Poking in Mesa…

Yesterday I had a look at the curent mesa sources…

And I found out that x86 and amd64 had plenty of optimized code, mostly hand made assembly!
What about ppc you’d ask? Well, there is an empty dir with the code to know if the cpu has altivec or not =_=…

I should study for my last exams so I won’t do much about it in the short time =/

Fixing errors on the toolchain

Recently got some reports about bizarre hangs on binutils on non ppc arches

Today eventually I fixed the (pretty dumb) error I made while preparing the patch, don’t ask me why I hadn’t tracked it before (working on semi-native environment shadowed it pretty well).

Hopefully today I’ll add a fix for the linux makefile trying to use uname -m while crossbuilding headers.

Japan!

As probably some of you already know (since I’m telling everybody about that) I’m about to spend 3 weeks in japan, with some luck I’ll get some random uplink now and then.

I’ll be in Tokyo for a week then I’ll wander for the Island by train for a week and then I’ll spend the rest of the time in Tokyo again.

First gentoo system image for Cell done

It is basically a standard ppc64 image with gcc-cell native, the spe-elf toolchain and libspe. I tested it with the sistemsym a bit today and seems working as should soon I’ll try to make a proper image for the simulator with a quite stripped down baselayout (since looks like it doesn’t like at all doing much I/O from and to the disk but is blazing fast on memory, out init system uses too much the disk…)

Screenshots and instructions about how to craft it soon =)

Plans

I’m still waiting for a new HD for my tibook and I’m still half productive (I’m at home 1/3 of my time and I can effectively use the pegII, that genesi gave me (than you so much, I would be dead without it) for developing altivec related code, just there. Hopefully I’ll have my nice laptop back the next month (looks like the HD I ordered is more or less a chimera…).

Here there are some of my plan for the next months about gentoo/ppc:

  • Get gcc-4 and test the autovectorization features and hopefully provide a new and sane set of cflags to use it safely
  • Update the rest of the toolchain (glibc 2.3.5 was recently released and hopefully will be in a good shape soon)
  • Test some laptop related kernel patches that are pending for inclusion in the mainline but looks quite interesting
  • Try to altivectorize snow in order to have it working at a decent speed here

Probably something else could happen but I think that I already put too much in the list.

How to file a bug

Recently I almost got overwhelmed by bugs and other issues in real life ( uni related stuff and some other stuff in some clubs I belong) and other opensource projects

The problem usually isn’t the number of bugs but how they are written.

If you want me not to lose the latest glimpse of sanity please follow those rules:

  • Always check if there is a duplicated bug with the EXACT same issue you are experiencing
  • Every time you fill a bug PLEASE use the keywords if applies (eg, for a new ebuild put EBUILD)
  • Please be specific when you fill the subject, please put the full version of the package that way we can filter it
  • Be as much informative as you can, within the limits (eg, sending me your hd image or attach it to bugzilla isn’t nice; attaching your config.log if configure refuses to work or your emerge info and .config if a kernel module doesn’t work is indeed the best practice)

Thanks

Snapshots, stuff and so on

My last posts dates long time ago. I guess that’s the right time for an update.

– real life:
I have to complete my thesis and courses within the next summer, will I manage to get that result, keep working on gentoo, have some of social life and retain my mental sanity? Nobody will know. What’s sure is that I had an ugly compulsory course about win32 internals and .net, I couldn’t think about anything worse than ill coded java or C++, now I can say that there is something wrong, is written in C and is the win32 api.
(Incidentally I should also write a screen/audio capture software in .not for the course and I don’t know how to start, if someone has some ideas feel free to mail me the suggestions)

– media-video:avifile vs mlt++ Who’s more broken?

I’m actively wiping any trace of avifile from our tree, it broken, unmantained and superceeded by other better tools, what is currently stopping me from close it’s coffin is jahshaka and mlt++, mlt++ is a wrapper for mlt (mlt.sf.net if you don’t know) and is currently used by that fine or less NLE as backed. mlt++ is broken and since I’m not so accustomed to C++ (sooner or later I’ll write yet another text about not using C++ if you don’t really want to spend the right time over it). mlt probably could have a release sooner or later and probably those bindings will be fixed, anyway I’m working on a snapshot of it, looks like that just the C++ wrappers are bogus =/ pity, no jashaka with mlt for now.

– media-video:mplayer and ffmpeg To snapshot or not to snapshot?
In those month many interesting new bugs got added, ahem, squashed and some features got in as well, I’m thinking about adding a new snapshot to match what the new xine-lib expects from ffmpeg and probably produce an mplayer snapshot to have most of the related bugs that actually are crawling in bugzilla wiped.

Since the server of mphq is currently about to die and the replacement in on route probably I could wait a bit more and ask upstream for a winter release.

– PPC:

I’m currently testing gcc-4.0.2 and the new glibc and they looks quite promising (and overally more or less bugfree), the only problem I have is about gdb not debugging so well and probably I should pay more attention at the issue.

– media-gfx:blender

I do hope the new, upcoming, blender release won’t spawn again lots of bugs, the new features added look nice, if only had the time to study it properly so I can do something else than a cookie-man (first example of the blender book)

– media-gfx:gimp

The new gimp 2.4 will be great, I’m testing (what an excuse to spend some time editing some old pictures and drawing something…) the latest 2.3.x releases and they look relatively stable (at least on ppc) and with some nice features.

– not really gentoo projects: xiph-rtp

As part of my thesis I’m pushing some draft about vorbis and theora in rtp towards rfc status (I do hope they reach it), so far the most annoying problem has been the vorbis bitstream internals (since is quite complex), ogg (since is too simple and doesn’t have all the informations I need ^^) and getting input for every people/association interested on the topic.

– not really gentoo projects: fenice and nemesi

Another part of my thesis is make fenice and nemesi support vorbis and theora. If you are wondering what they are http://streaming.polito.it
They are quite promising and, beside the Italian doxy comments still present, quite understandable. The main problem: too few people are working on it so the development isn’t as fast as it should.

I guess that’s long ago and I should go back coding/studying/writing/doing something productive

Plans for the week

I’m getting more and more involved in the radio project for the polito‘s webradio, I’ll be one of the people giving tech direction for the whole thing. Once the boxes will arrive I’ll have to figure how the best and fastest way to configure alsa and jack on m-audio. Given I haven’t had suck hardware will be quite interesting even if I should admit I’m a BIT afraid.

mlt hasn’t been released yesterday, BAD since now I have to wait a bit before killing avifile.

In the mean time, I’d like to kill the deprecated oggvorbis.

School is still present and I’m trying to follow lessons in a more regular way, no matter how boring they could be ^^;

More will follow, probably later today