plugwash
Forum Moderator
Forum Moderator
Posts: 3357
Joined: Wed Dec 28, 2011 11:45 pm

Re: Debian Hard Float (armhf) for RPi

Thu Mar 29, 2012 12:43 pm

mpthompson said:


BTW, I also have a Mele A1000 on order to see if I can get a basic port of Debian armhf running on it. http://www.cnx-software.com/20.....droid-stb/


Nice to hear people are trying to get debian armhf running on more hardware.


It runs the AllWinner A10 (Cortext A8) SOC which seems to be the hottest chip coming out of China these days.  Being half the cost of the iMX53, a 1 GHz processor, a decent 512M RAM and a real SATA interface, I'm hoping these can be converted into inexpensive build machines if there is enough interest in pushing through a full Debian armhf port for the Raspberry Pi.   A number of these along with a few iMX53 systems for heavy lifting could probably make a nice build cluster for Debian ARM.


The problem is if you have a mixture of "strong" and "weak" autobuilders then somone has to sort out which packages can be built on the "weak" autobuilders and which require the "strong" autobuilders.

I'd expect the time cost of sorting that out would outweigh the financial cost of running a homogenous build cluster.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Thu Mar 29, 2012 2:46 pm

Good point about sorting out strong and weak autobuilders.  Anyway, I'm getting way ahead of myself. It will be nice just to have another solid Debian armhf system to work with if I can get that far with the Mele A1000.

On the compilation front, I ran into a gcc compiler abort while building Perl.  The the error indicated problems in the intermediate RTL code produced by compiler.  I'll need to see if the same problems exist for the package under regular armhf (I suspect not) and then try to understand the best way to mitigate the issue so I can continue building.  It's these types of issues that will slow things down.

Of the dozens of packages created so far, this is the first time this type of problem has cropped up.  Hopefully it's rare.  I suspect the ARMv6+VPF with the hard-float ABI flags set for compilation haven't been really well tested and it's not unexpected that these types of errors might crop up.

plugwash
Forum Moderator
Forum Moderator
Posts: 3357
Joined: Wed Dec 28, 2011 11:45 pm

Re: Debian Hard Float (armhf) for RPi

Thu Mar 29, 2012 3:56 pm

The first thing i'd suggest trying when you run into internal compiler errors is playing with the optimisation level.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Thu Mar 29, 2012 4:24 pm

Yep, turning the optimization level from -O2 to -O1 fixed the compilation issue.  Looking in the logs for the Debian package I see that this has been done/undone various times for the arm port due to issues in various versions of gcc popping up.  Perl is back in the queue to be compiled.

Right now I'm building everything manually with debuild.  I guess I should look into automating this stuff more or I'll be spending a LOT of quality time staring at the compilation output as each package compiles.

BTW, I'm really learning to hate C++ compiled packages.  A package of say 100 straight C files will typically compile in 5 to 15 minutes or so.  C++ packages of about the same number of files, for instance aptitude which I'm waiting to compile now, could drag on for hours.  To be fair, the Gnome pieces of aptitude are being compiled which required me to install 100's of megabytes of dependent libraries -- all of which I'm crossing my fingers are dynamically linked.  Not sure how to pick an choose which packages I want made from a particular source package as it seems like an 'all or nothing deal', which is fair enough.

shirro
Posts: 248
Joined: Tue Jan 24, 2012 4:54 am

Re: Debian Hard Float (armhf) for RPi

Thu Mar 29, 2012 9:16 pm

Have you given Clang a go on some of those big c++ projects? I find it really nice and fast and it even has decent error messages. I am not sure if it is up to compiling a linux kernel but I imagine it is probably ok for lots of other things.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Thu Mar 29, 2012 9:39 pm

shirro said:


Have you given Clang a go on some of those big c++ projects? I find it really nice and fast and it even has decent error messages. I am not sure if it is up to compiling a linux kernel but I imagine it is probably ok for lots of other things.


I'm not familiar with Clang and I've been sticking pretty much to source packages as distributed.  If I make any attempt to modify the sources outside of the "debian" configuration directory, some part of the build process barks at me for changing sources and promptly aborts.  I certainly need to know more about the package build process to better understand how to go about bending the rules a bit.  I'm picking things up as I go, but the "school of hard knocks" is in full effect for me.

The problem with Perl is back even with the -O1 optimization level -- seemed to work outside the package build process, but within a package build some compiler flags must be being set that I have to figure out.  Also ran into aptitude compilation problems.  Some of the test code is failing and the build just hangs rather than aborting.  Oh well, I'll get through these.

The good news is that my fear that an overwhelming number of packages would be statically linked is unfounded.  If this were the case, it would have contaminated all the packages I'm building with ARMv7 code as right now I'm building on a mixed ARMv6/ARMv7 system so that I can pull down regular armhf packages to meet build dependencies.  I've only found very few instances of static links against libraries and in those few cases it's pretty easy to pull down the associated libraries and compile then ARMv6 as well.

shirro
Posts: 248
Joined: Tue Jan 24, 2012 4:54 am

Re: Debian Hard Float (armhf) for RPi

Thu Mar 29, 2012 10:06 pm

I hoped static stuff would be rare. Good to see that is working out.

After chasing my tail for a week I finally worked out that it didn't matter how many different kernel versions I tried to compile for the mx53, what patches, debugging or magical incantations. I wasn't going to get USB even in a kernel from the freescale git. Until I used the Freescale u-boot.

u-boot is totally evil. Just load the kernel dammit. I am so happy we will not need to deal with it on the Pi.

Looking forward to trying out the results of your hard work.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Fri Mar 30, 2012 12:47 am

shirro said:


u-boot is totally evil. Just load the kernel dammit. I am so happy we will not need to deal with it on the Pi.


I've had much the same experience as you.  I'm now on the Freescale u-boot and their Kernel.  No use battling the kernel when I have other work I want to get done.  I needed 1Ghz and the Freescale kernel was the only way I saw to get it.

Actually, I'm not sure it's u-boot that's evil.  To me it seems the kernel has certain expectations with regards to how the bootloader is leaving the system before passing control to it.  On an ARM system that doesn't have a standardized BIOS, this just seems plain wrong.  No matter what u-boot or any other bootloader does to get the kernel image in memory, the kernel should be able to reconfigure the system as needed.  At least that's the way I would like to see it.

Anyway, the iMX53 is a good warm-up for what are sure to be similar issues when I get my hands on the Mele A1000 I have on order.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Fri Mar 30, 2012 9:52 pm

I'm continuing to make good progress.  I've got two problematic packages that I'll examine later (perl and aptitude) and about 40 libraries left.  All the runtime applications and most of the libraries for debootstrap+build-essential are done.

On the library front, I've run straight into the need to compile three different versions of the Boost.IOStreams library which various packages are linked against.  Ugh!!! Being C++ code these are taking forever to compile and apparently I need to compile the FULL source just to get the IOStreams libraries.  Oh well, they will get made.  Just not quickly.

At the rate I'm going by early next week I should be able to use debootstrap to create a Debian ARMv6+VFP file system and have it boot under QEMU with an emulated ARMv6+VFP processor.  If that works, that should hopefully validate all the work I'm doing.  It's also the best I'll be able to do until I get my hands on real RPi hardware.

Assuming I reach that point, I'll figure out a way to share all the packages with other people interested in this work.  Also, I would be more than happy to help someone who has a Raspberry Pi beta boards to create their own SD image to boot from.  It would be very, I mean very, very, nice to run some floating point benchmarks on real RPi hardware and compare the results against the same benchmarks on less optimized releases of Linux.

----

Looking ahead and assuming my goals above are reached, it would be nice to continue the work to at least get to something more than just minimal Debian system -- perhaps something including a GUI.  However, I can't continue to manually compile the packages like I've been doing up to now.  I'll need to switch to the automated Debian package building tools such as sbuild.  However, with only a single iMX53 QSB to do the builds on, the required packages won't be popping out very fast.

I'm wondering if Amazon EC2 might provide a solution.  The idea I have in mind is to run EC2 instances of Linux+QEMU where QEMU would be running an instance of sbuild.  Although the packages would be built slower than they would be on my iMX53 QSB, if I got everything working right I could throw 10, 20 or more EC2 instances to build the packages.  Of course, there may be problems running QEMU in this manner, but it is something I'm pondering.

The other alternative would be to start building a cluster of iMX53s to pound out packages, but I'm not certain I'm ready for such a commitment without knowing if the results would be well utilized (it also has to pass the wife test).

I guess if things got that far I should seek help from Debian folks to see if there is interest in the Debian community to help push this project further.  I'm learning a lot as I go along, but I certainly could use the help of people already experts at using the tools I'm just learning to use.

Mike

shirro
Posts: 248
Joined: Tue Jan 24, 2012 4:54 am

Re: Debian Hard Float (armhf) for RPi

Sat Mar 31, 2012 9:11 am

When you have a proof of concept up and going on hardware I think it would be good if it became an official unofficial Debian port. You might be able to get some time on Debian's arm build infrastructure.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Sat Mar 31, 2012 3:56 pm

shirro said:


When you have a proof of concept up and going on hardware I think it would be good if it became an official unofficial Debian port. You might be able to get some time on Debian's arm build infrastructure.


I'm hoping things go that way.  I've been reading the archives with regards to how the armhf port got started just to understand how these things go.  Early on with armhf, it it was already anticipated that there would be flavors (or is it flavours) of Debian armhf for specific ARM processors just as I'm attempting to get started with the RPi.  What's unclear to me is how this fits into Debian proper.  The fairly rigid hierarchy and infrastructure of Debian is terrific from the standpoint in that it has allowed me to leverage the years of man effort put into it so that I can make significant progress on an RPi optimized Linux runtime.  On the other hand, it is a bureaucracy that I know little about.  I'll start asking questions as it become more clear that what I'm building is worthwhile.

On the building front, I've completed all the packages that should go into a basic bootstrap of Debian and the build-essential packages.  That leaves me with the six packages I'm having trouble building and will need to give special attention towards.  Of course, four of them are huge C++ builds that take anywhere from 2 to 6 hours to get to the point in the build that they fail.  I don't relish debugging those.

Once that is taken care of and I've done my own testing under QEMU I'll start asking on the forums for help from people with RPi beta boards.  At this point, even a smoke test with an image would be encouraging and let me know I'm on the right track.

Mike

dom
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 5243
Joined: Wed Aug 17, 2011 7:41 pm
Location: Cambridge

Re: Debian Hard Float (armhf) for RPi

Sat Mar 31, 2012 8:23 pm

I won"t have a huge amount of time available for debugging, but if you point me at an sdcard image, I can test it and let you know how far it gets.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Sat Mar 31, 2012 8:53 pm

dom said:


I won"t have a huge amount of time available for debugging, but if you point me at an sdcard image, I can test it and let you know how far it gets.


Thanks Dom, that would help a LOT.  I'll do my best to not waste any of your time.

Before I get you an image, let me first verify things by creating a generic image for QEMU -- probably in the next day or two as I finish of the last few troublesome packages.  I'll then attempt to create an SD image for RPi using the kernel from the Debian release from the middle of January.  I think I should be able to verify that under QEMU as well if I ignore the kernel messages about hardware issues.  At that point I'll contact you that an SD image is ready for you to try and where you can download it from.

Don't expect a lot from it, as it will be minimal.  However, it will have the build-essential tools so you can compile and run some benchmarks if you so desire.  Hopefully there won't be any debugging -- I say as I cross my fingers.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Mon Apr 02, 2012 4:34 am

OK, I'm getting very close. I can use the help of someone familiar with how debootstrap works and what the nature of the issue I'm running into below.

I now have all the packages compiled, put them into a local mini-repository and I'm now attempting to use them with debootstrap on an emulated ARMv6+VFP running Debian armel within QEMU.  Because the executables can't be run on the armel side of things, I'm using a deboostrap in a two stage mode with the --foreign flag.

The following command line is for the first stage of debootstrap:

---------------------------------------------

sudo debootstrap --arch=armhf --foreign --variant=buildd --verbose testing /home/mike/armhf file:///imports/rpi/mirror/debian

---------------------------------------------

That runs fine.  I use the following command to initiate the second stage debootstrap within the chroot environment:

---------------------------------------------

sudo chroot /home/mike/armhf /debootstrap/debootstrap --second-stage --verbose

---------------------------------------------

This runs fine for a while, but I run into the following error:

---------------------------------------------

I: Configuring e2fsprogs...
I: Unpacking the base system...
I: Unpacking gnupg...
I: Unpacking libusb-0.1-4...
I: Unpacking debian-archive-keyring...
I: Unpacking libstdc++6...
I: Unpacking gpgv...
I: Unpacking readline-common...
I: Unpacking libreadline6...
W: Failure while installing base packages.  This will be re-attempted up to five times.
W: Failure while installing base packages.  This will be re-attempted up to five times.
W: Failure while installing base packages.  This will be re-attempted up to five times.
W: Failure while installing base packages.  This will be re-attempted up to five times.
W: Failure while installing base packages.  This will be re-attempted up to five times.
---------------------------------------------

Looking in the debootstrap log file I see more verbose messages about the problem:

---------------------------------------------

...

Unpacking libstdc++6 (from .../libstdc++6_4.6.3-1_armhf.deb) ...
Unpacking gpgv (from .../gpgv_1.4.12-4_armhf.deb) ...
Unpacking readline-common (from .../readline-common_6.2-8_all.deb) ...
dpkg: regarding .../apt_0.8.15.10_armhf.deb containing apt, pre-dependency problem:
apt pre-depends on libstdc++6 (>= 4.6)
libstdc++6 is unpacked, but has never been configured.
dpkg: error processing /var/cache/apt/archives/apt_0.8.15.10_armhf.deb (--unpack):
pre-dependency problem - not installing apt
Unpacking libreadline6 (from .../libreadline6_6.2-8_armhf.deb) ...
Errors were encountered while processing:
/var/cache/apt/archives/apt_0.8.15.10_armhf.deb
Version 1.4.12-4 of gnupg already installed, skipping.
Version 2:0.1.12-20 of libusb-0.1-4 already installed, skipping.
Version 2010.08.28 of debian-archive-keyring already installed, skipping.
Version 4.6.3-1 of libstdc++6 already installed, skipping.
Version 1.4.12-4 of gpgv already installed, skipping.
Version 6.2-8 of readline-common already installed, skipping.
dpkg: regarding .../apt_0.8.15.10_armhf.deb containing apt, pre-dependency problem:
apt pre-depends on libstdc++6 (>= 4.6)
libstdc++6 is unpacked, but has never been configured.

---------------------------------------------

It seems that libstdc++6 has been unpacked, but that another package depends upon it being also configured, but for some reason it isn't.  Deboostrap sees this conflict and barfs.  I really don't understand why debootstrap isn't able to configure libstdc++6 if another package needs it configured.

I actually can start the second stage debootstrap again and it seems to work, but I would really like to figure out why this isn't working as intended.

Anyone have a clue as to what exactly is wrong and what a fix may be?

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Tue Apr 03, 2012 6:26 am

Still working my way through debootstrap issues, but I have a greater understanding of what the dependency problem might be.  Even if I don't figure out a specific fix, the second stage debootstrap will succeed the second time I run it.  This is not a showstopper.

Unfortunately, I do have a potentially more serious issue.  I've encountered two packages that have fairly serious compilation problems - perl and aptitude.  In both cases, the gcc 4.6 compiler dies of an internal compilation error (aka a bug in the compiler itself) and build of the package is terminated after spitting out RTL information.  Since the packages do compile with the regular armhf gcc compiler for ARMv7+VFPv3-D16+Thumb2 code, there seems to be a problem having the compiler instead output ARMv6+VFP code.  I can only imagine that the gcc compiler is less well tested creating ARMv6+VFP code.

Getting a fix for this could be an arduous process.  In the meantime, I'm going to attempt to build the gcc version 4.7 and then rebuild these two packages with that compiler.  I'm crossing my fingers that these two packages don't case the same internal compiler error in gcc 4.7.  Even if this works, it still bothers me that gcc 4.6 outputting ARMv6 code can't be trusted to compile the same packages if it outputs ARMv7 code.

User avatar
Jim Manley
Posts: 1600
Joined: Thu Feb 23, 2012 8:41 pm
Location: SillyCon Valley, California, and Powell, Wyoming, USA, plus The Universe
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Wed Apr 04, 2012 9:40 am

If you need a vote in support of the fantastic work you"re doing, my project to teach students how to develop a network-distributed, massively multiplayer on-line role-playing game (MMORPG) will need every ounce of every resource on the R-Pi (which is the only system it will run on, by design to further promote its use - as if that"s really necessary now ), especially floating-point and GPU graphics:

http://www.raspberrypi.org/for.....mes-mmorpg

Thanks, and keep up the great work!
The best things in life aren't things ... but, a Pi comes pretty darned close! :D
"Education is not the filling of a pail, but the lighting of a fire." -- W.B. Yeats
In theory, theory & practice are the same - in practice, they aren't!!!

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Wed Apr 04, 2012 3:19 pm

Hey, if someone wants to help in a big way of getting a port of Debian armhf (hard float) to the RPi, I could use some help tracking down some internal compiler errors.  Without fixing the compiler to build Debian packages for ARMv6+VFP in a reliable way we'll never get to automating the process which is an absolute requirement.

I've filed a bug for one problem including a reproducible test case GCC Buzzilla here: http://gcc.gnu.org/bugzilla/sh.....i?id=52855

I'm working on creating a test case against the umodified version of GCC that is standard with Debian wheezy armhf.  That probably stands a greater chance of being examined by the gcc TPTB.

Finding someone who knows something about the internals of GCC, the ARM code generation in particular, is probably a very rare thing, but perhaps someone on these forums knows a friend of a friend who could help answer some questions.

I suspect that for some reason, the gcc compiler is producing RTL code (the internal "register transfer language" that gcc uses to represent machine code in a generic manner) that can't be mapped into ARMv6 instructions.  But this is just an educated guess on my part.  The fix could be simple as a config option for gcc I'm overlooking.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Thu Apr 05, 2012 9:34 am

Making slow progress.  I now have a debug version of gcc/cc1 going under gdb and I'm examining a reproducible failure case.  My biggest problem now is getting a good understanding of what exactly the code is doing as it attempts to map RTL statements to ARM assembly instructions and why it might be failing.  Hopefully a fresh start tomorrow will help me make better/faster progress towards a fix.

chorlton
Posts: 50
Joined: Mon Feb 06, 2012 1:57 pm

Re: Debian Hard Float (armhf) for RPi

Thu Apr 05, 2012 10:39 am

I'm afraid I can't help but only offer words of encouragement. I'm finding this a fascinating read.

My final aim is to work through Cross Linux From Scratch for the Pi so all this talk of compile flags and hard vs. soft floats is great stuff. Unfortunately at the moment I'm struggling to get some plain LFS steps working on a vanilla iX86 architecture.

Good luck!

drewbharris
Posts: 20
Joined: Tue Mar 06, 2012 8:16 pm

Re: Debian Hard Float (armhf) for RPi

Thu Apr 05, 2012 1:57 pm

Just wanted to say that you're doing God's work.  Multiarch would make life SO much easier.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Thu Apr 05, 2012 5:24 pm

Quick update.  Just before turning in last night I found an obscure reference in a google search to a very similar bug from about a year ago involving ARMv6+VFP which suggested -O0 as a workaround.  I tried the -O0 flag and the darn file compiled cleanly.  I could swear I tried this numerous times without any option flag (which should match the default -O0) and it continued to fail. Ugh!!! I should do things when I'm so tired.

I believe when I tried this earlier I was attempting to use -O0 within the context of building the entire package and I probably made a mistake in setting things up (which is pretty easy to do with Debian packages) and the package build system was resetting the optimization to something else without me knowing about it.  Having a reproducible test case now allows me to tweak the options in very specific ways without the cruft of the package build system.

Well, it now seems that I have a clear path to building the last few remaining packages even if they won't be optimized.  I'll continue to try to understand the nature of compilations failing with -O1 or -O2 settings.  It's a little less of a priority, but something that should be addressed.

On to building more packages and the initial boot image.

shirro
Posts: 248
Joined: Tue Jan 24, 2012 4:54 am

Re: Debian Hard Float (armhf) for RPi

Thu Apr 05, 2012 10:58 pm

chorlton said:


My final aim is to work through Cross Linux From Scratch for the Pi so all this talk of compile flags and hard vs. soft floats is great stuff.


Good move. I am also running through an LFS except on ARM hardware for extra pain though I might retreat from that. I have done most of this before but not for over 10 years. Learning a bit about arm platforms as I go with the goal to build a fairly limited but interesting hardfloat image for the Pi when I am done.

I am looking forward to using Debian hf when it is done as well. I wish I knew something about gcc internals to help out.

Back to the compiler issues.  Have you tried cross compiling Perl with the same debian compiler and if it fails tried the code sourcery, linaro or the compiler the foundation/broadcom(?) have been using in the raspberry pi git (from memory the softfloat is 32bit bit but the hardfloat is 64bit which may cause some issues depending on your desktop). It is possible one of the big embedded contributors to gcc has the problem fixed already but it hasn't made it upstream. If that is the case probably all you have to do is wait. [edit: just read above and -O0 worked. Well done. Probably isn't a bad move on the Pi anyway]

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Fri Apr 06, 2012 12:02 am

I tried the latest sources to the Linaro gcc 4.6.3.  I'm pretty sure I built it right, but it gives the following error if I attempt to compile for armv6 and vfp at the same time: "sorry, unimplemented: -mfloat-abi=hard and VFP".  I don't get this problem with Debian gcc 4.6.3.

I haven't tried Code Sorcery, but I'll give it a try soon.  The frustrating thing is each time I build a new version of the compiler it takes about 4 to 6 hours build.  I've built the 4.6.2 compiler, 4.7.0 compiler, Linary 4.6.3 compiler, etc... It certainly tries my patience.  Fortunately, I can work with the gcc build system well enough now that I can tweak some of the files and get a quick rebuild when needed.

Now that I can compile the problematic packages with the -O0 option I'm going to put the compiler issues aside for a few days and work on getting an initial image to test out on an RPi beta board.  Earlier in this thread Dom indicated he would run a test for me and it would be useful to just make sure everything I've built run as expected on real hardware.  I'll try to have a download ready for him tomorrow (Friday).

If I get conformation things are working as I expect on real hardware I'm going to then start working on automating the entire build process so I can have my iMX53 QSB turning out packages as fast as possible.  We'll see what issues I run into getting that set up.

User avatar
mpthompson
Posts: 620
Joined: Fri Feb 03, 2012 7:18 pm
Location: San Carlos, CA
Contact: Website

Re: Debian Hard Float (armhf) for RPi

Fri Apr 06, 2012 8:36 pm

I've created the first SD test image from the hardfloat packages I've recompiled.  I'll reach out Boz to see if he can boot it on an RPi beta system as he indicated earlier in this thread (hopefully I didn't misunderstand he had a beta system).  I'm very interested in getting back the results of the benchmark test I've put on the image.  Hopefully all this work over the last few weeks hasn't been a waste.

Nr90
Posts: 216
Joined: Sat Nov 26, 2011 12:39 pm

Re: Debian Hard Float (armhf) for RPi

Fri Apr 06, 2012 9:23 pm

mpthompson said:


I've created the first SD test image from the hardfloat packages I've recompiled.  I'll reach out Boz to see if he can boot it on an RPi beta system as he indicated earlier in this thread (hopefully I didn't misunderstand he had a beta system).  I'm very interested in getting back the results of the benchmark test I've put on the image.  Hopefully all this work over the last few weeks hasn't been a waste.


Congratulations on the excellent work! Fingers crossed

Return to “Raspbian”