Stereo Tool

It appears to be very marginally faster on XP as well. That's in a stand alone test, I don't know yet how it performs in the whole program.

Edit: Oddly, with the old compiler the performance gain appears to be smaller than with the new compiler. With the new compiler my own code gets slower - and the library should not be affected. Which is really weird. I need to test it on another system (that's in my office and I'm at home now) - that's the system on which I did the previous comparisons. I hope to see better results tomorrow.

Edit #2: Indeed, code with the old compiler appears to be slower, at least on my laptop. The difference is extremely small - but it's there. It could be that VTune is unreliable (and part of the numbers reported for my code was actually coming from the library) but that doesn't seem very likely. I'll continue tomorrow, and post some files for comparison soon.

I tried this on 2 PC's and 1 laptop,

• On the PC's (i7 970 -6 cores -12 threads, 12GB mem, Realtek HD Audio, WIN7x64), (Intel QX6850 -4 cores -4 threads, 8GB mem, Realtek HD Audio, WIN7x64)
On both PC's the CPU load is between 6% to 7%, a 1% drop from the earlier 7.50 beta version
Using only the US, Stereo settings, these Presets will not work (Sounds like a cache error/problem) the audio stutters.
- Bob Hawkins (All)
- D.R.Z (All)
- Jereon Platenkamp (All)
- Jeff Bebelusa FM (US. Stereo)
- Bojcha - King's Fire. King's Fire 2, Sevdah FM, Thorium6 FM, Ibiza FM
I get the following error: EXCEPTION!
EXCEPTION OCCURRED IN Modify_samples call internal
(Exception 0xc0000005 at address 0x5e18009fa, read/write=0,r/w location=0x0)

• On the laptop (Toshiba i7 4700MQ -4 cores -8 threads, 12GB mem, Realtek HD audio, WIN8x64)
The CPU load is between 11% to 12%, a 1% drop from the earlier 7.50 beta version
Using only the US, Stereo settings, these Presets will not work (Sounds like a cache error/problem) the audio stutters.
- Bob Hawkins (All)
- D.R.Z (All)
- Jereon Platenkamp (All)
Same error as PC, but slightly different address #

Not sure if it is wishful thinking or actual, but the audio does sound cleaner on all my systems, with new release version, with settings that work.

Same here, some presets work, others don't...don't know why but error messages continu to fill my screen when some presets are selected.

_________________
----------------------------------------------------------------------------
AIRCHAIN-GURU professional independant airchain consultancy.
Orban/Omnia/Vorsis/DSPX/Aphex/Inovonics
----------------------------------------------------------------------------

I know, will check that soon. Problem is probably that some function somewhere is optimized too aggressively. That can easily happen with the aggressive optimization settings that I use. I have to turn them off for functions that break due to this.

Wooh! If I enable all the new optimizations for i3/5/7 processors, I can run a complete FM preset including composite clipper and even Stokkemask with a CPU load of 10%! (CPU: i7-4770S). Without Stokkemask 7%! Streaming preset 2%.

Can we expect new build soon? I wonder how it works on DualCore intel's

Hm... I think I need to make a different compile for those (with different settings). The i3/5/7 has AVX, which allows performing calculations with 256-bit values (which, if you're using 32-bit floats, means that you can do 8 calculations with a single instruction, making it 8 times as fast as 'normal' code). For the Core2 CPU's I think the best they support is SSE4, which is 128 bit. The question is: Does that also get faster, or (what some measurements seem to indicate) does it get a lot slower?

Have you tried the DLL that I posted above? I know it crashes in many cases... (that seems to be fixed now btw, on my end anyway). But, you can compare the performance with that.

Ah, i did not. I dot use DSP much. Will give it a try.

Ok tested.. it's using slightly more cpu.

Confirmed, slightly more CPU usage here also.

BTW> How much ST relies on CPU cache and/or system bus? I have multiple Intel CPUs that are head 2 head in terms of raw clock performance, yet the differences are clearly visible for example between E6400 and E2220 in favor of the E6400 which is slower in clock speed, but has significantly larger cache and system bus (I know they are "oldies" and don't support SSE4, for these tests I used other CPUs)

I know one time I managed to run ST 7.24 in full blast with B-Town4 preset on a Core 2 E6300 system (2x1,86GHz 2M 1066 FSB, 3Gb of 800MHz DDR2) on Win7x64 but in a "run but don't touch" manner. I mean, even a slight mouse move was eating buffer inevitably.

I think the problem is that Hans is using a very powerful Computer to test "loads" and this gives values which are very high range for the "old" CPUs. I mean 1% on his side is equal may be to 10% for some CPUs on other users. So that is why other users reports "Confirmed. Slightly more CPU usage here also".
May be optimizations are for a small range of CPUs around his CPU and for others this optimizations are not that effective.

I know that he operates his computer to the lowest possible frequency - he mentioned this a couple of times, but looks like it`s not enough for the differences .