#milkymist IRC log for Wednesday, 2012-06-06

wolfspraultoo bad that the webirc guy left without a trace03:03
wolfspraulangry webirc man - please come back! :-)03:04
sh4rm4he's probably as pissed as me when i got my C-one and found it needs one of those old 30 KHz screens. it's collecting dust since.05:29
wolfspraulC-one?05:30
sh4rm4the "commodore one"05:30
wolfspraulwell I would like to know more background info. where did he buy his m1? what was his expectation? which message on the box did he mean? and so on05:30
sh4rm4sure05:30
wolfspraulI didn't sell a single M1 in two months so there are not that many ways how he could have gotten one, if the story is true at all05:30
wolfspraulnot that I wouldn't be surprised about that kind of reaction, given that most things that people expect to work won't work05:31
sh4rm4but then, if you only sold one MM the last month, its easy to find out...05:31
wolfspraulbut anyway, first he needs to come back, in the meantime we make m1 better :-)05:31
wolfspraulbetter than that, I sold zero since March05:31
wolfspraulso I first want to know whether that is not just a simple troll :-)05:32
wolfspraulwebirc guy - please come back...05:32
wolfspraulsh4rm4: when you are out in the real world, selling real products for real money, you will quickly realize that the ratio of bullshit talk to 'just' talk to serious talk to action is about 100:1:0.01:0.000105:34
wolfspraulthat's my experience at least05:34
wpwrakif he's so quickly incensed, he may calm down as quickly as well :)05:35
rohmorning06:35
rohhm c-one needs a what? it says vga out in the web06:37
sh4rm4yeah but apparently needs one of those multiscan screens attached to that.06:40
Fallenoumorning07:42
lekernelhi08:59
lekernelso, let's see how we could encode video... hmm.... http://opencores.org/forum,OC%20H.264%20project ...oh, well.09:08
Fallenouopencores.org/project,intra_predection_h264 < empty :'09:13
wolfspraullekernel: thanks a lot for the links to xdl and s6bitstream the other day10:13
wolfspraulunfortunately I got a high fever for a few days, but finally it's gone :-) first in years though, good10:13
wolfspraulyes I knew xdl and s6b already, sure10:14
wolfspraulbitgen does not support going back from .bit to .ncd/.xdl, one problem10:14
wolfspraulplus I really just want to learn about the format and write a small converter both ways, so it's easy to convert a .bit to text, change something, and back to .bit10:14
wolfspraulI've made my way through the headers and commands, now looking at the actual frame data... a lot to learn10:15
wolfspraulone quick question: do you know any free software replacement or alternative to ise's fpga_editor ?10:16
lekernelof course not! this would require understanding of the fpga architecture.10:17
lekernelor well, there's rapidsmith - but it doesn't do everything10:18
lekernelI never managed to run it, either (it's java)10:18
wolfspraulit seems rapidsmith is unmaintained now?10:38
wolfspraulanother question: is there a tool that compiles verilog down to simple gates and netlist?10:41
lekernelthere used to be an option for that in icarus, but it didn't work very well and was removed10:42
wolfspraulthat's all?10:44
wolfspraulso rapidsmith is unmaintained?11:10
wolfspraullooks like they ran through their government money and then stopped somewhere? no way I will sink much time into that beyond looking at the screenshots and grepping throuhg their sources for any bits of wisdom11:11
wolfsprauland no verilog compiler (to gates) exists either, urgh11:12
wolfspraulfpga really seems to be the land of dead & abandoned projects11:12
lekernelnot fpga... relatively complex logic design in general11:42
wolfspraulrapidsmith is unmaintained?11:43
wolfsprauldo you know more?11:43
wolfspraullooks like it stopped over 1 year ago11:43
wolfsprauland for the removed icarus verilog compiler - any other idea, at all?11:43
wolfspraulhow was llhdl supposed to work?11:43
lekernelno, I don't know more about rapidsmith11:43
lekernelllhdl is verilog -> netlist as well11:44
wolfspraulcomplete verilog or just a subset?11:44
lekernelatm a very limited subset11:44
wolfspraulthen you need to implement the whole language and optimize away all sorts of inefficient stuff one *can* write in Verilog, no?11:45
wolfspraulyeah, that's what I thouht11:45
wolfspraulwell at least since we control all our sources we can easily compromise on the other side as well, and only support the subset we need11:46
wolfspraulmigen only creates the 'basic' verilog that a future llhdl could easily support?11:47
wolfsprauljust trying to understand how all this stuff might or might not fit together11:47
lekernel(migen) yes11:48
lekernelbut the main goal of migen really is to be able to build more complex/flexible/fast architectures...11:52
wolfspraulsure, understood11:54
wolfspraulthanks!11:54
wolfspraulthat is helpful11:54
lekernelnerd-fu, heh? :)11:54
rohwolfspraul: sounds like you do a inventory of whats there and missing ;) like it12:17
roh .oO(which is the first step on every sane battle-plan)12:18
kristianpaulbah trying autotools now after non sucess with cmake 2.8814:07
Fallenoulekernel : who is genesis on frlab ?14:35
AlarmHello all. Is it possible to easily change the colors of the interface configuration Flickernoise? Orgines colors gives a pale color on a video projector.14:38
lekernelyeah, edit the mtk source code14:43
wolfspraulFallenou: saw your long update mail, congratulations!14:45
Fallenouwolfspraul: because you did not post instantly about it I thought you were sick, I was right ^^14:46
Fallenouthx14:46
Fallenoustill a lot of work, but it's going in the right direction14:47
wolfspraulI was quite sick at the time so couldn't respond :-)14:47
wolfspraulyour working style & persistance and how you communicate it are quite amazing14:47
wolfspraulthat's the way to go14:47
wolfspraulyeah14:47
wolfspraulI was, pretty bad fever for 2 days14:47
wolfspraultoday is good though :-)14:47
wolfspraulI can imagine14:47
Fallenougood that you're OK now !14:48
wolfsprauldefinitely, yes. thanks.14:49
FallenouAs I make progress rather slowly, I try to communicate when I make progress, in order to show that things are not frozen14:50
lekernelthat's great14:50
Fallenoubut yes I'm kind of happy of the latest results14:51
Fallenoufeatures start to show up :)14:51
lekerneland you definitely make more progress than the other 89 persons on #frlab, for example :)14:51
Fallenouthat's a good point :p14:51
wolfspraulsure that's no problem at all14:52
wolfspraullook at kristian pauls gps work, and you will appear fast like a ferrari :-)14:52
wolfspraulbut I don't think speed is our problem right now. I am not sure what will help milkymist to greater success, but somehow I believe in high quality & reusable pieces of tech and code.14:52
wolfspraulI think it's more important that the eventual outcome is really high quality, low maintenance, very easy to use and reuse into the future. so it becomes a building block.14:52
Fallenoureuse <= MMU implementation is cpu dependant, but it should be reusable on any platform using lm32 :)14:53
kristianpaul:p14:54
wolfspraulkristianpaul: you know I really like your gps work, but one great thing about your work is - nobody has to be worried to be slow anymore!14:57
wolfspraulthat's good!14:57
wolfspraulI don't like this constant hectic stuff14:57
lekernelis that a comment about me? :)14:59
kristianpaulme14:59
wolfspraulthat was a praise of kristianpaul :-)15:02
Fallenouit's clear that for M1 (as a tech device as well as a product): the faster anyone here works, the better. But since we're all volunteers and most of us have day jobs we're going at the pace/rythm we can. I think everyone knows and understands this :)15:09
Fallenouand it's fine for anyone15:09
GitHub105[milkymist-mmu] fallen pushed 1 new commit to mmu: http://git.io/n_5xDw17:50
GitHub105[milkymist-mmu/mmu] Add new BIOS commands to play with MMU - Yann Sionneau17:50
Fallenoupage fault caused by a memory store was causing a SoC hang because of the write-through. I was still sending write instruction on the wishbone bus, even on dtlb_miss (page fault)18:05
Fallenoubut pfn was 018:06
Fallenouso it was accessing 0x0000YYYY :)18:06
Fallenouand boom "null pointer" hardware protection hangs the SoC18:06
FallenouI think I fixed it, resynthetizing18:06
wpwrakthe null pointer catcher - greatest debugging aid ever :)18:10
Fallenouyes but it's not very verbose :)18:11
Fallenouyou just see the SoC hanging18:11
Fallenoubut with ISim's help I saw d_adr_o with the incorrect value :)18:12
Fallenoueven though in simulation it was "working" (i.e not hanging) because I have no such null pointer catcher in my migen generated soc for simulation :)18:12
GitHub17[milkymist-mmu] fallen pushed 1 new commit to mmu-bios: http://git.io/5s6SYA18:14
GitHub17[milkymist-mmu/mmu-bios] Add support for writting to memory with MMU activated - Yann Sionneau18:14
Fallenouyes, fixed :)18:16
Fallenouhttp://pastebin.com/PsGDg3iA < now it does this instead of hanging the SoC :)18:17
GitHub0[milkymist-mmu] fallen pushed 1 new commit to mmu: http://git.io/2_NJNw18:19
GitHub0[milkymist-mmu/mmu] Fixed a bug upon page fault when caused by a memory store which hung the SoC - Yann Sionneau18:19
GitHub83[milkymist-mmu-simulation] fallen pushed 1 new commit to master: http://git.io/fGFOtQ18:22
GitHub83[milkymist-mmu-simulation/master] Fixed a bug upon page fault when caused by a memory store which hung the SoC - Yann Sionneau18:22
Fallenouhum cache seems to be messed up upon store :)18:24
Fallenougn8!20:31
kristianpauln821:00
--- Thu Jun 7 201200:00

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