SIPS v3.4 released

Main page▹Product Overview▹Software	30.6.2017 \|

SIPS v3.4 released

The single most important feature of SIPS v3.4 is the very significant speed up of some algorithms. Still, some new tweaks make astronomer's life more convenient, which is always welcomed — e.g. SIPS can handle yet another batch of FITS keywords and Photometry including light curve plotting can be performed in real time immediately after each image is read from the camera.

There are algorithms implemented in SIPS, demanding a lot of computer resources (CPU computational power, amount of memory). Of course the requirements depend on the images processed, but processing of hundreds of images, each consisting of millions (or tens of millions) of pixels, needs certain time and space. Probably the most demanding algorithm is finding star within an image. This algorithm is performed not only when calculating image astrometry or processing photometry series, but also when images are stacked together and need to be mutually aligned, when the telescope mount is guided (either by stand alone guiding camera or by main camera inter-image guiding) etc. So making this algorithm as fast as possible makes the whole SIPS more responsible.

As illustrated on the image above, SIPS v3.4 can perform Photometry processing of 231 4k × 4k images 3.3-times faster, this means the whole batch processing time is shortened from 1 hour and 26 minutes to only 26 minutes. The most significant 4.2-times speed up is in finding of stars, other algorithms are performed in more or less the same time.

The test was performed on the same hardware (Intel Core i7 based PC with 16 GB of memory), two released SIPS versions 3.3 and 3.4 (available on this web site for download) run the same data set.

Remark:

Ironically, the speed up was implemented as a result of a failed attempt to move the computation to GPU. Unfortunately, the algorithm can be well parallelized, but it almost cannot be vectorized. This means it can run in many parallel threads on different CPU cores, but individual threads often need to run completely independently (through different parts of the code), they also need to share common data blocks and to access them exclusively (they need to be synchronized) etc. All these requirements are difficult to implement in GPU or they bring significant performance penalties.

Another part, which runs faster in v3.4 compared to previous versions, is image display and zoom. Zooming and scrolling of images on display is much faster and needs less memory. This makes SIPS more responsible.

Whenever SIPS needed to work with image height above the horizon (e.g. when calculating air mass), it performed all calculations by itself from image equatorial coordinates, geographic location of the observatory and universal time. Now it can also accept FITS keywords AZIMUTH and ALTITUDE, if present. Or directly the keyword AIRMASS if the original software stored it.

Remark:

As usual in FITS “standard”, some software packages do not use natural AZIMUTH and ALTITUDE keywords, but TELAZ and TELALT variants. SIPS can read both variants.

SIPS can of course also write these headers, just check respective options in the New FITS Header tool.

One more new feature in SIPS v3.4 is particularly useful. Beginning with this version, SIPS uses the ability of Windows applications to integrate with the OS graphics shell and indicates all progress bars also in the SIPS icon in the task bar. Well, this feature was introduced in Windows 7 and it is a shame SIPS implements in only now, but better late than later.

The progress bar is updated whenever another progress bar is shown within the SIPS GUI. So many actions are indicated, e.g. progress of an exposure, progress of image processing, image batch matching, multiple images loading into set, image set calibration etc. Especially for tasks performed for long time, the user can use the whole screen do something else and only check the progress in the task-bar SIPS icon.

SIPS is a freeware and can be downloaded from the Download section of this web site.