I’ afraid using VM will drastically reduce performance though.
Is Qt version 5.5 specifically required, or would any Qt version 5.x work? I ask because the current version of lib Qt core for Ubuntu 14.04 LTS, after any package upgrades, appears to be version 5.2.1 (this is already installed on the JTK1 I’m testing with).
You mentioned not having physical access to the JTK1, so it may be important to note that if you use remote display from your desktop PC of a forwarded X11 program, then the Qt running the display is on the desktop host and not from the Jetson even if the program runs on Jetson. A VNC type remote login program would run Qt completely from the JTK1, but actual display would be drastically slowed if the network or desktop bottleneck data.
What is the URL you have for your specifically downloaded Qt 5 you are compiling (I might download and test config/compile)?
Thank you, linuxdev.
If you wouldn’t mind, please have a look and give some advice on the performing readme.txt installation steps using the readme file from the following link Dropbox - File Deleted
Sincerely yours,
Andrey Volodin
The Qt5 “run” file from the link requires i386 architecture…it isn’t source code, or at least not entirely (the installer portion itself cannot run on ARM CPUs). Although Qt5 already exists on L4T, I was unable to work around using native Qt5 instead of the provided Qt5…every time I did something to get a step further, something else more difficult to work around showed up. So Qt5 is already there, but other incompatibilities prevent the already-installed Qt from working.
I was advised to try guvcview for YUV streaming for a ESP870-based camera, but guvcview says FATAL: can not locate cpu MHz at /proc/cpuinfo
What software could I possibly use for YUV streaming at armv7l?
It was said:
“You can run the camera with guvcview or other video software to stream YUV video, but you cannot get the depth video without the SDK.”
cat /proc/cpuinfo
processor : 0
model name : ARMv7 Processor rev 3 (v7l)
Features : swp half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 idiva idivt
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x3
CPU part : 0xc0f
CPU revision : 3
processor : 1
model name : ARMv7 Processor rev 3 (v7l)
Features : swp half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 idiva idivt
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x3
CPU part : 0xc0f
CPU revision : 3
processor : 2
model name : ARMv7 Processor rev 3 (v7l)
Features : swp half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 idiva idivt
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x3
CPU part : 0xc0f
CPU revision : 3
processor : 3
model name : ARMv7 Processor rev 3 (v7l)
Features : swp half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 idiva idivt
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x3
CPU part : 0xc0f
CPU revision : 3
Hardware : jetson-tk1
Revision : 0000
Serial : 0177000044440000
Processor : ARMv7 Processor rev 3 (v7l)
FYI, i386 cannot run on ARMv7 without emulation, which would indeed slow things quite a bit. Compiling on ARMv7 with full source code would do the trick.
The “cannot locate cpu MHz” is an old issue which I believe has been solved, but I have not personally had to deal with it. Here’s a thread on the topic, perhaps someone who has had to deal with this can comment on workarounds:
[url]https://devtalk.nvidia.com/default/topic/765800/workaround-for-fatal-cannot-locate-cpu-mhz-in-proc-cpuinf/[/url]
Keep in mind when reading that thread that your particular program may share some libraries, and others may be completely unrelated to your particular situation. I would guess that people here are familiar with guvcview and may have advice.
probably luvcview will do the trick
http://elinux.org/Jetson/Cameras
Normally, guvcview can be used to adjust default camera settings, but it seems that the default guvcview is not working correctly on ARM due to a kernel change (c) Jetson/Cameras - eLinux.org
which nvidia cuda toolkit to install at jetson jtk1?
cat /etc/nv_tegra_release
R21 (release), REVISION: 4.0, GCID: 5650832, BOARD: ardbeg, EABI: hard, DATE: Thu Jun 25 22:38:59 UTC 2015
The following which is available seems to be for jtx:
Probably that one :
Am I right?
Andrey
cat /etc/nv_tegra_release
R21 (release), REVISION: 4.0, GCID: 5650832, BOARD: ardbeg, EABI: hard, DATE: Thu Jun 25 22:38:59 UTC 2015
CUDA 6.5 is the most recent 32-bit CUDA, and works with JTK1. Past that (7.0+) is 64-bit only (and a JTK1 is 32-bit). The most recent L4T release for 32-bit is R21.4; I believe both JetPack2 and 2.1 probably have R21.4 with CUDA 6.5 for a JTK1, mostly the 64-bit installs were changing when JetPack2.1 came out (JetPack2.1 added L4T R23.2 on the 64-bit side, but kept L4T R21.4 for the 32-bit side).
So…stick to CUDA 6.5 under R21.4.
Is there a downloadable link available somewhere for cuda 6.5 under r21.4 in the context of l4t?
Do I have to extract somehow the cuda from the jetpack?
Thanks
Oh,
Perhaps that one:
But the above is for r21.2
It seems that there is no r21.4 deb file exist .
I will try to extract the cuda from JetPack2.1 somehow .
at the page https://developer.nvidia.com/linux-tegra-r214
I’ve found the http://developer.download.nvidia.com/embedded/L4T/r21_Release_v3.0/cuda-repo-l4t-r21.3-6-5-prod_6.5-42_armhf.deb
But it is still different version
With help of EC2 Amazon got managed to get cuda-toolkit from jetpack 2.1. It is still r21.3.
What surprised me was option to install tegra using IP address , username and password, to the guest jetson jtk device.
Unfortunately It refused my attempts to install it to a guest via internet.
What port open does the jetpack require for installation over the Internet?
If it make any sense,though.
Perhaps I will install the above deb file which is for r21.3 at r2.4 tegra.
What OpenCV to install presuming I have ubuntu 14.04, but trying to follow the steps at
http://elinux.org/Jetson/Installing_OpenCV
Both?