Enable Hardware Acceleration for FFMPEG under MacOS Compiled Version?

[speedheathenULTRA]

Would it be possible to modify the program so that the user can add additional FFMPEG commands before the program runs? Alternatively, add buttons to enable Hardware Acceleration for those of us who have Metal Capable GPUs? i.e.: I have a Radeon RX580 which will both outperform my Dual Xeon 3GHz in my MacPro as well as free up all those system resources so I can run other processes simultaneously.

[theophanemayaud]

Hey @speedheathenULTRA I could do the former on my fork quite easily I think ! It should be more flexible to just allow custom commands ! I’ll check if it’s easy 😉

[speedheathenULTRA]

@theophanemayaud that's awesome, man! Thank you! I'm churning through almost 20TB's of video......SPEED IS KEY!!

:)

[theophanemayaud]

As I said before I’m not sure that it’s really the bottleneck, but it’s worth testing it out !

[speedheathenULTRA]

I’m running the fastest available Intel i9 with 256GB and a 4tb NVME RAID 0 running on a Highpoint Pcie card operating at pci 4.0x16. On black magic I can sustain over 6,000mbs. When I’ve tested the data throughput using the Radeon, I do experience fastest speeds. Even if they’re the same as when using cpu if I could free up my cpu to do all my other tasks while running Viddupe of the Gpu it would allow me to complete my tasks faster.

[theophanemayaud]

Oh yes, but I meant the bottleneck in Vidupe itself, not in your machine. FFMPEG is just used to extract two frames per video. Then we have to compute a hash of each image.
I haven't tested to see which part of the computation is really intensive, and I have to read about how FFmpeg handles hardware (gpu) acceleration. I seem to remember something about only being able to use it for very specific codecs, and operations. Do you know if it can be used for picture extraction ?

[speedheathenULTRA]

@theophanemayaud I can tell you this, running the MacOS ported version I can scan a full library of roughly 15TB's in about 45mins. To do that same thing while using the Windows version within Parallels it takes 4+ hours.

[speedheathenULTRA]

@theophanemayaud I'm not sure about extracting pictures, I just don't have that kind of knowledge.

[theophanemayaud]

I’ll read a bit more about this and see if it’s worth trying. Then we’ll see if it really makes a difference ;) Théophane Mayaud Le 22 févr. 2021 à 16:45 +0100, speedheathenULTRA <[email protected]>, a écrit :

…

@theophanemayaud I'm not sure about extracting pictures, I just don't have that kind of knowledge. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

[speedheathenULTRA]

I’ll read a bit more about this and see if it’s worth trying. Then we’ll see if it really makes a difference ;) Théophane Mayaud Le 22 févr. 2021 à 16:45 +0100, speedheathenULTRA [email protected], a écrit :
…
@theophanemayaud I'm not sure about extracting pictures, I just don't have that kind of knowledge. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or unsubscribe.

Oh dude, it is! :)

Just Google VidProc. It's the fastest growing basic end user video app and was the first to support full hardware acceleration for supported and or patched Macs/Hackintosh.