Topic for #milkymist is now Radical Tech Coalition :: Milkymist One, Migen, Milkymist SoC & Flickernoise :: Logs: http://en.qi-hardware.com/mmlogs
<wolfspraul>
ok, playing the unsuspecting m1 user...
<wolfspraul>
I got a free mouse with a notebook bag I bought over the holidays, tried it on m1, but the mouse moves so fast vertically and very slow horizontally
<wolfspraul>
(still on 07-13)
<wolfspraul>
I just bought a new rubber keyboard in Media Markt, works in my notebook, doesn't work on m1
<wolfspraul>
now... onto the update! :-)
<wpwrak>
if you upgrade from july, mice will regress a little and keyboards will get a little better
<wolfspraul>
let's see :-)
<wolfspraul>
I got 2 new devices (normally I keep a strict set of 'works with m1')
<wolfspraul>
plugging both in - both don't work
<wolfspraul>
it can't get worse :-)
<wpwrak>
excellent :)
<wolfspraul>
wow, I just notice the reflash_m1.sh script stands at 389 lines!
lekernel_ [lekernel_!~lekernel@g225032136.adsl.alicedsl.de] has joined #milkymist
<wolfspraul>
built urjtag from source (on fedora 16), worked with a few hickups
<wolfspraul>
ah yes, busname is in the command, "fjmem" is the bus
<wpwrak>
does help initbus mention fjmem ?
<wpwrak>
as in "fjmem FPGA JTAG memory bus driver via USER register, [...]"
<wpwrak>
ordering is only almost alphabetical. it would be too easy otherwise :)
<wolfspraul>
oh well
<wolfspraul>
when you are in the jtag console (just run 'jtag'), then 'cable milkymist', then 'detect' - what output do you see from the 'detect' command?
<wolfspraul>
somehow when I do it interactively it just returns without anything
<wolfspraul>
and subsequent commands say 'run detect first'
<wolfspraul>
it works in the scripts, argh
<wolfspraul>
good thing that I don't have to fix C compiler bugs before compiling urjtag before finding fjmem before reflashing before booting
<wpwrak>
one problem is that urjtag doesn't return all errors. so neither reflash.sh nor m1nor can actually catch problems happening inside urjtag (jtag comand)
<kristianpaul>
afai my laptop bios got crazy with F16 and installed the damn xubuntu alpha :/
<wolfspraul>
now my m1 suffers from d2/d3 dimly lit :-)
<kristianpaul>
*g*
<wolfspraul>
need to replug dc a few times...
<wolfspraul>
all normal
<kristianpaul>
thats souns like my rc2 ;)
* kristianpaul
loves fedora bugzilla
<wpwrak>
wolfspraul: what revision is it ? rc3 ?
<wolfspraul>
I found when I wiggle the power cable a little when inserting, I can make the dimly lit go away
<wpwrak>
eek
<wolfspraul>
rc2
<kristianpaul>
i knew it !
<wpwrak>
ah. before my time :)
<wpwrak>
rc3 never does "dimly lit" unless the NOR got corrupted
<wolfspraul>
yes! it renders!
<wolfspraul>
now...
<wolfspraul>
first the old mouse
<wpwrak>
drum roll ...
<wolfspraul>
works
<wpwrak>
whee ! no regression :)
<wolfspraul>
new mouse: same as before, unusable
<wpwrak>
:-(
<wolfspraul>
vertical is crazy fast, horizontal very slow
<wolfspraul>
plugged into my notebook - all fine
<wpwrak>
some mice are odd. not sure yet why. i have one that's slow on x and y. need to find out how linux gets it right.
<wolfspraul>
ok, will keep the old one and carry the new one around. probably dump with xiangfu who is collecting all the gear that doesn't work ;-)
<wolfspraul>
next, the new keyboard
<wpwrak>
how about the keyboard ?
<wpwrak>
yeah :)
<wpwrak>
from the silence, it probably detonated and left wolfspraul searching for his laptop in a smoke-filled room
<wolfspraul>
no no
<wolfspraul>
it worked
<wolfspraul>
but of course I ran into another problem
<wolfspraul>
after switching the gui from 1024x768 to 640x480, my projector couldn't pickup the signal right anymore
<wpwrak>
heh :)
<wolfspraul>
so the display was shifted (rolling around on top)
<wpwrak>
sounds interesting :)
<wolfspraul>
turning off the projector didn't help, strange
<wolfspraul>
I am rebooting everything now
<wpwrak>
did that keyboard work before ?
<wolfspraul>
he
<wolfspraul>
no it did not!
<wpwrak>
if you have a projector that fails to work, that may be VERY useful. that would be the first truly reproducible failyre.
<wolfspraul>
after rebooting everything from total power down, the m1 was in rendering mode
<wpwrak>
excellent. progress has been achieved ! :)
<wolfspraul>
and my projector first picked up the wrong vertical shift again!
<wolfspraul>
but after a few seconds, it's back to normal
<wolfspraul>
now GUI...
<wolfspraul>
yeah, normal in 640x480
<wolfspraul>
go figure
<wolfspraul>
I can probably reproduce this first going into the GUI at 1024x768 and then switching down to 640x480
<wpwrak>
nobody said it would be easy :)
<wolfspraul>
but given the number of small issues here and there, that's minor
<wpwrak>
well, it may be a lead to fixing the video issues
<wpwrak>
alright, now you should be able to use the icreativ. you probably won't need the keyboard. so connect it just there.
<wpwrak>
then go to the midi monitor and see if the icreativ is understood
<wolfspraul>
my keyboard comes out with a german layout, after full jtag reflashing?
<wolfspraul>
is that stored in some place that was not overwritten, or is that our default?
<wpwrak>
not sure how much got deleted in the "full reflash"
<wolfspraul>
anyway, I'm happy
<wolfspraul>
the keyboard works
<wolfspraul>
great
<wpwrak>
usb keyboards can actually tell you what layout they are :)
<wolfspraul>
since my old silicone keyboard broke
<wpwrak>
phew :)
<wpwrak>
so .. did you try the icreativ ?
<wolfspraul>
not yet, really too late
<wolfspraul>
I need to fix the problem on fedora, will check kpaul's links tomorrow
<wolfspraul>
the new mouse goes to xiangfu's pile
<wolfspraul>
new keyboard works - awesome
<wolfspraul>
tomorrow I continue with icreativ and the other controllers
<wolfspraul>
have to get a midi cable still
<wpwrak>
(kbd) let's see how long until i break it again :)
<wolfspraul>
oh wait, the usb-midi controllers don't work on m1 yet, do they?
<wpwrak>
they should
<wolfspraul>
ok
<wolfspraul>
I will just plug it together and watch in amazement what happens
<wpwrak>
didn't try exhaustively. but it recognized the one i plugged in
<wolfspraul>
ok
<wolfspraul>
so usb-midi works
<wpwrak>
surprisingly well so far, yes
<wpwrak>
i'm a little suspicious. if can't be *that* easy. it never is
<wpwrak>
s/if/it/
<wolfspraul>
I go to sleep happily now. this all worked better than expected.
<wpwrak>
despite all the troubles. pessimists have happy lives ;-)
<wolfspraul>
optimist, that's why I plug stuff in!
<wolfspraul>
let me do a quick check whether I can reproduce the vertical shift problem ;-)
<wolfspraul>
no
<wolfspraul>
well then
<wolfspraul>
rendering in simple mode feels faster than before
<wpwrak>
(shift problem) that'll haunt us for a while longer
<wpwrak>
(rendering) dunno
<wolfspraul>
oops, my m1 just froze
<wpwrak>
that's unusual
<wolfspraul>
I wouldn't say that, but we haven't done extensive testing on rc2
<wolfspraul>
I know of one guy in Stuttgart for sure who says his m1 (rc2) is freezing so often and so fast that it is basically unusable
<wpwrak>
naw, in my experience, it rarely freezes nowadays
<wpwrak>
aah .. maybe an rc2 issue
<wolfspraul>
I have been able to reproduce freezing every few hours with my rc2, and I believe xiangfu had it too
<wolfspraul>
just saying
<wpwrak>
how do you reproduce it ?
<wolfspraul>
it's not a crazy bad problem, and we are fixing bug after bug
<wolfspraul>
just let m1 run overnight
<wolfspraul>
mine is back rendering now
<wpwrak>
nope. that doesn't do anything to my rc3.
<wolfspraul>
yes sure
<wolfspraul>
I just respond to you saying 'unusual'
<wolfspraul>
freezing is not unusual
<wolfspraul>
but it's becoming far less common with recent hw & s
<wolfspraul>
sw
<wpwrak>
>= rc3, there's hope that it is :)
<wolfspraul>
kristianpaul: have you seen your m1 freeze?
<wolfspraul>
some of those freezes may also have been related to the L3 and L19 discoveries, I am not sure
<wolfspraul>
it doesn't matter. we just fix one bug after another, until it's perfect.
<wolfspraul>
wow 3am. that is no problem, but alarm at 7.45 am is. calling it a day, n8
<wpwrak>
oh, you still have unfixed L3/L19 ?
<wpwrak>
;-))
<wolfspraul>
mine should be fixed, I am not sure
<wpwrak>
sweet (and quick) dreams :)
<wolfspraul>
we have tried to apply these fixes whereever possible, at every opportunity
<wpwrak>
heh :)
<wpwrak>
at some point in time, you may just retire all the rc<old> you can get a hold of
<wolfspraul>
or offer upgrade, sure
<wolfspraul>
that's already happening actually
<wolfspraul>
I have bought back (!) several rc2, and upgraded to rc3
<wpwrak>
kewl :)
<wolfspraul>
I think Adam has published a hardware upgrade guide in the wiki
<wolfspraul>
one rc2 was donated back to me
<wolfspraul>
I will try to upgrade it to rc3 and make best use of it, maybe as review unit or so
<wpwrak>
i kinda doubt anyone will use that upgrade path. but it's nice to be there, just as a symbol :)
<wolfspraul>
so that is definitely happening, sure
<wpwrak>
L3/L19 should only matter when you connect things
<wolfspraul>
an easy 23 steps
<wolfspraul>
with pictures!
<wolfspraul>
anybody can do it
<wpwrak>
nothing there explains random freezes. kinda odd.
<wpwrak>
yeah, right :)
<wpwrak>
maybe it's the reset circuit. or whatever was there in rc2.
<wolfspraul>
I wouldn't say 'random' freezes
<wolfspraul>
just too little time and energy was spent in tracking them down, finding patterns etc.
<wolfspraul>
they just mostly (all?) went away with the various fixes in hw and sw
<wpwrak>
heh :)
<wolfspraul>
i'm really out now, my m1 rc2 renders overnight, we see... n8
<wpwrak>
in rc3, i only know of the NOR corruption, generally "dead" systems (catastrophic breakdown, like the flux thing), and then the rtems queue corruption
Technicus [Technicus!~Technicus@24-196-36-61.dhcp.stpt.wi.charter.com] has joined #milkymist
cocoadaemon [cocoadaemon!~cocoadaem@2a01:e35:8a99:e90:20d:93ff:fe3b:868c] has joined #milkymist
elldekaa [elldekaa!~hyviquel@abo-168-129-68.bdx.modulonet.fr] has joined #milkymist
kilae [kilae!~chatzilla@catv-161-018.tbwil.ch] has joined #milkymist
<lekernel_>
wpwrak: even with that overengineered and stupid protocol, most USB keyboards won't tell you their layout.
<lekernel_>
I've heard the reason is that adding a programmable part to store the layout would increase the cost compared to a mass-produced keyboard PCB that "works" with all layouts
<whitequark>
they are not present in Xilinx environment, are they?
<kristianpaul>
it depends if you mean build process for sinthesys, so yes
<kristianpaul>
there is one order, let see
<kristianpaul>
ngdbuild
<kristianpaul>
map
<kristianpaul>
par
<kristianpaul>
and bitstream generation, bitgen
<wpwrak>
whitequark: you probably need lekernel for a definite answer. i don't see us enabling IROM anywhere. (would probably be in boards/milkymist-one/rtl/lm32_include.v)
<kristianpaul>
indeed
<kristianpaul>
and yes,lattice ofer a full set of propietary IP cores for making a soc as well
<kristianpaul>
of course we just use the lm32 core :)
<kristianpaul>
also fyi and think you alreadt noticed the jtag part for lm32 was adapted to sparnta-6, plus thre is a debug rom written by mwalle and lekernel
<whitequark>
aha, thanks
<whitequark>
I'll look into it more closely
<whitequark>
that question was from a friend of mine
<whitequark>
he has ported lattice IP cores to use Xilinx library
<whitequark>
*changed reference
<whitequark>
you got the idea.
<kristianpaul>
yes
<kristianpaul>
you friend got a M1 too? :)
<whitequark>
nope
<whitequark>
just several Xilinx boards
<whitequark>
he didn't even knew about m1 until a few days from now
<kristianpaul>
cool!
<kristianpaul>
did you showed yours in action?
<whitequark>
I don't have an M1 (yet) :)
<kristianpaul>
still traveling??
<kristianpaul>
i mean shipping?
<whitequark>
no, I didn't bought it yet
<kristianpaul>
ah xD
<whitequark>
I think I should make everything work in simulator
<kristianpaul>
i imagined !
<whitequark>
because it's the way fpgas are debugged anyway
<kristianpaul>
well, you still need the silicon
<kristianpaul>
but yeah, a testbench can be veryhandy..
<whitequark>
(plus I've recently relocated and, for example, I still need a proper work table...)
<kristianpaul>
just is icarus could run a lm32 core... as it is
<whitequark>
wolfspraul convinced me to try making an MMU for lm32
<whitequark>
I don't need silicon for that
<kristianpaul>
indeed
<kristianpaul>
about mmu, where it goes in milkymist?
<kristianpaul>
because still the FML, so the mmu will be a master core for the conbus?
<kristianpaul>
the TLB is mandatory?
<kristianpaul>
ah yes
<kristianpaul>
what does mean that the cache is tagged?
<kristianpaul>
also this means there is a limit about the number of process can use mmu?
<whitequark>
kristianpaul: I don't know a lot about MMUs and caches, but I know a bit :)
<whitequark>
I can recommend you a book, "See MIPS Run Linux"
<whitequark>
I didn't know anything about caches back when I read it, and it was quite easy to understand, unlike some other material on the topic
<whitequark>
even the most stupid MMUs, like those in MIPS, do not limit the number of processes
<whitequark>
actually, they are not like mmus found in Intel chips
<kristianpaul>
what about arm?
<whitequark>
hm
<kristianpaul>
or sparc?
<whitequark>
I don't know a lot about ARM ones, but I've read about them. ARM ones (at least for Cortex-m3) are slightly more advanced than MIPS ones, but they're still very simple
<whitequark>
and I don't know anything about sparc
<wpwrak>
so what sort of MMU architecture are you aiming for ? MIPS R4000-like ?
<wpwrak>
i.e., virtual cache with physical tags. MMU running in parallel to the cache
<wpwrak>
or a physical cache, MMU between cache and RAM, like MIPS R2000/3000 ?
<whitequark>
and the answer is: I do not know yet. I'm not proficient enough in LM32 arch.
<wpwrak>
okay. not sure it depends a lot on LM32, though
<whitequark>
maybe
<whitequark>
I try to not make preliminary decisiona
<whitequark>
*decisions
kristianpaul [kristianpaul!~kristianp@unaffiliated/kristianpaul] has quit [#milkymist]
<larsc>
there is a discussion about the MMU topic in the MM ML archive
<wolfspraul>
in keyboards, there may be a few more that already work today
<wolfspraul>
I'm in the patch editor, and can't get out. there is some text, and when I move the mouse, it's always in selection mode
<wolfspraul>
cannot press the close button in the top left corner, nor any other button. it's always moving the selection
<wolfspraul>
already tried all sorts of pressing and unplugging/replugging action, no way out yet...
<wolfspraul>
somehow it release the selection mode :-)
<wolfspraul>
released. was able to close the window
<lekernel>
whitequark: we're indeed not using the IROM, but in every case, this proprietary memory description can be easily replaced (as I did for the caches)
<whitequark>
lekernel: okay, got it
<wpwrak>
wolfspraul: hmm, may be slight report-misinterpretation. or a low battery :)
<whitequark>
why they need an explicit memory description at all? Xilinx does a good job guessing it from my Verilog
<larsc>
whitequark: today
<wolfspraul>
wpwrak: sorry don't understand? which report? which battery?
<wpwrak>
wolfspraul: (report) that protocol used for HID devices. we have a few hardcoded formats but there could be a lot more variations.
<wpwrak>
(battery) if the mouse is battery-powered. sometimes, they start acting funny when low on power
<wolfspraul>
not battery powered afaik. totally regular cheap high-volume mouse
<wolfspraul>
free give-away with a notebook bag
<wolfspraul>
the vertical resolution seems 10* faster than the horizontal one. other than that the behavior is quite stable.
<wolfspraul>
it's mostly unusable though because the super sensitive vertical means that you first have to scroll scroll scroll to get the horizontal right, and then the vertical very carefully. unusable.
<wolfspraul>
not very important for me right now, I have another one that works
<whitequark>
larsc: today ?
<larsc>
whitequark: today the tools do a good job at guessing the memory type
Gurty [Gurty!~princess@mtg95-7-78-233-26-36.fbx.proxad.net] has joined #milkymist
<lekernel>
warning: 600-euro registration fee, including (and especially) if you speak. the joys of academia ...
<lekernel>
I wonder how strongly they enforce this, though - I've crashed a few times into expensive academic conferences without paying them a cent, and no one said anything
<wpwrak>
oslo in summer is nice
<wpwrak>
booze is expensive, though. maybe that's where the price tag comes from :)
<lekernel>
phew, most academic conferences have > 300E registration fees
<lekernel>
it's within the "normal" range
<lekernel>
they had FPL2011 in Greece, with 420E "early registration" fee
<wpwrak>
EUR 600 seems a bit high. don't remember how much they were in my days of conference hopping, but i think i would remember anything > EUR 400. so adding inflation, EUR 600 is still high.
<wpwrak>
maybe it depends on the sector and FPGA folks are known to be wealthy ;-)
<lekernel>
it's quite funny how this kind of stuff varies widely depending on the conference
<lekernel>
academic conference: you pay expensive registration, travel, accommodation, food
<lekernel>
security conference: you don't pay any entrace fee, and they often cover travel, accommodation and/or food, and sometimes you even get a speaker fee on top of that
<wpwrak>
yes, but the non-speakers pay for you :)
<lekernel>
yes, of course
<wpwrak>
see, at the academic conferences, you're one of the customers
<wpwrak>
you go there because you need a conference and publication for your curriculum
<lekernel>
oh, and I forgot that in the case of the academic conference, they make you sign off your copyright (for free) and sell your paper later on
<lekernel>
this stuff is crazy
<wpwrak>
why do fat cats like fat cat business ? :-)
<wpwrak>
but don't worry, open access will kill them
<lekernel>
that's why I have absolutely no moral issue with crashing in and not giving them a cent
* whitequark
has memorized that.
<larsc>
lekernel: do you tell them you left your ticket at home or do the just don't check the ticket?
<lekernel>
they don't check
<lekernel>
in many conferences, they just have a registration desk somewhere, which you simply ignore and walk past :)
<lekernel>
in extreme cases, attendes even pick up their badge themselves on a table ...