IRC logs for #openrisc Thursday, 2014-02-06

--- Log opened Thu Feb 06 00:00:27 2014
stekernblueCmd: I'm going to steal your atomic syscall patch and apply that to eco32 (with proper credit), hope you don't mind02:38
stekernI just reached a whole new level of geekness, sitting by the pool hacking08:08
stekernthe internet connection seems to be a lot faster here...08:09
jonibostekern: could you provide some feedback on blueCmd's atomic syscall patch... I'm not in a position to review it properly right now and would appreciate the help... if you put your Reviewed-by tag on it, I'll push it upstream... note that pushing upstream makes the ABI "unchangeable" so I'm largely interested in knowing that its fundamentally sound -- this, of course, if you have time, as I know you're on vacation and all... ;)13:37
stekernjonibo: no problem, I'm looking at it right now while translating it to eco32 asm, as far I can tell it's good. there's some minor whitespace issues in it (if that wasn't a result from me copying the patch as is from the web browser that is).14:55
stekernfundamentally, I think you could do all of them with the cmpxchg, but it's less overhead doing them straight up like this I assume, and it's no big cost in the kernel to have the implementations15:12
olofkWho's stupid idea was it to implement orpsocv3 in a weakly typed language?15:50
stekernolofk: I think it was that longhaired guy with a beard16:32
olofkstekern: Fuckin' hippie!16:46
stekernthe whitespace damage was just due to my cop-paste, sorry for wrongly accusing you blueCmd ;)16:46
olofkIs anyone still using coff files, or can I remove support for that whenever I see it? (looking at orpsoc testbenches)16:49
stekernI don't think anyone does16:50
olofkBetter not ask jeremybennett though. I know what he usually has to say about ancient hardly ever used features ;)16:51
maxpalnOk, feeling like a complete GIT newbie - I am looking for the wb_sdram_ctrl that olofk_ suggested for debugging my new DDR3 controller. I amnot sure what to look for (a verilog file, a whole system). Can anyone point me int he right direction?17:18
olofkBut to make things slightly more complicated, the BFMs used in the test bench are stored here:
maxpalnok, this is making a little more sense17:20
maxpalnI obviously haven't looked at it yet but before I do the obvious question - will this work as a standalone test for my DDR3 Controller?17:21
maxpalnI was hoping to have something that I could use to verify my controller under a bunch of different Wishbone transactions - will this help?17:22
olofkmaxpaln: That's what it does17:22
maxpalnooops - -wrong window in focus17:22
olofkIt's a bit rudimentary though. Haven't implemented any fancy ways to shape the traffic. You set a number of transactions and a maximum burst length, and it will generate all kinds of requests for you17:23
maxpalnok, so looking at the wb_sdram_ctrl_tb.v - I just need to swap out the instantiation of the 'wb_sdram_ctrl' with my controller17:24
maxpalnI say 'just' I will bevery happy if its that simple - but that looks like a good place to start17:24
olofkI guess you will need some kind of DDR3 model as your backend as well, but otherwise yes17:26
olofkThis is your workhorse. You can check the parameters to see how to configure ut17:27
maxpaln:-) great - I'll dive into that and see if i can't get something working17:27
olofkInstantiate one of these and call me in the morning ;)17:28
maxpaln:-) yes doc!17:31
stekernolofk: I was thinking the exact same thing when I wrote that sentence, he has a good point though ;)17:33
olofkWhy won't the old c++ elf loader from orpsocv2 load my .elf files?17:57
stekernc++ elf loader?18:04
olofkYes, from the original verilator test bench18:04
olofkIt's got a complete elf parser there18:04
olofkthat silently fails after reading a bit of the header18:05
olofkHopefully we can throw it all out when _franck_'s improvements in the libelf-based loader is in place18:07
olofkShould e_shnum be greater than 0 in an elf file?18:10
olofkhmm... this elf reader is broken. Just don't know how yet :(18:16
stekerndidn't we just use .vmems with orpsocv2?18:17
stekern...perhaps because the elf loader was broken...18:17
olofkI thought it was mainly to have the same mechanism for all simulators18:17
olofkBut maybe you're right. Perhaps it never worked18:18
olofkBut it looks like jose managed to load some elfs18:18
stekernI don't think I've ever used it at least18:18
olofkPerhaps he cheated18:18
stekern...reminds me, I need to check the debug issues in mor1kx he pointed out..18:19
stekernI'm too slow at context switching18:19
olofkFinally closed that pull request18:33
--- Log closed Fri Feb 07 00:00:28 2014

Generated by 2.15.2 by Marius Gedminas - find it at!