GIT checkin includes Cakewalk update + limited specialized hardware acceleration options.
Today's GIT checkin includes: - Cakewalk update - Nvenc option for encoding/rendering hardware acceleration for computers with Nvidia graphics for h264/h265. - Cuda build option for systems that already have the Cuda software loaded - with 2 demo plugins available. ---------------------------------------------------------------------------------------------------------------------------------------------- More info on the hardware acceleration options follow. Some testing done on Fedora 29 and Leap 15, but not all possibilities have been tried. I would expect to see some fallout. ---------------------------------------------------------------------------------------------------------------------------------------------- NVENC/NVDEC for Nvidia graphics board encoding and decoding will automatically be built by way of the configure scripts. But first, before anyone gets too excited, let me just warn you that the ONLY encode codec Nvidia chose to support is H264 and H265 with H265 available on just the latest graphics boards (supposedly at least the ones with the Maxwell chip). Also included in the ffmpeg build is "nvdec" but we have not figured out yet how this affects the hardware acceleration already in place with the use of CIN_HW_DEV=vdpau. Some information that was posted online 11/04/2018 -- "VDPAU falling away into irrelevance but no clear statements that consumers should switch to nvdec or that nvidia will advance VDPAU feature parity. ...All of these things make it hard to see what the correct way to approach nvidia support is, and even if you pick one, you can't actually integrate in a simple way." The easiest way to determine if nvenc is working for you is just to do a short render using h264_nvenc.mp4 or h265_nvenc.mp4 and see if it works or just gives you an error message. Error message for computers without nvdia cards (or not implemented) is: Cannot load libcuda.so.1 on the window from where your started cin. BTW this libcuda is part of nvidia libraries in the Operating System and does not mean you have CUDA in. A 2 minute demo from Big Buck Bunny 3804x2160, took 6 minutes 3 seconds with nvenc to render as opposed to 22 minutes 30 seconds without nvenc -- so not quite 4 times faster and I had no plugins added on a 4 core Intel laptop with an Nvidia 950M graphics board. ------------------------------------------------------------------------------------------------------------------------------------------- CUDA® is a parallel computing platform / programming model that provides big increases in computing performance through use of the GPU, graphics processing unit. It was first introduced in about 2006 for applications in computationally intense fields such as astronomy, biology, chemistry, and physics. The use of Cuda is not going to improve the playing and rendering of video in Cinelerra EXCEPT in the case where you use a specific CUDA-enabled plugin that is computationally intense - sadly, most of what Cin does, Cuda will not help. Cuda is mostly a "block oriented algorithm" which works well for such things as "a flock of birds all flying next to each other". If you have already installed the Cuda software on your computer -- and let me warn you, this is about 3GB of additional space usage -- the default build in the configure script is for cuda is "auto". Otherwise it will NOT be included. Obviously, you must have the Nvdia drivers for your graphics board installed too. Cuda is available to install on your computer only for a specific set of Operating System distros. Go to the following website: https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_... and gg says follow the very good set of directions to install around 3 GB. It installs repos by package. Also, install the "fusion repo" -- we do not know if this needs to be installed or not, but all of our test included it. I will attempt to create a demo a little later and there will most likely be a Ubuntu16 and Mint18 build available soon for testing purposes. gg/Phyllis
Update: there are 3 test builds for NVENC only (no Cuda) Ubuntu 16 and 18 and Mint 18. There will be no demo because I do not have the record facility loaded on a nvenc/Cuda/Nvidia computer. But for those who have CUDA in, please check out the 2 CUDA plugins, N_body and Mandelbrot. - Cakewalk update
- Nvenc option for encoding/rendering hardware acceleration for computers with Nvidia graphics for h264/h265. - Cuda build option for systems that already have the Cuda software loaded - with 2 demo plugins available.
https://www.cinelerra-gg.org/download/testing/cinelerra-5.1-ub16-x86_64-stat... https://www.cinelerra-gg.org/download/testing/cinelerra-5.1-mint18-x86_64-st... https://www.cinelerra-gg.org/download/testing/cinelerra-5.1-ub18-x86_64-stat...
Hi Phyllis, My computer is equipped with a Nvidia GTX-750-Ti video card. It was one of the first two nvidia cards to use the Maxwell architecture. It includes 640 Cuda cores. I use Linux Mint 18.3 Mate 64bit (which is built from a 16.04 kernel from Ubuntu). The nvidia 384.130 driver is installed. I never installed anything particular for the Cuda cores and I don't know if I would be able to successfully install the CUDA Toolkit from https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_... But if I can be useful in testing your development, I'm ready to try. Pierre On 19-06-16 16 h 07, Phyllis Smith wrote:
NVENC/NVDEC for Nvidia graphics board encoding and decoding will automatically be built by way of the configure scripts. But first, before anyone gets too excited, let me just warn you that the ONLY encode codec Nvidia chose to support is H264 and H265 with H265 available on just the latest graphics boards (supposedly at least the ones with the Maxwell chip). Also included in the ffmpeg build is "nvdec" but we have not figured out yet how this affects the hardware acceleration already in place with the use of CIN_HW_DEV=vdpau.
If you have already installed the Cuda software on your computer -- and let me warn you, this is about 3GB of additional space usage -- the default build in the configure script is for cuda is "auto". Otherwise it will NOT be included. Obviously, you must have the Nvdia drivers for your graphics board installed too. Cuda is available to install on your computer only for a specific set of Operating System distros. Go to the following website:
https://developer.nvidia.com/cuda-downloads?target_os=Linux&target_arch=x86_...
and gg says follow the very good set of directions to install around 3 GB. It installs repos by package. Also, install the "fusion repo" -- we do not know if this needs to be installed or not, but all of our test included it.
Pierre, I would not bother installing CUDA but if you could just do a render using the h264_nvenc.mp4 to see if you get a speed up over using h264.mp4 that would be appreciated. You will find h264_nvenc.mp4 in the render menu when you click on the Video wrench and use the down arrow after setting File Format to ffmpeg / mp4. We did not have an opportunity to test on Mint at all. Thanks, gg/phyllis
I tried three times to render in h264_nvenc.mp4, each time CinGG refused to proceed. Before that I had done the rendering without problem with h264.mp4 pierre@i7-3770k ~ $ /home/pierre/Cinelerra-GG_5.1/cin Cinelerra Infinity - built: Jun 16 2019 14:06:46 git://git.cinelerra-gg.org/goodguy/cinelerra.git (c) 2006-2018 Heroine Virtual Ltd. by Adam Williams (c) 2007-2018 cin5 derivative by W.P. Morrow aka goodguy Cinelerra is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. There is absolutely no warranty for Cinelerra. RenderFarmClient::main_loop: client started Render::render_single: Session finished. ** rendered 1370 frames in 111.717 secs, 12.263 fps Render::render_single: Session finished. ** rendered 950 frames in 111.186 secs, 8.544 fps [h264_nvenc @ 0x7f28601e7a00] Driver does not support the required nvenc API version. Required: 9.0 Found: 8.0 [h264_nvenc @ 0x7f28601e7a00] The minimum required Nvidia driver for nvenc is 390.25 or newer FFMPEG::open_encoder err : Fonction non implantée int FFMPEG::open_encoder(const char*, const char*): échec de l'ouverture h264_nvenc:/media/principale/Pierre/RENDUS_de_CINELERRA-GG_5.1/Test_rendu_visuel.mp4 Render::render_single: Session finished. Session time: 0:28:40 Cpu time: user: 0:26:50.903 sys: 0:00:15.527 pierre@i7-3770k ~ $ pierre@i7-3770k ~ $ /home/pierre/Cinelerra-GG_5.1/cin Cinelerra Infinity - built: Jun 16 2019 14:06:46 git://git.cinelerra-gg.org/goodguy/cinelerra.git (c) 2006-2018 Heroine Virtual Ltd. by Adam Williams (c) 2007-2018 cin5 derivative by W.P. Morrow aka goodguy Cinelerra is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. There is absolutely no warranty for Cinelerra. RenderFarmClient::main_loop: client started [h264_nvenc @ 0x7f14a01ee500] Driver does not support the required nvenc API version. Required: 9.0 Found: 8.0 [h264_nvenc @ 0x7f14a01ee500] The minimum required Nvidia driver for nvenc is 390.25 or newer FFMPEG::open_encoder err : Fonction non implantée int FFMPEG::open_encoder(const char*, const char*): échec de l'ouverture h264_nvenc:/media/principale/Pierre/RENDUS_de_CINELERRA-GG_5.1/Test_rendu_visuel.mp4 Render::render_single: Session finished. Session time: 0:01:27 Cpu time: user: 0:00:05.223 sys: 0:00:03.225 pierre@i7-3770k ~ $ pierre@i7-3770k ~ $ /home/pierre/Cinelerra-GG_5.1/cin Cinelerra Infinity - built: Jun 16 2019 14:06:46 git://git.cinelerra-gg.org/goodguy/cinelerra.git (c) 2006-2018 Heroine Virtual Ltd. by Adam Williams (c) 2007-2018 cin5 derivative by W.P. Morrow aka goodguy Cinelerra is free software, covered by the GNU General Public License, and you are welcome to change it and/or distribute copies of it under certain conditions. There is absolutely no warranty for Cinelerra. RenderFarmClient::main_loop: client started [h264_nvenc @ 0x7fde749c4b40] Driver does not support the required nvenc API version. Required: 9.0 Found: 8.0 [h264_nvenc @ 0x7fde749c4b40] The minimum required Nvidia driver for nvenc is 390.25 or newer FFMPEG::open_encoder err : Fonction non implantée int FFMPEG::open_encoder(const char*, const char*): échec de l'ouverture h264_nvenc:/media/principale/Pierre/RENDUS_de_CINELERRA-GG_5.1/Test_rendu_visuel.mp4 Render::render_single: Session finished. Session time: 0:03:13 Cpu time: user: 0:00:10.186 sys: 0:00:06.002 unjoined tids / owner 3 00007fde7d7fa700 / 00007fde927fc700 12RenderEngine 00007fde7f7fe700 / 00007fde9bff7700 12RenderEngine 00007fde90ff9700 / 00007fdeb3fff700 12RenderEngine pierre@i7-3770k ~ $ On 19-06-16 17 h 51, Phyllis Smith wrote:
Pierre, I would not bother installing CUDA but if you could just do a render using the h264_nvenc.mp4 to see if you get a speed up over using h264.mp4 that would be appreciated. You will find h264_nvenc.mp4 in the render menu when you click on the Video wrench and use the down arrow after setting File Format to ffmpeg / mp4. We did not have an opportunity to test on Mint at all. Thanks, gg/phyllis
- My nvidia 460M doesn't support nvenc and in fact trying to render it makes a mistake (same errors as Pierre). - F_Mandelbrot does not work: PluginFVClient::activate() F_mandelbrot err: Function not implemented PluginFVClient::process_buffer() F_mandelbrot err: Operation not permitted I remember several months ago, when I tried it out, it worked properly. - I don't know N_body. PS: Damn! Just as I was moving to an AMD GPU for my future PC, increase support for expensive Nvidia. :-)
participants (3)
-
Andrea paz -
Phyllis Smith -
Pierre autourduglobe