пт, 6 февр. 2026 г., 02:14 Terje J. Hanssen <terjejhanssen@gmail.com>:


On 2/5/26 8:52 PM, Phyllis Smith wrote:
As time permits, try this AppImage and please let me know if Vulkan and Onevpl capabilities are there.  I cobbled together the binary on the CinGG website Download page in the testing subdirectory: vulkan_onevpl-x86_64.AppImage.  It was built on kernel version 6.18.5-100, libc version 2.41.

Tested the vulkan_onevpl-x86_64.AppImage on two Intel machines , Coffee Lake / UHD 620 iGPU and Alder Lake / DG2 dGPU, both with current Slowroll.
Unhappily the actual rendering methods failed:
VPL: cant find codec h264_qsv or hevc_qsv.

Sad. :(

I wonder if it was just path mismatch like with vaapi, or something else?

Can you unpack appimage, delete onevpl lib,  and let it link to system's onevpl loader, like with vaapi before?

Vulkan: compression was not available. (tried to adapt and edit tje h264_vaapi preset to cin_hw_dev=vulkan which didn't work).
you need to change actual codec name, too.

ffmpeg on termux says something like

ffmpeg -encoders | grep vulkan

V....D av1_vulkan           AV1 (Vulkan) (codec av1)
 V....D ffv1_vulkan          FFmpeg video codec #1 (Vulkan) (codec ffv1)
 V....D h264_vulkan          H.264/AVC (Vulkan) (codec h264)
 V....D hevc_vulkan          H.265/HEVC (Vulkan) (codec hevc)




Tested also my own appimage (20241120) where still VPL works on Leap15.6



On Wed, Feb 4, 2026 at 9:29 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:


чт, 5 февр. 2026 г., 03:51 Phyllis Smith <phylsmith2017@gmail.com>:
FYI, I did get --with-vulkan to compile but not --with-onevpl. Still working on and got errors when making AppImage so am not done yet.
root@fedora:~# dnf install libvpl-2.16.0-1.fc42
Updating and loading repositories:
Repositories loaded.
Package "libvpl-1:2.16.0-1.fc42.x86_64" is already installed.

Try libvpl-devel ?





Nothing to do.
root@fedora:~# dnf install vulkan
Updating and loading repositories:
Repositories loaded.
Package "vulkan-loader-1.4.313.0-1.fc42.x86_64" is already installed.

Just in case you can add


?

But may be they were already pulled in ... if you say --enable-vulkan works.


Nothing to do.

Problem --with-onevpl:
configure: error: requires onevpl support.
make: *** No targets specified and no makefile found.  Stop.

On Mon, Feb 2, 2026 at 2:13 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:


пн, 2 февр. 2026 г., 22:59 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, added "--with vulkan --with-onvpl" to the configure line. This is what I get in the log:
checking for vaGetDisplayDRM in -lva-drm... yes
checking for MFXCreateSession in -lvpl... no
checking for vkQueueSubmit in -lvulkan... no
configure: error: requires vulkan support.
make: *** No targets specified and no makefile found.  Stop.

AND configure: error: requires onevpl support.


you need to install onevpl-devel package, and vulkan-sdk ..... (on your build machine)



On Sun, Feb 1, 2026 at 4:39 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:


пн, 2 февр. 2026 г., 01:59 Phyllis Smith <phylsmith2017@gmail.com>:
Andrew, did I miss something?
- I checked into GIT, the 0026-Attempt-at-fixing-prefix-system-install-with-libzmpe.patch, with the changes to configure.ac and Makefile.am.
- Was there another specific patch for this? Quote="I think we traced failure back to wrong shell used for our configure (mksh vs bash)." If so, I missed it.

No, it should be enough, just T2's recipe fetch *release* tarball, not git commit, so install still fails there, even if we fixed that in git.


- I will gladly make a new test AppImage with enabling Vulkan, but I thought that it had to be built on a computer that had specific hardware capabilities.  Am I wrong? or can I just add "--enable-vulkan" to the "configure" line? Same question about ONEVPL?

Well, good question!

While building cinelerra-10 on T2SDE I noticed ffmpeg 5.1 for example outright fail to build with new vulkan headers (no configure check tight enough, I guess).

Depending on what kind of vulkan sdk your build OS for new appimage have it may or may not work with different distros ... I think only way to test is try!

You also can try to enable openvpl, and see if it get pulled into appimage so at least it starts on system without OneVPL installed? Mismatches still possible, with all those moving parts ... 



On Sat, Jan 31, 2026 at 9:45 PM Andrew Randrianasulu <randrianasulu@gmail.com> wrote:


сб, 31 янв. 2026 г., 04:41 Andrew Randrianasulu <randrianasulu@gmail.com>:
On Fri, Jan 30, 2026 at 10:24 PM Andrea paz via Cin
<cin@lists.cinelerra-gg.org> wrote:
>
> My Arch has gcc 15.2.1. The build from git was successful without errors.
> Do I need to do any specific tests?

Thanks.

I will try to replicate this error, just for now T2 does not build on
my Slackware, but hopefully should build on itself in qemu VM ...


I think we traced failure back to wrong shell used for our configure (mksh vs bash).

Dear Phyllis, I hope everything is more than less ok on your end?

Can we make new release so T2 picks our fix for non-thirdparty  install? While at it you can try to enable Vulkan for new appimage so Terje will have chance to test it?





>
> _______________________________________________
> Cin mailing list -- cin@lists.cinelerra-gg.org
> To unsubscribe send an email to cin-leave@lists.cinelerra-gg.org