[01:17] <mjeamiguel> good evening guys
[01:17] <persia> hey mjeamiguel
[01:17] <mjeamiguel> hi persia
[01:17] <mjeamiguel> i just have a noob question
[01:17] <persia> which one :)
[01:18] <mjeamiguel> ok, I downloaded this file dhrystone v2.2
[01:18] <mjeamiguel> it runs when sh is invoked
[01:18] <mjeamiguel> then produces 3 files
[01:18] <mjeamiguel> I am testing it on my x86 machine
[01:19] <mjeamiguel> now my question is, do i need edit the file if i decide to run it on ARM?
[01:19] <mjeamiguel> like cc options edit?
[01:19] <mjeamiguel> i can pastebin the drystone now..gimme a sec
[01:21] <mjeamiguel> here is the file  http://paste.ubuntu.com/524783/
[01:23] <persia> Oh, my.  What an amusing way to write C code.  I7d sugget just trying it.
[01:23] <persia> If it doesn't work, then it might be an issue, but there's a good chance it will work.
[01:23] <mjeamiguel> but yeah
[01:23] <mjeamiguel> looking into the code
[01:23] <mjeamiguel> the first few lines are to compile the dhrystone right?
[01:24] <mjeamiguel> but without optimizing it for arm, would it be possible to get proper results?
[01:24] <persia> Yeah, it's a shell script *AND* C code, so the shell gets escaped when the C is compiled, and the C gets ignored when the shell is run.
[01:24] <mjeamiguel> it worked on my x86 laptop (dual core)
[01:24] <persia> And the shell compiles the C.
[01:24] <mjeamiguel> yes :)
[01:25] <mjeamiguel> is this script good enough to benchmark a beagleboard?
[01:25] <mjeamiguel> or I need to modify it?
[01:26] <persia> Depends what you're trying to achieve.
[01:26] <persia> If you believe that code accurately reflects some workload you intend to run, then use it unmodified.
[01:26] <persia> If you believe it doesn't, it makes for a poor benchmark.
[01:27] <mjeamiguel> oohh ok..any suggestions for armv5 benchmark scripts?
[01:27] <persia> Why do you want a benchmark?
[01:27] <persia> (and Ubuntu is ARMv7a these days)
[01:28] <mjeamiguel> yeah im sorry bout that, I wanted to try it both
[01:28] <mjeamiguel> its a school project and at the same time, my curiosity
[01:29] <persia> OK.  Let's start at the beginning :)  What are you trying to accomplish?
[01:30] <mjeamiguel> cool :) I wanted to compare  the architecture (5 and 7) through bm
[01:30] <persia> Ah, OK.
[01:30] <mjeamiguel> yes sir :)
[01:30] <persia> Looking to compare different hardware running the same code, or different code opimisation on the same hardware?
[01:30] <mjeamiguel> I havent started yet..i just want to know where Im starting and heading eventually
[01:32] <mjeamiguel> 2 different hardware, same codes, speed of execution for compiling ..etc
[01:32] <mjeamiguel> for now, just contrasting both
[01:32] <persia> Are both hardware solutions ARMv7a compatible?
[01:33] <mjeamiguel> yes its beagleboardXM
[01:33] <mjeamiguel> my teacher told me that 1 board is already running ..but uses armv5
[01:33] <persia> Hrm.  OK.
[01:33] <mjeamiguel> more like optimized for armv5
[01:33] <mjeamiguel> yes
[01:34] <mjeamiguel> but yeah
[01:34] <mjeamiguel> if i can only
[01:34] <mjeamiguel> get a script for benchmarking
[01:34] <mjeamiguel> i just needed data comparisson..
[01:34] <persia> So there's some stickiness if you're running an ARMv5 kernel and make certain sorts of syscalls from ARMv7a+Thumb code, but I think that's mostly the Thumb stuff.
[01:34] <mjeamiguel> yeah..im more like into arm codes
[01:35] <persia> You can probably use your dhrystone script if you just want an artificial benchmark, but be aware it may not be meaningful for any real workload.
[01:35] <mjeamiguel> so no need to edit the script for now right?
[01:35] <persia> You will want to modify the top of it to force different ISA compliance.
[01:35] <persia> But I'm not convinced you can safely run ARMv5 code in Ubuntu: I'd probably use Debian for such a comparison.
[01:35] <mjeamiguel> will look into that :)
[01:36] <mjeamiguel> ill try that and fedora
[01:36] <persia> I wouldn't send you that far away :)
[01:36] <mjeamiguel> its just a little confusing for me, considering its my first time with a project
[01:36] <mjeamiguel> hahahah :) thanks
[01:36] <mjeamiguel> one last question
[01:37] <persia> Understood.  Folk here are likely to be willing to help with some stuff, but nobody will be willing to actually solve some of the basic issues for the project (but they may be willing to help with issues you discover in the base environment whilst working on your project)
[01:37] <persia> No limit :)
[01:37] <mjeamiguel> i also wanted to compile a program lets say written for x86
[01:38] <mjeamiguel> i jsut needed to change the gcc options right?
[01:38] <mjeamiguel> in the make script?
[01:39] <persia> Best effort is made to have sane defaults for gcc so you don't have to do that in the common case.
[01:39] <persia> If you're working on something involving comparisons of various compiler flags, the best effort defaults won't necessarily help :)
[01:40] <mjeamiguel> lol..im kinda lost there..can you please explain further thanks persia :)
[01:40] <persia> OK, so for each architecture, we ship the compiler with default settings that tend to just work.
[01:40] <mjeamiguel> yes
[01:41] <persia> So one can generally get nearly best results for "int main(int argc, char* argv[]){printf("Hello World\n");}" or slightly more complex code.
[01:42] <persia> However, if you're working on an academic project to compare different compiler flags, it won't help much, because you need to select specific flags, rather than using the defaults.
[01:43] <mjeamiguel> i guess im not so close to that, thanks for explaining :)
[01:43] <mjeamiguel> any good resource for armv7 readings?
[01:44] <mjeamiguel> i already checked the mailing lists
[01:47] <persia> I think you'll have to wait for someone else.  I just run the software, and try to fix bugs that aren't that ISA-specific.  The few things I know about the ISA changes are mostly hearsay or based on other folks solutions to other, similar bugs.
[01:47] <mjeamiguel> persia: I really really apreciate you helping me tonight..thanks so much!
[01:49] <mjeamiguel> I'm to present about my project soon, and will let my fellow students how this channel and you helped me out a lot
[01:49] <mjeamiguel> thank you :)
[01:51] <persia> Thanks for helping advertise us.
[01:57] <mjeamiguel> you guys deserve :) bye!
[08:22] <hrw> re
[09:14] <persia> amitk, Did you get the kernel source links I posted a couple days ago?
[09:20] <amitk> persia: I did not, could you post again?
[09:21] <persia> http://netbook-remix.archive.canonical.com/updates/pool/public/l/linux-fsl-imx51/
[09:22] <persia> I believe "araneo" is for the PC-Z1 and "sendai" for the PC-T1.  I have no idea about the 2.6.31 source.  I can only test for the PC-Z1, but can probably find some PC-T1 testers if you need.
[09:22] <amitk> persia: what is PC-Z1 and PC-T1?
[09:24] <persia> amitk, PC-Z1 is my little laptop.  PC-T1 is a tablet with nearly identical guts, also distributed by Sharp.
[09:25] <amitk> persia: ok, got the araneo sources now
[09:27] <persia> I think it was the araneo sources you dug into last year to upstream stuff.  The more recent updates should just be security maintenance and the like.
[09:28] <amitk> persia: I dug into the BSP from Freescale for the babbage baord, not Pegatron's kernel
[09:28] <persia> PC-Z1 wasn't Pegatron.
[09:29] <persia> I heard PC-T1 was, but some of my comments in early planning phases were not as commercially sound as they might have been, so I didn't end up working on that project.
[14:44] <hrw> ok, who here uses ubuntu on panda?
[15:03] <rsalveti> hrw: I do, why?
[15:04] <hrw> rsalveti: how to get more then 684MB ram?
[15:04] <rsalveti> hrw: you can get by giving mem=1g at the kernel cmd line, but the system turns unstable
[15:04] <hrw> default has 460+256 ram and 32 for vram, when I dropped it I got 512
[15:04] <rsalveti> there's still an unknown bug
[15:05] <hrw> thx
[15:09] <sebjan_> rsalveti: wasn't this issue hidden by the 2G/2G memory partitionning/
[15:09] <sebjan_> ?
[15:09] <rsalveti> sebjan_: nops, gets better, but you still face the issue sometimes
[15:10] <sebjan_> rsalveti: hum, ok...
[15:10] <rsalveti> and now I need to find why my both pandas got stuff while building qt =\
[15:10] <rsalveti> *stuck
[15:11] <sebjan_> rsalveti: out or memory? (ho, you already have some swap setup, right?)
[15:11] <sebjan_> s/or/of/
[15:11] <rsalveti> yup, first time was lack of memory to run ld while building webkit
[15:11] <rsalveti> then I added swap and I couldn't access it remotely
[15:12] <rsalveti> so probably a kernel bug
[15:13] <sebjan_> rsalveti: it can help to not start gdm, to get some extra memory for building :)
[15:13] <rsalveti> sebjan_: yeah, one of my builders is running with a minimal image, on a usb-disk :-)
[15:15] <sebjan_> rsalveti: I like this setup :)
[15:53] <hrw> http://hrw.pastebin.com/QH29C3zp - uboot found just 512MB ram on panda - normal?
[15:55] <topfs2> iirc there is the highmem problem so only 512 is used to not get some problems
[15:55] <rsalveti> hrw: iirc this was just a hard coded value
[15:55] <hrw> ok
[15:56] <rsalveti> topfs2: you can use until 768MB
[15:56] <topfs2> ah ok
[15:56] <topfs2> 512 is more than enough for me atm :)
[15:57] <hrw> builds takes mem... so more == better
[16:16] <hrw> http://hrw.pastebin.com/30ABmFSu ;(
[16:19] <rsalveti> ouch, that's weird
[16:20] <hrw> that 4->10s jump is because hdd has to start rotating
[16:21] <hrw> worked before ;(
[16:21] <hrw> have to test one thing
[16:22] <hrw> not related to 512<>1024MB ram
[16:22] <hrw> worked before ;(
[16:49] <vstehle> Hi rsalveti, would you please have some time to discuss this pvr-omap4 / mesa dev conflict issue mentioned on the pandaboard ML?
[16:49] <rsalveti> vstehle: yup, just replying the ml :-)
[16:50] <vstehle> Oh. Nice. I don't see an easy solution right now.
[16:50] <rsalveti> vstehle: sorry not answering before, still catching up on my emails
[16:50] <rsalveti> vstehle: in case you want a dev package for sgx, you have to create your own
[16:50] <rsalveti> and conflict/replace the mesa ones
[16:51] <vstehle> Hum... Ok, I'll do that, then.
[16:51] <rsalveti> but the problem is that the library lacks a proper soname
[16:51] <rsalveti> then it's not 100% compatible with the mesa ones
[16:51] <rsalveti> for example, if you end building something using the sgx package, you'll be unable to run with the mesa ones, unless you install the mesa-dev packages
[16:52] <rsalveti> because your bin will be linked with the lib.so
[16:52] <rsalveti> that mesa provides only at the -dev package
[16:52] <vstehle> The packages provide libX.so, libX.so.1 and libX.so.1.y; isn't this sufficient?
[16:53] <vstehle> Or do I need to specify this soname thing somewhere?
[16:53] <prpplague> hey pandaboard users
[16:53] <prpplague> i'm kicking around some ideas for expansion boards for the pandaboard, any thoughts on an expansion board the exact same pcb size as the pandaboard that adds 4 additional usb ports, rtc with battery, eeprom, seconday sd/mmc slot, secondary uart, and LVDS lcd interface
[16:54] <rsalveti> vstehle: it provides the proper links, but when you link something with it, it'll link against the lib.so, instead of lib.so.1, for example
[16:54] <rsalveti> the library doesn't export the proper soname
[16:55] <vstehle> Oh, that is in the .so itself? I did not realize that.
[16:55] <rsalveti> you need to change that in build time
[16:55] <rsalveti> in the sgx lib build time
[16:55] <vstehle> I can do that, but I don't know where to begin (Makefile or debian/* :)
[16:56] <rsalveti> probably at the sgx build scripts/makefiles
[16:56] <rsalveti> we have the same issue for the omap 3 packages
[16:56] <vstehle> Oh: http://en.wikipedia.org/wiki/Soname
[16:56] <rsalveti> that's why I didn't provide a -dev package for it
[16:57] <vstehle> Ok, I need what to do this week :) Thanks! First, -dev, then fix soname.
[16:58] <rsalveti> I'd prefer first soname, then -dev, but it's your call :-)
[16:58] <rsalveti> then it can be compatible with the mesa ones
[16:58] <rsalveti> and the replace/conflict will be 100% correct :-)
[17:02] <hrw|gone> http://hrw.pastebin.com/K7zW3RrL - problem is not rootfs related ;(
[17:04] <topfs2> btw vstehle and rsalveti I built xbmc with mesa-dev and it runs perfect with sgx
[17:04] <topfs2> incase its of interest :)
[17:04] <rsalveti> topfs2: yup, that's expected :-)
[17:04] <rsalveti> the other way around that's not
[17:04] <vstehle> Hi topfs2. Good to know.
[17:04] <topfs2> ah ok :(
[17:05] <topfs2> While I have about nil knowledge on the matter couldn't it be done by using the mesa-dev package?
[20:27] <rsalveti> sebjan_: can you try the proposed package for the bug 665039 and report back the result?
[20:27] <ubot2> Launchpad bug 665039 in linux-ti-omap4 (Ubuntu Maverick) (and 1 other project) "Crash at boot on blaze board on wlan driver init (affects: 1) (heat: 306)" [Undecided,Fix committed] https://launchpad.net/bugs/665039
[20:28] <rsalveti> then we can close this bug
[20:28] <rsalveti> I don't have a blaze to test
[20:57] <orbarron> got a quick ?? --> can someone point me to the ubuntu's OMAP4 kernel location for pandaboard?
[21:01] <rsalveti> orbarron: http://kernel.ubuntu.com/git?p=ubuntu/ubuntu-maverick.git;a=shortlog;h=refs/heads/ti-omap4
[21:02] <orbarron> rsalveti: thanks