IRC logs for #openrisc Tuesday, 2014-08-12

--- Log opened Tue Aug 12 00:00:01 2014
poke53282stekern: Good morning. I have something for you. load and type "prboom"00:38
poke53282Also for you blueCmd00:38
poke53282scummvm works too00:40
poke53282the keybindings are a little bit weird00:43
poke53282the framerate is limited to 20FPS. No matter what you do00:45
poke53282This is the manual to compile it
juliusb_olofk: although, having levelled accusations of bias against him, he clearly knows what he's talking about. although, the comment about saying open CPU implementations being useless (because only 100 companies in the world are capable of manufacturing competitive SoCs, which is probably about accurate) and it's good that they all use ARM is a fairly one-eyed view of things. the whole point of this work (and I thought was a point made in the article itself) is that if companies weren't paying 100s of $k for the RTL plus royalties for low-end ARM CPUs, then chips would be cheaper, thus the items they are in would be cheaper01:50
juliusb_plus more open, plus hopefully benefiting from the bazaar model of open source development01:51
juliusb_he is well aware the point of this is to eat some ARM's lunch, and no doubt they want to rubbish it01:52
juliusb_s/some/some of/01:52
sb0poke53282, how do you load the different files? (basic X-windows, wayland, etc.)03:39
poke53282This is disabled for the moment.03:56
poke53282Take the official version http://jor1k.com03:56
stekernpoke53282: way cool! ;)03:59
sb0yeah, nice :)04:01
sb0has anyone actually run x-windows on a fpga board?04:01
stekernpoke53282: and this is with musl?04:01
poke53282sb0: Can't remember. Ask stekern.04:02
poke53282stekern: yes04:02
poke53282with musl04:02
poke53282doom is almost playable. At least on my core i7.04:04
stekernwe even got it on video ;)04:04
stekernpoke53282: it's 'playable' on my crappy i3 laptop as well04:05
poke53282While correcting the keyboard controller for this game I found a severe error in my emulation.04:09
poke53282What is the correct behavior, when the PICSR is written?04:13
stekernto clear on '1'04:15
stekernbut it only matters for edge interrupts04:15
poke53282so if an hardware interrupt line is still on high, it should not change the PICSR?04:16
poke53282I build an fifo in my keyboard controller and I wanted to clear the interrupt only when the fifo is empty.04:16
stekernit should, if it's a level interrupt04:16
stekernfor level interrupts, writes to picsr are basically no-ops04:17
poke53282But the picsr was cleared, so the interrupt was cleared, but the fifo was still full.04:17
poke53282Ok, and this was my error.04:17
poke53282For my emulation picsr should be always  a no-op.04:18
stekernyeah, that makes most sense04:18
stekerndebugging failing spinlocks is loads of fun...04:49
stekern...especially when it's probably down to a hardware bug04:50
stekernand you have to do magic rituals to even make them appear (i.e. simple stress-testcases can't reproduce them)04:51
poke53282Hmm, a specific tool from valgrind could help here.04:52
poke53282But you have to adapt valgrind first to run on openrisc.04:52
stekernyes, and software tools like that usually make assumptions that the hardware do sane things ;)04:52
stekernmy theory is that l.swa is determined as succeeded somewhere, but the store doesn't actually go through04:53
poke53282My l.swa is still wrong implemented.04:59
poke53282It's on my to do list. Any other game, that would be worth to compile?04:59
poke53282or tool05:07
stekernvalgrind? ;)05:15
poke53282a tool, which wouldn't take month to alter.05:20
stekernthe 'good' news is that I can reproduce this bug with dcache disabled05:46
stekernat least one parameter less to take into account05:46
stekern_franck__: I just found a bug in your or1k-tcltools07:03
stekernthey are still comin to use ;)07:03
_franck_stekern: good to know that is still useful07:30
HeshamDoes or1ksim/gdb supports 64-bit now?08:11
HeshamI want to use uint64 for timestamp usage (as RTEMS is doing that for all CPUs)08:21
stekernHesham: I'm not sure how does to sentences are related?08:55
stekernwhat did I just write?08:56
stekernHesham: I'm not sure how *those two* sentences are related?08:56
blueCmdpoke53282: haha, that's awesome! nice job :)10:09
blueCmdHesham: or1ksim has supported 64 bit for quite some time10:10
blueCmdI don't see why GDB wouldn't support 64 bit10:10
HeshamSo, I can use uint64_t normally?10:10
blueCmdas for uint64 that's not dependent on 64-bit support for either gdb nor or1ksim10:10
blueCmduint64 (uint128, uint256 if there is such a thing) should just work, it's just a different data type10:11
blueCmddoesn't require anything extra10:11
HeshamDoes not it relate to the HW register size (i.e, adding two 64 bit values)?10:12
blueCmdHesham: hm? the CPU is still 32 bit and has 32 bit registers10:17
HeshamAnd 64 bit math operations are done in gcc SW library?10:18
LoneTechit's still a mystery to me why MCU compilers so rarely have an int_least24_t.10:27
HeshamblueCmd: Thanks10:34
-!- Hesham is now known as hesham12:48
-!- simoncoo1 is now known as simoncook13:14
-!- Hesham1 is now known as hesham13:45
n2nnHello :-)14:16
poke53281So, prboom runs also in some sort of window-fullscreen. Press the fullscreen button. What is missing is sound :)21:15
poke53281Ah, hi dalias. Did you see my musl-powered machine yet?21:28
poke53281start "prboom" on the console and press fullscreen.21:29
daliasi'm guessing you need some crazy fast system for that to be practical... :-p21:30
poke53281It runs pretty well on my core i7.21:30
pecastroWhat's the thing with all the ^M in the end of the lines ? You changed editor in between ?21:34
pecastros-macke latest commits ...
poke53281Yes, I changed the editor in between.21:50
poke53281where do you see it?21:51
poke53281I change all the time between, nano, joe, Notepad++ and the WinSCP editor :)21:53
pecastroIn a linux shell by using the following ... git log -p -w --color21:53
poke53281Yes, I created a new file in notepad++ and didn't check the eol21:54
pecastroApparently traditional git clients are good at hiding those subtleties ... even github doesn't show them :(21:54
pecastrovery nice work btw. I've been playing with it since last year and find it inspiring. Well done.21:56
poke53281there are five files or so with wrong eol.21:58
poke53281I correct this.21:58
poke53281thanks pecastro21:58
poke53281The question is, how to make some really useful out of this?21:59
n2nnguys, do you think the OR softcore will get faster with time? e.g., instead of 100mhz now, say, 400mhz in some years?22:01
n2nnhow to run doom on the jor1k? :D22:02
n2nni see the wad, but no executeable.22:02
poke53281type "prboom"22:02
poke53281I will edit some help program with a menu in future.22:02
poke53281But it is still an unofficial website. My online testing site for the moment.22:04
pecastroI guess it depends on the meaning of *useful* :) , I had a blast reading some bits of it. It's not every day you can look at the inside of the CPU :p22:06
poke53281try prboom on a modern xeon CPU and you should get a very good framerate.22:06
n2nnso nice!22:08
n2nnbut i dont get it ... javascript runs the core?22:08
poke53281It's a full emulation,22:09
poke53281of the cpu instruction level.22:09
poke53281you can get up to 200 MIPS. That's close to an AMD K6 or Pentium 2 or so22:11
poke53281And still I have three ideas to increase the speed further.22:11
daliasi get 2-3 mips...22:11
n2nnhm --- that is very interesting. because i was always puzzled how to debug and see the output of bare metal apps.22:11
poke532812-3 MIPS on our iphone probably.22:12
poke53281Or stop using the Internet Explorer :)22:12
n2nnthat prboom runs about the speed it used to run on my 386. it was playable. kind of.22:12
pecastropoke53281: Did you see the lowRISC announcement? What's your thoughts on it ?22:12
poke53281you have to run something.22:12
poke53281Otherwise it is in idle mode.22:12
poke53281check while running prboom.22:13
daliasthat's what it got running prboom22:13
daliasit wasn't playable :(22:13
poke53281which CPU? Try to restart chrome.22:13
poke53281try firefox22:13
daliasatom n28022:14
poke53281pecastro: difficult. In some way I like it. In some way I don't linke the spreading of the community instead of working on one project together.22:14
poke53281On the other hand, there is no 64 Bit openrisc CPU as far as I know.22:15
daliasi'll try later on my 64-bit system22:15
poke53281dalias: Ahh, that's explains it.22:15
poke53281At least it works on it. I get even 2 MIPS on my ipad 2.22:15
poke53281pecastro: So I see the lowRISC cpu more like openrisc 2.022:16
pecastro32it was the thing putting me a bit off previously, yap, fair enough.22:17
poke53281Well, without 32 Bit, jor1k would never work. Or maybe at 5 MIPS maximum.22:17
daliasi'm getting up to 17 on here22:18
poke53281They write a little bit about the openrisc cpu in theirs docs. Yes, openrisc has it's flaws. But they are not so bad. Some of the critics are not valid in my opinion.22:18
daliasin firefox22:18
daliasi don't have chrome on here22:19
poke53281Let me guess. 64-Bit Atom CPU?22:19
poke53281Firefox is good.22:19
poke53281You should get on a Core-i5 around 50-60 MIPS.22:19
daliasno, celeron 84722:19
poke53281Ok, almost an Atom. ;)22:20
daliasand firefox is massively swapping :-p22:20
poke53281512MB RAM probably.22:20
n2nni think i got a core i5 on here22:20
n2nnhow to check mips?22:21
poke53281As I said, I saw jor1k on a XEON running with 200 MIPS.22:21
poke53281There is a number written under the terminal22:21
n2nngoes up to 45 on here.22:22
n2nnon prboom22:22
poke53281Ok, that is a good value22:22
n2nnmy boss was quite nice with my laptop22:22
poke53281Keep in mind, that the MIPS value here includes also memory accesses. The MIPS number written on Wikipedia are the optimum values, which you normally never reach.22:25
poke53281pecastro: I would like to see any open source CPU built as an ASIC. No matter if openrisc or lowRISC.22:26
poke53281lowRISC seems to be more complicated.22:27
pecastroIt can only be more fun then ... :)22:31
n2nnwhat is lowrisc?22:36
ysionneau[6~/win 1822:38
poke53281n2nn: google is your friend22:50
poke53281Page 322:52
poke53281In some way a competitor or not. Not sure.22:53
poke53281I still wonder how expensive it is to build your own ASIC, buy stuff like an USB controller, memory controller and so on.22:59
poke53281We should add an bitcoin miner ASIC next to it ;)23:00
--- Log closed Wed Aug 13 00:00:04 2014

Generated by 2.15.2 by Marius Gedminas - find it at!