From: Milton Aupperle <milton@outcastsoft.com>

Date: March 15, 2008 10:42:34 AM MDT

To: Astro_IIDC@yahoogroups.com

Subject: Re: [Astro_IIDC] Re: Astro IIDC 4

Hi Chris;

On 15-Mar-08, at 10:10 AM, Christoph Bosshard wrote:

Hi Milton

Congratulations and thank you for the new version!

Thanks. It was a tonne of work.

I have just upgraded to V4 and I am test driving right now.

When recprocessing a Saturn image I found the speed on an Intel based

machine a big gain over V3.

The native version is anywhere from 30 to 70% faster, depending on what options your using for stacking.

Unfortunately none of the Altivec code for pixel comparison could be moved over to x86, as SSE3 simply is missing too many features to make it worth while porting. In some cases there are no equivalent functions like no such thing as bit level testing or permute for SSE3, which Altivec has in many different ways. And in others it would take 10 to 30 SSE3 instructions to replace a single Altivec instructions or I would be forced to switch back to scalar code and then have to re-load everything back into SSE3. Once you start adding up the total code your replacing, it just didn't make any sense perfrormance wise.

The new feature to manually select the

best images is improving the results dramatically.

Alan seems to like it too. The Sharpness code does a reasonably good job of weeding out really bad frames, but it can be fooled by high noise. And it's pretty ewasy to step through the list of frames one at a time using the keyboard keys and then tap the mouse button to reject a frame.

I guess there are more new features, just need some time to play with

them all.  ;-)

Yep - have fun :)

Milton J. Aupperle