IRC logs for #openrisc Friday, 2014-03-28

--- Log opened Fri Mar 28 00:00:40 2014
stekernblueCmd: haha, you make it sound like I'd be a huge fotball supporter, drinking beer in front of the games ;)03:52
stekernwhen the truth is, the only fotball games I watch are the ones that a particular team play, and only the ones where the ones born 2005 or 2007 play03:54
stekernand the beer picture even had an OpenRISC connection, it's the brand that juliusb bought for refreshments last orconf ;)03:55
stekernolofk: I think I've found some kind of problem in the arbiter that is used in wb_intercon_gen04:13
stekernI've got the one port on the memory controller shared between or1k and the debug port, and only the debug port is asserting wb_cyc. but the cyc is still not coming through wb_intercon04:15
stekernwhy this only happens on the atlys board (synthesis difference? or the root cause is in the memory controller?) is still unclear04:16
stekernI'm doing a test run without the or1k connection now, thus completely taking out the arbiter of the picture to get some more info04:18
stekernblueCmd: since you looked closer at my photo stream, you didn't miss the snotty dog, did you?
stekernThose danish people are crazy!04:26
stekernor the infamous error I encountered at the finnish post office:
stekern(to all others, once again, sorry about Swedish oriented jokes)04:44
stekernnah, it's not the arbiter, it still happens with that disabled...05:57
stekernI bet this will turn out to some stupid error on my behalf, like undeclared wired or so... even though I've double-checked that a million times05:58
stekernhmm, definetely something strange going on, if I only connect the debug port to the memory slave (i.e. no other slaves) wb_intercon_gen doesn't connect the two at all06:33
stekernbut that seems like an unrelated bug to my original problem06:42
olofkstekern: Looks like someone displaced the dots in forsook and put them somewhere else07:00
stekernyou mean the 'pricks'?07:04
stekernbut the finns don't only have problems with the swedish language, they need to learn to hyphenate better as well:
stekernkukkia = flowers07:08
olofkhahaha . My girlfriend found a finnish children book yesterday. There were a lot of mutta and kuken in it07:10
olofkI was planning to ask you about that07:10
stekernheh, yeah. mutta = but. it's not only finnish children book authors that have a dirty mind, this was a gem I found during my study times:
stekernthat's the end of the swedish internal jokes, I promise ;)07:13
_franck_web_stekern: would it be possible to finish an instruction cache refill if you detect a branch outside the cache line address range before you finished the refill ?08:14
_franck_web_it's just I thought I had while watching the cache refill during my ddr controller debug session :)08:15
stekernpossible, yes. worth the trouble.. /me shrugs08:17
stekernbesides, there is a slight possibility that you will need the things you are fetching into cache (consider jal)08:19
_franck_web_ok that's what I thought. However, I have now 128 bytes on cache in my memory controller. So i thought I could increase my icache line size08:20
stekern32 byte is the max the arch spec dictates anyway ;)08:20
_franck_web_(about jal) but if you know that your destination address is outside your current line ?08:21
_franck_web_32 bytes limitation ? arf08:21
stekernyes, but you'll jump back to what you are fetching at some point08:21
stekernif the data in cache have been replaced by then, that's of course a possibility08:22
_franck_web_anyway, I'm fighting with my memory rigth now08:23
stekernand there is of course a possibility that the jal and the delay slot is just at the end of a cache line08:23
stekernbut in that case, I don't think "false" data will be fetched08:24
_franck_web_the controller passes olofk bfm transactor test, my board runs barebox, it runs linux until I have a dtlb miss and then it fetches wrong datas...08:25
_franck_web_I debug it with or1k-downloader and signaltap :)08:25
stekernthat's how real men do it08:25
_franck_web_12 minutes to download vmlinux :)08:26
stekernexcept, I don't have any signaltap, only diila in my current setup08:26
_franck_web_I was so exited I frame buffer was working and then I figured out my board wasn't able to ru Linux !!! aaahhh !08:28
olofkI like that my bfm transactor test is being used. I just hope it's doing the right thing :)08:30
_franck_web_I also used write functions:
blueCmdstekern: haha09:08
olofk_franck_web_: Cool. Does it work?09:33
olofkI think stekern did some fixes a while ago, but I haven't tested it since then09:33
_franck_web_olofk: I had to add a wb_cyc_o <= 1'b0 in wb_bfm_master.v (task next, after wb_stb_o <= 1'b0)09:36
olofk_franck_web_: Ah ok. I haven't noticed that. Is this only a problem for non-burst acesses?09:40
_franck_web_AFAIR yes09:43
stekernI've pinpointed my debug port problem to that it fails when I add the SPI slave11:46
stekernlooking at the code doesn't make me go 'aha' (yet)11:46
stekernhmm, xst decides that all wbs_ack_i[>=4] in are not used...12:26
stekern"Input <wbs_ack_i<5:4>> is never used."12:27
stekernwhen num_slaves = 612:27
stekernand "Input <wbs_ack_i<4:4>> is never used." when num_slaves = 512:27
stekernis it some implicit bit-size that is being overflowed?12:28
stekernon e.g. idx or something like that12:29
blueCmdI'm really excited on how the Debian port is coming along, I have a few thousand packages and when I do 'apt-get build-dep X' it can build a lot of stuff!13:14
olofkstekern: I vaguely remember that I hardcoded something there that really should be dynamic13:41
olofkHmm.. but I can't see anything now. It's not a huge amount of code in there13:42
olofkYou could try to rewrite the block that I changed because of the verilator bug13:45
stekernyes, I did that, and it works in verilator13:46
stekernbut... it seems like xst goes into an infinite loop on the rewritten code...13:47
stekernpile of turd...13:47
stekernmaybe that was unrelated to the change, now it didn't crash14:05
stekernbut the change was ineffective, it still claims that the bits are never used...15:11
blueCmdstekern: I'm soon at the stage where you were in september I think :P15:59
blueCmdwith X1116:00
blueCmdI have twm and building mesa currently16:00
stekerncool, I just used poke53282's build scripts back then to run stuff at fpga16:04
stekerni.e. most credit goes to him ;)16:05
blueCmdah, right16:06
blueCmdstekern: I bet you didn't have to compile OpenLDAP, Heimdal and systemd for Or1k then though :P16:06
blueCmdDebian's packages are a bit .. feature-rich16:07
blueCmdbut on the bright side we can now brag that we can run the world's most open source LDAP server16:08
szabolcsberkican anybody help me with orsoc?16:29
szabolcsberkii need some help in crosscompiling with gcc16:30
stekernsomebody might be able if you state your problem16:44
szabolcsberkiwell my problem is that i have an altera developer board.i downloaded the virtual box image linux from the orsoc homepage16:48
szabolcsberkiuploaded the hw and on that the linux kernel16:48
szabolcsberkiall is working but i want to write some program for it16:48
szabolcsberkiand i cant compile with crosscompiler cause it doesnot find the header files.16:49
szabolcsberkifor example16:49
szabolcsberkii am really new to gcc and i cant find the problem16:49
stekernI assume there's a or32-linux toolchain on that image and you are using that17:30
szabolcsberkii use that17:58
szabolcsberkii wrote in in the terminal this17:59
szabolcsberkior32-linux-gcc xxx.c -o xxx.elf17:59
szabolcsberkiand it worked when i only included stdio.h17:59
szabolcsberkibut when i try to include module.h it says its missing18:00
szabolcsberkii rtied to add search path for the gcc with -B but it doesnot seems working18:02
stekernwhy do you need to include module.h? are you creating a kernel module?18:34
szabolcsberkiactually not18:51
szabolcsberkii am trying to make a simple ethernet program18:51
szabolcsberkijust to see it is working18:52
szabolcsberkiand the one example program i found included that18:52
szabolcsberkiu think i dont need that?18:52
stekernI don't18:53
szabolcsberkiactually i have never made program that uses ethernet sso i dont know what headers i have to include .especially in the case of a linux kernel18:59
analognoiseHi all20:24
analognoiseszabolcsberki: Having a fine Friday?20:37
szabolcsberkiwell actually i am strugling a bit:S21:02
szabolcsberkitrying to get along with the crosscompiler of the orsoc developement board21:03
szabolcsberkibut maybe i ll continue it tomorrow:s21:06
szabolcsberkign all21:06
stekernthe verilog tools really put up a fight against my $clog2 workarounds...22:05
analognoisewhat's the $clog2 workaround?23:01
_franck__3 hours to finally realize that the instruction at that address changed because our Kernel has some self modifying code...23:39
-!- FreezingAlt is now known as FreezingCold23:49
--- Log closed Sat Mar 29 00:00:41 2014

Generated by 2.15.2 by Marius Gedminas - find it at!