- Cin - Cinelerra-GG Mailing Lists

Autovectorization works! (resend)
by Andrew Randrianasulu 12 Feb '19

12 Feb '19

resend without attached image ------------------ At last for my machine ..not sure if it will break some plugins? from http://gcc.gnu.org/projects/tree-ssa/vectorization.html " To enable vectorization of floating point reductions use -ffast-math or -fassociative-math." I added -ffast-math -ftree-vectorizer-verbose=2 to CFLAGS (global, but I recompiled and installed only color3way plugin) and now we have tons of assmebly! objdump -M intel -d /dev/shm/tmp/cinelerra-goodguy-20190211/cinelerra-5.1/plugins/color3way/i686/color3way.o | ./opcodes.sh -s SSE2 -s SSE3 -s SSE -s AVX -s FMA | less --------------------- 16e: c5 f8 77 vzeroupper -- 427: c5 f8 77 vzeroupper -- 6cc: c5 f8 77 vzeroupper -- 751: c5 f8 77 vzeroupper -- 80e: c5 fa 10 81 00 00 00 vmovss xmm0,DWORD PTR [ecx+0x0] -- 816: c5 fb 10 89 00 00 00 vmovsd xmm1,QWORD PTR [ecx+0x0] -- 824: c5 fa 10 12 vmovss xmm2,DWORD PTR [edx] 828: c5 ea 5c 10 vsubss xmm2,xmm2,DWORD PTR [eax] 82c: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 830: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 834: c5 f9 2f d1 vcomisd xmm2,xmm1 -- 83e: c5 fa 10 52 0c vmovss xmm2,DWORD PTR [edx+0xc] 843: c5 ea 5c 50 0c vsubss xmm2,xmm2,DWORD PTR [eax+0xc] 848: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 84c: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 850: c5 f9 2f d1 vcomisd xmm2,xmm1 -- 85a: c5 fa 10 52 18 vmovss xmm2,DWORD PTR [edx+0x18] 85f: c5 ea 5c 50 18 vsubss xmm2,xmm2,DWORD PTR [eax+0x18] 864: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 868: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 86c: c5 f9 2f d1 vcomisd xmm2,xmm1 -- 876: c5 fa 10 52 24 vmovss xmm2,DWORD PTR [edx+0x24] 87b: c5 ea 5c 50 24 vsubss xmm2,xmm2,DWORD PTR [eax+0x24] 880: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 884: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 888: c5 f9 2f d1 vcomisd xmm2,xmm1 -- 892: c5 fa 10 52 04 vmovss xmm2,DWORD PTR [edx+0x4] 897: c5 ea 5c 50 04 vsubss xmm2,xmm2,DWORD PTR [eax+0x4] 89c: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 8a0: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 8a4: c5 f9 2f d1 vcomisd xmm2,xmm1 -- 8ae: c5 fa 10 52 10 vmovss xmm2,DWORD PTR [edx+0x10] 8b3: c5 ea 5c 50 10 vsubss xmm2,xmm2,DWORD PTR [eax+0x10] 8b8: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 8bc: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 8c0: c5 f9 2f d1 vcomisd xmm2,xmm1 -- and so on! Now I have like 15 fps for same 1280x720x30 h264 video + color3way plugin (one track)

1 0

Autovectorization works!
by Andrew Randrianasulu 11 Feb '19

11 Feb '19

At last for my machine ..not sure if it will break some plugins? from http://gcc.gnu.org/projects/tree-ssa/vectorization.html " To enable vectorization of floating point reductions use -ffast-math or -fassociative-math." I added -ffast-math -ftree-vectorizer-verbose=2 to CFLAGS (global, but I recompiled and installed only color3way plugin) and now we have tons of assmebly! objdump -M intel -d /dev/shm/tmp/cinelerra-goodguy-20190211/cinelerra-5.1/plugins/color3way/i686/color3way.o | ./opcodes.sh -s SSE2 -s SSE3 -s SSE -s AVX -s FMA | less --------------------- 16e: c5 f8 77 vzeroupper -- 427: c5 f8 77 vzeroupper -- 6cc: c5 f8 77 vzeroupper -- 751: c5 f8 77 vzeroupper -- 80e: c5 fa 10 81 00 00 00 vmovss xmm0,DWORD PTR [ecx+0x0] -- 816: c5 fb 10 89 00 00 00 vmovsd xmm1,QWORD PTR [ecx+0x0] -- 824: c5 fa 10 12 vmovss xmm2,DWORD PTR [edx] 828: c5 ea 5c 10 vsubss xmm2,xmm2,DWORD PTR [eax] 82c: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 830: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 834: c5 f9 2f d1 vcomisd xmm2,xmm1 -- 83e: c5 fa 10 52 0c vmovss xmm2,DWORD PTR [edx+0xc] 843: c5 ea 5c 50 0c vsubss xmm2,xmm2,DWORD PTR [eax+0xc] 848: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 84c: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 850: c5 f9 2f d1 vcomisd xmm2,xmm1 -- 85a: c5 fa 10 52 18 vmovss xmm2,DWORD PTR [edx+0x18] 85f: c5 ea 5c 50 18 vsubss xmm2,xmm2,DWORD PTR [eax+0x18] 864: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 868: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 86c: c5 f9 2f d1 vcomisd xmm2,xmm1 -- 876: c5 fa 10 52 24 vmovss xmm2,DWORD PTR [edx+0x24] 87b: c5 ea 5c 50 24 vsubss xmm2,xmm2,DWORD PTR [eax+0x24] 880: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 884: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 888: c5 f9 2f d1 vcomisd xmm2,xmm1 -- 892: c5 fa 10 52 04 vmovss xmm2,DWORD PTR [edx+0x4] 897: c5 ea 5c 50 04 vsubss xmm2,xmm2,DWORD PTR [eax+0x4] 89c: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 8a0: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 8a4: c5 f9 2f d1 vcomisd xmm2,xmm1 -- 8ae: c5 fa 10 52 10 vmovss xmm2,DWORD PTR [edx+0x10] 8b3: c5 ea 5c 50 10 vsubss xmm2,xmm2,DWORD PTR [eax+0x10] 8b8: c5 e8 54 d0 vandps xmm2,xmm2,xmm0 8bc: c5 ea 5a d2 vcvtss2sd xmm2,xmm2,xmm2 8c0: c5 f9 2f d1 vcomisd xmm2,xmm1 -- and so on! Now I have like 15 fps for same 1280x720x30 h264 video + color3way plugin (one track)

1 0

gcc's autovectorization
by Andrew Randrianasulu 10 Feb '19

10 Feb '19

Not sure if this article will be useful for anyone, but I was trying to understand why some color effects are ..well, slow. I saw all my four cores ( AMD FX(tm)-4300 Quad-Core Processor) used, but resulting framerate for single track, 1280x720x30 fps h264 video + "color 3 way" effect was just ..7 fps?! I found this https://locklessinc.com/articles/vectorize/ "Auto-vectorization with gcc 4.7" with interesting conclusion --------------- Summary gcc is very good, and can auto-vectorize many inner loops. However, if the expressions get too complex, vectorization will fail. gcc also may not be able to get the most optimal form of the loop kernel. In general, the simpler the code, the more likely gcc is to give good results. However, you cannot expect gcc to give what you expect without a few tweaks. You may need to add the "--fast-math" to turn on associativity. You will definitely need to tell the compiler about alignment and array-overlap considerations to get good code. On the other hand, gcc will still attempt to vectorize code which hasn't had changes done to it at all. It just won't be able to get nearly as much of a performance improvement as you might hope. However, as time passes, more inner loop patterns will be added to the vectorizable list. Thus if you are using later versions of gcc, don't take the above results for granted. Check the output of the compiler yourself to see if it is behaving as you might expect. You might be pleasantly surprised by what it can do. -------- I thought CinGG as of today might not use any those nice SIMD extensions, but right now I haven't cin-gg freshly copiled, for checking object files ..currently compiling latest git. (32-bit build with gcc 5.5, but without -march=native or something.. I should try this ...with package version bump!)

1 0

Re: [Cin] MKV containers
by Benrob0329 10 Feb '19

10 Feb '19

I use the end-of-the-month builds, and it doesn't render any background frames in the specified /tmp directory. I do not experience slowdowns in playback (except when specifying either render every frame or applying some rather heavy effects) however I have noticed that MKV containers take significantly longer to seek than MP4s. I took a large constant frame-rate screen capture I had and converted it from an h264/aac MKV to an h264/aac MP4 (copying the streams with ffmpeg) and the seek times at a cut were greatly reduced (Those files will still send any NLE through a loop though because of how ungodly large they are). So I believe that MKV may have some extra overhead or lack of information that makes seeking through them harder for editors.On Feb 10, 2019 2:28 PM, Andrea paz <gamberucci.andrea(a)gmail.com> wrote: > > Hi Benrob0329. > Can I ask if you compile from GIT or use the end-of-month arch build? > Apart from background rendering, do you find slowdowns in playback? > For the background rendering it just doesn't work for me and the proof > is that it doesn't create still images. Does it do that to you too? > @IgorBeg. > Thank you for your patience. I converted an .mp4 file to .mpg. I went > from 12MB to 160MB but the GOP went from 250 to 12. But nothing has > changed in playback in CinGG. Both mp4 and mpg went at 30 fps normally > and dropped to 13 with effects.

1 0

Re: [Cin] MKV containers
by Benrob0329 10 Feb '19

10 Feb '19

I have also had problems getting background rendering to work on Arch. Not sure if it's specific to Arch Linux or if its a Cinelerra bug.On Feb 10, 2019 10:25 AM, Andrea paz <gamberucci.andrea(a)gmail.com> wrote: > > Some time ago I tried to use the background rendering to get the list > of png images. Now, using it on a test movie doesn't seem to affect > achivied frames (which always remain 14 fps dropping to 4 fps with > plugins). But the most disconcerting thing is that you do not create > still images: the folder I had indicated remains empty, even in /tmp I > can't find anything. In short, background rendering doesn't seem to > work on my system (as if that wasn't enough!). I've tried both to > enable the render farm in localhost and to disable it. Same result. > On the arch forum I don't find any problem with the graphic libraries, > the versions are the same. > The only thing that comes to mind is a formatting of the notebook and > restart from 0. But I'll wait until after finishing the translations > for the website. > -- > Cin mailing list > Cin(a)lists.cinelerra-gg.org > https://lists.cinelerra-gg.org/mailman/listinfo/cin

1 0

Turn this mailing list into a searchable archive?
by Terje J Hanssen 09 Feb '19

09 Feb '19

Hi, Is it an idea to make the growing cinelerra-gg mailing list into a searchable archive, available also for non-list members ? As a related sample below, the (Old) cinelerra(a)skolelinux.no<mailto:cinelerra@skolelinux.no> list is searchable, and many interesting Cinelerra subjects from the past can be looked up quick and easy. On the other hand the Cinelerra-cv mailing list looks to be missing there. If the list manager want to do this, here is the guide: ------------------------------------------------ [The Mail Archive] turns your mailing list into a searchable archive Search the archived lists Search 148855091 archived postings, 3539 active mailing lists. Archive your mailing list Looking for an easy way to turn your mailing list into a searchable archive? Just add The Mail Archive as a member to your mailing list as described in the how-to-guide. https://www.mail-archive.com/ ----------------------------- Samples: (Old) cinelerra(a)skolelinux.no<mailto:cinelerra@skolelinux.no> https://www.mail-archive.com/cinelerra@skolelinux.no/ [Bug 5093] Review Request: cinelerra-gg - A non linear video editor and effects processor https://www.mail-archive.com/rpmfusion-developers@lists.rpmfusion.org/msg23… [Bug 118] Review Request: cinelerra-cv - Advanced audio and video capturing, compositing, and editing https://www.mail-archive.com/rpmfusion-developers@lists.rpmfusion.org/msg16… Terje

2 2

AV1 codec and WebP are now part of GIT checkin today +
by Phyllis Smith 02 Feb '19

02 Feb '19

You can now decode *AV1* and encode AV1 in Cin-gg if you build from the GIT repository. Obviously it takes advantage of the ffmpeg 4.1 capability. Sam had announced the release of AV1 codec in the News section December 14 at cinelerra-gg.org. The encode opts file for rendering is under ffmpeg/webm as av1.webm. Most important to know is that it is very slooooooooooooooow. A 90 second video with 2 audio channels took 1 hour and 23 minutes to render BUT the resulting file is impressively small. It seemed to use only 1 - 3 threads (this is known and will be improved in the future). WebP is now also available for encoding/decoding. *WebP is an image format.* MatN had asked about this in the wiki MantisBT issue #97. The ffmpeg opts file is of type webp. OpenJpeg2 is another new format and is the selection jp2. The following libraries have been upgraded to include: *x265 to 3.0* (as Terje notified), fftw,opus,ogg,vorbis,tiff, and openjpeg. Non-ffmpeg ogg/vorbis were removed as there were several errors in them. Use ffmpeg instead as that version is maintained well. Issue #123 shuttle fixes are in. Some cleanup has been done to include; ydiff ffmpeg tweak, Makefile clean tweak, delete old CHANGELOGS, update downloads.txt, and comments changed in some ffmpeg opts files. gg/Phyllis

1 0

Play every frame
by Andrea paz 02 Feb '19

02 Feb '19

I'm working on very big files and for the first time my hardware can't keep up with them. (i7 272QM 2.2 MHz, Nvidia 460M, 12 GB RAM and SSD). The best solution I found was to disable "Play every frame". There is really a lot of difference to playback with or without this option, especially with many plugins. What does it mean to keep it disabled? (except for plugins that require it).

3 4

Cinelerra-GG new monthly builds ready for download
by Phyllis Smith 02 Feb '19

02 Feb '19

The new monthly Cin-GG builds are ready for download. Visit the website and while you are there, check out some changed pages as Sam keeps things fresh: cinelerra-gg.org The shuttle connections have been tested most on Fedora, a little on Mint 18, and barely on Ubuntu 16 and Leap 15. Do not hesitate to contact us if problems as we could not test all distros and you might want to wait for feedback from Pierre who will be doing real use testing before ordering a shuttle. Features5.pdf includes section 40.29 on usage and initial setup. Only a 1 time setup per Operating System should be required for non-root usage. > As root, just copy a file that provides the necessary permissions to use > the shuttle. A sample copy would be: > # cp {cindat_path}/doc/99-ShuttlePRO.rules > /etc/udev/rules.d/99-ShuttlePRO.rules > then the next time after you reboot, the permissions should be correct. > Copy of this month's release notes follow. gg/Phyllis ------------------------------------------------------ Cinelerra-GG (Version Infinity) Release Notes for 01/01/2019-01/31/2019 for builds 1. The O/S for all of the distros that are used in creating the monthly builds have been updated. This takes an intense 2 days of work for a little over 20 distros and usually gets done quarterly. 2. x264 and x265 libraries have been updated to the latest version at the time. And it has been recently brought to our attention that if using x264, you can do 10-bit as well as 8-bit with the standard monthly builds (thanks to Terje notification). For rendering, the opts video file for x264- 10bit.mp4 has been added to facilitate usage. 3. Cin now supports the ShuttlePROv2 and ShuttleXpress devices and can be user modified. 4. Drag Handle usage has been reworked. Besides Ripple, Roll, & Slip, there is now Slide and Edge. 5. Usability Improvements Add double edge insert and overwrite with highlight. “Fast” drag mode now allows for moving the effects, labels, keyframes, and autos along with the clips and honors the Settings pulldown checked items. Reset buttons have been added to 23 more plugins: Brightness/Contrast, Gamma, Hue Saturation, ReframeRT, Sharpen, Polar, Wave, Whirl, Blur, Radial Blur, Linear Blur, and Zoom Blur, Oil painting, Unsharp, Rumbler, YUV, YUV411, YUVShift, Swap Channels, ShiftInterlace, RGBShift, Translate, and Swap Frames. Contribution from another programmer. You can now dump the last 32 un-do’s in the undo stack on your terminal window if start cin there. Because cut and paste behavior was new last month, there has been some minor changes and additions to fine-tune specific areas. Groups can now be single source dragged. 6. Bugs/Issues Fixed Thanks this month to all of the users who took time to report bugs in MantisBT. It takes your time to do that but everyone benefits. The Dump Assets in File pulldown was crashing but is now fixed. Delete_edits optimize was causing a crash in drag and drop mode. Code cleanup to combine 2 separate color header files. Updated CineRmt which provides a remote control for using cinelerra on broadcast TV. A bug in ReframeRT with the settings has been fixed. Fixed bug in Rotate plugin of an issue when you move the draw pivot out of the canvas on bottom. Rework paste edits to accommodate new capabilities. Removed “code for lost auto” put in February 26, 2018. Changed build scripts to not allow an mjpeg v4l, non-version 2, build. Fixed Tip of the Day to rotate through the tips instead of always showing the same one. Create separate groups instead of adding to an existing group. Expanders.txt has had an additional time plugin added. Batch Render has been fixed to report the filename when it gets an error (was null). Render Farm had an additional field (now 6) added awhile back that caused the batch render columns to be incorrectly recorded because it only had 5 columns. This has been fixed. Users who change from batch render to render farm MAY BE IMPACTED but the workaround is to delete the $HOME/.bcast5/Cinelerra_rc file. DVB channels/tv fix ups as prompted by work of MatN. Removed build requirement of ESD sound driver. Batch render segv on boot defaults fixed. Blue Banana plugin bug fixed where a draw was happening inside destructor code. A potential hang when freeing locks has been corrected. Mjpegtools has been modified because there was a typo. The default ffmpeg audio and video icons were corrected with Sam’s logo (a mistake here). Shortcuts.html accessible from Cin has been updated. Differences.pdf was updated on the Server.

1 1

Configurations for Shuttles....
by Pierre autourduglobe 01 Feb '19

01 Feb '19

Hi Phyllis, I am very happy that both Shuttle models are now supported on Cinelerra and I look forward to testing them and reporting my observations to the Forum. But for that I have to be able to apply the configurations indicated in article 10.29 of Features5. For me, who am an illiterate of the linux code and the terminal, the line: # cp {cindat_path}/doc/99-ShuttlePRO.rules /etc/udev/rules.d/99-ShuttlePRO.rules For me, it could be written in "Chinese", it would be the same... Is it enough for me to simply paste as is in the terminal: # cp {cindat_path}/doc/99-ShuttlePRO.rules /etc/udev/rules.d/99-ShuttlePRO.rules without changing anything, and it will be settled? I guess it's not that simple, but I have no idea what to do... Pierre

2 2