Thanks in advance!
Thanks in advance!
I've been asking in this group a few weeks ago on how to use the VCE of AMD graphics chips to do hardware based h.264 encoding. As I've finally gotten it to work (although I still have to use gst-launch...), I thought it might be a good idea to share how I did it.
There are several components required to utilize it, and all of them have to work together correctly:
o) The VCE circuitry on the GPU requires the appropriate firmware file. On most distributions it's part of the firmware-linux-nonfree package, and if your distribution is using initramfs, it'll likely work already. If not, you'll need to make sure that the appropriate file is available at boot time, by either putting it in the initramfs, or packing it directly into your kernel image. If unsure, grep your dmesg for "VCE". If there's an error, the most likely reason is that the firmware file couldn't be loaded. No problem though, as dmesg should also tell you the filename of said firmware.
o) In userspace you'll need a mesa build with the OpenMAX state tracker enabled. On Gentoo this can be done by enabling the "openmax" use flag for mesa. I don't know about other distributions though.
o) An integration layer for the OpenMAX state tracker has to be installed as well. The one I've had success with (and the only one I know of for mesa's OpenMAX state tracker) is omxil-bellagio. But installing it isn't enough, one also has to run "omxregister-bellagio" for every user who wants to use hardware encoding. Best run it with the "-v" option, to see immediately if the mesa OpenMAX state tracker is being detected.
o) The next component one needs is the gstreamer plugin for OpenMAX. The version for gstreamer 1.0 is called gst-omx. If your distribution doesn't have packages, the source code can be found here: https://gstreamer.freedesktop.org/releases/gst-omx/1.0.0.html This gstreamer plugin again has its own pitfalls that caused me quite some headaches. It expects a configuration file to be present in /etc/xdg/gstomx.conf, but the one it ships with, even if one specifically builds it to support omxil-bellagio, is registering unavailable functions, and doesn't list those of mesa at all. The solution I found is to get the /etc/xdg/gstomx.conf file from gearsongallium (here: https://github.com/pontostroy/gearsongallium/blob/master/cd-image/source/root/etc/xdg/gstomx.conf ). With this file in place, gst-inspect-1.0 should be able to detect your gst-omx plugin successfully without blacklisting. If it gets blacklisted, running gst-inspect-1.0 on the gst-omx plugin shared library with GST_DEBUG being set to 4 or above will help identify the issue.
Now, with a bit of luck, you'll be in principle able to use gst-launch to set up a gstreamer pipeline using omxh264enc to encode your videos using your graphics card's hardware. If you figure out how to properly use gstreamer that is (I'm still struggling to get it to decode my mpeg4 files fast enough to actually benefit from hardware based encoding, and I haven't even tried to mux the result with the audio track...).
Does anyone have experience with this, or could recommend specific software?
It's nice having functionality available through an API, but for me as an end user it'd be preferable to have a program meant for, well, end users that can utilize said API, as I hardly have the time to write one myself...
Meanwhile I've found out that gstreamer is capable of using OpenMAX ( https://gstreamer.freedesktop.org/releases/gst-omx/1.0.0.html ), but I've never found gst-launch to be user friendly (I know that the gstreamer pipelines are incredibly powerful, it's just that they are a pain to write...), and would strongly prefer a program that's as easy to use as ffmpeg or mencoder...
If there'd be a GUI tool that'd be able to utilize gst-omx without requiring the user to manually write the gstreamer pipeline (because then I could just as well use gst-launch...), that would be even better.
Allow me to use this case to make a general observation: I find it increasingly "interesting" to see which security stories get lots of coverage - and which don't. Is this a bad bug? Sure. Is it worse than any of the four (!) root exploits Google has fixed in the last Nexus update? No. It's not even clear this one is exploitable on most Android devices at all. So why does it get so much coverage? If we are all honest it simply is because the security vendor cared enough to write a detailed blog post about it. (And yeah this obviously also is some sort of self-criticism, we are all part of this circus)
I've stumbled across a nice logic circuit simulator, Atanua, and just wanted to share how I managed to build it on Linux (as the Linux builds on their download page are from older versions):
First, one needs to get the Windows version from the official website to obtain the required data files. The needed folders are called "data" and "tests" in the zip file. Without these files, Atanua will crash on startup.
Then, on the download page, there's a link to the source code git repository. The included makefile is outdated, so my solution was to, instead of editing that thing by hand, just adding all files to Qt Creator and tweaking the automatically generated qmake project file afterwards. Here's the result.
I think it's rather self explanatory - just save this as a .pro file in the top directory of the source code, (above the "src" directory). I prefer out of tree builds, so I then created a new directory, and just ran qmake <path-to-project-file> there, followed by make. A huge pile of warnings later, a Atanua binary was created. Now all I had to do was to copy over this binary, and the "tests" and "data" folders from the Windows version to a new folder, and launch Atanua from that directory (paths to data files are hard coded...).
There's one thing that I found annoying though, and that's that the person who assembled it obviously didn't read the manual, which clearly states that one should use the bottom RAM slot if one only installs a single module below the keyboard... That made my RAM upgrade a matter of trial and error. So, if anyone else wants to upgrade a P50, the RAM slots are wired in the following way:
Slot 1 is the bottom slot under the keyboard
Slot 2 is the top slot (the one farther away from the battery) on the back side
Slot 3 is the top slot under the keyboard
Slot 4 is the bottom slot on the back side.
If one buys a device with only one memory module installed, one therefore would expect it to be in Slot 1 (as stated in the manual), meaning for Dual Channel Mode one could either plug the additional module to Slot 3, or Slot 4. As Slot 4 is easier to access than Slot 3, that's likely the best option for most users. Anyhow, I'd run CPU-Z (or some other diagnostics tool that shows which memory slots are in use) before opening the case, just to check where to mount the additional module in order to get dual channel working.
(Oh, and don't get fooled by Lenovo Support claiming that you won't see the difference between single and dual channel - it's performance critical with integrated graphics chips, see for instance http://goo.gl/qlHqIL and also certain memory heavy applications can benefit greatly from having dual channel available.)
Sadly it's in German...
... Hudson: Zentralbanken sind stark von Geschäftsbankenlobbys beeinflusst und vertreten deren Interessen – nicht die der Bevölkerung. Sie erzeugen kein Geld für Investitionen und Lohnzahlungen. Das Geld wird Geschäftsbanken gegeben und diese verleihen nur gegen Sicherheiten wie Anleihen, Aktien oder andere Finanzprodukte. Die Preise von Finanzprodukten steigen durch die Zentralbankpolitik, nicht aber Löhne und Investitionen. Mittlerweile werden jeden Tag am Finanzplatz New York Transaktionen im Wert des jährlichen US-Inlandsprodukts durchgeführt. ...
- Johannes Kepler Universität LinzPhD student (employed), 2010 - 2015PhD thesis on optical and magnetic properties of transition metal doped gallium nitride
- Johannes Kepler Universität LinzPostdoctoral researcher, 2015 - 2016Research on Hot Wall Epitaxy of oligothiophenes, -phenyls, and purine alkaloids.
- stillalive studios GmbHGame developer, 2016 - present
- HTL Vöcklabruck1999 - 2004Engineering school with focus on informatics in industry
- Johannes Kepler Universität Linz2004 - 2015Diploma and PhD in Physics
- Magic Spin Group Homepage (current)
law blog» Archiv » Adressbuch-Abzocke: Anwälte drehen den Spieß um
Normalerweise ärgert man sich ja nur über den dreisten Abzockversuch, wenn mal wieder ein dubioses Adressbuch-Angebot eingeht. Die Anwälte d
Google Maps für Android jetzt mit Fahrradnavigation
Die Nächste rechts bitte: In der neuesten Version von Google Maps für Android gibt es jetzt eine Fahrradnavigation mit Abbiegehinweisen. Vor
Australischer Geheimdienst soll Hackerbefugnis erhalten
Das australische Justizministerium will es dem Nachrichtendienst ASIO erlauben, IT-Systeme Unbeteiligter zu knacken, um von da aus Zugang zu
Studie: US-Cloud-Überwachung steht EU-Datenschutzreform im Weg
In einer Untersuchung zu Strafverfolgung und Datenschutz in der Cloud weisen Forscher darauf hin, dass der Datentransfer in die Cloud und ex
Kritik an geplanter Vorratsdatenspeicherung von Fluggast- Informationen
Erfasst werden sollen neben Namen, Adressen auch Essensvorlieben
Google spendet 15.000 Raspberry Pis für Schulkinder
Gemeinsam mit Raspberry Pi Foundation sollen Kinder in Großbritannien ausgewählt werden
"Spark": 200-Euro-Tablet mit "Plasma Active" - Linux/Unix - derStandard....
Basierend auf Linux- und KDE-Plattform - Echte freie Software statt "abgeschlossene Gärten"
"Das Schwarze Auge": Comeback der Nordlandtrilogie
Launch von "Schicksalsklinge" für zweites Quartal vorgesehen
Umweltgifte: Pestizid-Verbot soll Bienen retten - SPIEGEL ONLINE
Clothianidin, Imidacloprid und Thiamethoxam sind drei hochwirksame Pflanzenschutzmittel - und womöglich Bienenkiller. Die EU-Kommission will
Frankreich: Google zahlt für Durchleitung von YouTube-Videos - SPIEGEL O...
Die Kunden des französischen Providers Orange rufen häufig YouTube-Videos auf, die Datenpakete machen einen Großteil des Traffics aus. Nun p