PixInsight 1.8.0 RC3 'Ripley' Released

[astroedo]

Thank you for really HUGE job!

I've installed RC3 and get it a very fast try.

I have Windows 7 64 bit so i've installed 64bit version of PI 1.8.
I have a little problem with toolbars:

When i undock a toolbar for re-position it in a different position i cannot dock it again in any position and remains floating.

Closing and reopening PI makes the floating toolbars disappear.

View-Toolbars - Toolbar name  menu item shows the toolbar again, but floating and undockable.

I hope i was clear 

it isn't a big problem, but is quite annoying

Looking forward for final release

Edoardo

[red-shift]

I've upgraded to RC3 and been trying it for a few days. It's really faster and much less resource hungry -- the best proof is that my music player no longer gets choked anymore when PI runs (even in processing > 30 frames).  Nice job PI team!

However I did run into two problems so far. One is that ImageSolver script doesn't work (don't know if it worked in 1.7 as I just learned its existence). The other is that DBE got stuck in a weird state that I had to restart PI to make it work again. The weird state was that, though the auto sample generation created tens of good background sample points, DBE kept complaining it couldn't find more than three samples.

Anyone had the same problems?

[Gary]

I had the less than 3 dbe samples warning yesterday. Scratched my head realised I hadnt cropped all the black borders away from my image, soon as i did this it worked fine.

[georg.viehoever]

Try adjusting the tolerance parameter.

[DaveS]

I also had  the same 3 dbe samples problem with RC3, but lowering the 'Tolerance' from its default value, resolved the problem. Albeit, I never met this problem with V 1.7

Dave

[astroedo]

Thank you for really HUGE job!

I've installed RC3 and get it a very fast try.

I have Windows 7 64 bit so i've installed 64bit version of PI 1.8.
I have a little problem with toolbars:

When i undock a toolbar for re-position it in a different position i cannot dock it again in any position and remains floating.

Closing and reopening PI makes the floating toolbars disappear.

View-Toolbars - Toolbar name  menu item shows the toolbar again, but floating and undockable.

I hope i was clear 

it isn't a big problem, but is quite annoying

Looking forward for final release

Edoardo

Solved!
As shown in the attached image I had too many Explorer window visible on the left bar, so they overlapped and toolbars was undockable, hiding one or more windows i created enough space to host all the windows and now the toolbars behave normally

[cs_pixinsight]

* A new cache system in the ATrousWaveletTransform and MultiscaleMedianTransform tools. Thanks to this cache, these tools are now fast and accurate when used with partial previews and with the real-time preview window.

* Redesigned/reimplemented real-time preview functionality in all standard PixInsight tools. The real-time preview is now much faster and does not block the graphical interface, as happened with versions 1.7 and older.

Juan, any chance this will also apply to StarMask?  Having it work the same on the full size image as on a small preview and having a real-time preview would be most helpful.

Craig

[jtalbot]

Juan,

Thanks for the nice job on RC3.  Its been stable as far as I can tell on Win 7 64 and Mac Mountain Lion.

I will be showing off some of the new features at the WSP next week.  Look forward to the layers functionality in 1.8 final.

Jon Talbot

[derrickf]

I loaded both 64 bit and 32 bit versions of 1.8-RC3 onto separate machines respectively running Windows 7 64bit Ultimate and Windows XP-SP3 today and both run considerably more slowly than version 1.7.

For example on the 64 bit machine (AMD 6100 6 core 3.3MHz with 16GB RAM) using the same image:

• create star mask: V1.8 = 190 seconds.    V1.7 = 37 seconds
• Realtime preview ATWT noise reduction  using a preview: V1.8 = 20 seconds for cache build + 15 seconds.     V1.7 < 3 seconds

Each time I toggle the preview button on the ATWT dialogue, it takes another 15 seconds to update the preview on 1.8 and <3s on 1.7
Similar diffrences in processing time apply between 1.8 and 1.7 in all the functions I tried. It is the same story with the 32 bit system which uses a dual core AMD processor.

I ensured that in each case PI was using all available processors (confirmed via the process console at startup)

It seems that others are seeing a speed increase with 1.8, any suggestions why I might be seeing the opposite?

I have reverted to 1.7 for the time being since 1.8 makes it next to impossible to run resource intensive processes like Image Integration, Deconvolution etc.

[gianpri]

I installed RC3 in Vista 64bit. All good, but the script "batchFormatConversion" is not running. The file ".Fit" always R: 0.000 G: 0.000 B: 0.000 or K: 0000, Why? With Version 1.7 was all OK. thanks
John

[AstroScience]

Hi,
change output sample format to 32 bit.

[Juan Conejero]

Hi John,

I installed RC3 in Vista 64bit. All good, but the script "batchFormatConversion" is not running. The file ".Fit" always R: 0.000 G: 0.000 B: 0.000 or K: 0000, Why? With Version 1.7 was all OK.

I have just tested the BatchFormatConversion script with a combination of FITS, TIFF, JPEG, DSLR raw files and different sample formats, and it works correctly in all cases.

Could you upload a bunch of files, so I can try to reproduce the problem?  Thanks!

[gianpri]

Thanks AstroScience, in fact works with 32 bit output. But why with version 1.7 also worked 16-bit?

Juan how can I send a picture. "Cr2" to 7 MB?

Sorry for my English

John

[Juan Conejero]

I loaded both 64 bit and 32 bit versions of 1.8-RC3 onto separate machines respectively running Windows 7 64bit Ultimate and Windows XP-SP3 today and both run considerably more slowly than version 1.7.

Thank you for your report. We have detected a performance regression in version 1.8.0 RC3, which affects irregularly to some tools. I have already identified and fixed this problem today, and right now I'm preparing a new 1.8.0 RC4, which hopefully will be the last one before the final 1.8.0 release.

As for the performance regression in RC3, some tools are more affected than others, and StarMask is indeed one of the worst cases. However, the numbers you're showing are much more severe (more than twice) than what we have measured on Intel-based machines. One possibility is that the performance loss be hardware-dependent, being worse on AMD processors. Not too surprising because the problem is related to atomic operations (reference/dereference operations in multithreaded code).

As for ATrousWaveletTransform, take into account that the previews (including real-time previews) of this process in version 1.7 were so inaccurate that they were basically useless. They are accurate in version 1.8, but this requires computing the transform for the whole image, not just for the preview or the real-time preview. Once the whole transform has been cached, if you avoid modifying parameters that invalidate the cache (as resetting the tool, changing the number of layers, or the scaling function), previews are (especially once we have fixed this performance regression in RC4) considerably faster than in version 1.7. The same applies to MultiscaleMedianTransform, where I have implemented the same cache system.

[Juan Conejero]

Hi John,

Juan how can I send a picture. "Cr2" to 7 MB?

You can request a PixInsight Files account. See this thread for more information.