IRC logs for #openrisc Saturday, 2013-08-31

--- Log opened Sat Aug 31 00:00:14 2013
--- Day changed Sat Aug 31 2013
hansfbaierthe or1k-gcc toolchain has build errors :(00:46
poke53282which ones. Four weeks ago was my last test run and I had no problems.01:38
poke53282I know that there could be one problem with texinfo and the gcc documentation.01:40
hansfbaierpoke53282: I used this script from the wiki:02:11
hansfbaieror should I just plain vanilla ./configure && make02:12
poke53282What is the error message?02:15
poke53282At which stage of the script it fails?02:15
andresjkare you in a 32 bit machine/os?02:16
hansfbaierandresjk: yes 32bit02:19
poke53282Could it be that you missed changing the user rights of /opt/or1k-toolchain ?02:19
poke53282# NOTE: on 32-bit machines --disable-werror is needed due to an enum acting as bit mask is considered signed02:19
andresjkyes, it could be the flag02:20
hansfbaierpoke53282: Yes, I got that error.02:20
poke53282add "--disable-werror" in each configure line02:20
poke53282I am wondering why -j12. Do you have one of those super expensive 6 core (+Hyperthreading) Intel processors? :)02:21
hansfbaierpoke53282: No 4 core02:22
hansfbaierI observed if I only use -j4 the cpu isn't really utilized well, since there is quite a bit io involved too02:22
hansfbaierwith -j12 I get pretty good steady load02:22
poke53282Yeah, this could well be. On a SSD or Ramdisk the result with -j12 could be devastating.02:23
hansfbaierI also get this: checking for suffix of object files... configure: error: in `/data/home/jack/HDL/openrisc/alt-toolchain/bld-or1k-gcc/or1k-elf/libgcc':02:24
hansfbaierconfigure: error: cannot compute suffix of object files: cannot compile02:24
hansfbaierSee `config.log' for more details.02:24
hansfbaiermake[1]: *** [configure-target-libgcc] Fehler 102:24
hansfbaiermake[1]: Das Target »all-target« wurde wegen Fehlern nicht aktualisiert.02:24
hansfbaiermake[1]: Verlasse Verzeichnis '/data/home/jack/HDL/src/openrisc/alt-toolchain/bld-or1k-gcc'02:24
hansfbaiermake: *** [all] Fehler 202:24
poke53282Uhmm, don't know. Would need the config.log02:27
poke53282Does this happen during the first compilation of gcc or the second?02:29
hansfbaierrunning it without -j ATM seems to work, still running....02:35
hansfbaierpoke53282: on the second I get:02:37
hansfbaiergcc -g -O2   -I. -I../../or1k-src/gdb -I../../or1k-src/gdb/common -I../../or1k-src/gdb/config -DLOCALEDIR="\"/opt/or1k-toolchain/share/locale\"" -DHAVE_CONFIG_H -I../../or1k-src/gdb/../include/opcode -I../../or1k-src/gdb/../opcodes/.. -I../../or1k-src/gdb/../readline/.. -I../bfd -I../../or1k-src/gdb/../bfd -I../../or1k-src/gdb/../include -I../libdecnumber -I../../or1k-src/gdb/../libdecnumber  -I../../or1k-src/gdb/gnulib/import -Ibuild-gnulib/import   -DT02:37
hansfbaiermake[2]: *** Keine Regel vorhanden, um das Target »../libgui/src/libgui.a«,02:37
hansfbaier  benötigt von »gdb«, zu erstellen.  Schluss.02:37
hansfbaiermake[2]: Verlasse Verzeichnis '/data/home/jack/HDL/src/openrisc/alt-toolchain/bld-or1k-gcc/gdb'02:37
hansfbaiermake[1]: *** [all-gdb] Fehler 202:37
hansfbaiermake[1]: Verlasse Verzeichnis '/data/home/jack/HDL/src/openrisc/alt-toolchain/bld-or1k-gcc'02:37
hansfbaiermake: *** [all] Fehler 202:37
hansfbaiersorry german, but LANG=en_US doesn't help too.02:37
hansfbaierthat means: no rule to build target.....02:37
poke53282Koi Problem. I schwaetz au Deutsch, wenns denn sei muss.02:40
hansfbaierMeine Mutter ist auch schwäbischer Herkunft02:41
poke53282Well when I read the name I guessed already south Germany. Bavaria especially02:42
poke53282But to respect the others we should speak English here.02:43
hansfbaierpoke53282: True, but my ancestors are from Czechoslovakia actually (Sudetendeutsche)02:47
hansfbaierstekern: (or who knows about mor1kx)03:02
hansfbaierI get those errors:03:02
hansfbaierwhen swapping in mor1kx03:02
hansfbaierswapping = removing or1200 and adding all verilog files from mor1kx/rtl/verilog03:02
hansfbaierthere is an `ifdef, sorry03:04
hansfbaierjust saw it now03:04
hansfbaierHow can i trim down the cache sizes of the mor1kx?03:14
hansfbaierMy EP4CE10 does not have enough block RAM03:14
hansfbaierwith or1200 it was just a couple of defines03:14
poke53282I think you have to wait. Around 5 AM in Europe. And most people in this forum are living in Europe.03:15
poke53282Can`t help you with this03:16
hansfbaierah maybe the parameters03:17
hansfbaierin mor1kx.v03:17
stekernhansfbaier: in orpsoc_top, change the OPTION_ICACHE_BLOCK_WIDTH, OPTION_ICACHE_SET_WIDTH and OPTION_ICACHE_WAYS to suit your needs03:45
stekernand same with DCACHE03:46
stekernblock width can be 4 or 503:46
stekerncache size is ways * 2^(block+set)03:48
hansfbaierstekern: What are the minimum values for OPTION_ICACHE_SET_WIDTH ?03:52
stekernthere isn't really a minimum03:52
hansfbaierstekern: What would the best tradeoffs be?03:53
hansfbaierstekern: What sacrifices would have the least impact on performance/03:54
stekerndepends on your application03:54
stekernbut since you are size constrained (logic wise), perhaps using a 1-way setup makes sense03:55
stekernI think I have used block=4, set=8 and ways=103:57
stekernsome time, so that's tested at least ;)03:57
stekernthat'd be a 4kb cache03:57
stekernI'm trying to understand the Linux clock framework03:58
stekernIt looked so simple, but even if I think I'm doing everything right, it just returns zero when I try to 'clk_get_rate' on the clock03:59
hansfbaierstekern: 37.39 MHz37.39 MHzaltera_reserved_tck04:13
hansfbaier40.06 MHz40.06 MHzclkgen0|pll0|altpll_component|auto_generated|pll1|clk[1]04:13
hansfbaierthats the FMax summary from quartus04:14
hansfbaierno difference to the or120004:14
stekernbut that's with 25MHz clock?04:17
hansfbaierstekern: yes.04:19
hansfbaierShould I try use 50?04:19
hansfbaierstekern: Can you help me out for a moment: I try to compile ext2fs into the kernel, but I can't find it anywhere.04:19
hansfbaierThat's weird04:20
hansfbaiernot under Filesystems04:20
hansfbaierbecause I want to use sdcard/SPI04:20
hansfbaier# cat /proc/filesystems04:21
hansfbaierthat's what is built in by default04:22
hansfbaierI need ext3 or vfat04:22
hansfbaiercant find it in the kernel menuconfig, maybe I am just too tired...04:22
stekernit depends on BLOCK04:23
stekernaha! openrisc doesn't select COMMON_CLK04:44
stekernno wonder it just returns 0...04:45
stekernit == clk_get_rate04:45
hansfbaierstekern: which source file?04:50
stekernhansfbaier: does what?04:51
hansfbaieryour post before, I am not shure what you are talking about, exactly04:52
hansfbaierjust thinking out loud?04:52
stekernah, yes04:53
stekernirc is my notepad ;)04:54
hansfbaierstekern: same thing @60MHz with caches disabled05:01
hansfbaiermemory checker runs05:01
hansfbaierbarebox hangs05:01
hansfbaierstekern: Another idea I got today about the enc28j60-Interrupt that hangs the system:05:04
hansfbaierwhen it goes active=low the system hangs05:04
hansfbaierand linux is trapped in interrupt handling code05:04
hansfbaierIf I artificially put the line high again05:04
hansfbaierlinux wakes up sometimes05:04
hansfbaiershowing a gazillion of interrupts on the line05:05
hansfbaierso here's my little theory:05:05
hansfbaiermaybe the driver assumes edge triggered, but in OpenRISC it's level triggered?05:05
stekernyes, that sounds plausible05:09
hansfbaierstekern: reporting back: synthesized @50MHz with CACHE("DISABLED")05:12
hansfbaierbut still hangs05:12
hansfbaierat barebox05:12
hansfbaierbarebox 2013.08.0-00167-g14ac2e4-dirty #8 Sat Aug 31 12:03:05 WIT 201305:12
hansfbaierBoard: or1k05:12
hansfbaierfrequency adapted and recompiled05:12
stekernwith mor1kx?05:13
stekernhansfbaier: with mor1kx?05:43
hansfbaierno or120005:44
hansfbaierwhat a shame ext3/4 have compile errors in the linux kernel05:45
hansfbaierntfs/vfat compile well05:45
hansfbaier1:0 for Microsoft05:45
hansfbaierstekern: looks like missing includes05:45
hansfbaierweird it is defined in the header and header included05:49
hansfbaierstekern: aaaah05:50
hansfbaiergo it05:50
hansfbaierit uses test_bit_le05:50
hansfbaierle = little endian05:50
hansfbaiernot defined in OpenRISC05:50
hansfbaier=> no compile05:51
hansfbaierfor ext2/3/405:51
hansfbaier_franck_: Hi, barebox is really neat06:25
hansfbaierI like it06:25
hansfbaierbetter than grub06:25
-!- Netsplit *.net <-> *.split quits: poke53282, blueCmd, X-Scale, enghong07:11
-!- Netsplit over, joins: poke5328207:17
-!- X-Scale` is now known as X-Scale07:38
juliusbolofk: I paid 74.02 GBP all up for my de0 nano from Farnell07:46
juliusbstekern: as for hotels, let me do a little bit of research for you07:46
stekernjuliusb: appreciate it!07:59
stekernI looked around a bit myself, but didn't find very much08:00
stekernmany seemed full-booked at those dates, which is a bit worrying and others had night rates at £300-£400, that's a bit pricey08:01
stekernbut maybe I was looking in all the right places08:02
stekernjuliusb: also, I will be arriving already on friday night, will you be available to show me around cambridge then or do I have to explore by myself?08:06
stekernthat's of course directed to any other early arrivers too08:07
juliusbstekern: i'll be around!08:23
juliusbyes I didn't see much available when I looked, but there's things like B&Bs which aren't on the usual hotel sites08:23
stekernjuliusb: yes, that's what I was thinking too08:45
juliusbOK, strangely the Travelodge on Newmarket road (only a few months old) isn't showing up on kayak when I search08:45
juliusbit's 50 pounds per night, and appears to have availability08:46
juliusbit's about a 20 minute walk from the town center, but it's cheaper than the other hotels which are available08:46
juliusboh, actually, it's 50 pounds on Friday, then 86 on the Saturday night08:47
stekernthat sounds good08:47
juliusbso 136 pounds all up, 160 euro08:47
stekern(even with the 86 on saturday)08:47
stekernyup, already got there via google08:47
stekern20 minute walk, is that the same time measurement as in my phone?08:49
juliusbactually, it's not in such a bad spot, 2 minut walk to the river, and most importantly, about <5 mins to some of the best pubs in town (and of course, I live in that area, too, about a 2 minut walk from that hotel)08:49
juliusbhaha,well it depends where you're walking to, maybe 25 minutes if you want to go tot he dead center of Cambridge08:49
juliusbKing's College or something08:49
stekernit has to be for elderly people, yesterday it said 3.5 km would take 1 hour....08:49
juliusbah yes, 25 minutes walk to King's college, according to google! I was about right08:50
juliusbso yeah, that's a nice new hotel08:50
stekernyup, sounds good08:50
stekernolofk: are you reading this?08:51
juliusbwell, the only niceness would be ascribed to its newness hehe08:51
juliusbI'll put this information on the wiki08:52
stekernI don't need anything fancy, at least not when travelling alone and I'm paying for it ;)08:52
stekernhmm, the common clock framework doesn't really work as I expected it to do08:56
stekernwhy does something like fixed-clock need some board code registering it?08:56
stekernat least that's how I understand how it works atm...08:57
stekernthis thread also backs my understanding and confusion:
stekernbut it ends just when it started to get interesting...08:58
juliusbstekern: there's also this,Cambridge,CAM,United-Kingdom-c17146-h80251/2013-10-04/2013-10-0609:02
juliusbthat's a bit closer to town09:02
juliusband are quoting about 150GBP all up for 2 nights09:03
juliusbOk, documented09:07
stekernah, yes, that I actually found  myself when I was looking09:10
stekernhmm, multiple choices, you're making it hard...09:10
juliusbthe Travelodge is about 100 years newer I think hehe, but a bit further away and about 20 GBP cheaper09:10
stekernwonder if the bedspreads are less ugly in the travelodge...09:11
juliusb(closer to my place though, so more likely for me to drag you to pubs on the way home hehe)09:11
juliusbArundel is definitely a classic English hotel... :)09:11
stekernhmm, the latter is a clear pro for the travelodge09:11
juliusbI'd hope nothing there was Fawlty09:11
stekernmight be that on sunday it will turn out to be a con though ;)09:12
juliusbhaha, always09:12
stekernok, I booked a room at the travelodge now10:14
stekernmaybe I just have to prepare a patch for the common clock framework to have some way to automatically register a fixed clock10:55
stekernthis is what I'd like to do in the dts:
stekernand it make no sense at all for us to have any board code just to register that clock, since everything else we need can be described by the dts10:59
LoneTechyes, that looks about right11:14
olofkHmm.. seems like travelodge is full now11:18
olofkThe rumours about the Finnish OpenRISC celebrity who will be staying there must have spread fast11:25
stekernah, no sorry, I always book hotels Obama-style11:30
stekernLoneTech: were you commenting on my things, or something else?11:31
LoneTechthe dts fragment with the clock11:35
stekernbut don't worry olofk, I ordered the breakfast, and there was a "KIDS EAT FREE" bonus on it when ordering online, so perhaps if we pretend you're a kid, you might tag along on that ;)11:35
stekernLoneTech: yeah, the fragment is correct, the problem is that there is no generic code in the kernel that register the "fixed-clock" (at least to my understanding)11:36
stekernthat's up to the board code to do11:36
stekernand that's IMO a bit silly11:36
LoneTechI agree11:53
LoneTechought to have one case for fixed clocks, and possibility of drivers for programmable ones11:54
stekernso, I'm looking into what it'd take to make some generic code that would register it11:54
stekernperhaps the best forward is to just create a clk-generic-fixed driver that does the boilerplate that all the boards do12:00
_franck_juliusb: is that better to be at 10:25 at Gatwick or 10:55 at Luton ?17:54
juliusb_franck_: good question17:55
juliusbGatwick is much better connected by train, but is the wrong side of London. Luton is the right side of London for Cambridge, but you will have to get a bus. it's about 2 hours best case all up to get the train from Gatwick (train to Kings Cross St. Pancras and then train to Cambridge). The national express bus goes from Luton to Cambridge, but stops via Stansted and probably also takes 2 hours, but let me confirm that17:56
juliusbon Saturday October 5, leaving Luton, there's only a bus at 11.20, arriving in Cambridge at 13:00, or a bus at 13:20, arriving at 15:0017:58
juliusbYou definitely won't make that 11.20 one17:58
juliusbThe other option is a taxi to Hitchen station (about 20 pounds) and the train to Cambridge - they come twice an hour, at :22 past (taking 32 minutes) or :47 past the hour (taking 40 minutes)18:00
juliusbso if you could get to Hitchin for the 11:47, you'd be in Cambridge by 12:30, or for the 12:22 you would be in Cambridge by 12:5418:01
_franck_the train from gatwick seems not so bad18:03
juliusbFor Gatwick, there's an 11.17 train to London St Pancras, getting in at 12:04, and then it's a 5 minute walk to King's Cross Station (they are really the same place but different names) and there's a 12:15 train to Cambridge, and you'll get in at 13:0218:03
juliusbah, maybe that page says this18:03
_franck_yes, 10:57 or 11:1718:04
juliusbI think you could almost certainly make the change from St Pancras to King's Cross in 11 minutes18:04
juliusbYou'd be lucky to get the 10:57 if landing at 10:25 at Gatwick I think. I could be wrong, though.18:05
juliusbI guess it is half an hour18:05
_franck_I think it's easyjet18:05
juliusbBut yeah, Gatwick would be fine, too, but all in all you'll be arriving in Cambridge around 1PM either way. Getting to Gatwick is generally easier if you don't do the Hitchen-Taxi-To-Luton thing, which is also pretty easy18:06
juliusbOtherwise Luton is a right pain from Cambridge. The bus takes forever and is expensive.18:06
_franck_yeah, I choose Gatwick18:07
juliusbmmm, I don't like easyjet anymore. I flew with them to Haifa this week - we boarded and then left us in our seats for *4* hours while they fixed a technical problem and changed the flight crew18:07
juliusbthen the 5 hour flight to Tel Aviv! So 9 hours in an easyjet seat is enough for me for a while18:07
_franck_:) I don't like them neither but there is not so much other choices for me18:08
juliusbyeah, it's like that usually18:11
juliusbjeremybennett_: thanks for updating the schedule on the wiki, btw18:11
juliusbknz: hi18:12
knzI see discussions about travel/accommodation18:12
knzORCONF I assume18:12
knzI have registered, would like to come too18:13
_franck_juliusb: I was oly looking at opodo for flights...I'm looking now at another one and found much better...I'll ask you some more later18:14
_franck_thanks anyway18:14
juliusbkayak I usually use18:14
juliusbknz: oh great!18:14
juliusbthe more the merrier18:15
juliusbyou've been hacking on some OpenRISC stuff?18:16
knz"hacking stuff" is quite the generic phrase, isn't it :)18:17
knzI am pushing openrisc to my students this year18:17
juliusbvery nice18:18
knzand I may want to use one of the current core designs in my research18:18
knzI'm just not sure which one, hence the need tot alk18:18
juliusbI'm an OpenRISC pusher myself :)18:18
juliusboh, allow me to push mor1kx, then.18:18
juliusbthat's a good core that one18:18
knzmy priorities are 1) small in area 2) modular sources18:19
juliusbhowever in the interest of full disclosure, I'm the original author, but stekern has been honing it into a fine, Linux running, coremark thrashing machine, of late18:19
knzgood work18:20
juliusbhah, I initially started the mor1kx for exactly those 2 things18:20
knzthat is a pleasant coincidence18:20
knzhow small is "small" in your case?18:21
knzeg how many xilinx slices for the smallest configuration?18:21
juliusbsee my presentation from last year18:21
juliusbother links from the presentation are here:
juliusbI have been esentially developing a version which is as small as possible18:22
juliusbI'm part-way through a version which you just attach to a ROM and let her rip18:22
juliusbintended for use in an ASIC18:22
juliusbstekern has been getting all the attention though (I think he's addicted to celebrity to be honest) by making the cappuccino pipeline run Linux and generally kick ass18:23
knzyes I have followed these discussions18:23
knz2.4-3.4k slices18:24
knzthat was what I was looking for18:24
juliusbI can't tell you exactly how small it gets right now, but it's certainly smaller than the or1200 in its smallest configuration18:24
knz1023kgates @ 40nm is aboutum218:27
knz24kgates @ 40nm is about 9um218:27
knzthats what I meant18:27
knzthat's nicely small18:27
juliusbindeed :)18:28
knzif I show up at orconf with a de0 nano, any chance I could get some help setting up some openrisc cores on it?18:36
juliusbsee 5PM on Saturday in the schedule:
juliusbalso for some DIY18:37
knzdepending on my progress so far I'll be also interested to share some results from my research18:39
knz(assuming interest)18:39
juliusbwhat is it?18:39
knznew micro-architecture18:43
knzhardware multithreading and dataflow scheduling18:44
juliusbare you a PhD?18:44
juliusbok, nice18:44
knzwho's the most knowledgeable user of the de0 nano here?19:17
juliusbwhat is your question?19:17
juliusbwe've all dabbled19:18
knzI am looking for I/O options19:18
knzthe board itself is quite bare19:18
juliusbthere's all of those headers...19:18
knzthree of them19:18
knzI'm looking for a LCD19:19
juliusbOK. stekern is your man, I believe. He's been mentioning something about this a little while back19:19
knzand some networking option, anything from UART to Ethernet19:19
juliusbyou can search the irc logs at
juliusbah, Embecosm gave out nice little USB-UART boards at chiphack, it's likely they'll do the same at the ORCONF workshop if we ask nicely...19:20
poke53282One question. How much logical elements does the current implementation of opencore need?21:50
juliusbpoke53282: what do you mean by opencore?22:13
poke53282the FPGA implementation of OR120022:14
juliusbORPSoC, then?22:15
juliusband by logical elements (LE) you mean implemented on an Altera FPGA?22:15
poke53282Yes, Altera for example22:16
poke53282So the DE0-Nano board has 22,320 logical elements.22:17
poke53282And I would like to know how much you currently need.22:19
juliusbwell, it really depends on what you have in the system (if you're talking about the whole ORPSoC)22:19
poke53282:) This was the expected answer. "It depends". Let's say I am talking about the whole ORPSoc.22:20
juliusbI built this one, it was basic, on the DE0 nano:
juliusbsome version of the mor1kx, though, pronto espresso pipeline, not or120022:22
juliusbso that's 34% of total LEs22:23
juliusbbut i imagine you could blow it out to maybe 50% with a bigger CPU that could run Linux22:23
poke53282So this one has no MMU and no cache?22:24
juliusbbasically, yes22:24
_franck_juliusb: I'll be at Gatwick at 8:00, that's much better ;) Flight booked22:30
juliusboh nice one :)22:35
juliusbsounds like an early start, though :0/22:35
poke53282I am just curious. Don't have any experience in FPGA programming. What repositories do I need to be able to run Linux on it? The toolchain stuff and kernel sources excluded.22:57
hnopoke53281, the orpsocv2 I built most recently for ordb2a (same FPGA, differend board) had ; Total logic elements               ; 18,160 / 22,320 ( 81 % )23:04
hnothat's orpsocv2, or1200, and a bunch of peripherals.23:05
hnofull fitter summary at
poke53282Thanks, so the limit of the smallest FPGA chips is almost reached.23:08
poke53282Is there some compatible FPGA emulator to try orpsoc?23:11
hnowell, as juliusb said it depends a bit on what you need. There is no such thing as "the orpsoc". This one is fairly bloated in features with sdcard, ethernet, spi, i2c, mmu, caches, and a bit more.23:11
hnopoke53282, you can either emulate the FPGA using suitable verilog tools (including Altera modelsim), or emulate the or1200 if you are more into Linux than the FPGA.23:13
hnoopenrisc emulation is available in qemu and also in or1ksim.23:14
hnoemulating the full FPGA is very very slow.23:15
poke53282I am interested in learning a little bit how FPGAs work. The emulation on the machine language level I know already pretty well.23:15
poke53282And then I am interested in buying one. But probably a little bit better than the de0-nano or the sold out ORPSoc board.23:22
hnoit's a pretty steep price increase in getting dev boards with somewhat larger FPGAs.23:34
poke53282Yes I have seen it. Up to $2000023:40
poke53282This one would be in my budget23:40
poke53282For some reason they put an ARM CPU in the chip. Not really useful if you plan anyhow to build a whole SOC with an FPGA.23:43
--- Log closed Sun Sep 01 00:00:36 2013

Generated by 2.15.2 by Marius Gedminas - find it at!