Quote:
Have had a "veg out" day...where I haven't really wanted to think about anything. Will do more tests tomorrow.
I've done some tests myself, and the CPU load difference on my system is actually bigger than what you are measuring.
Still measuring now, will add the remaining results here later.
Preset Bojcha's webradio:
1:24 - 6.10
1:12 - 6.20 BETA 125 WITHOUT 'rrr rrr rrr' filter
1:09 - 6.20 BETA 125 (22% slower than 6.10)
1:11 - 6.20 BETA 125 without AEHP
1:14 - 6.20 BETA 125 without AEHP and STRICT BASS clipping
So:
- rrr rrr rrr filter costs 3 seconds (4%)
- AEHP costs 2 seconds (3%)
- STRICT BASS CLIPPING costs 3 seconds (4%)
Without clipper:
2:05 - 6.10
1:53 - 6.20 BETA 125
11% slower.
FM output + oversampling:
1:16 6.10
1:04 6.20 BETA 125
19% slower.
[Odd. a bit later I'm doing the same measurement and now I'm getting 1:18 vs. 1:02 -> 25% increase.]
So the CPU load appears to be around 20% higher than that of version 6.10. Caused by 3 things: Multiband, pre-ringing protection and clipping.
Edit: Now I also measured the difference in Multiband without the pre-ringing protection filter. And it's about 5% heavier than in 6.10. That must be the voice protection and band 1+2 link. In total this difference is small (about 1%).
--> Odd. The 2nd measurement of BETA125 FM above was done with these changes, and the CPU load is suddenly 6% higher than before. Might indeed be a memory/caching issue.
I will do a quick attempt to improve the performance of the pre-ringing protection (that should be relatively easy), but don't expect too much from it (because it only uses about 6% of the total, max gain would be around 2-3%). And I'll look into strict bass clipping and AEHP. But I won't spend more than a few hours on this.