Stereo Tool

Update:
- CPU load is 2.5% lower
- Fixed a bug in multiband that caused occasional glitches.
- Small change in the audio (not noticeable - it's just that if I subtract the output of the two versions I have a remaining signal at about -70/-80 dB, due to a slightly changed behavior in multiband compression and Loudness filter).

Update: A new version with a HUGE performance increment.

As someone here suggested a while ago, I've replaced my own FFT implementation by a commercial library. After looking at some benchmarks I found out that my own implementation wasn't that bad (actually it would be in the top 3 out of 15 implementations that were tested on that benchmark site) - but still there are two that are a lot faster, so much that I've stopped working on my own implementation - I don't think I'll ever reach the speeds that they do, and for Stereo Tool users faster is better.

Please let me know how well this new version performs. I'm especially interested in the performance on AMD PC's - I'm testing on Intel here and I know now that I'm going to ditch my own implementation.

The links again:
Winamp DSP plugin: http://www.stereotool.com/download/dsp_ ... BETA3A.exe
Stand alone version: http://www.stereotool.com/download/ster ... BETA3A.exe
VST version: http://www.stereotool.com/download/vst_ ... BETA3A.dll

good news ...

Previous version from 11.feb. is cleaner then version from 08.feb!
I'll test this new Beta3A on my stream pc (amd X2)....


edit:
on my E8400 (default 3GHz) from ~25% cpu usage goes to ~17% (winamp dsp st).. wooow... awesome
.. now amd test ..

beta3a (11.feb) ST Standalone - AMD X2

beta3a (13.feb) ST Standalone - AMD X2


nice.

Hm, AMD difference isn't as big as I had hoped (I'm using an FFT library created by Intel... Which might explain it). Anyway, on my Intel I'm seeing a drop from 19% to 13%, which corresponds to your 25->17.

Previous version contains a bug that introduces some distortion-like artifacts (very soft, probably not or hardly noticeable). I'm now building a fix now.

Wow , i have a clean sound with the last version, and the cpu is lower on my cpu pentium dual core!

good work hans

Maybe dumb question...
Is guys from AMD have some library optimized for AMD cpus?
Then it will be good to use it for StereoTool ..so thru installation detect which CPU is used.

And what about multicore optimization ?

I hope to be able to remove the need for multicore optimization . By the way, even the Intel FFT library doesn't use multiple cores because the calculation is faster when it's done on a single core

Anyway, with the biggest CPU hog out of the way, I'm now looking at some other parts of the code that might be optimized. I'm going to post a new version in about an hour where the CPU load is again 7% lower. (Some time ago I wrote a filter especially for loud trumpet sounds combined with bass. Now I just noticed that it was using almost 10% of the total Stereo Tool CPU usage. I've lowered it a lot, the audio that comes out is different when I look at the waveforms, but it should sound nearly identical. (I was looking ahead to the next 512 to see what's coming every 64 samples, now I'm doing it every 256 samples, which reduces the amount of calculations by almost a factor 4).

Uploaded! BETA3A again, CPU load (with everything enabled) should be about 6-7% lower.

On My intel now from ~17% cpu usage ~15%

On AMD x2 from ~29% to ~28%.