Eliosound AirEQ CPU Optimization differences
- KVRian
- 936 posts since 29 May, 2002 from UK
So far I've been using the AirEQ in 'generic processor' mode. The cpu hit in this mode has been nice and low, along similar lines to Sonalksis.
You know how on installing it it asks you to choose between 'SSE' 'SSE2' and 'Generic'
Well just now I decided to uninstall it and re-install, selecting SSE2 instead, thinking this would be even more optimal and run with lower cpu.
To my surprise the cpu usage went up!
This is on an AMD 4400+ X2, any thoughts on this?. Seems generic mode runs better.
Anyone find similar findings with Intels?
I then tried 'SSE' mode, and that was very close to generic on cpu usage, perhaps just a fraction more. But SSE2 was the worst on the AMD.
Load a few of them up to see the accumulative difference.
You know how on installing it it asks you to choose between 'SSE' 'SSE2' and 'Generic'
Well just now I decided to uninstall it and re-install, selecting SSE2 instead, thinking this would be even more optimal and run with lower cpu.
To my surprise the cpu usage went up!
This is on an AMD 4400+ X2, any thoughts on this?. Seems generic mode runs better.
Anyone find similar findings with Intels?
I then tried 'SSE' mode, and that was very close to generic on cpu usage, perhaps just a fraction more. But SSE2 was the worst on the AMD.
Load a few of them up to see the accumulative difference.
Last edited by Arksun on Wed Sep 27, 2006 12:33 am, edited 1 time in total.
- KVRist
- 154 posts since 23 Feb, 2006
Arksun wrote:Well just now I decided to uninstall it and re-install, selecting SSE2 instead, thinking this would be even more optimal and run with lower cpu.
To my surprise the cpu usage went up!
It seems to be the same thing on Intels, but it may vary for different CPUs. We are looking at this issue these days, but this kind of SSE development is unfortunately more like "let's try and see what's happening".Arksun wrote:This is on an AMD 4400+ X2, any thoughts on this?. Seems generic mode runs better.
Anyone find similar findings with Intels?
I then tried 'SSE' mode, and that was very close to generic on cpu usage, perhaps just a fraction more. But SSE2 was the worst on the AMD.
Load a few of them up to see the accumulative difference.
But I promise that we will work on this issue, and we'll maybe have to buy more efficient profiling tools.. So please be patient !
I can say that AirEQ will be constantly improved regarding CPU usage, with standard algorithm optimization (even if it's already highly optimized, we can do more.. but it will take some time !), SSE , SSE2, 3 & 4 optimization..
Since AirEQ is intended to be a Track EQ, it has to be the most efficient possible.
Fabrice,
Eliosound
- KVRAF
- 19156 posts since 13 Feb, 2003 from Vancouver, Canada
So if you are experiencing less CPU with the non-SSE setting, should you leave it there? Is there a sound quality difference?
- KVRian
- Topic Starter
- 936 posts since 29 May, 2002 from UK
Given its such a low hit on generic mode anyways its no biggy, just thought I'd mention the increase on SSE2 incase people were defaulting to that without realising it might run even better on generic 
Let us know if/when SSE2 is more optimised!
Let us know if/when SSE2 is more optimised!
- KVRAF
- 19156 posts since 13 Feb, 2003 from Vancouver, Canada
Well, I noticed that the last update, while having some great new features, felt a little more "sticky" than before. Maybe that's it.
- KVRist
- 154 posts since 23 Feb, 2006
Absolutely, for the moment the non SSE setting is the one you may choose without any further test. The algorithm remains exactly the same, in fact this is only specialized instructions for DSP processing that are computed more efficiently.bduffy wrote:So if you are experiencing less CPU with the non-SSE setting, should you leave it there? Is there a sound quality difference?
For sure I'll do !Arksun wrote:Let us know if/when SSE2 is more optimised!
Fabrice,
Eliosound
- KVRAF
- 19156 posts since 13 Feb, 2003 from Vancouver, Canada
Awesome! That's what I thought.Eliosound wrote:Absolutely, for the moment the non SSE setting is the one you may choose without any further test. The algorithm remains exactly the same, in fact this is only specialized instructions for DSP processing that are computed more efficiently.bduffy wrote:So if you are experiencing less CPU with the non-SSE setting, should you leave it there? Is there a sound quality difference?
- KVRist
- 154 posts since 23 Feb, 2006
For the moment the SSE optimizations are for PC only, you are rignt, but with the coming MacIntel version, there will also be some SSE optimizations on Mac, for MacIntel only.OMU wrote:This does not have anything to do with the mac version, does it?
I didn't noticed anything like that.
Fabrice,
Eliosound
- KVRist
- 154 posts since 23 Feb, 2006
OMU wrote:OK, by the way, did you heard the news?
I'm your new customer, starting yesterday night
So I say welcome, and please drop your whish list here for the next AirEQ updates !
OMU wrote:AirEQ works just fine, the WaveBurner problem is still there though.
All the best!
I'll keep you informed about that !
Fabrice,
Eliosound
-
- KVRian
- 755 posts since 12 Mar, 2004
Hi Fabrice and thank you.
My first wish is to keep it optimised with the ppc architecture and not go just for the intel optimisations in the future. I'm stil on a g4, and, as it works great, I don't intend to change it soon. But I do want to have your EQ up to date!
I really like AirEQ just like it is; it's very well thought and I can't really say that I wish it to be different.
All the best!
My first wish is to keep it optimised with the ppc architecture and not go just for the intel optimisations in the future. I'm stil on a g4, and, as it works great, I don't intend to change it soon. But I do want to have your EQ up to date!
I really like AirEQ just like it is; it's very well thought and I can't really say that I wish it to be different.
All the best!
Last edited by OMU on Wed Sep 27, 2006 9:00 am, edited 2 times in total.
