linux

Open Source Graphic Cards — A reality?

Hi All,

Some of you, like me, really curse both ATI and NVidia about not being more open with their drivers.  I’d dearly love Intel to release some AGP graphic cards that I could use on the non-Intel motherboard in my desktop PC.  Another dream, was a truly open video card.  Some of you may also have seen the discussions on linux-kernel, regarding an open-source friendly graphic card.
Whilst chatting in (irc.freenoden.net), one of the other chatroom members mentioned this page.

The prototype, is a PCI based FPGA development card with dual DVI output, TV-out and three 300MHz Analog to Digital Converters. After it finishes stress testing, Traversal the company founded by members of the Open Graphic Project will not only be selling this board for those interested in a serious FPGA development tool but will offer the card at a great discount for the Open Graphics Project. Under the Open Graphics Project this development board will continue to grow until it becomes a graphic card in it’s own right. However the project does not finish there, because we will need to move from the standard PCI interface onto other faster interfaces. And after the engine and drivers have been finished we then want to investigate moving the graphic’s engine into an ASIC, which will enable our design to reach even greater speeds and very reasonable prices.

The card is powered by two FPGAs and provides dual DVI monitor outputs along with S-Video for TV output as seen in the photo below.

ODG1 preview

Apparently, the target is to achieve 20~30 frames per second in Quake III at 1280×1024 — quite an admirable target for a FPGA-based design.  I will be keeping a close eye on the development of this card, as this is looking very interesting.  I wish the people involved the greatest success. 🙂

KDE 3.80.2 Test Run

KDE 3.80.2 DesktopHi All…

Well, after much waiting, I managed to get KDE 3.80.2 to build and run. What can I say? Right now it’s not much to look at (low-res screenshot 800×600 100kB, full-res screenshot 2MB 2048×1576), and there are loads of sharp edges, but this is to be expected for a pre-alpha release.

My impressions:

  • The new build system — This is going to take a bit of getting used to, but cmake seems to do a nice job building KDE. And there’s progress displayed as each package is built — very handy with long-haul builds.
  • Build Time was significantly lower in KDE 3.80.2, compared to KDE 3.5 releases
  • Some of the core apps (Konqueror, Konsole…etc) seemed to work fine, although there were some instability issues. (not suprising) Konqueror also seemed to like picking up the resources from KDE 3.5.5, leading to some oddities in the interface.
  • KDE Control Centre appeared to be broken, in that it would not display applets. kcmshell worked however.
  • Some of the kicker Panel applets failed to work (e.g. the clock)
  • The startkde script locks up half-way (loading the Window Manager). In the screenshots above, I had to run kdesktop, kicker and kwin manually from xterm.
  • There appear to be DBUS communications issues that need to be resolved.

In short though… it’s early days, but it looks like we’ll be seeing a lot of behind-the-scenes action in this release. I didn’t see anything that resembled the ideas for Plasma, but perhaps that will come later. It’ll definately be worth having another look in a few months.

Now I know a number of you have been gagging for a guide on how to install KDE 3.80.2. Here’s a very rough guide.

Part 1: KDE Dependancies

For this release, you need a number of things:

cmake is fairly trivial, just add the appropriate line into your /etc/portage/package.keywords to allow you to install version 2.4.3 (currently unstable on x86) and run emerge cmake.

For the others… you’ll need to unmask both ebuilds in package.mask, and will need to hack the Qt ebuild to enable D-BUS support. D-BUS 0.95 is masked at the moment, pending updated mono bindings. Qt 4.2 is masked, pending D-BUS 0.95. You’ll need to uninstall earlier versions of D-BUS in order to install the new release. Be prepared, this will break HAL, avahi, and anything else you have which uses D-BUS, so prepare to have a big rebuild task ahead.

To enable support for D-BUS in the Qt ebuild you’ll need to change the following:

  1. Add dbus to the IUSE variable:
    IUSE="accessibility cups dbus debug doc examples firebird gif glib jpeg mng mysql nas nis odbc opengl pch png postgres sqlite xinerama zlib ${IUSE_INPUT_DEVICES}"
  2. In DEPEND, put in the D-BUS dependancy (it’ll be commented down below):
    glib? ( dev-libs/glib )        input_devices_wacom? ( x11-drivers/linuxwacom )        dbus? ( >=sys-apps/dbus-0.93 )"
  3. Update the qt_use function to pass -qdbus to the configure script (it’ll be commented out). Also, comment out the line below as shown here:
    use dbus        && myconf="${myconf} -qdbus" || myconf="${myconf} -no-qdbus" #      myconf="${myconf} -no-qdbus"

Once the updates are complete, run ebuild qt-4.2.1.ebuild digest. Then, install all of the above: emerge dbus qt.

Part 2: Installing KDE by hand

As mentioned earlier, this new release uses the cmake build system, which is a bit different to get used to. Some of the docs (namely the ones for kdebase) haven’t been written yet, but in short, the procedure is as follows:

  1. Unpack the source tarball (tar -xjvf /path/to/sourcepkg-3.80.2.tar.bz2) into a temporary directory.
  2. Create a build directory (mkdir sourcepkg-build), then change into it.
  3. Run cmake -DCMAKE_INSTALL_PREFIX=/place/where/you/want/kde/3.80.2 -DCMAKE_BUILD_TYPE=debug /path/to/sourcepkg-3.80.2
  4. Run make
  5. (As root) Run make install

The tricky bit, is the arguments to cmake:

  • -DCMAKE_INSTALL_PREFIX= specifies the directory in which you want KDE installed. I put my installation in /usr/kde/3.80.2 so that it was placed alongside my KDE 3.5 install (but not in a place where it would get clobbered). This location is entirely up to you.
  • -DCMAKE_BUILD_TYPE= specifies the level of debugging to include. Valid values are debug (normal debugging), debugfull (full debugging) and profile (full debugging with test coverage).
  • The last argument, points to the directory containing the sources.

The KDE components need to be installed in the following order:

  1. kdelibs
  2. kdepimlibs
  3. kdebase
  4. other packages

Those three above mentioned packages, will get you the desktop shown in the screenshots. I have not yet installed the remaining packages, but these three will get things up and running.

I don’t recommend people mess with this release, unless they know what they are doing… As I say though, I’ll be keeping a close eye on this release, as it looks like there’s going to be lots of goodies going in behind the scenes for us KDE users come release time. 🙂

Taking KDE 3.80.2 for a test drive

I’ve been a KDE user since installing SuSE Linux 5.3, and seeing KDE 1.0 in all its glory. I later tried Gnome 1.0/Enlightenment, when Red Hat 6.0 was released, and numerous other desktops, but I always found myself moving back to KDE. Prior to this, I had been using window managers like AfterStep, WindowMaker, FVWM and Red Hat’s FVWM95 clone, AnotherLevel.
The other day, I was looking around on the KDE news site, and noticed KDE 3.80.2 had been released for testing. This is a prerelease version of the upcomming KDE 4 desktop, intended for developers to take the desktop for a spin, port their applications to it, etc… Given my interest in the desktop, I figured I’d download a copy and take it for a spin.

The new KDE uses a different build system to previous releases. KDE 3.5 and earlier used GNU autotools, but not anymore. You’ll need cmake 2.4.3 or above to compile KDE 4. You’ll also need Qt4.2, and dbus, which means fun-and-games with masking, modifying ebuilds to disable some tests, and in general, lots of nasty hacks. Unless you know what you’re doing, I wouldn’t recommend anyone try it.

KDE 3.80.2 Build SystemAnyway… after much fiddling, I managed to get it building. Already, I notice something new. The new build scripts actually show the build progress, as KDE builds. This is handy if you’re running on slow machines, and the first time I’ve seen such a feature in use. Here (right) is what the start of the build looks like (low resolution, full resolution).
I shall upload some pics of the desktop the moment I get kdebase installed, but so far, this is prooving very different to previous releases of KDE.

Gentoo/MIPS Cobalt: n32 Stage 1 complete

Hi all…

Yep, the first 64-bit userland stageballs for Cobalt are well underway. A stage 1 tarball is being uploaded right as I type this, and Catalyst is just starting up the bootstrap script now to get the stage 2.

I also began a build for µClibc, but I’ve hit technical issues with Portage there, which I need to work through before I can make stages with it.

I won’t mention the link where to download them at this early stage, because these are über-experimental. They’re totally inappropriate to be used as a root userland, even on a crash-and-burn box at this stage, and will likely be that way until at least 2007.0 or even 2007.1. For those who know what they’re doing, I’m happy to provide the links to these experimental builds, ping me on IRC/IM, or shoot me an email.

On other news… the stable 32-bit glibc builds for 2006.1 will soon be moving to a mirror near you.
As always, I’ll keep you posted on the progress. 🙂

Gentoo/MIPS Cobalt: µClibc stages on their way

Hi All…

Whilst forging ahead into new territory for the Gentoo/MIPS project on the n32 front, I also figured now would be a good time to update the µClibc stages for mipsel.

I’ve completed a cross-compile from x86 of an entire mipsel environment, and I’m now just about to set this up to use as a seed stage in Catalyst.  Once this is done and I’ve bootstrapped a system with it, I should be set to build a full mipsel environment.  This will be compiled for MIPS-1 class CPUs, and thus should be compatable with embedded devices such as wireless routers.  Note: This does not mean we will officially start supporting wireless routers.  They are still officially unsupported, and users are very much on their own when troubleshooting problems on these devices.

I’ll keep you all posted on the progress. 🙂

Gentoo/MIPS Cobalt: n32 Stages just a little closer to reality…

Hi All…

Yep, I dusted off my n32 chroot again this morning (it’s 1:22am as I type this), determined to talk some sense into Portage. I figured I’d give it one last time before I invested the time into trying out Paludis (which I may still do yet, I’m hearing lots of good things about it).

So, what’s been holding me up? Well, the issue has been this nagging bug that I couldn’t figure out. Cobalt doesn’t have any n32 stages, let alone NPTL n32 stages. For the most part, I was able to nick the settings out of default-linux/mips/2006.1/generic-be/n32/* copying this into the default-linux/mips/2006.1/cobalt/ directory, and replacing mips64 with mips64el in the CHOST variable.

This worked quite well, but there was still one nagging issue that cropped up when trying to compile various packages, particularly portage itself:

These are the packages that would be merged, in order:

Calculating dependencies... done!
Traceback (most recent call last):
File "/usr/bin/emerge", line 3316, in ?
mydepgraph.display(mydepgraph.altlist())
File "/usr/bin/emerge", line 1650, in display
verboseadd += create_use_string(key.upper(), cur_iuse_map[key], cur_use_map[key],
KeyError: 'elibc'

On further investigation, I noticed that on all the working environments, there was a USE-expand flag: elibc_glibc. This is susposedly set in the base profile, but for whatever reason, my sub-profile transplant seems to have lopped this flag off. Portage would see this, and b0rk when it didn’t know which libc to use. Thus, I tried something… I hacked around it by setting USE="elibc_glibc" in /etc/make.conf then gave it another try. Sure enough, emerge --info now listed the illusive USE flag, and packages started compiling once more.

Right now, I’m rebuilding all the system packages in my chroot (which also has lib64 stuff floating in it). This will hopefully get me to the point of producing a first seed-stage for Catalyst, and will allow stagebuilds to be done for n32 at long last. As for n64? Well, time will tell… it’s certainly a possibility. I’d like to first discover why this USE flag is getting dropped… as setting it in make.conf is not an acceptable workaround IMHO, but it’s better than nothing.

I shall keep you all posted on my progress. 🙂

Gentoo/MIPS: New Cobalt Stage 3 Uploaded, and Documentation to review…

Hi All…

A few days ago I mentioned I’d be uploading the new stage 3 tarball for Gentoo/MIPS… I’ve now done this, and for now, you can find it in my devspace. Cobalt users… please give this stageball a try and report back… I’d like to get things checked out early before I get it pushed out to the mirrors.

Also needing attention, is the handbook. My draft version has been updated with the new profiles I mentioned in my last post, as well as other corrections. Any typos, mistakes, ommissions… please let me know. 🙂

Regards,
Stuart Longland

Gentoo/MIPS Cobalt 2006.1 Stage 3

I’ve finally completed the Gentoo 2006.1 stage 3 tarball… I’ll hopefully be uploading this on Tuesday when I’m next at uni.

Over the next week, I’ll be updating the handbook, and putting together some new netboot images. Things to look out for — the profiles have moved somewhat since the last release.

The profiles are now arranged as follows (Comments mine):

stuartl@beast /usr/portage/profiles/default-linux/mips $ find . -type f -name make.defaults -exec dirname {} \;

Outdated 2006.0 profiles
./cobalt/2006.0
./mips64/ip28/2006.0
./mips64/multilib/2005.1
./mips64/multilib
./mips64/multilib/2006.0
./mips64/n32/2006.0
./mips64/2006.0
./2006.0

New profiles
./2006.1/cobalt/o32/nptl Cobalt with NPTL (not ready for production use)
./2006.1/cobalt/o32 Cobalt (using linuxthreads, recommended for Cobalt users)

Generic Big Endian profiles (including SGI Indy, Indigo2 R4x00, O2)
./2006.1/generic-be/n32/nptl
./2006.1/generic-be/n32
./2006.1/generic-be/n64/nptl
./2006.1/generic-be/n64
./2006.1/generic-be/o32/nptl
./2006.1/generic-be/o32

Profile for SGI Origin 200/2000
./2006.1/ip27/n32/nptl
./2006.1/ip27/n32
./2006.1/ip27/o32/nptl
./2006.1/ip27/o32

Profile for SGI Indigo2 Impact R10000
./2006.1/ip28/n32/nptl
./2006.1/ip28/n32
./2006.1/ip28/o32/nptl
./2006.1/ip28/o32

Profiles for SGI Octane/Octane2
./2006.1/ip30/n32/nptl
./2006.1/ip30/n32
./2006.1/ip30/o32/nptl
./2006.1/ip30/o32

You’ll also notice… we’re working on NPTL support on MIPS. The new glibc-2.4 no-longer includes support for linuxthreads, and thus all architectures are migrating across. NPTL support requires gcc-4.1, kernel 2.6.17, and use of the NPTL profiles. This is a work in progress and is not recommended unless you know what you’re doing. It is recommended people stick to using linuxthreads until things are stable using NPTL.

As mentioned earlier… this will be going into the handbook. If people have any questions… feel free to drop in, we lurk in -mips on irc.freenode.net.

Gentoo 2006.1 for Cobalt: Preparing for another build

Well, it’s that time again… time to wind up the Qube2, and get it building a seed stage in preparation for yet another release of Gentoo.

Then it’ll be a case of waiting for the official 2006.1 snapshot to be released, and the full stagebuilds will begin in earnest. I may release pre-2006.1 stageballs, we’ll see how everything pans out.

I’ll keep you all posted. 🙂

Gentoo Linux/MIPS for Cobalt — The beginnings of a 64-bit port

Hi All,

Just to let you know, things haven’t been totally silent over here… Lately, I’ve been messing about with a mips64el tarball I got from Joel Martin. This is a full multilib environment (o32/n32/n64), which was cross-compiled from x86 via some build scripts based around portage and crossdev. There was however, a snag … there were artifacts of cross-compilation lurking about, some i386 binaries (cc1, cpp1 and tbz2tool being a few), and some even more bizzare “n32 MIPS32” binaries (which I’d have normally considered impossible).

This afternoon, after some tinkering with crossdev, I discovered I was able to get gcc-4.1.0 to build enough to patch up the broken files, to the point I was able to build packages like rsync (although it failed to package due to tbz2tool being broken). I’ve since managed to rebuild things like libtool, and have Python 2.4 well underway as I type this post.

What does this mean? Well, in the distant future (probably in the next few months following my end-of-semester exams), there’s the possibility of providing n32 and n64 (i.e. 64-bit) stages for Cobalt hardware. These will probably be as broken as their big endian counterparts (probably more so), and won’t be officially supported by the MIPS team, but are provided for those who know what they are doing. Those who want stable stages, just stick with the 32-bit (o32) ones we officially support. 🙂

It’ll be a while before I make 64-bit netboot images available… it won’t be anytime soon… if people want a 64-bit userland, it’s best at this point to use it in a chroot environment (as I am doing now), using it for a root environment could lead to all sorts of avoidable headaches.

Also in the plans, as soon as I have the time to do so, I’ll look at a LiveCD for netbooting Cobalt servers. This will be a minimalist x86 LiveCD, which when booted, turns a standard x86 desktop into a BOOTP/NFS server required for booting Cobalt hardware. Initially this will be exclusively for MIPS-based Cobalt servers, but may later be extended to x86-based Cobalt hardware, when suitable netboot images become available. SGI hardware is also a possibility that I may consider. At the moment, it’s little more than some thoughts in my head … I’ll keep you all posted on what happens here.

Anyways, I best get some sleep (1:19am right now), and get some university work done … I’ve got a very busy week ahead. 😐