IRC logs for #openrisc Thursday, 2013-10-24

--- Log opened Thu Oct 24 00:00:53 2013
hansfbaierpoke53281: What background would I need to apply finite element analysis? what parts of num. math, PDEs? What would be a good introduction to FEA?01:40
poke53281The finite element books usually start with the partial differential equation laplace U = rho02:16
poke53281So, you should understand what this means.02:16
poke53281Then of course integrals and linear algebra.02:16
poke53281Here you can find literature02:17
poke53281including german books.02:18
poke53281Otherwise I had a very nice online script where I learned how this works. But this over 6 years ago. I can't remember.02:18
poke53281It's not difficult when you understand the main idea.02:20
TW__Hi. Working through the ORCONF2013 Workshop ORPSoC on DE0 Nano.  Built everything nicely, ran all 3 bare-metal examples on ORPSoC, including UART.  Excellent instructions.02:20
TW__On last step when I downloaded and loaded vmlinux-de0_nano, seems to load nicely but after reset, no response.02:21
TW__Have I missed something?  LEDs just stay steady on and nothing on JTAG or UART.02:21
poke53281Sorry, can't help you. You have to wait a little bit02:22
stekernTW__: does it work if you try the prebuilt fpga image?02:50
TW__stekern:  I am using the prebuilt image that I downloaded from the link on the workshop page.  No apparent functionality after reset.03:02
TW__I can load the bare-metal images reliably so JTAG/OCD connectivity is working03:05
stekernok, then that's odd03:06
stekernhow are you loading it with openocd?03:07
stekernI assume with the command in the workshop material03:09
TW__stekern: yes, I am using the OCD command string from the workshop03:10
stekerndoes it make a difference if you add a resume to the end?03:12
TW__I just tried again with the downloaded (prebuilt) de0_nano_orconf2013.sof.  I have tried reset and resume and changing program counter to 0x100 before resume.03:13
TW__cannot see any activity on LEDs or UART03:13
stekernand, if you load the baremetal image after you've loded the linux image, does that run then?03:13
TW__stekern:  Yes, I can just do halt, load timerled, and it starts right up.03:15
TW__load_image seems to wok for vmlinux:  2087404 bytes written at address 0x00000000 2400012 bytes written at address 0x001fe000 downloaded 4487416 bytes in 19.256693s (227.570 KiB/s)03:17
TW__sizes and addresses make sense03:18
stekernsounds weird03:32
stekernI'll test03:32
stekernbut I think everybody at the workshop got linux working03:33
stekernyup, worked here03:34
stekernbut actually, it seems like the uart output is missing from the early bootup for some reason...03:36
hansfbaierpoke53281: Thanks.04:14
hansfbaierTW__: stekern: I sometimes have it hang, then I do something weird: I kill openOCD, and reprogram the FPGA with ./orpsoc pgm de0_nano. When it is done programming, linux boots on its own in the serial console. Only sometimes needed, usually boots right after gdb loads/continues.04:16
hansfbaierpoke53281: I don't know about PDEs yet, will I need additional PDE readings for the Großmann? Integrals and Linear Algebra is fine with me.04:19
hansfbaierstill need to refresh my vector analysis though.04:19
hansfbaierpoke53281: You probably use FEA a lot, don't you?04:19
poke53281hansfbaier: I don't know. Maybe not. But then you should start with the Finite difference methods. Finite Elements are more complicated.04:22
hansfbaierpoke53281: I also dont know about function theory/functional analysis04:22
hansfbaieris that required?04:22
poke53281I don't know you what you mean if I know "a lot". I know enough to work with it and to understand the big tutorial of deal. II04:23
poke53281Probably not. You don't need function theory.04:24
hansfbaierthe großman explains finite differences, fine04:24
poke53281Yes. The principles of finite differences is easy.04:24
poke53281With finite difference you approximate the differential equation. With FEA you approximate the solution.04:25
hansfbaierpoke53281: ok thanks. Looks like another huge time sink :)04:27
hansfbaierThanks for the links04:28
poke53281Well, the source code could be very small for finite difference. To solve Laplace U = rho takes only 10 lines of code.04:29
poke53281With finite difference you need a few hundred.04:29
poke53281The deal.II tutorial is very difficult. It is more or less a summary of all project which were done with this program.04:30
poke53281They assume already that you know what how finite elements work.04:30
poke53281So the website is a bad example.04:30
poke53281Sorry the second sentence should be finite element, not finite difference04:32
hansfbaierpoke53281: But the tutorial looks well written. Great resource, though.04:35
poke53281Yes, but when it comes to the mathematics and why they doing this and this you are lost. Ok, the tutorial I have read 5 years ago. So, there could be some nice additions04:36
poke53281And again 3 hours wasted for an Heisenbug. The emulated image is crashing randomly. Sometimes with an illegal instruction, sometimes with a wrong memory access. But only if I remove all debug messages and singlestep mode.06:33
PowermaniacHave any of you guys worked at MIT?06:55
PowermaniacLong shot but might as well ask06:55
PowermaniacFriend sort of nudged me in the direction of applying there, just for the hell of it so...06:56
olofk_I want to pull in the SIGINT fix from _franck_, but my tree has commits made after the pull request. How can I apply the SIGINT fix on top of my tree without having to do a strange merge? There shouldn't be any conflicts08:38
olofk_Is that cherry-pick?08:38
hansfbaierolofk_: If there are no conflicts cherry-pick should work08:48
olofk_Can I cherry-pick directly from another repo?08:50
hansfbaierolofk_: If you added it as a tracking branch, it should work.08:52
hansfbaiergit remote add a_fancy_name http://an.ur.l08:53
olofk_ah ok08:53
hansfbaierI then usually also do08:54
olofk_Is this _the_ way to do it when I want to add a patches that has an older HEAD than my local tree? Or should I do a pull and somehow reparent the extra commit to my head?08:55
hansfbaierIf you want, you can also create tracking branches with git branch --track08:55
hansfbaierolofk_: Good question, I'm rather inexperienced in distributed git workflows.08:56
hansfbaierolofk: Probably depends on the desired workflow somehow and/or your taste08:56
olofk_Because, the branch I'm adding as a tracking branch will probably be a throw-away branch, so I would like to remove it after the cherry-pick08:57
hansfbaierolofk_: The remote or the tracking branch?08:57
hansfbaierolofk_: You can keep the remote, you don't have to fetch from it.08:57
hansfbaierolofk: This might be interesting:
hansfbaierI'll read it to, when I'm less tired.09:00
hansfbaierSee you around!09:01
olofk_I hope I never have to work with John, Josie and Jessica.09:04
PowermaniacDon't mind me just being nosy...09:05
olofk_Powermaniac: It's the three characters from the GIT link that hansfbaier sent me09:11
PowermaniacOh right09:11
olofk_Looks like I'm looking for this: git rebase --onto <new-parent> <old-parent>09:11
olofk_or perhaps git pull --rebase09:14
olofk_Or is that the other way around?09:15
PowermaniacDon't ask me xD09:15
_franck_olofk_: I would do like hans told you, git remote add then cherry-pick09:23
olofk__franck_: Why? It sounds more complicated than a rebase?09:29
stekernolofk_: you can just run git fetch and then cherry-pick from FETCH_HEAD09:39
stekernI don't know if thats *the* way, but that's how I do it09:40
stekernpull -r is the wrong way, youll get your commits ontop of francks09:44
olofk_git fetch... I understood it differs from git pull in some way, but I haven't had to figure out the difference yet09:47
stekerngit pull is syntactic sugar for: git fetch; git merge09:47
stekernand fetch just fetches data, it doesn't do any changes to your current branch10:05
--- Log closed Fri Oct 25 00:00:55 2013

Generated by 2.15.2 by Marius Gedminas - find it at!