Hello:
I used to run ASTAP routinely, as far as I can remember on my previous MacBook Pro. I'm now running an M1, and installed the designated package, and execute the codesign command in terminal. But the program is glacially slow, takes a crazy long time to load a file, and to compute the image tilt (my main use of ASTAP for the time being). Any help with this would be greatly appreciated!
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Analysing step on M1 is very slow. It takes all do to do it. I loaded 60 darks at 2:30 and by 5 it was still running. They are not large files either to at only 11MP or 23MB
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
That looks abnormal to me. At the moment my new camera is producing files of 50 mbytes. Yes the processing is slow but not like yours. Is there any difference with the older versions?
I will work here first on running the sigma clip in memory. Now it loads the files 3 times. Using the memory this can be reduced to once. Potential speed improvement maybe a factor 2. Assuming there is 8 gbyte of memory spare, it should be able to hold about 240 images in memory.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Under Windows, analysing darks of 50 mbytes takes about one second per dark. Combiningn to a master darks take also about one second. I will check on my old Mac.
Grimstad, ideally you share some darks so I can have a look what happening.
Howard, share some images which are slow.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Keeping the images in memory did not give an improvement.
However changing the array structure from
array{nrcolors,width,height}
to
array{nrcolors,height,width}
gave a very significant improvement in stack time. I assume the 3d arrays are stored as several 1D arrays. Processing rows rather then columns makes a significant difference.
First impression:
stack average 30% faster.
stack sigma clip 45% faster.
Hopes this stays this way. Very big modification. Every routine/procedure requires retesting.
A test version should be available in one or two days.
Han
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
For the Mac only the executables are updated and available in zip files. You have to move the executable to the app to test it. Installers I will update if no bugs are found..
New ASTAP development version v2023-09-18 was released.
This version will be faster then previous versions. This required a lot of code changes in almost every routine so this could have introduced a new bug. Please report any problem.
During one test, I noticed the following reduced processing times:
Average stacking. About 30% faster.
Sigma Clip stacking. About 47% faster.
Solving for near 2 degrees radius, about 23% faster. For large radius no difference.
Han
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hello:
I used to run ASTAP routinely, as far as I can remember on my previous MacBook Pro. I'm now running an M1, and installed the designated package, and execute the codesign command in terminal. But the program is glacially slow, takes a crazy long time to load a file, and to compute the image tilt (my main use of ASTAP for the time being). Any help with this would be greatly appreciated!
Hi Howard,
Can you share one of the images which are processed slowly?
Han
Last edit: han.k 2023-08-26
Analysing step on M1 is very slow. It takes all do to do it. I loaded 60 darks at 2:30 and by 5 it was still running. They are not large files either to at only 11MP or 23MB
That looks abnormal to me. At the moment my new camera is producing files of 50 mbytes. Yes the processing is slow but not like yours. Is there any difference with the older versions?
I will work here first on running the sigma clip in memory. Now it loads the files 3 times. Using the memory this can be reduced to once. Potential speed improvement maybe a factor 2. Assuming there is 8 gbyte of memory spare, it should be able to hold about 240 images in memory.
Under Windows, analysing darks of 50 mbytes takes about one second per dark. Combiningn to a master darks take also about one second. I will check on my old Mac.
Grimstad, ideally you share some darks so I can have a look what happening.
Howard, share some images which are slow.
I did a number of experiments to speed up.
Keeping the images in memory did not give an improvement.
However changing the array structure from
array{nrcolors,width,height}
to
array{nrcolors,height,width}
gave a very significant improvement in stack time. I assume the 3d arrays are stored as several 1D arrays. Processing rows rather then columns makes a significant difference.
First impression:
stack average 30% faster.
stack sigma clip 45% faster.
Hopes this stays this way. Very big modification. Every routine/procedure requires retesting.
A test version should be available in one or two days.
Han
For the Mac only the executables are updated and available in zip files. You have to move the executable to the app to test it. Installers I will update if no bugs are found..
New ASTAP development version v2023-09-18 was released.
Download link:
http://www.hnsky.org/astap.htm#beta_version
This version will be faster then previous versions. This required a lot of code changes in almost every routine so this could have introduced a new bug. Please report any problem.
During one test, I noticed the following reduced processing times:
Average stacking. About 30% faster.
Sigma Clip stacking. About 47% faster.
Solving for near 2 degrees radius, about 23% faster. For large radius no difference.
Han
Ill give the new one a try. Stacking is faster in the last build I used but mosiac took all day to get to 24%. Hopefully this speeds things up.
The mosaic routine has also changed. The canvas size is now sized automatically. But yes more then four tiles makes it slow.
Han