IRC logs for #openrisc Tuesday, 2013-08-13

--- Log opened Tue Aug 13 00:00:09 2013
-!- taotetek is now known as Guest5894001:09
stekernthis is starting to waste more time than I'm willing to spend on this, I think I just hook up a normal ps2 port to the pmod header03:30
stekernit's not even sending anything on the ps2 lines with neither of the mouse I'm connecting03:31
stekernno fun staring at the scummvm setting screen without being able to move the mouse :(04:45
stekerncan't you just start a game directly from the command line, without loading it from a start screen?05:34
stekernyes you can05:35
poke53281And there is probably some mouse emulation implemented via keyboard05:49
poke53281During that time not everyone had a mouse05:50
olofkOh yes. I've been playing monkey island before I had a mouse :)05:54
stekernyeah, maybe05:57
stekernit doesn't seem to grok the path I'm giving it though...05:57
stekernmaybe I should test with the mi demo05:58
stekernI see that scummvm segfaults if you hit ctrl-c in jor1k too, I was afraid it'd be a mor1kx bug06:01
stekerna bug nevertheless06:03
stekernI have a bit of extra debug printks in my kernel, so I can tell it's an unaligned access06:03
stekernmi2 is running!06:12
* stekern got no enter key...06:13
olofkYou can have mine06:14
stekernwell, more fun in the evening, some soldering of ps2 interfaces and I should be ready to go06:15
stekernI'm surprised how painlessly SDL compiled for openrisc06:18
stekernpleasently surprised, I might add06:18
olofkAre any patches needed?06:18
stekerna simple './configure --host=or1k-linux-uclibc && make' is enough06:20
stekernwell, + the usual search&replace of or32 to or1k too06:21
stekernnext after scummvm to try is this then:
olofkLoneTech and I were talking about running quake with aalib a while ago06:21
olofkBut all the kids seem to have those fancy color graphics nowadays, so it might be a bit overkill06:22
stekernnext after that: LAN parties with multiplayer DOOM on openrisc ;)06:23
olofkThat would be so cool :)06:23
stekernfeels like mor1kx should have enough oompf for doom06:26
stekernthe only problem might be driving the framebuffer fast enough06:26
stekernand minimum resolution for hdmi is 640x48006:27
stekernso to get less, I have to do some upscaling06:27
stekernwhich shouldn't be hard though06:27
poke53281opensdl does indeed compile pretty well, but no 3D and no sound.06:28
stekernwell, sound is just a matter of installing alsa, no?06:28
poke53281Yes, just installing alsa and you are done.06:28
stekernI've had alsa running in an old linux image06:29
poke53281But without 3D, libSDL is nothing more than a function wrapper.06:29
poke53281So in the end I was not wondering.06:29
stekernwith 3d it's still nothing more than a function wrapper ;)06:30
poke53281directfb has a lot more complexity inside.06:30
poke53281Unfortunately there is no 3D support for framebuffer. There are some patches lying around somewhere, but they were never implementecd.06:30
poke53281You can compile Mesa and do it yourself by render into a pixmap.06:31
poke53281Actually I wanted to try it.06:31
stekernah, cool06:31
poke53281So, everything I could compile. Source code is somewhere in the web to do it this way. Someone has to try it.06:32
olofkWe had some thesis workers who did a 3d engine. I wonder how hard it would be to add some driver support to make it run opengl or something like that06:38
poke53281I am getting an unknown opcode error when running ts_test from the tslib package.06:38
olofkor at least kms and dri06:38
poke53281Still have to figure out why06:38
poke53281Well, you have to emulate a graphics card06:38
poke53281And write an opengl wrapper which is send outside to a real graphics card06:39
olofkYes, I imagine it would be a huge job to do it in jor1k. I was mainly thinking about FPGA implementations06:40
poke53281Unfortunately graphics driver development for the newest opengl specifications belongs to the most complicated things you can imaging. E. G. You have to write your own compiler06:41
stekern"Unfortunately", I guess it all depends what wags your tail ;)06:42
poke53281The graphics card hardware would be much more complicated than the openrisc core.06:42
olofkIt doesn't have to be, though. I guess you can do a lot of operations in sw, and only accelerate some parts. Isn't that what glade did?06:43
poke53281But if I could write a OpenGL wrapper for jor1k to send it to WebGL, I would definitely get a news on every IT (nerd) news site on this planet.06:45
olofkAnd yes, I think you're right. Be prepared for some serious slashdotting in that case06:47
stekernman, that'd be cool06:47
poke53281github should be prepared :)06:49
olofkhaha. I see why you keep it on github now ;(06:50
olofkHmm.. I don't think sed wouldn't like that expression06:50
olofkHas anyone looked at MemoryLoad.cpp in the orpsocv2 sysc testbench? I would like to know if that's a decent elf loader, or if I should use something else06:55
poke53281I am indeed a little bit afraid. If I am continue with the project and implement a lot of fun stuff there will be sooner or later a big news site report about it. At this moment the size of the image is 7MB. I don't know if my 10Euro/month server could handle this.06:58
poke53281Well, I will reduce the size to 2MB+x in the future by loading on demand.06:58
olofkYou can start the framebuffer early and show ads while the kernel is booting :)06:59
stekernhmm, fbgrab isn't really endian aware07:51
olofkIs it possible to tunnel all data from a bluetooth device over an ethernet connection in linux. Does anyone know?08:25
stekernI would be surprised if it wouldn't be possible.08:50
stekernI'm of no help how to do it though08:51
stekernpoor Bernard have nothing to do without anyone giving him directions:
stekernI'm as eager starting playing it now as I was when I first downloaded it with a 9600 modem =P10:12
olofkJust let him stay there for 200 years10:12
olofkYeah, I know. I got the same feeling. Can't wait to get home and put my baby to sleep :)10:13
juliusbolofk: MemoryLoad.cpp was ripped out of or1ksim by me... it appears to work OK10:42
juliusband... I like this talk of games on OpenRISC systems10:42
juliusbI haven't been keeping up enough - have people ported games like mi2 and doom to Linux?10:43
juliusbI guess the source for these has been out a while?10:43
juliusbI always really wanted to run redneck rampage on FPGA for some reason... haha10:43
juliusbbut doom or quake would be fine too :)10:43
stekernjuliusb: poke53281 started the avalanche by adding scummvm support to his jor1k, being a huge lucasart adventure game fan, I couldn't resist taking up the challenge to get it up and running on mor1kx10:45
juliusbahh, I see it's some lucasarts-specific thing.10:46
juliusbfull throttle was my first exposure to those games, I got a copy only a few years back and played it again, I love it10:47
stekernyeah, it's a script interpreter for the scumm script language used to create those games10:47
stekernwell, actually, it supports non lucasart games too nowadays (like the sierra series, larry, pq etc), but that's how it started out10:48
juliusboh wow ok, so they were interpreted? not binaries?10:53
stekernI don't know the details, but I guess "scripted on a very high level"10:54
LoneTechiirc it's bytecode, so pre-parsed, and the VM had primitives for graphics and such10:54
juliusbah ok10:54
LoneTechbasically an extension of the classical text adventure system to do animated illustrations and point and click10:55
stekernfull throttle was fun, but it was the beginning of an end of the golden era, I remember how I despised the motor cycle chase in it10:57
juliusbwell, I've got no idea even how the classical text adventure systems work, but I guess it's some engine that you script10:57
juliusbstekern: oh, how that required a bit of twitch gaming?10:57
stekernyes, IMO that has no place in the adventure games10:58
stekernmaybe I'm just slow, and that's why I love those games so much =P10:59
juliusbI know what you mean10:59
stekernno stress, just relaxing brain gymnastics10:59
LoneTechI think one of the nicest recent releases I've seen is Brothers. It's not a point and click, but a two-stick platformish puzzle adventure11:00
LoneTechnot much replay value to it, possibly, but wonderful scene design, like Ico11:00
LoneTech(which I loved except for the battle system, which was just tedious)11:01
juliusbLoneTech: you're talking about full throttle?11:01
stekernI've missed that (brothers)11:02
juliusbsorry, ignore me.... I misread something11:02
juliusbthat's an impressive list of games:
stekernthe tales of monkey island series was actually pretty good too11:03
stekernbetter than the escape from monkey island flop11:05
olofkI love both monkey island games11:24
olofkjuliusb: MemoryLoad.cpp looks quite simple. Is that really all that's needed to decode an elf file?11:25
stekernolofk: there are 4 monkey island games in the original series + the 5 episodes in the tales of mi11:28
olofkstekern: No! There are two!11:28
olofkThere are also only two11:29
stekernI can go as far as counting away efmi, but I claim curse of monkey island is fullworthy!11:29
olofkYeah ok. mi3 was good too11:30
olofkEven though it was in SVGA11:30
stekernthey really lost it's when they went 3d11:30
olofkJust like Simon the Sorcerer :(11:30
stekerngrim fandango actually had a pretty good story, but it was so butt ugly that I almost had to force myself through it11:31
olofkDiscworld Noir was good though, even if it was 3d11:31
olofkI have Grim Fandango at home, but I haven't played it. People say it's fun, but that the controls are a pain11:31
olofkI used to have scummvm on my n900 so I could play on the bus. God, I miss that phone :(11:33
stekernme too11:33
olofkI got my Jolla t-shirt last week11:33
olofkNow I'm just waiting for the phone. I don't except a keyboard, but at least it runs glibc11:34
stekernthe 10" tablet I've got is actually pretty optimal for scummvm11:34
olofkHow does that work when you use your fingers? I thought you more or less had to have a pen to do some of the pixel hunts11:34
olofkoh... is scummvm ported to android btw?11:35
stekernand on a 10" screen it actually works very well with fingers11:35
olofkHow does that work? Do they have to redo the whole engine in java, or can they recompile the c code for android somehow?11:36
stekernyou can do native code in android11:36
olofkBut does that work for android on MIPS and x86 too? There aren't only arm androids, right?11:37
stekernthere are c compilers for mips and x86 too, right? =P11:37
stekernyes, it works11:37
stekernthe android sdk comes with gcc and clang/llvm toolchains for mips and x86 too11:38
stekernyou have to do seperate packages for x86/mips/arm if you have native code, naturally11:38
olofkBut how does it know that when I download from play store? Is it a fat binary with all versions, or do they select one from the server?11:38
olofkah ok11:39
stekernI think your device says "hello, I'm MIPS" to play store11:39
stekernnot sure though11:39
olofkMakes sense11:39
olofkI always thought you had to redo the whole thing in java, and I was so impressed that a lot of projects got their android ports running so quickly. Seemed like massive work to do the rewrite11:40
stekernI haven't done any serious android hacking, but from what I've seen, it's more of a "real" Linux than I thought11:40
olofkHow can I tell verilator that I have several C files that I want to compile? It seems like it thinks that all files except for the last one in the args list is verilog files11:45
olofkahh.. or do I make separate objects from the RTL and the C code and link them together afterwards?11:48
LoneTechthere are other options for native code as well. if you use renderscript, the android device itself compiles to native12:43
olofkverilator support was a lot more work than I had hoped for13:03
juliusbolofk: I'm not sure I've ever done it, I just looked through rdiez's stuff and he only ever used a single c++ file with the verilog13:20
juliusbcan't find the compile command in a makefile anywhere, though...13:20
juliusbI would be inclined to compile the other C files first, though13:21
juliusbactually, no, he passed multiple .cpp files:
olofkI've benn looking at the orpsocv2 makefile. It looks like it's creates a bunch of .o files. Turns most of them into a .a archive and finally links it together with the OrpsocMain object file. Does that seem correct?13:32
juliusbit's a bit over the top, but I'm not sure if we have to do it that way. I basically took all of that set up from Jeremy Bennett's app notes on using Verilator13:35
olofkI'll probably have to rewrite some pars to make it a bit more modular, but hopefully I can reuse a lot of the orpsocv2 stuff13:39
olofkEnlighten me... can I link multiple .a files, but only one .o file+13:43
olofkI'm looking at the last rule in the generated makefile for verilator, and I'm thinking that I can build all the cpp stuff and set VK_USER_OBJS to those objects/libraries. That would make the workflow a bit nicer13:44
olofkHmm.. I'm starting to understand now. If --exe and one cpp file is supplied to verilator, it creates makefile rules to build an executable of the verilated RTL code together with the cpp file14:27
olofkIf not the --exe switch is set, it just stops after creating <model>__ALL.a and excepts the user to do the rest14:28
olofkOh well. My time of paid orpsoc development is over now :(14:29
juliusbolofk: that rings a bell14:32
juliusbverilator has a pretty got set of docs on its website regarding the different ways of compiling it14:32
arokux1hi guys, do you have an usb host controller IP?14:55
stekernthere is one at opencores that have been used14:59
arokux1stekern, is it using wishbone or ahb?15:01
arokux1(maybe it could be interesting for you guys, if you didn't know it yet: this is a GPL IP library)15:06
stekernarokux1: wishbone16:44
stekernand yes, I think most of the people here are aware of gaisler16:46
mor1kx[mor1kx] skristiansson pushed 11 new commits to master:
mor1kxmor1kx/master 1966946 Stefan Kristiansson: immu: prevent tlb_miss from being signaled when busy is asserted19:24
mor1kxmor1kx/master f49885e Stefan Kristiansson: dcache: prevent dbus_we_o from toggling independent from dbus_req_o19:24
mor1kxmor1kx/master 62c3967 Stefan Kristiansson: cappuccino/lsu: prevent exceptions on pipeline_flush_i...19:24
olofkWhat's really the difference between an object file and an archive?19:51
stekernan archive (can) contain several object files20:50
juliusbbut an archive is not a library, right?21:00
olofkSometimes I wonder how I got this far without properly learning C basics :)22:24
olofkBut can you also combine several object files into one object?23:02
--- Log closed Wed Aug 14 00:00:10 2013

Generated by 2.15.2 by Marius Gedminas - find it at!