IRC logs for #openrisc Sunday, 2014-04-06

--- Log opened Sun Apr 06 00:00:53 2014
stekernolofk: sure, will do06:54
stekernmight take a day or two though, got preoccupied with a 8-bit mcu project yesterday06:55
stekernthe non-leisure type of project, sadly they take precedence06:57
stekernah, you have fancy tcl generation ;)07:01
stekernolofk: what is the advantage with not letting the subclasses accessing the dictionary direcly?07:15
stekernin the "convert quartus" I mean07:16
stekernseems to me that it would be extra overhead to have to add them in in the QuartusSection07:19
olofkstekern: Conformity mainly. I don't think overhead is a problem anyway08:18
olofkAnd you will get some automatic checks for invalid items in the section08:19
olofkAnd a basic string representation for free as well08:19
olofkI'm about to make a wishbone rom generator, similar to the one in orpsocv2. What source format is most sensible to use? elf, bin, C, asm, all of them?08:24
stekernok, fair enough, I noticed the extra sanity check myself after I wrote that08:45
stekernI think elf and bin makes most sense08:47
stekernasm might be nice, but that adds dependency on as08:47
stekernthat can be solved with normal Makefiles and pre-build script08:49
olofkFun article I found linked from a slashdot article
olofkBut someone should tell him that he's doing high-level stuff10:46
olofkWe have _real_ problems10:46
stekernhaha, yeah, debugging kernel stuff while being fairly certain that the compiler and cpu is bug free is for pansies11:51
Findetonyeah like that baby12:12
blueCmdstekern: when running X, did you use the framebuffer directly or did you have some sort of video driver as well?12:59
stekernblueCmd: frambäebuffer directly13:43
blueCmdstekern: roger13:46
Findetontiger tiger...13:53
olofkLooks like all the synthesis tools support $readmemh, which will make things much nicer15:58
olofkHmm.. maybe I can even use my new wishbone RAM component with wb_we set to 016:01
stekernisn't there even a verilog backend in binutils?16:25
blueCmdstekern: currently gcc pads the structure "struct a { int16_t a[3]; }" to 8 bytes. is that needed?17:24
stekernno, and IMO wrong17:27
blueCmdthen I'll fix that17:27
stekernclang doesn't do that17:27
stekernit also makes a struct describing a ethernet header not 14 bytes17:27
blueCmdyes, it sucks17:30
blueCmdI sort of understand why it was done this way if accesses on 4 byte boundaries are quicker, but it seems intrusive to have as a default17:56
_franck__anyone has an idea why in Linux, do_bus_fault doesn't print anything ?20:17
_franck__It does go in do_bus_fault then it goes to die put no prints20:17
olofkstekern: It works just as you said. or1k-elf-objcopy -O verilog gives me a readmemh-compatible file. That's pretty cool20:40
olofkBut what if I don't want 8-bit widths?20:47
-!- Netsplit *.net <-> *.split quits: bentley`, dalias, wkoszek_, Limb, mboehnert20:48
-!- Limb_ is now known as Limb20:48
-!- Netsplit over, joins: dalias20:54
-!- Netsplit *.net <-> *.split quits: dalias21:05
-!- Netsplit over, joins: dalias21:10
--- Log closed Mon Apr 07 00:00:54 2014

Generated by 2.15.2 by Marius Gedminas - find it at!