After being highly impressed by my wimpy little netbook screaming along at warp 20 on video recoding of shows digitally broadcast over the air, something like 4x faster than any of my ancient towers can do, they switched to 1080I. Things slowed down.
Hmm, what is wrong? It used to be terribly fast?
I then found out 1080I isnt as quick and easy to do as 720P.
Removing garbage commercials and the excess time I program into the tuner card when I record the broadcasts.
Delete all commercials in a frame accurate way. without decompressing the file into some huge monstrosity sized one.
Get rid of I turn that to P.
If needed, on some shows resize to remove the way they change the original image to be 16:9.
Then compress all that into something more sane, like DivX or H264.
It was taking around a whole day to do one little hour of the video. I was used to getting all that done in 3 hours. Including the idea of demuxing the audio, so that I could shift it in time to restore lip sync.
Enter a nice little software named VideoDeMux.
Makes frame accurate cuts, simple, fast, doesnt screw up the audo sync either.
But, what im writing about is how to max-out the use of how ever many cores you have on your microprocessor.
No matter what ive tried to do,since broadcast TV upgraded again, I could not manage to force the silly netbook (ACER 2 core Atom) to process faster.
It would sit there at 30% CPU and no more.
I went Googling about super computer clusters. Basically, that is what multicore is. Just smaller. lol
Windows, nor Linux will sit there and find 2 cores, nor 1,000,000 other 'cores' as a cluster of something and use that to speed things up in general.
Either specialized software must be written to process everything in parallel (if you want the solution to PI in under one picosecond), or you have to manually assign multiple instances of the same application working on different files at the same time.
Just for fun...
I tried the second option. I let go one instance of virtualdub do one file.
Opened another virtualdub to do another file, then a third file.
Now the CPU is at 97ish percent.
The estimated time till done is the same as if I did only one, except once that same time passes, I will have three done.
Mmmmm!
Very nice!
Since all are set to IDLE priority, I can just muck around and do any other thing I wish on the PC without any lag or delay.
I now understand the glory of multicore microprocessors.
I also now want about 20 cheap pcs running along in a cluster. Call it a supercomputer...
The only problem with the nice 4 core ICs is its pricey. 8 and I think 16 core is available too, just add more money (a lot more money).
Certain few software is able to do this parallel thing out of the box, photoshop for one, another is one that draws out the manelbrot? set. For those ones all resources available are used and things happen scary fast.
But, even if your using software that has no idea how to go parallel. Just run multiples of the software. It will get things done at warp .... (how many pcs do you have networked) speed. lol
I learned myself a new thing that creates blazing speed and am excited about it.
Maybe cause im getting over the evil flu? lol
Mind you, one file will recode just as slowly as it always has, but if your doing 3 at the same time, it doesnt slow the system or framerate of recode of any of the others. So, in my case, its now getting 3 times the work done in the same amount of time.
I want a Cray now. Dammit.
