IRC logs for #openrisc Monday, 2013-08-12

--- Log opened Mon Aug 12 00:00:08 2013
stekernpoke53281: I've got scummvm compiled and ready on my nfs root fs now, so in the evening I hope to be able to test on my atlys board ;)04:18
stekernright now it complains about "Could not initialize SDL: Unable to open a console terminal" though04:18
stekern...because I had no vt enabled in the kernel05:22
poke53281There is one small problem with the sound output.05:43
poke53281ScummVM has an assert in one function which stops the programs.05:43
poke53281Simply put a return in the function instead.05:44
stekernok, thanks for the hint05:45
stekernin the long run I want to enable sound output too05:45
stekernpoke53281: I just noticed your, you should make some more noise about that05:47
poke53281Here is the new version05:48
poke53281Well, I have send a link here in the chat :)05:48
poke53281Unfortunately I have check again everything from the beginning if everything is consistent.05:49
stekernI must have missed it, I know you were speaking about "your make scripts", but didn't know you had published them05:50
poke53281Unfortunately approximately every month there is some patch which prevents it from being build completely.05:50
poke53281No problem.05:50
poke53281So, the current one which will be published soon are in the or1k-toolchain.tar.bz2 link05:50
poke53281The last three weeks I had absolutely no time. Vancouver->Saskatoon->Vancouer->Stanford->San Franscisco->Stanford->Vancouver. A lot to travel :)05:51
stekernsometimes it's like that05:51
stekernpatches to what breaks your build scripts? in external projects?05:52
stekern(and by external, I mean non-openrisc projects)05:53
poke53281External projects not so much. Well, the Downloadlinks are not implemented yet.05:53
poke53281What breaks are especially my patches. So the dependencies are not well implemented.05:54
poke53281The other ones are the openrisc patches. Especially the Linux patches05:54
poke53281.config must be redone every time.05:55
poke53281And then there are some patches not yet in mainstream. Or maybe .... have not checked since three weeks.05:55
stekernah, yes05:57
poke53281So, escpecially the kernel fails to compile almost every two weeks.05:59
stekernbut jonibos tree isn't usually even updated that often06:00
poke53281Often enough. There are also changes in every rc.06:00
poke53281This week, something terrible happened. Firefox 23 were published. And it is 5 times slower without asm.js. Don't know what they did.06:02
stekernanyways, it serves me just fine, it does the things I usually do by hand because I've been too lazy to do build scripts to automate it06:03
poke53281Yes, the reason why I did the scripts.06:04
poke53281See, one change in gcc and you begin from zero.06:04
poke53281and compiling X without scripts you would die before.06:04
stekernheh, yeah I could imagine that06:05
stekernI'm using chrome06:08
stekern(on the ff comment)06:08
poke53281So, there is still a lot to do to stabilize this build-scripts. But so far they are also doing a great job for me. 3-4 hours to recompile everything and in average only 3-4 manual interventions.06:12
stekernyeah, usually waiting time isn't a huge deal for me, I've got enough other things to do =P06:13
stekernat least grabbing the framebuffer shows that scummvm is showing the start screen06:14
stekernso it's not crashing violently on mor1kx so far06:15
stekernmore intensive testing in the evening (and hopefully some playing too!) =P06:15
poke53281You need touchscreen support.06:16
poke53281You have compile tslib and calibrate with ts_calibrate06:16
poke53281and then set also two environment variables06:16
stekernwas it because of heritage you chose the amiga version of mi?06:16
stekernyes, I saw the two SDL_ exports you had06:17
stekernbut I'm planning on using a PS/2 mouse06:17
poke53281Oh, I didn't want the EGA version. And I don't think a Monkey Island Demo in VGA exists.06:17
stekernok, that makes more than sense06:18
poke53281Amiga and or1k have the same endianess. Maybe, that helps.06:18
stekernif I can get DOTT running at decent speed, that'll make my day06:21
poke53281How many IPS manages your FPGA?06:21
poke53281DOTT should be possible without problem. I can  say more about the speed when the real time timer is implemented.06:22
stekernmor1kx is pretty much 1 IPC and the SoC is currently clocked at 50 MHz06:23
stekernmemory writes is the current biggest bottle neck in mor1kx06:23
stekernthey are minimum 2 cycles06:23
poke53281Ok, so this is very close to the asm.js implementation.06:23
stekernand coincidently the memory writes create a constraint on the maximum frequency too when the DMMU is enabled06:24
stekernso high up on my todo-list is to implement a store buffer06:25
poke53281you mean a cache06:27
stekernnot exactly, more a fifo of writes06:29
stekernit already has a (writethrough) data cache06:30
stekernone option would be to implement that as a write back cache06:30
stekernbut that has it's drawbacks, it only helps when the cache is enabled for instance06:31
stekernand a storebuffer would be a lot simpler to implement06:33
stekernI'm thinking a really simple one, without any logic accounting for what's in the store buffer06:33
stekernany incoming load when the store buffer is not empty will stall the pipeline until the store buffer is empty06:34
poke53281Understood I think I have to buy an FPGA to experiment. FPGAs are uncharted water for me06:37
stekernit's load of fun!06:38
olofkWoohoo!! I can do orpsoc development at work for two days. This is like the best thing ever :)08:08
amsolofk: woo!08:09
stekernyou lucky b*stard =P08:53
olofkHow to make an open source developer jealous. Lesson 1 :)09:10
stekernI fear lesson 209:11
olofkLesson 2 is that you get paid for it as well :)09:11
amsLesson threee is that you do it all day, everyday.09:13
stekernah, but I thought that was implied in lesson 109:14
juliusbyou have made me jealous olofk :)09:18
ams1) Make juliusb jelous.10:08
ams2) ???10:08
ams3) Profit!10:08
olofkHow does VCD dumps work in verilator? Can I turn it on from the cpp top level?14:10
olofkAnd why doesn't the eval function finish?14:11
olofkHmm... and how do I turn on the debug prints in the verilated model?14:29
olofkGot two of three questions answered by myself. Help with debug prints is still welcome14:37
* ams sends invoice to olofk for 2/3 questions.14:45
* olofk is now only 1/3 happy14:46
stekernolofk: -l14:50
stekernI'm not sure if I ever done the 'nice' trace logs14:50
stekernI use -l --log-noregs14:51
olofkOn the executable?14:51
stekernand a hacked up version that prints time elapsed instead of instructions executed14:51
stekernyes, Vorpsoc_top -l --log-noregs14:52
stekernhmm, fb output doesn't work in my 3.1114:55
olofk3.11 is old now. I use Linux 9514:56
stekernhah, you fool! don't you know Linux ME is out?14:56
stekernit doesn't even produce output that my screen would recognise as dvi input14:57
stekernwhich is weird, because the line in the driver where the fb core is enabled is executed14:57
stekernat least I get the: 'ocfb: enabling framebuffer (640x480-16@60)'14:58
olofkI think I have read that they are doing a lot of changes around fb and display stuff14:58
stekernand that printk is right above the enable14:58
olofkGreat. My verilated models at least runs through cycles now after some RTL changes14:59
stekernthe only thing I can come to think of that would cause this is if the timing values I read out of a generic code module have changed14:59
stekernno.. the timing values looks ok15:34
stekerndoh, of course it was something silly...18:20
stekernjonibos tree is missing an endian aware config switch on the reg read/writes18:21
stekernwe should probably use the generic CONFIG_FB_FOREIGN_ENDIAN knob in that driver instead of CONFIG_WISHBONE_BUS_BIG_ENDIAN18:23
stekernbah, I remember now that this usb->ps2 converter on the atlys board is complete crap20:25
--- Log closed Tue Aug 13 00:00:09 2013

Generated by 2.15.2 by Marius Gedminas - find it at!