Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

New stacking software project "PlanetarySystemStacker"

astrophotography imaging moon planet solar
  • This topic is locked This topic is locked
881 replies to this topic

#551 Euripides

Euripides

    Ranger 4

  • -----
  • Posts: 326
  • Joined: 15 Aug 2019
  • Loc: Thessaloniki - Greece

Posted 03 June 2020 - 04:25 PM

Hello Rolf,

 

I'll go backwards:

 

If I load my tiff file and press Start, the app crashes immediately, with the same error as above.

 

 

About the *ini file, yes it is there in the home folder hidden. I deleted and create a new one and now it seems to work properly (with a new video, or just with a tiff file).

 

The only "major" difference between the 2 *ini files is the last part, the missing lines about the post layer.

[Hidden parameters]
current directory = /Users/euripides/ASICAP/Moon/2020-06-01Z
main window x0 = 76
main window y0 = 77
main window width = 1159
main window height = 771
main window maximized = False

[Global parameters]
version = PlanetarySystemStacker 0.8.0
protocol level = 1
write protocol to file = True
store protocol with result = True
buffering level = 1
include postprocessing = True
image format = tiff
parameters in filename = True
stack number frames = True
stack percent frames = True
ap box size = True
ap number = True

[Frames]
gauss width = 7
debayering default = Auto detect color
normalization = True
normalization threshold = 15
add selection dialog = True

[Align frames]
mode = Surface
automation = True
rectangle scale factor = 1.7857142857142858
search width = 102
average frame percent = 5

[Alignment points]
half box width = 54
search width = 14
structure threshold = 0.04
brightness threshold = 10
frame percent = 10

[PostprocessingInfo]
version selected = 1

The new one has also those lines at the end

[PostprocessingVersion 1 layer 0]
postprocessing method = Multilevel unsharp masking
radius = 1.0
amount = 0
luminance only = False

I've tried different combinations with the new *ini file to recreate the crash, but I've failed. So, it is not crashing anymore. That's good :-) The bad thing is that I haven't found the source of the issue but still, now I know where to look for first if this happens again.

 

PS1. I think that this started when I was playing with the Grid size...

 

 

PS2. Rolf, take a look in this video. It's about the Restore Standard Values button.

 

https://www.dropbox....es_bug.mov?dl=0

 

Is this correct?


Edited by Euripides, 03 June 2020 - 04:37 PM.

 

#552 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 04 June 2020 - 04:23 AM

Hi Euripides,

 

Then at least we know why PSS crashed. The .ini file was incomplete! I have no idea how this could have happened, and I have not heard from anybody else that this problem turned up. I think we will just wait and see if it occurs again.

 

I had a look at the video, and this is very strange as well! I immediately tested the button in my Windows installation, and it worked fine. Tonight I will try on Ubuntu 20.04LTS as well. What is strange is that this button is nothing special, just a push button. I will have a look at the code to find out if I'm overlooking something.

 

All the best,

 Rolf


 

#553 Der_Pit

Der_Pit

    Vanguard

  • *****
  • Posts: 2,187
  • Joined: 07 Jul 2018
  • Loc: La Palma

Posted 04 June 2020 - 04:48 AM

Hi Rolf,
 
sorry for my intermittent communication, work keeps me busy...
 

Regarding the file browser, the problem is that PSS lets you choose either a file (video with extension such as .avi) or a directory (containing still images of the same shape). Therefore, in the latter case the browser cannot stay outside the directory because the directory could contain an .avi which you might want to select. I would need to make the browser smarter, i.e. the browser would need to inspect the directory if there is any sub-directory or video file, and only in this case enter. I don't know an easy way to accomplish that.


Would it be possible to only cd into a directory on double click? The user usually does know what he's looking for and can decide by himself to either click once and select the directory as target, or double click to look for single files within that directory...
 

Congratulations to your first successful PSS processing! The moon image looks good, including the postprocessing. If you took the picture with your 140mm f/6.5 Apo refractor, the Baader FFC would be a very useful accessory. I use an FFC with my 130/1200mm LZOS Apo, and it works just great. There is no trace of lateral color or any other aberration.


Thanks you smile.gif
Yes, that was the 140APO.  With the FFC in lowest setting (3×) my ASI1600 would do fully critical nyquist sampling - you confirmation of it's quality makes this even more tempting....
 

Just one remark on 8bit vs. 16bit: In the past I took my moon SER files in 16bit mode as well, but then I realized that 8bit often give better results. After all, the files are halved in size, and the frame rate goes up. So, as compared to 16bit you can capture twice as many 8bit frames. If you take very many frames, the inevitable noise in dark regions and statistics in the end lead to a summation image with much more than the 8bit dynamic range of the individual frames. All the moon panoramas I took in recent years were shot at 8bit.

Hmm, I'll play with this a bit more again.  Maybe the first SER I took had focus issues, or the seeing wasn't as good.  The result was much worse than what I posted here, and trying to sharpen it mostly ended up showing noise....

Not sure if framerate is really an issue for the moon that doesn't change fast (at that resolution).  The data for the image I had taken in single frame readout, with short exposure time, but at ~0.7fps....  But that's likely an OT topic in this thread :)


 

#554 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 04 June 2020 - 06:13 AM

Hi Pit,

 

Hi Rolf,
 
sorry for my intermittent communication, work keeps me busy...
 


Would it be possible to only cd into a directory on double click? The user usually does know what he's looking for and can decide by himself to either click once and select the directory as target, or double click to look for single files within that directory...
 


Thanks you smile.gif
Yes, that was the 140APO.  With the FFC in lowest setting (3×) my ASI1600 would do fully critical nyquist sampling - you confirmation of it's quality makes this even more tempting....
 

Hmm, I'll play with this a bit more again.  Maybe the first SER I took had focus issues, or the seeing wasn't as good.  The result was much worse than what I posted here, and trying to sharpen it mostly ended up showing noise....

Not sure if framerate is really an issue for the moon that doesn't change fast (at that resolution).  The data for the image I had taken in single frame readout, with short exposure time, but at ~0.7fps....  But that's likely an OT topic in this thread smile.gif

What you describe as the desired behavior of the file chooser is exactly what happens at least on Windows: Double click to enter a folder, single click on a file or folder to just select it. Therefore, I don't understand what you are asking for.

 

Obviously you see a different behavior on your OpenSuse system. If on single-clicd the file chooser enters the folder already, how can you select a folder then?

 

All I can say is that I didn't write this code. I derived the file chooser from a standard file chooser and only added PSS-specific functionality. I did not touch the navigation part at all. Of course, QT adapts to the surrounding OS environment. After all, it would be bad if file navigation worked different in this widget as compared to the OS file explorers. Perhaps this is the cause of the different behavior?

 

All the best,

 Rolf

 

PS: I was tempted to answer on the imaging things SER vs. AVI, importance of frame rate (if you capture a 100 tile moon panorama) ... But I don't want to hijack my own thread.  smile.gif


 

#555 Der_Pit

Der_Pit

    Vanguard

  • *****
  • Posts: 2,187
  • Joined: 07 Jul 2018
  • Loc: La Palma

Posted 04 June 2020 - 06:46 AM

Hello Rolf,
 

What you describe as the desired behavior of the file chooser is exactly what happens at least on Windows: Double click to enter a folder, single click on a file or folder to just select it. Therefore, I don't understand what you are asking for.
 
Obviously you see a different behavior on your OpenSuse system. If on single-clicd the file chooser enters the folder already, how can you select a folder then?


Then this seems some difference between Windows and Linux - I strongly doubt it would be OS specific.  So I'd be interested in experience from users like John Duchek on Fedora, if they see the same issue.

What I have to do is, as described in my previous post, go to the directory that contains the folder I want to process, then click in the 'File name:' field and type the folder name.  Then the 'Open' button gets active, and I can use the selection by either pressing 'Enter' or clicking the button.

I looked in the file browser for some options regarding single or double click, but couldn't find any.  I'm aware PSS will just use the system defaults file browser (in my case, file.so from the kio plugins).  I'll see if I can find something about configuring it - a quick search didn't succeed :(
 


 

#556 Gucky

Gucky

    Viking 1

  • *****
  • Posts: 537
  • Joined: 18 May 2015
  • Loc: Switzerland

Posted 04 June 2020 - 07:53 AM

Then this seems some difference between Windows and Linux - I strongly doubt it would be OS specific.
...
I looked in the file browser for some options regarding single or double click, but couldn't find any.

In Linux this depends on the settings of the desktop environment you are using (you can't change this setting in the file browser).
I'm using Kubuntu with KDE as desktop manager. Here I have system settings -> Hardware -> Input Devices -> Mouse.
Then I can select "double click" or "single click" to open a folder or file.


Edited by Gucky, 04 June 2020 - 07:54 AM.

 

#557 Der_Pit

Der_Pit

    Vanguard

  • *****
  • Posts: 2,187
  • Joined: 07 Jul 2018
  • Loc: La Palma

Posted 04 June 2020 - 07:58 AM

blush.gif

So I found it.  This was a general configuration option in Workspace Behavior, 'Single-click to open files and folders'.  I first thought that I had set it at some point, but it seems to be the default setting of Plasma (I'm using 5.18).  So in doubt all KDE users might face this issue.

 

Just seeing Guckys posting:  Seems the location has changed in more recent versions, my 'Mouse' section doesn't have it anymore.

 

@Rolf: Would that be worth a footnote in the documentation, so users know where to look?


 

#558 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 04 June 2020 - 09:25 AM

Hi Pit and Peter,

 

Thank you very much for the clarifications. I agree with you that it is a good idea to add a note on this issue in the user guide. The only problem is that few users read it.  frown.gif

 

I had a look at the code which handles the "Restore Standard Values" button. I could not find anything that would special or risky. If the button is pressed, some variables are set and the sliders are adjusted. I have no idea why this goes wrong in this place, while the same operations seem to work properly for other buttons. It would be good to hear from other users (in particular on MacOS) if they experience the same problem. As I wrote before, I will check on Ubuntu 20.04LTS tonight myself.

 

All the best,

 Rolf


 

#559 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 04 June 2020 - 09:52 AM

Hi Pit,

 

Now I checked on Ubuntu 20.04LTS: The "Restore Standard Values" button works there perfectly well, as on Windows. The only explanation I can imagine is that this again is some QT issue on MacOS. We had seen a similarly strange thing before with the "start/cont" button not being rendered as active, but if one clicked on it, it just worked nevertheless. By some magic this issue went away with a different software version.

 

Does anybody else see the problem with the Restore button as well?

 

All the best,

 Rolf


 

#560 Euripides

Euripides

    Ranger 4

  • -----
  • Posts: 326
  • Joined: 15 Aug 2019
  • Loc: Thessaloniki - Greece

Posted 04 June 2020 - 09:53 AM

 

Then at least we know why PSS crashed. The .ini file was incomplete! I have no idea how this could have happened, and I have not heard from anybody else that this problem turned up. I think we will just wait and see if it occurs again.

 

 

Hi Rolf,

 

If I go to Post, remove all the layers and save it,  the *ini removes also the last lines. Tried it right now and again the app crashes instantly.


Edited by Euripides, 04 June 2020 - 09:54 AM.

 

#561 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 04 June 2020 - 10:12 AM

Hi Euripides,

 

Thank you very much for this test! You are right, the same happens on my Windows system.

 

Now I understand what happened. I had not anticipated that a user would remove all layers from a postprocessing version instead of deleting the whole version. But I agree that PSS should handle this strange user input more gracefully. It should definitely not crash!

 

I suggest that I change the code as follows: When there is only one layer left, and the user removes this layer, PSS deletes the postprocessing version. This should make the issue go away, and from the user's point of view removing all layers should be the same as deleting the whole version. What do you think?

 

All the best,

 Rolf


 

#562 Der_Pit

Der_Pit

    Vanguard

  • *****
  • Posts: 2,187
  • Joined: 07 Jul 2018
  • Loc: La Palma

Posted 04 June 2020 - 10:34 AM

Removing the layers itself is not the problem, as long as version selected = 0.  At least it doesn't crash for me then, and I can add layers and process etc....

 

Uhh, no, correction.  I can not add layers in that situation anymore confused1.gif

 

Edits 2:  Adding layers only works if I first click on 'New' (which automatically adds a first layer).  Than I can add additional ones....


Edited by Der_Pit, 04 June 2020 - 10:50 AM.

 

#563 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 04 June 2020 - 12:42 PM

Hi,

 

I think I fixed the postproc problem: If one tries to delete the last layer of a version, the layer is reset to "neutral" instead (radius = 1, amount = 0). So, the version is kept and can be populated with layers again. I found this solution better than removing the version without the user having asked for that.

 

@Pit: Please note that version 0 is always special. It is the original input file without any postprocessing. So you cannot add layers to it. When you start the postprocessor, and no versions were restored from a previous session, PSS starts with this version 0 and an initialized version 1 (i.e. with a single neutral layer).

 

The nice thing about PyPI is that you can try my bug fix immediately. I just uploaded a bug fix release 0.8.1 to PyPI which you can upgrade to with

 

   pip3 install --upgrade planetary-system-stacker

 

Do you agree that this bug fix solved the problem?

 

All the best,

 Rolf


 

#564 Euripides

Euripides

    Ranger 4

  • -----
  • Posts: 326
  • Joined: 15 Aug 2019
  • Loc: Thessaloniki - Greece

Posted 04 June 2020 - 04:01 PM

So far so good for me 👍

 

Only a small GUI issue on a 13 inch display.

 

 

GUI.jpg


 

#565 CarlGreen

CarlGreen

    Explorer 1

  • -----
  • Posts: 70
  • Joined: 16 Feb 2020
  • Loc: Northamptonshire, UK

Posted 05 June 2020 - 12:15 AM

Morning Rolf

 

Is there any way, for a non computer person like me, that there could be an "Update" button (For example in the about section) could be added?

 

That way to check I have the latest available release, I just click the Update and it automatically checks.

 

Just a thought.

 

Thanks

 

Carl


 

#566 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 05 June 2020 - 04:06 AM

Hi Euripides,

 

So far so good for me 👍

 

Only a small GUI issue on a 13 inch display.

 

 

attachicon.gifGUI.jpg

Well, that seems to be another strange thing to be filed under "QT on MacOS". In my Windows installation, the widget has the same width as the ones above it:

 

screenshot.png

 

Very strange, indeed!  frown.gif

 

All the best,

 Rolf


 

#567 Der_Pit

Der_Pit

    Vanguard

  • *****
  • Posts: 2,187
  • Joined: 07 Jul 2018
  • Loc: La Palma

Posted 05 June 2020 - 04:20 AM

On my computer, while having the same size as the ones above, the text also doesn't fit the box.

I'm quite sure this is an issue of HiDPI displays (I'm at 210ppi).  

I think box sizes are computed (fixed) on standard font size, but on HiDPI the system does upscale fonts, and things don't fit anymore.  I basically have similar issues with many applications.


 

#568 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 05 June 2020 - 09:49 AM

Hi Pit,

 

On my computer, while having the same size as the ones above, the text also doesn't fit the box.

I'm quite sure this is an issue of HiDPI displays (I'm at 210ppi).  

I think box sizes are computed (fixed) on standard font size, but on HiDPI the system does upscale fonts, and things don't fit anymore.  I basically have similar issues with many applications.

I fully agree! The problem is that the relation between point size and pixel size is not fixed. On HiDPI displays a point "has more pixels" than on an ordinary display. Since widget sizes are in pixels, the fonts don't fit any more. I have not found any satisfactory solution to this in a GUI application.

 

Alll the best,

 Rolf


 

#569 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 06 June 2020 - 03:27 AM

Hi Carl,

 

Morning Rolf

 

Is there any way, for a non computer person like me, that there could be an "Update" button (For example in the about section) could be added?

 

That way to check I have the latest available release, I just click the Update and it automatically checks.

 

Just a thought.

 

Thanks

 

Carl
 

Of course such an automated update could be implemented. The only question is who would do it. As long as I'm the only software developer on the project, I have to prioritize the many possible (or nice-to-have) additional features. An automatic update mechanism, I'm afraid, is not very high on that list.

 

That said, if you install PSS in the portable way, i.e. install Python 3.6 and then enter "pip3 install planetary-system-stacker" in a terminal, you have the requested update mechanism already now. All you have to do is to open a terminal and enter "pip3 install --upgrade planetary-system-stacker". The system will tell you if you already have the latest release, or otherwise it will do the update automatically. As you could see above, the other day I distributed a minor bug fix this way, and users could immediately check if it solved their problem.

 

Personally, I find the new installation procedure much better than the traditional one using the "all-inclusive" Windows installer. It minimizes the data transferred and installed on your computer, and it makes updates much more flexible. It is even possible to update libraries PSS depends on, without updating PSS itself. I only kept the single file Windows installer for people who don't want to care about installing third-party software such as Python 3.

 

All the best,

 Rolf


 

#570 CarlGreen

CarlGreen

    Explorer 1

  • -----
  • Posts: 70
  • Joined: 16 Feb 2020
  • Loc: Northamptonshire, UK

Posted 06 June 2020 - 04:08 AM

Hi Carl,

 

Of course such an automated update could be implemented. The only question is who would do it. As long as I'm the only software developer on the project, I have to prioritize the many possible (or nice-to-have) additional features. An automatic update mechanism, I'm afraid, is not very high on that list.

 

That said, if you install PSS in the portable way, i.e. install Python 3.6 and then enter "pip3 install planetary-system-stacker" in a terminal, you have the requested update mechanism already now. All you have to do is to open a terminal and enter "pip3 install --upgrade planetary-system-stacker". The system will tell you if you already have the latest release, or otherwise it will do the update automatically. As you could see above, the other day I distributed a minor bug fix this way, and users could immediately check if it solved their problem.

 

Personally, I find the new installation procedure much better than the traditional one using the "all-inclusive" Windows installer. It minimizes the data transferred and installed on your computer, and it makes updates much more flexible. It is even possible to update libraries PSS depends on, without updating PSS itself. I only kept the single file Windows installer for people who don't want to care about installing third-party software such as Python 3.

 

All the best,

 Rolf

Hi Rolf, thanks for the reply and completely understood re the prioritization of things.

 

I will stick to the windows installer method as being a novice, the only Python I have any knowledge of eats small animals and slithers!.

 

I think you are doing a marvelous job and don't envy the amount of work it must take.

 

Thanks again

 

Carl


 

#571 airscottdenning

airscottdenning

    Viking 1

  • *****
  • Posts: 536
  • Joined: 22 Aug 2008
  • Loc: Colorado

Posted 12 June 2020 - 08:58 AM

Hi Rolf,

 

I have been happily using PSS on my Mac in Catalina since you showed me the proper setup, but every time I start the program I get the following warning in the terminal window:

 

07-52-12.6 Warning: mkl_rt.dll / libmkl_rt.so does not work (not a Windows or Linux system, or Intel Math Kernel Library not installed?). dlopen(libmkl_rt.so, 6): image not found

 

This is despite the fact that I do indeed have mkl-2019.4 installed in python 3.7 via conda:

 

Downloading and Extracting Packages
certifi-2020.4.5.1   | 159 KB    | ##################################### | 100%
openssl-1.1.1g       | 3.4 MB    | ##################################### | 100%
ca-certificates-2020 | 132 KB    | ##################################### | 100%
mkl-2019.4           | 155.2 MB  | ##################################### | 100%
conda-4.8.3          | 3.0 MB    | ##################################### | 100%
Preparing transaction: done
Verifying transaction: done
Executing transaction: done

 

I still get great results for my Moon images from PSS, but I wonder if the program would be dramatically faster if I had the Intel Math Kernel Libraries properly linked.

 

Do you have any suggestions about how I might modify the code or the setup to get this resolved?

 

Thanks as always for a fantastic software effort to support our community!
 

Scott


Edited by airscottdenning, 12 June 2020 - 09:22 AM.

 

#572 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 12 June 2020 - 12:59 PM

Hi Scott,

 

First of all, I'm glad that you are getting great results for your Moon images from PSS. Thank you very much for this encouraging remark which makes me feel like all the progamming effort was no wasted time. smile.gif

 

The libmkl_rt.so issue is really difficult to solve. Euripides in this post told me the location of that library on his Mac. As I wrote in my reply, I expected the "2020.1.216" in the path to be highly version-dependent, so in PSS I use the search pattern "/opt/intel/compilers_and_libraries_*/mac/mkl/lib/libmkl_rt.dylib". Obviously, on your Mac the library is stored somewhere else, most likely because you installed Python with Anaconda. It would be nice if you could report back where on your system you find a library with the name "libmkl_rt.dylib". I could then put this location in the search path of PSS as well, so that the warning would go away with the next release.

 

It is a good question how much the actual performance depends on whether PSS finds the library or not. It's purpose is to tell the "Math Kernel Library" explicitly how many threads it should use, in order to make best use of the processor. On my Windows laptop, however, I did not see any noticeable performance difference. I just don't know how much it would help on your Mac. Therefore, I don't think it is a real problem if PSS does not find the library.

 

All the best,

 Rolf


 

#573 airscottdenning

airscottdenning

    Viking 1

  • *****
  • Posts: 536
  • Joined: 22 Aug 2008
  • Loc: Colorado

Posted 12 June 2020 - 01:16 PM

Hi Rolf,

 

I understand now ... there's no way you can know all the possible locations of the dylib file, especially because conda stores it in user space.

 

For what it's worth, here's the location of my copy:

 

/Users/denning/miniconda3/pkgs/mkl-2019.5-intel_281/lib/libmkl_rt.dylib

 

Perhaps you could allow the user to specify this in a config file someplace?

 

Anyway it seems it won't matter much that PSS doesn't find the library. Thanks for your reassurance on this.

 

My Mac has a 6 core/12 thread CPU and I find PSS to be pretty fast and painless to use. A HUGE improvement on my older workflow of Parallels -> Windows 10 -> AstroStakkert!3 -> Resgitax6 -> back to Mac -> PixInsight. It's *wonderful* to be able to process the data from raw video to finished image in a single modern and maintained application.

 

Thanks again,

Scott


 

#574 Der_Pit

Der_Pit

    Vanguard

  • *****
  • Posts: 2,187
  • Joined: 07 Jul 2018
  • Loc: La Palma

Posted 13 June 2020 - 06:43 AM

It is a good question how much the actual performance depends on whether PSS finds the library or not. It's purpose is to tell the "Math Kernel Library" explicitly how many threads it should use, in order to make best use of the processor. On my Windows laptop, however, I did not see any noticeable performance difference. I just don't know how much it would help on your Mac. Therefore, I don't think it is a real problem if PSS does not find the library.

Hi Rolf,

 

that's a point I am wondering about, too.  Is mkl supposed to run stuff multi-threaded?  Because so far I can always only see PSS use one CPU on heavy stuff....

 

If I use the 'lsof' command I can see that there are several threads having, e.g., libmkl_core.so open, but it's always only one CPU active.  Do you have any idea if that is "as it's supposed to be", or suggestions how to debug this if its not?

 

This is linux, latest 0.8.1 version of PSS.

woodstock:~% pip show mkl
Name: mkl
Version: 2019.0
Summary: Math library for Intel and compatible processors
Home-page: http://software.intel.com
Author: Intel Corporation
Author-email: scripting@intel.com
License: Proprietary - Intel
Location: /home/pit/.local/lib/python3.8/site-packages
Requires: intel-openmp
Required-by: planetary-system-stacker

woodstock:~% pip show intel-openmp
Name: intel-openmp
Version: 2020.0.133
Summary: Intel(R) OpenMP Runtime Library
Home-page: https://software.intel.com/en-us/articles/empowering-science-with-high-performance-python
Author: Intel Corporation
Author-email: scripting@intel.com
License: Proprietary - Intel
Location: /home/pit/.local/lib/python3.8/site-packages
Requires: 
Required-by: mkl



 

 

#575 Rolf

Rolf

    Viking 1

  • -----
  • topic starter
  • Posts: 508
  • Joined: 25 Apr 2016
  • Loc: Cologne, Germany

Posted 13 June 2020 - 09:15 AM

Hi Pit,

 

My problem is that I have not written that part of the code myself. It is burried deep in the Numpy and OpenCV routines which are implemented in C++. They are supposed to use multithreading, with the parameters set in libmkl_rt, but I made the same observation as you that there is no real concurrency. It could be, however, that this is problem-dependent. Some user reported for a very large example using many thousands of alignment points that PSS completed after 2 hours, three times faster than AS!3, which is very unusual. The only plausible explanation to me was that concurrency was active in this case. I did not check if that was true, though.

 

This issue is definitely worth investigating. After all, PSS could be made a lot faster with real multithreading! I will put it on my to-do list. Better still, if some MKL expert reads this and tells us what to do. smile.gif

 

All the best,

 Rolf


 


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics





Also tagged with one or more of these keywords: astrophotography, imaging, moon, planet, solar



Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics