/regression then04:33
n0carri3rhmm04:33
wolfspraulwe'll figure it out04:33
wolfspraulbbiab04:33
n0carri3rahh seems to be filezilla04:47
n0carri3ri can connect with another software04:47
n0carri3rstrange!04:47
n0carri3rhmm but the files seem to be read-only, so i cant download stuff from the SSD now04:51
n0carri3rnow it seems to be working04:56
n0carri3rwith transmit for OSX04:56
n0carri3rbut not with finder or filezilla04:56
xiangfu<n0carri3r> yes. that is a bug07:26
xiangfu<n0carri3r> for now ftp is not work with some GUI ftp applicatioins. command line tools 'ftp' always working.07:26
xiangfu<n0carri3r> i will fill a bug about this.07:26
lekernel_we have not changed the root, the FTP configuration is exactly as before08:24
lekernel_I simply think RTEMS introduced some bugs in their FTP server08:25
lekernel_the only difference between the 1.0rc1 (July 13) software and the later versions is that RTEMS has been synced with upstream, as far as FTP is concerned ...08:26
wpwrakyou'd think after all those years, people would get FTP right ;-)08:28
xiangfu(simply think RTEMS introduced some bugs in their FTP server) this is true.08:36
wpwraki suspect it may be quoting. well, or just some general string operations issue08:39
wpwraknice variable name: "wave_mystery"08:54
xiangfuScopeuk, Hi09:12
xiangfuScopeuk, about those DMX devices. :) the 'DMX desk' under flickernoise. works just fine.09:13
xiangfuI can connect those DMX devices to Milkymist one. and use 'DMX desk' control them.09:13
xiangfuI still not test connect more then one Devices. because I only have one DMX cable. I order 5 more.09:14
xiangfushould be ok.09:16
xiangfuand flickernosie have 8 output channels, that is great.09:17
xiangfuall those device have 'sound active mode' but I don't think it works well. because the FAN nosies inside is louder then music a lot.09:18
wpwrak"sc->p->pervertex_prog" sweet ! ;-)))09:18
xiangfuwpwrak, I will try to do same thing you did with MIDI controller. but I will  use one DMX controller. :)09:20
xiangfuwpwrak, should be ok. since they are all variables under flickernosie.09:20
wpwrakyup, that should work fine09:21
wolfspraulxiangfu: sorry I didn't understand our OSC 'address' problems09:41
wolfspraulso M1 has an IP address, and OSC works over one (or several?) ports?09:42
wolfspraulwhat's the address issue?09:42
wolfspraulis there some sort of OSC namespace that we have to fit into?09:42
xiangfuwolfspraul, wait one moment. back in ~20 minutes. sorry.09:44
wolfspraulsure, take your time09:44
aw_wpwrak, according to pericom's tmds trace length @( 280 mil = 7.1 mm ) 1/16 wavelength and @( 370 mil = 9.3 mm) 1/3 length of rising edge of signal, those results derived and based on FR4.10:09
aw_wpwrak, I am still reading that application note to digest.10:13
lekernel_what's wrong with sc->p->pervertex_prog ?10:14
aw_you mentioned fpga pins reassignment is inevitable. especially to use bank2 with differential tmds signals. hmm ... the wires from bank2 now is quite messy with usb/flash/video-in.10:20
lekernel_wolfspraul: http://liine.net/en/products/lemur/10:22
wolfspraullekernel_: interesting, nice11:23
wolfspraulso yes, they focus on software and use the ipad11:23
wolfspraulI think that move makes sense if you have an existing business to protect11:23
wolfspraulI mean just looking at all this from the distance11:23
wolfspraulyou cannot compete with the ipad head on, better to use it, then think about next hardware steps, if any :-)11:23
wolfspraulaw_: I think we plan to avoid moving currently assigned pins12:25
wolfsprauldefinitely for now. it would cause difficulties to build and test bitstreams.12:25
aw_until now, directly wiring 14 pieces differential signals from messy 'overlayed' area from bank2 seems is a must(supposedly), I'm still reading from http://lists.milkymist.org/pipermail/devel-milkymist.org/2011-October/001966.html12:30
wpwrakaw_: (pericom) eh, what's the context ? :)12:31
wpwrakaw_: (inevitable reassignment) what i meant is that i'm certain that, at some point in time, we will have to reassign pins. what i don't know is when that time will come :)12:32
aw_I'd like to manually draw a proposed fpga pins of assignments to see how signals/complexities house may face up to firstly.12:32
wolfspraulyes ok12:33
wolfspraulbut we don't want to move existing pins12:33
wpwraklekernel_: (pervertex_prog) i just find it funny that it has "pervert" as prefix :)12:34
aw_wpwrak, wolfspraul have had this topic goal (don't move existing pins) but at some points. Yes we don't know yet to if having a reassignment soon.12:34
aw_but I'll bring house's idea about layers/difficulties  v.s. cost12:35
wpwrakwolfspraul: (osc) to transport OSC over UDP, you need IP and UDP port. but then, inside the OSC message, you have a string that identifies what entity you want to talk to. e.g., we have one that's called /midi12:36
aw_wpwrak, (pericom) : page 2 from http://www.pericom.com/pdf/applications/AN204.pdf12:36
wolfspraulaw_: we know now that we don't want to move existing pins, at this point12:36
wpwrakwolfspraul: seems that there's no universal convention for these names, so everyone has their own crazy little set. they're basically like path names for files. maybe they took the idea from dbus12:37
aw_wolfspraul, yes, i know. just study only and try to bring more info then plan to meet with house. :-)12:37
wolfspraulaw_: sure, study is always good12:38
wpwrakaw_: (pericom) hmm, looks hairy12:39
aw_wpwrak, check both pericom and this: http://www.xilinx.com/support/documentation/application_notes/xapp495_S6TMDS_Video_Interface.pdf (see page 11) no matter placing hdmi receiver or transmitter, the tmds signals are much close to connector. as a matter. i.e. tmds signals should as closer as possible to dvi-i connector.12:48
lekernel_they conveniently do not mention 1080p12:50
wpwrakaw_: JA (the one directly connected) doesn't seem be extremely close ?12:52
wpwrak(the others have buffers, so that may change the rules)12:52
aw_wpwrak, (the buffers) it probably be the goal. so if directly wiring from fpga, it seems that we misunderstood at all. If we are now looking at that (pericom) theory shows.12:54
wpwraklekernel_: maybe they follow buddhism: "In the case of words that the Tathagata knows to be factual, true, unbeneficial, unendearing and disagreeable to others, he does not say them." ;-)12:54
wpwrakhttp://en.wikipedia.org/wiki/Noble_Eightfold_Path#Right_speech12:54
aw_wpwrak, i just not sure if my study/understand is correct or not. so just I just pointed up.12:55
aw_so if yes, adding buffer plus moving vga encoder are also a super tough work under current board size. If I read misunderstood, please correct me.12:57
lekernel_why is Adam still discussing adding a buffer? we won't.13:07
wpwraki think he's just looking at how other people do this and why they make certain choices13:09
kristianpaulfrom llhdl mail list in case not all people here is suscribed14:35
kristianpaulhttps://blogs.msdn.com/themes/blogs/generic/post.aspx?WeblogApp=satnam_singh&y=2011&m=01&d=18&WeblogPostName=reconfigurable-data-processing-for-clouds&GroupKeys=14:35
kristianpauloops14:36
kristianpaulhttp://blogs.msdn.com/b/satnam_singh/archive/2011/01/18/reconfigurable-data-processing-for-clouds.aspx14:36
kristianpaulthis looks prety interesting http://legup.eecg.utoronto.ca/14:51
kristianpaulalso uses llvm, lekernel_ may take a look ? ;)14:52
kristianpaulOnly non-commercial, not-for-profit use of this software is permitted, bah :(14:52
kristianpaulbah..14:53
wolfspraulkristianpaul: :-)14:53
wolfspraulamazing, eh?14:53
wolfspraulthere are more stylish ways to express that you feel good in the past, but some people lack that style so they do this -nc stuff instead :-)14:55
kristianpauluniversities funded reserach, with its "excuse" of _only_ educational stuff, waiting for the best bif from a big corp14:56
kristianpaulis like a myth, i have a frien work on a social project for a public univrsity here, i asked for lincese of his work, and he reply all copyright and right belong to university, (even with this is for public common wealth??)14:57
kristianpauloh well14:57
lekernel_c-to-verilog is a dead end14:58
kristianpaulmay be support in c for verilog based libraries is a better aprouch15:00
kristianpauljust like your libfpvm15:01
wolfspraulkristianpaul: I think academia has missed the free culture boat15:01
lekernel_also, most academic tools do not work and/or are written using inane languages such as slowc++ (aka java)15:01
kristianpaulhe, i agree15:01
wolfspraulwe are sailing on our own now, but I'm not looking back15:01
kristianpauljava, yeah :( first language i was tought at university :(15:02
wolfspraulthat's certainly true for computer science, haven't made up my mind yet about electronic engineering/ic design etc.15:02
lekernel_it should be called "slowc++" :) decribes pretty well what's it's all about: combining the unfriendliness of C++ with the slowness of interpreted languages15:03
wolfspraulthere's too little free culture in those areas right now (which hopefully Milkymist will change), so I cannot tell15:03
wpwrak"--c" ? :)15:28
wpwrakwolfspraul: i don't think CS has missed free culture. lots of projects are open source, including big ones. you should have seen the sorry state of things in the 80es and early 90es.15:30
wpwrakwolfspraul: e.g., in OS research, the standard way of doing things was to get a source license from Sun, then hack your stuff, and distribute object files. fast-forward to the late 90es and beyond: if what you do isn't opens source (in the case of linux, with the right terms conveniently being enforced by the GPL), you'll simply be ignored15:32
wpwrakwolfspraul: you may still be able to collect enough grant money to stay afloat, but that's about all you can hope for15:33
kristianpaulhey, xiangfu bought a fpga miner board :)15:37
kristianpaulwow dual spartan6 !15:39
GitHub34[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/1b637cea618828f85ba6ee29ba6fb0178d4607b515:39
GitHub34[migen/master] Instance support - Sebastien Bourdeauducq15:39
rojasHello16:54
wpwraklekernel: after switching to libc, make in software/libfpvm/ fails, as expected. shall i redirect it to lm32-rtems/, since this probably what people want ?17:17
wpwrakanother unhappy day for those who don't like patches in their mailbox17:33
lekernelwpwrak: yes17:47
lekernelactually, remove the lm32-rtems folder, and build everything at the root (no redirect), since this is the only supported configuration17:48
lekernelthough we can keep one subfolder for x86 tests ...17:49
lekerneland i'll remove the demo firmware, don't send a patch for this (it'd be a mess :p)17:49
GitHub165[migen] sbourdeauducq pushed 2 new commits to master: https://github.com/milkymist/migen/compare/1b637ce...c43f3da17:52
GitHub165[migen/master] Simple bus base class - Sebastien Bourdeauducq17:52
GitHub165[migen/master] Wishbone declarations - Sebastien Bourdeauducq17:52
GitHub133[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/ed05ec5f6a2be6eeb332e9040b0a2124a8348e5218:00
GitHub133[migen/master] instances: signal override - Sebastien Bourdeauducq18:00
wpwrak(mess) oh yes ;-)18:03
GitHub11[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/5720a51dad05bd26c51558605dc85d32d16e824618:13
GitHub11[migen/master] wishbone: add missing SEL - Sebastien Bourdeauducq18:13
GitHub121[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/7c99e51b907e76553b1474d18a4434c5e1950fd718:20
GitHub121[migen/master] Named buses - Sebastien Bourdeauducq18:20
wpwrakhmm. changing the makefile breaks makedepend18:32
wpwrakor, rather, makes it a little bitchy18:34
wpwrakwell, let's worry about that later18:35
lekernelhttp://pastebin.com/buSGD6vT19:47
GitHub154[migen] sbourdeauducq pushed 3 new commits to master: https://github.com/milkymist/migen/compare/7c99e51...84eb96420:20
GitHub154[migen/master] corelogic: round-robin module - Sebastien Bourdeauducq20:20
GitHub154[migen/master] verilog: fix unary operator conversion - Sebastien Bourdeauducq20:20
GitHub154[migen/master] fhdl: support negation operator - Sebastien Bourdeauducq20:20
GitHub114[migen] sbourdeauducq pushed 2 new commits to master: https://github.com/milkymist/migen/compare/84eb964...5034af320:29
GitHub114[migen/master] corelogic: MC divider module - Sebastien Bourdeauducq20:29
GitHub114[migen/master] Corelogic conversion example - Sebastien Bourdeauducq20:29
GitHub82[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/512655c108a072773fb910193261c03844300b0820:32
GitHub82[migen/master] fhdl: improve automatic signal naming - Sebastien Bourdeauducq20:32
kristianpaulhow looks verilog code generated from that?21:03
wpwrakkristianpaul: after the ==============21:12
kristianpauloh lol yes21:13
kristianpaulnv, i tought that stil python, too much tabs xD21:14
GitHub141[clang-lm32] jpbonn pushed 150 new commits to master: http://git.io/4w_x0A21:50
GitHub141[clang-lm32/master] Implement name hiding for declarations deserialized from a non-visible - Douglas Gregor21:50
GitHub141[clang-lm32/master] Driver/Darwin: Add ASAN runtime library link support. - Daniel Dunbar21:50
GitHub141[clang-lm32/master] When we're passing a vector with an illegal type through memory on x86-64, use byval so we're sure the backend does the right thing.  Fixes va_arg with illegal vectors and an obscure ABI mismatch with __m64 vectors. - Eli Friedman21:50
GitHub125[llvm-lm32] jpbonn pushed 150 new commits to master: http://git.io/HmbX5A21:52
GitHub125[llvm-lm32/master] LSR: prune undesirable formulae early. - Andrew Trick21:52
GitHub125[llvm-lm32/master] Rename "Trap Values" to "Poison Values", to better reflect their - Dan Gohman21:52
GitHub125[llvm-lm32/master] Line up the comments in a code example. - Dan Gohman21:52
GitHub18[migen] sbourdeauducq pushed 3 new commits to master: https://github.com/milkymist/migen/compare/512655c...c1041b922:10
GitHub18[migen/master] verilog: handle default in case statements - Sebastien Bourdeauducq22:10
GitHub18[migen/master] corelogic: multimux module - Sebastien Bourdeauducq22:10
GitHub18[migen/master] simplebus: export GetSigName function - Sebastien Bourdeauducq22:10
wpwrakthe example looks fairly cryptic. at least i can figure out what it does in the verilog form :-)22:13
GitHub155[migen] sbourdeauducq pushed 1 new commit to master: https://github.com/milkymist/migen/commit/5c7131dc86f6ea84a66e37c4371ca3b1c902704d22:25
GitHub155[migen/master] wishbone: arbiter - Sebastien Bourdeauducq22:25
lekernelcryptic? well, it's concise and you probably have to dive a bit in the code to understand it :)22:29
lekernelhow about this one? http://pastebin.com/LGsRADak22:29
lekernel(ofc the [cpu.ibus, cpu.dbus] list can be arbitrarily extended and it does the right thing)22:29
wpwrakand how does that "glue" translate to verilog ? (i support sharedbus and arbiter themselves incorporate major functional blocks and thus produce a huge amount of verilog ?)22:36
kristianpaulcan it generate FSMs?22:37
wpwrakand what's a "fragment" ? what would be the corresponding whole ?22:38
lekernelit generates code comparable to this little mess: https://github.com/milkymist/milkymist/tree/master/cores/conbus/rtl22:38
lekernelyes, it can generate FSM22:38
lekernela "fragment" is a piece of logic (combinatorial and/or sequential)22:39
lekernelthe exact generated code here is http://pastebin.com/Jrw5CN1B22:43
wpwraklooks like a case for macros ;-)22:44
lekernel?22:45
lekernelamong many other limitations, you can't automatically generate case values with macros22:45
wpwrakthe .v examples (with a lot of redundancy) look like things a macro processor could handle well.22:46
wpwrakwhy not ?22:46
lekernelwell, the verilog preprocessor doesn't handle it22:46
wpwrakthe generated code looks different - seems to be mainly "library" and little "glue", though ?22:47
lekernelof course you can have your own macro preprocessor... or something more powerful :)22:47
lekernelhere there's a quite bit of "library", but in the previous example you'd have a hard time using macros22:47
wpwrakyeah. if the standard macro preprocessor ain't good enough, do your own :)22:47
lekernelbtw, the library here uses the code from the previous example, only with n=222:48
lekernelalso, macros are bound to a language, here generating VHDL is semi-trivial22:48
kristianpaullast pastebin doest look that bad compared to conbus indeed :)22:48
wpwraki see one limitation of macros would be that you have "N of these" constructs. so you'd need some kind of iterator in your macros. doable, of course.22:48
lekernelhere I have python iterators. much more powerful, and I'm not reinventing the wheel :)22:49
wpwrak(n=2) yes, that's why the generated part doesn't look very redundant22:49
lekernelredundancy isn't a problem as long as it's machine generated :)22:50
wpwrakyeah :)22:52
wpwrakwhat's what we have those machines for :)22:53
ThihiDarn.23:25
ThihiThe person who sent me my review milkymist didn't send the euro area ac adapters.23:26
ThihiWell, I guess I'll go buy a few uk -> euro adapters.23:27
--- Fri Dec 9 201100:00

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