IRC logs for #openrisc Sunday, 2014-03-23

--- Log opened Sun Mar 23 00:00:32 2014
stekernblueCmd: finally! that's the killer app or1k has been waiting to have!05:14
stekernare data resizing handled by wb_intercon_gen now?05:59
_franck__stekern: yes it is. However, data signals of "resized" bus are still declared as 32 bits07:52
stekern_franck__: what do you mean, "declared as 32 bits"?08:47
stekernas far as I could see, the right signals were exported08:48
analognoiseHowdy Openrisc10:36
stekernwb analognoise10:50
analognoisety10:51
-!- Netsplit *.net <-> *.split quits: unknown_lamer, trem, wkoszek, chad__, vxe11:30
blueCmdstekern: there is _a lot_ of crap in Debian :)11:45
stekernI can imagine that, but what we consider crap is for sure someones treasure12:20
blueCmdyes12:23
blueCmdI'm a bit disapointed that binutils-atmel didn't build automatically - that would have been a nice thing.12:24
blueCmdoh, 'verilator' is queued. that opens up for some philosophical discussions I bet12:28
stekernheh, running verilator in a verilated environment12:35
_franck__stekern: I thought the generator generates 32 bits wires connected to 8 bits ports12:52
stekern_franck__: ah, yes, I see what you mean. but that's just a slight oversight. it will still work13:11
_franck__stekern: yes it works...13:21
stekernwoho, an atlys.bit15:22
blueCmd:D15:50
blueCmdstekern: do you know of https://github.com/Wolfgang-Spraul/fpgatools ?15:50
stekernblueCmd: yes, it's a pretty cool initiative. unfortunaly a bit stalled15:57
blueCmdyes, that's unfortunate15:58
stekernolofk: I know you suggested to use --dummy as an argument to get help on the sub-parsers and noted that you wanted to fix that, but I think it just would make sense to document that you can get help from the subparsers by e.g.: fusesoc build atlys -- -h16:49
stekernthe -- will prevent the parent parser from catching the arguments16:49
blueCmdstekern: so, I have two packages where GCC crashes if you try to use -O219:54
blueCmdthat's pretty exciting19:54
stekernblueCmd: sure is, do you think you can get a small test case out of the fails?20:12
stekern_franck__: your turn to do a fusesoc pull-request20:13
blueCmdstekern: I'm confident that I can - I have written down what I need to reproduce it for later20:13
stekernsince we seem to take turns by looking at the pipeline of pending requests ;)20:13
stekernblueCmd: excellent! *in Burns voice and fingertips together*20:14
blueCmdoh. another one, but not helped by turning off optimization20:16
blueCmdhttps://github.com/bluecmd/or1k-debian/issues/1020:17
stekernhmm, how does it crash? segfaults and those values are from the kernel printout?20:21
blueCmdstekern: I don't know actually, that's what qemu prints when or1k crashes20:22
blueCmdI should probably say that this is not cross-compiling, but natively20:23
blueCmdor, under qemu20:23
stekernlike 'full' qemu?20:24
blueCmdno, userspace20:25
stekernok20:25
blueCmdi'll try to do a binary search in the C code and see if I can narrow it down20:28
stekernmight be worth running that single compile in or1ksim or something20:30
stekerntoo20:30
blueCmdyes20:30
stekernjust to rule out qemu20:31
blueCmdPyLong_FromString(char *str, char **pend, int base)20:31
blueCmdthat's the function that causes a crash20:31
stekernoh, that looks like a small and simple function ;)20:35
blueCmdyes!20:36
stekernespecially when a three page comment was warranted for it20:36
blueCmdI'm down to 1/3 of the function now20:37
blueCmd            log_base_PyLong_BASE[base] = (log((double)base) /20:39
blueCmd                            log((double)PyLong_BASE));20:39
blueCmdthat's the statement20:39
stekernhmm, so probably something related to floating point20:39
blueCmdstekern: http://3a3ceb0a4450e3f9.paste.se/20:47
blueCmdthat crashes gcc20:47
blueCmdI haven't tried it outside of qemu yet20:47
blueCmdmy cross compiler doesn't complain20:48
stekernthat makes it more fun to debug then ;)20:49
blueCmdhaha20:50
stekernbut at least the testcase is nice and small now20:51
blueCmdyep20:52
blueCmdI'm surprised I could get this far without thinking about floating point really20:53
_franck__stekern: I have pull requests in stock but I don't want to give to much work to our dear olofk :)21:04
stekernhaha, yes, I have a couple in queue as well21:11
stekernmost are for orpsoc-cores though, and they depend on the fusesoc ones21:11
blueCmdor1k debian now contains 3000 packages21:38
blueCmdseems we need to port libatomic-ops22:51
blueCmd*sigh* oh well22:51
--- Log closed Mon Mar 24 00:00:34 2014

Generated by irclog2html.py 2.15.2 by Marius Gedminas - find it at mg.pov.lt!