IRC logs for #openrisc Thursday, 2015-02-12

--- Log opened Thu Feb 12 00:00:37 2015
stekernblueCmd: was the offer for a git janitor perhaps? ;)00:07
stekernah, ah git clone of or1k-gcc had time to finish at 40k/s during the night...00:07
stekernolofk: it should be in here:
stekernolofk: I just built a microblaze-elf (in paralell with an or1k-elf) toolchain according to the updated instructions on opencores.org02:26
stekern(iow, they seem to be solid)02:37
-!- Netsplit *.net <-> *.split quits: antgreen, rhythmx, FreezingCold02:52
-!- Netsplit over, joins: FreezingCold02:59
wallentothe musl-cross tool does not provide the option to build from the git repositories, correct?07:53
olofkstekern: Hey! There should be a clause that you can't use the build instructions for evil :)08:02
olofkwallento: I think you can override some stuff. Did you look at stekern's clone of musl_cross?08:03
olofkAnd I saw that the newlib patches were applied btw08:03
wallentono, I will have a look at it08:05
olofkI think that it only supports cloning musl from git.08:11
olofkBut it gets an archive from github for gcc if that helps08:11
wallentoah, okay08:11
wallentothats the trick08:11
wallentodo I understand it correctly, that musl support is not upstream in gcc?08:13
wallentommh, I will manually trigger builds for the moment :)08:18
olofkI don't know if gcc requires some patches to work with musl. Never considered that there would be any interaction between those09:06
wallentoI just followed the instructions so far and made it only for releases. is there any active work on musl at the moment, by the way?09:08
wallentookay, works:
wallentonice filenames: or1k-linux-musl_1.1.6-bintuils_2.25-gcc_4.9.1-linux_3.15-ubuntu-14.04-amd64.tgz09:20
wallentothe musl build does not do multilib builds?09:22
stekernolofk: it's actually for a good cause, I'm porting a little kernel to or1k, and there's a microblaze port to copy off10:11
stekernso I wanted to know that it at last compiles and boots10:11
stekerns/a little kernel/little kernel10:11
blueCmdstekern: no, it was for DevOps for a high-frequency trading thingy10:32
stekernah, oki10:39
olofkstekern: Cool. Still waiting for the Windows CE port though10:49
olofkI couldn't figure out what should go in the init-valid file for the icache, so I just initialized it to all zeroes10:51
juliusbthe moxie stuff looks fun - what infrastructure exists for in way of debugger, etc?11:22
antgreen`I have firmware with a tiny gdb stub in it11:23
-!- antgreen` is now known as antgreen11:23
antgreenbut no jtag debugging .  I'm not sure how that works yet.11:24
stekernolofk: joking aside, windows ce is one of the least crappy things that has came out of that company. especially in the incarnation of win mobile 6.111:25
juliusbOK cool. But no hardware debugging?11:25
antgreenthat's right11:26
antgreenonce I tried to implement the gdb remote protocol in hardware... . It's not complete.  Someday I may finish it.11:27
juliusbCool. It looks like an ISA that'll result in very compact implementations.11:27
juliusbGot a LUTs/LEs number for FPGA implementation?11:27
antgreennot handy.  I'd generate some now, but I'm under pressure to get something else done for work right now!11:28
juliusbYeah I saw someone who did that for the RISC-V I think, literally accepted the RSP characters over UART and did all of the processing in hardware IIRC. Quite neat :)11:28
juliusbno problems, just curious11:29
antgreenoh, really?  pointers?11:29
juliusbjust looking...11:29
antgreenoh ,man, that was my idea!  Too bad they beat me to it.11:30
juliusbPage 5 of
juliusbTo  be honest I'm not sure if it's exactly what I said it was now, at least it doesn't look like it on that slide, but I'm pretty sure that's what they claimed to have done.11:31
juliusbThe other bad news is that it looks like if it exists, it'd be in Bluespec :(11:31
antgreenit loos like exactly what I did.  a11:33
antgreen'gdb stub' in hardware that hooks directly into the CPU11:34
olofkHmm.. would it make sense/be possible to implement the or1k debug protocol in Moxie? Then we could reuse adv_debug_sys and OpenOCD and things directly11:36
olofkstekern: I feel sorry for WindowsCE sometimes. It's mostly used as a joke11:37
juliusbolofk: the advanced debug sys stuff (and even the original one) is independent of the processor it's attached to I reckon.11:40
antgreenwhere can I read about all of that?11:41
juliusbSo you could attach it to anything and write the appropriate bit of OpenOCD to translate GDB's "read my PC" into the appropriate register read on the CPU and you're about done11:41
juliusbantgreen: andvanced debug sys?11:41
olofkjuliusb: Yeah, that was what I was hoping for. Just a set of instructions, like read pc, stall CPU, reset and stuff11:42
olofkantgreen: Does mox125 work btw? I never got it to move the PC. Looked like icache is the culprit, never reporting valid to the CPU11:44
antgreenolofk: no! you caught me in the middle of adding the new icache11:44
antgreenI'll have it working again in  a couple of days11:44
antgreenalthough it should report valid to the CPU.  Just give me a little more time...11:45
juliusbolofk: sure, I'm not so familiar with the adv_debug_sys but it is very similar to the original one. Basically you select the "slave" (ie cpu0, cpu1, or wishbone bus) in a config register, load in the address and whether it's a read or write, press go, and then poll another bit, wait until it looks like the access has completed, and if it's a read, shift out the read data, otherwise move along11:45
olofkantgreen: Yeah. I probably chose a bad time to play around with Moxie. No stress. Just let me know if you want some help testing stuff11:46
olofkjuliusb: Makes sense. I've never looked at the details, so it's a bit of a black box for me11:47
olofkjuliusb: You had a good time on the prison island btw?11:49
wallentookay, there we are:
juliusbolofk: always :)12:26
juliusbironically I consider visiting the ex-British prisoner colony as a brief release from the confines of the UK12:27
stekerncool, lk boots now14:57
ysionneauwhat's lk ?15:00
olofkstekern: Nice. How big is the binary?15:01
stekernmm, wait15:02
ysionneauah some embedded kernel15:05
ysionneauis the application (payload) linked with the kernel?15:05
ysionneauor is this loading the application via ethernet/serial ?15:05
stekernI think it's used as a bootloader on some android systems15:05
stekernno, the applications are linked in (afaict)15:06
ysionneauok, like RTEMS then15:06
stekernI just started to look at it today ;)15:06
ysionneauand it already works? :p15:06
ysionneauthat's productivity!15:06
stekernwell, it boots... I don't evn have the timer running ;)15:06
ysionneaucool anyway!15:07
stekernbedtime here now, will try to sleep with the *very* loud thai karaoke garden party going on down-stairs...15:09
wallentortems toolchain also builds automatically now:
wallentonext will be rtems and linux temselves15:23
* juliusb applauds15:24
juliusbthat's pretty cool!15:24
wallentofor both musl and rtems there is nothing like continuous integration at the moment15:28
wallentobut I think having on-demand builds for certain versions/version combinations is sufficient for now15:29
ysionneaugreat :)15:36
Me1234poke53282: uses linux from, which is unavailable.16:52
juliusbwallento: that is beer-worthy18:17
olofkstekern: 14 branches in your orpsoc-cores repo. I'm impressed :)21:24
-!- FreezingAlt is now known as FreezingCold22:33
--- Log closed Fri Feb 13 00:00:39 2015

Generated by 2.15.2 by Marius Gedminas - find it at!