Stereo Tool
https://forums.stereotool.com/

Stereo Tool 6.10
https://forums.stereotool.com/viewtopic.php?t=3065		 Page 194 of 213
Author:  hvz [ Mon Sep 26, 2011 12:48 pm ]
Post subject:  Re: Stereo Tool 6.10
@Luke: Not really an option. Because the idea was that countries each used a different code, but in practice many stations use very different codes.

I found a few interesting links:
http://www.w9wi.com/articles/rdsreverse.htm - PI calculator for the US! Apparently there it's regulated such that the PI code is based on the station name!
http://radio-tv-nederland.nl/rds/rds.html - A list of Dutch radio stations' PI codes. No clear use of a country code here either.
Author:  eldoradofm [ Mon Sep 26, 2011 12:54 pm ]
Post subject:  Re: Stereo Tool 6.10
If i remember correctly the country option is in http://www.diffusionsoftware.com/airomate.php
Yes: PI and ECC can be selected by country from a list.
Author:  Brian [ Mon Sep 26, 2011 1:01 pm ]
Post subject:  Re: Stereo Tool 6.10
Quote:
Actually this whole cache thing doesn't make much sense anymore since I 'removed' Natural Dynamics. The only thing that uses a lot of memory is Multiband (because it separates the data into 10 bands per channel = 20 * 4096 floats = 320 kB, plus extra buffers for other data which might mean a doubling of this number).

The declipper and clipper however do things with very little memory (one channel at once, and only a few buffers = a few times 4096 floats = a few times 16 kB).

I'm actually starting to wonder if it could be something else. For example:
- Code cache vs. data cache.
- Cache lines issue (for example, 64 kB multiples of addresses cannot be in the cache simultaneously. This can give HUGE performance effects.)

Could someone who has cache problems do the following tests?
1. Try what happens if you turn Multiband off.
2. Try what happens if you lower the latency by one step (or multiple steps). Each step causes a reduction of memory usage by half! (The memory will stay allocated, but it's no longer used, hence no longer placed in the cache).

If lowering the latency does NOT help, the issue is probably code cache instead of data cache. If it's that I can try building a version that's more optimized for size than for performance.
These tests are with the DSP version and Winamp. With all of the tests, the vast majority of the CPU % was noted as coming from the in_mp3.dll instantiated thread of dsp_stereo_tool.dll.

Also, note that by "Multiband ON", I'm meaning just loading my standard preset with everything I normally have on. What I normally have on is:

Stereo Image: Azimuth correction and Stereo Boost (without the reverb option)
AGC (3-band, no bass AGC)
Multiband
Bass Boost
Noise Gate
Bandpass (Phase Linear Highpass, both options checked)
Singleband limiter
Final Limiter's volume increase
Loudness (Dirty bass, Improved Bass, Strict, Deep Bass, deesser, no AEHP, CPU shortcuts checked)
Hard Limit

4096 samples:

Normal Preset with GUI UP = 56-65%
Normal Preset with GUI in System Tray = 43-52%

Multiband ON with AGC off and GUI UP = 43 - 55%
Multiband ON with AGC off and GUI in System Tray = 36-40%

Multiband OFF with GUI UP = 32-38%
Multiband OFF with GUI in System Tray = same

Multiband and AGC OFF = 34-38% (no change in GUI UP vs. Tray)

2048 samples:

Normal Preset with GUI UP = 55-62%
Normal Preset with GUI in System Tray = 15-22% :!:

The other tests show similar reductions, with the final test of AGC OFF and Multiband OFF = 8-12%, GUI UP or in Tray

Per our PM discussion, you felt the data collection routines for the GUI updating could be expensive. I'd suggest looking into that, as the most significant reduction was when I lowered the number of samples AND put the GUI to the tray.
Author:  Luke [ Mon Sep 26, 2011 1:15 pm ]
Post subject:  Re: Stereo Tool 6.10
Quote:
@Luke: Not really an option. Because the idea was that countries each used a different code, but in practice many stations use very different codes.

I found a few interesting links:
http://www.w9wi.com/articles/rdsreverse.htm - PI calculator for the US! Apparently there it's regulated such that the PI code is based on the station name!
http://radio-tv-nederland.nl/rds/rds.html - A list of Dutch radio stations' PI codes. No clear use of a country code here either.

Wow, now that is just crazy. Talk about all the agreements that were made thrown away and ignored by even official radiostations.
It's like the PS that should have been a fixed code to keep the driver of the car focusing on the road instead of the radio.

I should check my documents, if I still have them after .... what 18 years (djeeeeeez). I think I made a list somewhere of all the countries and the types.
Author:  Luke [ Mon Sep 26, 2011 1:41 pm ]
Post subject:  Re: Stereo Tool 6.10
Quote:
If i remember correctly the country option is in http://www.diffusionsoftware.com/airomate.php
Yes: PI and ECC can be selected by country from a list.

Aha, that was it ECC

ECC - Extended Country Code: RDS uses its own country codes. The first most significant bits of the PI code carry the RDS country code. Their four bit coding structure only permits the definition of 15 different codes, 1 to F (hex). Since there are much more countries to be identified, some countries have to share the same code which does not permit unique identification. Hence there is the need to use the Extended Country Code. The ECC consists of eight bits.
Author:  hvz [ Mon Sep 26, 2011 2:03 pm ]
Post subject:  Re: Stereo Tool 6.10
Quote:
2048 samples:

Normal Preset with GUI UP = 55-62%
Normal Preset with GUI in System Tray = 15-22% :!:
Wooh! Now that's a BIG difference. And on my system there is nearly no difference at all (all the different latencies use the same amount of processing +/- 5% or so).

Just to make sure that this is not one of these Task Manager issues: Could you try the following?

1. In Winamp, select output to disk (I think it's called out_wav.dll, you could also use out_mp3.dll).
2. Then let Winamp run for a minute and check how much data it has processed.

This gives much more reliable numbers.

Could you also check this at latency 1024?

Like I said, on my PC I don't see anything like this _AT ALL_ so, this could really be the explanation of the performance issues on AMD's.
Author:  DJ-DOGGY [ Mon Sep 26, 2011 2:28 pm ]
Post subject:  Re: Stereo Tool 6.10
I`m sorry but i`m with AMD too and no differences like that
like Hans said just arround 5%
Author:  Brian [ Mon Sep 26, 2011 3:10 pm ]
Post subject:  Re: Stereo Tool 6.10
Quote:
I`m sorry but i`m with AMD too and no differences like that
like Hans said just arround 5%
It is very important for you to state your EXACT processor, as if you have Phenom (K10 architecture), you have a shared 2MB L3 cache, and if you have a newer Phenom II (also K10, but with changes), you have a shared 6MB L3 cache, while K8-based AMDs, like what I have, only have a maximum of 1MB L2 cache per core and no L3 cache.
Author:  Brian [ Mon Sep 26, 2011 3:12 pm ]
Post subject:  Re: Stereo Tool 6.10
Quote:
Quote:
2048 samples:

Normal Preset with GUI UP = 55-62%
Normal Preset with GUI in System Tray = 15-22% :!:
Wooh! Now that's a BIG difference. And on my system there is nearly no difference at all (all the different latencies use the same amount of processing +/- 5% or so).

Just to make sure that this is not one of these Task Manager issues: Could you try the following?

1. In Winamp, select output to disk (I think it's called out_wav.dll, you could also use out_mp3.dll).
2. Then let Winamp run for a minute and check how much data it has processed.

This gives much more reliable numbers.

Could you also check this at latency 1024?

Like I said, on my PC I don't see anything like this _AT ALL_ so, this could really be the explanation of the performance issues on AMD's.
I'll do comparisons later today. I have to take care of some real life stuff for a while... BTW, yes, I was a bit extra-grumpy over the addition / change of the GUI displaying the ligher shading (amplification values) because of knowing about this GUI up vs. tray issue... The increase in load from BETA120 to 123 was 3-5% (absolute value of increase, not delta... the delta from 120 to 123 would be +5% to +10%) at 4096 samples.
Author:  hvz [ Mon Sep 26, 2011 4:17 pm ]
Post subject:  Re: Stereo Tool 6.10
@Brian: It gets weirder and weirder. Actually I had expected a VERY small CPU load increase from the voice protection filter in Multiband - but that was finished already before BETA120. So then it must be the band 1+2 link code - that does almost nothing (but almost probably isn't "nothing" enough). Will check if I can improve it; I hope that I'll be able to see the same performance difference.

The new display should not really make any difference for the performance - but the band 1+2 lock might. (I use some extra data for it. Only a few single values, but it might just be too much).
Page 194 of 213 All times are UTC+02:00
Powered by phpBB® Forum Software © phpBB Limited
https://www.phpbb.com/