IRC logs for #openrisc Tuesday, 2016-09-13

--- Log opened Tue Sep 13 00:00:09 2016
olofkok, I finally bit the bullet and started a new fusesoc-cores repo that will replace orpsoc-cores03:25
olofkSlooooowly migrating things there03:25
shorneolofk: yeah, I saw that. Are you planning on separating all code out of fusesoc-cores and using it only for .core/.system files?08:58
olofkshorne: Yes. I thought I'd take the opportunity to do a cleanup09:00
olofkI realized a while ago however, that non-local systems don't work, so I need to fix some things on the FuseSoC side first09:01
olofkAlso, I plan to deprecate the .system files and move the contents into the .core file instead09:01
olofkHave a half-finished patch for that09:01
shornecool, I moved my projects to fusesoc and kind of noticed an strange overlap between .system and .core09:02
olofkshorne: Yeah. It turned out to not be a good idea09:07
olofkDid you move any publicly available projects to FuseSoC?09:07
shornethose are just some toy projects09:09
shorneI was thinking to change them to use cores from orpsoc-cores and use a wb_bus09:10
olofkYou should :)09:10
shornemy problem is that in my sdram_controller.core I put the whole toplevel in there09:11
shorneso when I try to use from another project it doesnt really work09:12
shorneI need to either09:12
shorne1. separate out 2 cores in sdram_controller, i.e. sdram_controller.core, sdram_controller_toplevel.core, sdram_controller_toplevel.system09:12
shorne2. Create separate github projects for each core09:13
shorneI was kind of leaning towards 1, but in the end I was planning to just tranplant everything with orpsoc-cores09:14
olofkIf see you're using the [verilog] sections instead of the newer filesets09:14
olofkNot sure I understand you correctly, but would it help to just mark the sdram_controller top-level private?09:14
shorneI guess I copied from old examples09:14
olofkThen it would be used when you build sdram-controller, but not if sdram-ctonroller is used as a dependency09:15
shorneThat sounds ok, Would that work? Basically I want a core where the (core) files are exported and re-usable.09:16
shorneBut in the same project I want a system that can run on a device to showcase the core09:16
olofkYep. That's exactly what it's for09:17
shornemaybe on one project have multiple system, i.e. xilinx abc, altera xyz09:17
olofkThat way you can hide things like top-levels and testbenches from other cores, but still share the rtl files and test utilities09:17
shorneI see, I thought that for testbench, I didn't realize the toplevel stuff was the same09:18
shorneLet me try that out now... also I can move to filesets09:18
olofkIn lack of better documentation09:19
shorneolofk: looks like I got it to work with fileset and scope=private ( I didnt find scope documented )09:49
shornebut its very useful09:49
shornestekern: I made some notes on the patches10:33
shorneI remember icache sync change also wanted to go it... but it looks to depend on smp10:34
shorneI guess its only an smp issue10:35
shorneAlso, I did notice the based on your coversation with Alan Cox you started to work on lwa/swa emulation10:36
shorneSo I added that10:36
stekernah, yes, that is probably more important than the code that makes use of them10:39
stekernbecause userspace may use lwa/swa and it'll be broken without the emulation on implementations that don't support them10:39
shorneright, it looks like the patch needs a bit more work10:40
shorneYou have a fixme on the delay slot, also I am guessing we can remove the Kconfig option10:41
stekernyeah, the delay slot fixme isn't that urgent. there is a bunch of other code that assumes delay slot in the kernel atm10:42
stekernthe comment is more "be aware, this makes this assumption"10:43
shorneAlso, for the support and initramfs.devnodes, I am thinking housing these in a separate project10:48
shornelike openrisc-os, as kind of a distribution with build instructions would be good10:49
shorne+ or1ksim configs etc10:49
stekernmmm, that might be a good idea10:52
shorneIn the future people could just download the mainline kernel + openrisc-os (initramfs, boardconfig) and build11:02
shornejust an idea now11:03
--- Log closed Tue Sep 13 11:12:39 2016
--- Log opened Tue Sep 13 11:12:47 2016
-!- Irssi: #openrisc: Total of 48 nicks [0 ops, 0 halfops, 0 voices, 48 normal]11:12
-!- Irssi: Join to #openrisc was synced in 14 secs11:12
-!- Netsplit *.net <-> *.split quits: clopez, mithro, ZipCPU, _franck_, eliask, andrzejr, SMDhome, jonmasters, rohitksingh, stekern, (+33 more, use /NETSPLIT to show all of them)15:17
-!- Netsplit over, joins: zama, julzmb, ssvb, mithro15:20
-!- Netsplit over, joins: stekern, blueCmd, SMDwrk, bentley`, jeremybennett, mwfc, nurelin, olofk, SMDhome, eliask (+4 more)15:20
-!- Netsplit over, joins: Shentino15:21
-!- Netsplit over, joins: Finde, ZipCPU15:22
-!- Netsplit over, joins: kc5tja15:22
-!- Netsplit over, joins: wallento, jonmasters, andrzejr, Neuron1k, clopez15:22
-!- Netsplit over, joins: rokka15:23
-!- Netsplit over, joins: rah15:23
-!- Netsplit over, joins: rohitksingh, shorne, fotis215:23
-!- dalias_ is now known as dalias15:28
olofkshorne: Yeah, sorry about the lack of documentation. You can however build html documentation of the available options in the .core files15:57
olofkcd doc && make15:57
olofkRequires asciidoc15:57
olofkshorne: Also, yes, I think that's a great idea to move out initframfs and stuff from our kernel tree. There was some troubles building busybox IIRC, but I got that working eventually15:58
olofkDon't think I ever got around to document it though15:58
olofkThe reason was to see how it would work with only upstream kernel+busybox, just as you propose15:59
-!- Netsplit *.net <-> *.split quits: rah, oleg-nenashev17:27
-!- Netsplit over, joins: rah17:28
-!- Dan_ is now known as Guest8245718:12
-!- Guest82457 is now known as ZipCPU22:57
--- Log closed Wed Sep 14 00:00:11 2016

Generated by 2.15.2 by Marius Gedminas - find it at!