IRC logs for #openrisc Friday, 2014-08-15

--- Log opened Fri Aug 15 00:00:07 2014
poke53281stekern: What was you favorite DOS game?00:11
poke53281Forget it. It's too slow. You get just an image from me as proof: http://jor1k.com/dosbox.png00:53
stekernolofk: sometimes, we used to have a lot of problems with spurious interrupts from child threads.02:41
stekernpoke53281: :) wouldn't really expect that that would have been usable02:44
stekernbut it's still cool that it compiles and 'works02:44
stekern''02:44
stekernto answer your question, apart from the lucas art games, I think my favourite dos game was pinball fantasies02:48
stekernand http://en.wikipedia.org/wiki/Command_%26_Conquer:_Red_Alert02:50
poke53281Yes, wouldn't have expected it either. Some very early DOS Games could run smoothly.03:05
poke53281But hey. Finally we can run Windows 95 on the cpu.03:06
poke53281never played pinball fantasies. I played also red alert, but the game I remember most from that genre is Dune 2.03:08
stekernI never played that a lot03:11
stekernsb0: how do you see the fmax from the vivado timing reports? or do you just have to calculate it from the negative slack?03:11
sb0yes, from the "max period"03:12
sb0or timing constraint minus slack03:12
stekernright03:13
stekernI got 178 MHz in my test run03:13
stekerncan't remember what kind of mor1kx I have lying around there though, might be some non-functional proof-of-concept hacks in that03:13
stekernlet's see what I get with a clean checkout from git03:19
stekern178MHz from git head too04:04
stekernmaybe I have some parameters setup differently04:05
stekernone snooping bug caught (aka 'the spinlock issue'), at least one to go....07:35
stekernit's tricky, the problem is that the lock get locked more times than there are CPUs07:37
stekernI can trap the situation where the lock is getting locked the third time, but that doesn't seem to be the place where the error happen07:38
stekernand the situation where the lock is 'held' twice (well one holding the lock, the other spinning for it) is of course a valid situation in most cases07:40
-!- Netsplit *.net <-> *.split quits: aburgess, ssvb13:20
-!- Netsplit over, joins: aburgess13:27
stekern[A[A[A[B[A[A[B13:40
stekernoops13:40
stekernsb0: with the 'vanilla' parameters + the p_FEATURE_CARRY_FLAG="NONE" I get 185MHz from git head when synthesizing for kc70513:50
stekernstill not 200MHz, but that number makes sense13:52
stekernthere's only one commit difference from the test you made last, and I can't imagine that making this much difference though13:53
stekernI'll do a test run with that commit too and see if I get the same (or similar) result as you got.13:54
stekern...obviously that one commit did that much difference, I got 141MHz when stepping back to the commit before that...17:03
sseitzhi, i'm currently trying to build a new kernel for this very awesome jor1k emulator. I'm following the instructions at https://github.com/s-macke/jor1k-toolchain-builder (well, that's obviously outdated, so i switched over to the README.MD inside the referenced toolchain.tar.bz2). Except there's no eglibc repo at github anymore (I think this isn't necessary since I only want to build a new kernel) all works as expected ... except the freshl17:49
sseitzStarting emulationmaster.js (Zeile 143) 8master.js (Zeile 143)Abort execution.master.js (Zeile 143)Received stop signalmaster.js (Zeile 136)Error at http://jcloud.uubi.net/js/worker/system.js:257: TypeError: this.cpu.r is undefinedmaster.js (Zeile 25)TypeError: this.cpu.r is undefinedsystem.js (Zeile 257)17:49
sseitzmay I ask for assistance or help? :)17:49
sseitzin jor1k all of my DIY kernels are booting until ttyS1 is initialized. Using the "safe core" javscript throws an "Error in WriteMemory8: RAM region 0x90A00004 is not accessible"18:14
poke53281Hi sseitz18:29
poke53281unfortunately the toolchain is indeed outdated.18:29
sseitzhi poke53281, nice to see the inventor of that awesome software18:29
poke53281Hehe, thanks18:29
poke53281I am switching to musl currently. Get rid of uclibc18:30
sseitzwhat i initially wanted to do was adding ipv6 in the kernel .config ... later on i stumbled upon different kinds of toolchains :)18:30
poke53281Ok, so the current version is just online, because it is the only one more or less stable currently.18:31
poke53281Under jor1k.com/jor1k/ is the current development version.18:32
sseitzjo do not accidently built an vmlinux.bin with ipv6 which i could use?18:34
poke53281Oh, no problem.18:34
poke53281Ok, a small problem. I switched to Linux 3.15 and the network is my network emulation is not working anymore.18:35
poke53281Wanted to fix it soon.18:35
sseitzthat sounds promising18:36
poke53281compiling18:38
poke53281What do you want to try with ipv6?18:40
poke53281the problem with such toolchains is, that they have to be updated more less every week.18:40
poke53281when they are not stable.18:41
sseitzwell we thought about building some kind of cloud :) mayby as an art project :) ... but also for our internal trainings your emulator could quite come handy ... just pressing f5 and get a fresh machine deployed18:41
poke53281Yes, you are right. That would be awesome. But the relay currently used doesn't support ipv6 as far as I know.18:43
poke53281So you have to build your own one.18:43
sseitzif you put a recent version of your toolchain in github one could make forks and pull requests ... :) ... or depend on stable versions? kind of portage / ports ?18:43
sseitzoh, the relay is layer2. i didn't dug too deep into it, but my tap already has a radvd advertising ... by now i can't investigate the traffic inside the guest18:45
poke53281jor1k.com/vmlinux.bin.bz218:48
poke53281you can try this one18:48
poke53281and the most recent toolchain is this one: jor1k.com/or1k-toolchain.tar.bz218:49
poke53281But if you only want to compile Linux, take this: https://github.com/skristiansson/musl-cross18:50
poke53281and that https://github.com/bluecmd/or1k-linux18:51
sseitzgreat! thanks! ... i moved over to jor1k.com/jor1k/ as an educated guess (as of 404)18:52
poke53281Yes18:52
sseitzthanks so far mate! ... i'll try this out and keep you updated! have to move from office to home first :)18:53
poke53281I changed basically everything18:53
poke53281Ok18:53
poke53281blueCmd: Can you provide me with a working linux kernel for qemu and the proper command line for the 9p/virtio filesystem. I run into problems and want to compare with the output from qemu.18:58
poke53281by the way, to put qemu-or1k-static into /usr/bin is very unhelpful for cross compiling.19:00
poke53281the configure script do some funny stuff.19:01
poke53281"Checking whether we are cross compiling ...   No" :)19:01
poke53281sseitz: I have updated my toolchain. Try http://jor1k.com/or1k-toolchain.tar.bz219:11
poke53281make info && make init && make fetchtoolchain && make musl_toolchain && make linux19:12
stekernsb0: I get lm32 chopped off at 185 MHz when I try to synthesize for that19:21
stekernhttp://pastie.org/pastes/9476333/text19:24
stekernxilinx timing reports are just random noise...19:41
--- Log closed Sat Aug 16 00:00:09 2014

Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!