IRC logs for #openrisc Friday, 2016-03-18

--- Log opened Fri Mar 18 00:00:39 2016
juliusbso, yes the searech function has stopped working since my server switch-over, I think the difference is I'm on apache 2.4 now and the permissions thing I copy-pasted from don't work anymore06:05
juliusbis anyone an apache guru and care to help me out?06:06
juliusbIn my log I get this: [Fri Mar 18 06:00:09.482658 2016] [authz_core:error] [pid 11767:tid 140008417621760] [client] AH01630: client denied by server configuration: /usr/local/bin/irclogsearch, referer:
juliusb.. then I added this to my apache2.conf:06:07
juliusb<Directory /usr/local/bin> Options All AllowOverride All Require all granted06:07
juliusb... based on someting I saw on the internets. but, that results in showing you the contents of the Python it's trying to execute, not actually executing it06:08
olofkjuliusb: Not my area, but could it be a python 2 vs 3 problem, rather than apache?06:30
olofkDoes your apache user (www-data?) have access to /usr/local/bin/ ?06:30
juliusbokay, fixed it! Just had to enable CGI on the server, which apparently doesn't get done by default07:25
juliusbolofk: yes also had to add permissions to do CGI-binning in /usr/local/bin07:28
juliusbalright, I promise not to touch the server for another 5 years07:28
wallentoshorne: Would you create a "Build Linux" for the tutorial?07:57
wallentostekern: I am building all toolchains and put them up as releases on Github11:09
wallentofor musl I forked musl-cross and started a branch plus tags/releases for ork1, is that good?11:10
wallentoand I added travis:
wallentoI have also updated the newlib release with a gcc 5.2 version11:13
wallentoblueCmd and stekern: I pushed gcc 4.9.3 as the latest 4.9.x12:34
olofkwallento: Great12:42
wallentostrangely I stumbled over this for the first time now:
wallentoadding the prototype to or1k-proto.h solves it14:28
olofkwallento: Are you trying to break the world record of most releases in a day :)15:39
stekernwallento: yeah, that sounds good16:08
shorneI need to remove the reset button on my machine. Daughter pressed it last night so I got logged out :)16:57
shorneanyone looking for me ?16:57
olofkshorne: juliusb got the logs back up, so you can check for yourself now :)17:04
shorneolofk: great17:09
shornewallento: Ill create a build linux, sure.  I was playing with that a lot 2 days ago...17:13
shorneI think I found that loading the linux image via gdb doesnt really work, not sure where it fails17:14
shorneif I load and boot in openocd it boots17:14
shorneif in gdb, I dont have luck, it starts and I can step through it, but when I let it run it crashes with some errors17:15
shorneanyway... Ill try some more thing later.17:15
olofkshorne: Did you see the linux document I put up in the tutorials repo?17:18
shornewallento: in terms of building linux I thought I saw something, maybe this?
shorneoh... maybe that was olofk17:18
shorneyeah, that was it.  wallento want more? I could add a few notes17:19
shorneolofk: you did memset, did you have a look at memcpy?17:22
shorneI am looking at it now. there are actually a ton of different implementations in the kernel already17:23
olofkshorne: I never got around to do memcpy. It was slightly more complicated than memset, so I let it be17:26
shorneolofk: many implementations just do something like this...
shorneloop unrolls17:27
olofkYeah, I think I based my memset on the microblaze c algorithm, but wanted to do asm to shave off a few more cycles17:28
olofkBut just doing that C algorithm will make it ~4 times faster on larger chunks17:30
shornemicroblaze. this one
shornethe microblaze looks the same as a kernel default17:31
shorneanyway, the main goal with assembly will be to use the word memory transfers17:31
shorneoh, they do that with c in the memset17:32
shorneok, cool17:32
shornelet me play with that, I was seeing that during boot it takes about 5 seconds to load my initramfs17:32
shorneIf I can get it 4x faster that would be nice17:33
shornebut maybe its gzip17:33
olofkFour times faster memcpy should probably help a bit in many cases18:12
olofkI noticed that the improved memset only used 90% as many instructions at boot. That's a 10% improvement, and probably more than that, since we reduce the number of memory accesses a lot too18:14
wallentoshorne: I thought you needed to do adopt some of the build instructions to get it running18:58
wallentomissed the one thats already there :)19:01
wallentoolofk: just want to get all releases straight to bring to a good state19:01
olofkwallento: It's great that you do that19:05
olofkI'm playing with yosys and icestorm. Almost done with the FuseSoC backend19:05
shornewallento: Ill update it if I see anything I do different19:12
shorneSo I am playing with memcpy, for now I think the microblaze one is a bit too much and I just dont want to copy. Also it doesnt unroll loops19:12
shorneSo I make one that assumes word alignment, and if not if falls back to by copy with loop unrolls19:13
shornewhich should be pretty good, and Usually we would have word alignment19:13
shorneall in C for now19:14
shornehah, its slower19:17
shorneslower than the byte copies19:17
shornemithro: good writeup, I didnt see the openrisc gsoc proposal19:46
shornewhere is that?19:46
shornemithro: do you know a C braken at work?19:47
mithroName sounds familiar19:52
shornelast I talked to him he was working on dart, if you see him say hi :)19:56
mithroPlease do share that blog post around, want as many potential GSoC students to see it19:57
--- Log closed Sat Mar 19 00:00:41 2016

Generated by 2.15.2 by Marius Gedminas - find it at!