DocScrutinizer05 changed the topic of #neo900 to: http://neo900.org | conversations are logged to http://infobot.rikers.org/%23neo900/ and http://irclog.whitequark.org/neo900 | 2013-11-04 - the day our fundraiser reached its goal | 2014-05-01 360 devices 75k€| 0712 183 ~30k | 0810 300 ~49k | 0914 346 ~56k
<bencoh> 23:28 < freemangordon> but as there is no gcc4.7 in the repos... :)
<bencoh> I guess we could make that happen though
<bencoh> prolly not the hardest part
<bencoh> actually ... since sb is running in qemu-user and all ... dont we just need to build it as a regular package first ?
<bencoh> and find some way to glue it with sb_gcc_wrapper
<bencoh> (uh, okay, looks like I didnt read part of the backlog properly ...)
<bencoh> (but I'm still interested in the reason why it doesnt compile/is complicated)
norly has quit [Remote host closed the connection]
jonwil has quit [Quit: ChatZilla 0.9.91 [SeaMonkey 2.30/20141013232806]]
xes has quit [Remote host closed the connection]
nox- has quit [Quit: Leaving]
sixwheeledbeast has quit [Ping timeout: 265 seconds]
sixwheeledbeast has joined #neo900
jonwil has joined #neo900
Kabouik has quit [Ping timeout: 255 seconds]
jusa_ has quit [Ping timeout: 255 seconds]
trench has quit [Ping timeout: 255 seconds]
jason25262 has quit [Ping timeout: 240 seconds]
trench has joined #neo900
j4s0nmchr1st0s has quit [Ping timeout: 265 seconds]
jusa_ has joined #neo900
jonwil has quit [Remote host closed the connection]
sparetire_ has quit [Quit: sparetire_]
jonwil has joined #neo900
<freemangordon> bencoh: imagine gcc running in qemu. she will get old while she waits for some big project to compile :).
ecloud is now known as ecloud_wfh
Oxyd76 has quit [Remote host closed the connection]
Oxyd76 has joined #neo900
j4s0nmchr1st0s has joined #neo900
Oksana has quit [Read error: Connection reset by peer]
Oksana has joined #neo900
che1 has joined #neo900
<bencoh> freemangordon: so ... how does it work atm ?
kolp has joined #neo900
Kabouik has joined #neo900
<freemangordon> bencoh: it is x86 compiled and statically linked with the needed libs
<jonwil> so does xprot work with gcc 4.2.1 or not?
<freemangordon> it does not, gcc 4.2.1 miscompiles a_xprot_lfsn_mono()
<freemangordon> jonwil: ofc we can throw gcc intrinsics away and use plain assembly. But I am not sure this is the right way :)
<jonwil> We should throw away intrinsincs in the specific case it miss-compiles
<freemangordon> jonwil: so, why it is better to use plain asm instead of a pre-compiled .a?
<freemangordon> did you see my latest commits?
<jonwil> yes I have
<freemangordon> also, most of the other neon stuff compiled with 4.2.1 is 20=50% slower, while the one compiled with 4.7.2 is about 10-30% faster
<freemangordon> compared to stock that is
<jonwil> anyhow, did you end up talking to jusa_ about the things you didn't know about?
<freemangordon> he seems busy lately
<jonwil> ok
<freemangordon> jonwil: the only sane way to use asm instead of intrinsics is to tell gcc to produce asm output and to make the functions out of it
<freemangordon> but that's an enormous job I am not going to do :)
j4s0nmchr1st0s has quit [Remote host closed the connection]
<freemangordon> jonwil: take that https://gitorious.org/pulseaudio-nokia/pulseaudio-nokia/commit/1d3448594cee7077333de9147621675aa0474a5b for example. gcc and intrinsics make it > 25% faster. I won't give up on that just to keep the things politically correct (i.e. no .a files on gitorious) :)
j4s0nmchr1st0s has joined #neo900
<freemangordon> and I even didn;t play much to optimize, just a couple of hours, I can bet if we want we can squeeze another 10-20%
<freemangordon> which is simply not possible with asm, no way I am going to learn cortex-a8 pipeline
<freemangordon> jonwil: so, I will, put precompiled libs on gitorious, unless you threat to rage-quit the project :P
<freemangordon> if I do so that is
j4s0nmchr1st0s has quit [Remote host closed the connection]
j4s0nmchr1st0s has joined #neo900
<freemangordon> oh, great sailius is again active on linux-omap ML
astr has joined #neo900
che1 has quit [Ping timeout: 272 seconds]
Pali has joined #neo900
j4s0nmchr1st0s has quit [Ping timeout: 240 seconds]
j4s0nmchr1st0s has joined #neo900
Kabouik_ has joined #neo900
Kabouik has quit [Ping timeout: 240 seconds]
j4s0nmchr1st0s has quit [Ping timeout: 255 seconds]
j4s0nmchr1st0s has joined #neo900
nicksydney_ has joined #neo900
nicksydney has quit [Ping timeout: 244 seconds]
nicksydney has joined #neo900
che1 has joined #neo900
nicksydney_ has quit [Ping timeout: 244 seconds]
Kabouik has joined #neo900
Kabouik_ has quit [Ping timeout: 255 seconds]
<Pali> we have maybe chance to see working n900 (front) camera with upstream kernel...
Kabouik has quit [Read error: Connection reset by peer]
<jonwil> Damn, my house is being invaded by flying beetles :(
<DocScrutinizer05> Pali: on N900 the front camera is muxed with main camera to same interface
<Pali> DocScrutinizer05: I know, but problem is with isp camera subsystem in kernel
<DocScrutinizer05> aaah
<Pali> it is broken since 3.5
<Pali> we have patches for mux
<Pali> but only green grass is visible :D
<DocScrutinizer05> >> The status is now that images can be captured using the Nokia N9 camera, in which the sensor is connected to the CSI-2 interface.<< made me suspect different
<DocScrutinizer05> patches for mux sounds good, it's not exactly simple thing to handle, since I guess several things that usually get done in module init now have to get done on switching mux, and one module needs to export *2* interfaces for two cameras
<DocScrutinizer05> typical case for epic fail of DT handling real life complexity
<DocScrutinizer05> well, maybe the module alrady was smart even for single cam, and dit whole cam power up and init only on open() to cam device
<DocScrutinizer05> did*
<DocScrutinizer05> so module_init would only expose the device nodes
<jonwil> great to see kernel progress is being made :)
<Pali> in DT should be description of that two GPIOs and (ISP) driver should handle it
che1 has quit [Ping timeout: 264 seconds]
<Pali> anyway, tomorrow I'm in Krakow... at ACM ICPC (regional programming contest in algorithm)
che1 has joined #neo900
<DocScrutinizer05> >>regional programming contest in algorithm<< oooooOOOOooohhhh! 8-)
<bencoh> :)
<DocScrutinizer05> will it be like "A competes B in chess played by minimax algos written by the combatants" ?
<Pali> no, this is something other... there is no battle between teams like playing chess... each team will get set of problems and want to solve as most as possible in less time
<Pali> you have problem description and you need to program algorithm which solve this problem for every possible input (= write correct output)
<Pali> once you think you have working program you will sent it for testing and automatic system start it on set of input data and check if output of program is correct for tested input data
jonwil has quit [Quit: ChatZilla 0.9.91 [SeaMonkey 2.30/20141013232806]]
<Pali> example of (easy) problem: you have N towns and on stdin is specified which two towns are directly connected (with distance). On stdout you should output minimal path cost between towns A and B (or write that there is no possible path between A and B)
freemangordon has quit [Ping timeout: 245 seconds]
freemangordon has joined #neo900
obsed has quit [Ping timeout: 245 seconds]
j4s0nmchr1st0s has quit [Quit: j4s0nmchr1st0s]
j4s0nmchr1st0s has joined #neo900
sparetire_ has joined #neo900
freemangordon_ has joined #neo900
<freemangordon_> Pali: great, finally some progress on ISP
<Pali> yes :-)
<freemangordon_> which most probably means I will have to RE omap3camd after PA :(
che1 has quit [Ping timeout: 245 seconds]
j4s0nmchr1st0s has quit [Quit: j4s0nmchr1st0s]
Kabouik has joined #neo900
paulk-collins has joined #neo900
paulk-collins has quit [Ping timeout: 265 seconds]
paulk-collins has joined #neo900
nox- has joined #neo900
<freemangordon_> LOL
<freemangordon_> doc, you're getting better, on the second picture it is harder to say what's wrong :P
<DocScrutinizer05> :-)
Kabouik has quit [Ping timeout: 255 seconds]
<freemangordon_> hmm, nothing wrong with it
<DocScrutinizer05> sure
<freemangordon_> besides it is tiny
<DocScrutinizer05> then how about this http://wstaw.org/m/2014/11/14/plasma-desktopXx1987.png
<freemangordon_> sorry, my eyes are not what they used to be :)
* DocScrutinizer05 considers publishing a "The Holy Bible" set of issues, with one char per page
<freemangordon_> test
<freemangordon_> shit!
<freemangordon_> another test
Kabouik has joined #neo900
chainsawbike has quit [Quit: yep... i broke it good that time...]
arcean has joined #neo900
paulk-collins has quit [Read error: Connection timed out]
paulk-collins has joined #neo900
paulk-collins has quit [Quit: Quitte]
astr has quit [Ping timeout: 265 seconds]
astr has joined #neo900
Pali has quit [Remote host closed the connection]
j4s0nmchr1st0s has joined #neo900
Kabouik_ has joined #neo900
Kabouik has quit [Ping timeout: 245 seconds]
xes has joined #neo900
freemangordon_ has quit [Quit: Leaving.]
che1 has joined #neo900
jonwil has joined #neo900
j4s0nmchr1st0s has quit [Quit: j4s0nmchr1st0s]
arcean has quit [Read error: Connection reset by peer]