IRC logs for #openrisc Tuesday, 2015-10-20

--- Log opened Tue Oct 20 00:00:34 2015
cobalt60Any recommendation on an FPGA and board?01:14
_franck_andrzejr: don't know. You can try a smaller memory space to see if it fails again at the end04:42
GeneralStupid!rst did not work (clkgen) i think last time it worked the pin was not wired...05:30
GeneralStupidprocramcounter is 100...05:42
andrzejrI did an overnight simulation of barebox booting in icarus, dumped waveforms while executing cache_init, and... nothing. It all works in simulation.08:14
andrzejrSo, it may be something in DDR2 i/f (icarus uses a BFM model of xilinx DDR2 ctrl), an initialization issue, or a HW debug issue. Any other ideas?08:17
_franck_andrzejr: you should use chipscope or to see what is happening before you get an unaligned access10:16
_franck_or you could write some logic to save in some registers the instruction value and address and some other signals10:18
_franck_when the unaligned access occurs. Then you dump registers information using openocd10:19
olofkok, so who will look at the or1200 pull request? I promised myself to never touch that code again after fixing the issues yesterday13:46
_franck_olofk: close your eyes and apply the patch14:16
GeneralStupidhi, what should i try next? the reset signal is ok, if i revert it (Or press the button) i cannot connect to jtag.17:43
olofkGeneralStupid: You could try SignalTap and probe some signals in the CPU and the instruction bus17:45
GeneralStupidolofk: SignalTap?18:08
GeneralStupidolofk: one thing, which openocd version do you use?18:24
olofkGeneralStupid: SignalTap is Altera's logic analyzer18:42
GeneralStupidolofk: wow18:42
olofkI've mostly used openocd 0.8 I think, but now I'm using some 0.9 dev version18:43
GeneralStupidolofk: 0.7.0 dev...18:44
GeneralStupidolofk: my debian has 0.8.0 but he does not like my tcl fil18:49
olofkGeneralStupid: I always build it myself. The one in gentoo doesn't have support for all the necessary devices18:51
GeneralStupidolofk: i think thats the problem with mine18:52
olofkHere's my makefile if you want to try and build it
olofk(it assumes the source is in a directory called openocd)18:53
GeneralStupidolofk: for 0.9?18:53
olofkI think it works for both 0.8 an 0.918:55
GeneralStupid0.9 dev works fine for you?18:55
olofk0.9.0-dev-00346-g3e1dfdc (2015-03-16-12:03) to be more exact18:56
olofkI think that the OpenRISC stuff that _franck_ did was added to 0.8, so 0.7 might not work... but I'm not completely sure18:57
GeneralStupidolofk: i remember i had two directories of openocd, one was working, one not. I may just deleted the wrong directory...18:58
GeneralStupidolofk: .... loading, waiting...19:11
GeneralStupidohh sourceforge? really?19:12
olofkI think their git repo is somewhere else19:12
andrzejrI use git://
andrzejr0.10.0-dev-00002-g79fdeb3 (2015-07-20-18:52)19:14
GeneralStupidhmmm your makefile does not work :/19:18
GeneralStupidolofk: everything fine :D19:18
GeneralStupiddo you use altera-dev.tcl?19:28
_franck_GeneralStupid: I use something like this ./src/openocd -f ./tcl/interface/altera-usb-blaster.cfg -f ./tcl/board/or1k_generic.cfg -s ./tcl19:31
GeneralStupid_franck_: hmmmm19:34
GeneralStupid_franck_: ... runs... wait19:35
GeneralStupidok, seems like jtag was not the problem, OR1200 was not the problem,...19:36
andrzejr_franck_, how to access spr registers in new openocd/gdb? I channel logs I found: "LoneTechlooks like recent openocd has an addreg command to register SPRs"19:36
GeneralStupidcan you told me how the Tagsomething binary is called?19:37
_franck_andrzejr: in openocd just do "reg your_reg"19:37
_franck_in gdb you can print group of regs. Groups can be print with (IFAIR) "maintenance print reggroups"19:38
_franck_then "info registers *your_group*"19:39
GeneralStupidokay, if i change the address of the npc register it is changed in openrisc, so its not the reset ...19:39
_franck_andrzejr: I think you can also do "print $your_reg"19:40
GeneralStupidcan anyone give me a hint how i can get any data out of signaltap?19:41
andrzejrdoes "reg iccr" work for you? It returns "Burst read timed out; Burst read failed; Error while reading spr 0x00002000; iccr (/32): 0x00000000" here19:42
andrzejr"reg sr" indeed works19:43
_franck_andrzejr: I don't have a working platform here. But you may have found a problem in openocd. I think you might be the first to read iccr :)19:44
andrzejrso far only "sr" worked (all other spr regs I tried failed)19:45
andrzejrincluding vr, upr etc19:46
_franck_oh, weird19:47
GeneralStupidis there any already compiled elf file ?19:47
GeneralStupidhow do you compile?19:51
GeneralStupidwith -mBOARD Option?19:51
GeneralStupidand which version19:53
olofkGeneralStupid: I don't usually set -mBOARD. It should probably work without it19:57
GeneralStupidolofk: i dont get it...19:58
GeneralStupidolofk: it does not work but it should19:59
olofkGeneralStupid: You could try this one
olofkI haven't tested it on hw, but I'm pretty sure it should work19:59
GeneralStupidwhat does it do?20:00
GeneralStupid(correct english? sounds wrong=20:00
GeneralStupidshould the core be "mor1k" if i switched that?20:02
GeneralStupidChip is or1200.cpu, Endian: big, type: or1k20:02
GeneralStupidok npc does not count20:15
olofkGeneralStupid: It's just a hello world20:46
Lukan27What a warm welcome20:47
olofkLukan27: Hello :)20:47
Lukan27I just discovered OpenRISC20:47
olofkGeneralStupid: Can you send over the the .map.rpt file from the build? I would like to rule out any problems with unconnected pins and so on20:49
Lukan27How's the project going?20:50
olofkLukan27: It's going quite well. As with any open source project, there are tons of more things to be done, but it's pretty stable at this point20:54
Lukan27Well, that's good. Any chance we can buy an open source CPU/PC any time soon? ;)20:55
olofkLukan27: I don't think that's ever going to happen. This is mainly for embedded applications20:56
Lukan27Aw, what a shame.20:57
wbxi am tyring to play with qemu, but I don't get more than a shell. all forking after the first process is in a endless loop. what is the preferred kernel version and qemu version to use or1k?20:57
olofkwbx: I think poke53281 or blueCmd_ might be your best sources for qemu stuff21:04
wbxone of them done the musl port?21:06
olofkstekern did most of the or1k musl port and dalias is the main author of musl21:07
daliaswbx, you need the or1k-qemu:
daliasupstream qemu does not implement the atomics21:09
wbxdalias: so you get a working system with that?21:09
daliassomeone should work to get this merged21:09
olofkdalias: We talked a bit about that at orconf21:09
daliasi'm not sure. i've only used app-level emu, not system-level21:09
olofkblueCmd_, stekern : What was the status with the qemu patches? I already forgot :)21:10
wbxi think I already tried this qemu fork without success. but with 1.1.12 i get no panic from init anymore with qemu, so may be I should give it a try again21:10
olofkoh well. That's it for me. Good night21:14
andrzejr_franck_, thank you for the suggestion. I've got diila working and managed to capture waveform causing the exception.22:34
andrzejrit looks like it is the memory after all - in one of the bursts it returned data than actually stored. Debug interface returns correct values.22:36
--- Log closed Wed Oct 21 00:00:35 2015

Generated by 2.15.2 by Marius Gedminas - find it at!