Release of NeXus v1.3.0 with data compression

[Brian Tieman]

Mark Koennecke wrote:

>
>   Well, what is disappointing about this. A compression ratio of about 5
>   is very good for lossless compression. Apparently you also get the
>   necessary speed. BTW: what kind of machine did you use for the test?
>   As compression is CPU bounded the next generation processor will
>   probably give you 8hz.
>

The tests were done on a 500 MHz PIII machine.

The compression ratios are very acceptable and the speed is as good as to be
expected.  What's a little dissapointing is that I think I'm looking for a silver
bullet.  We will almost certainly start to compress data during the next run to start
gaining real world experience.

>   Users don't need to know about compression. But I think the instrument
>   scientist should be able to select a compression scheme based on some
>   tests.
>

For one of our projects, the data is consitant enough that this is certainly true.
Unfortunately, it's also the project that has the least user involvement.

The other project is a general purpose imaging system which can be used for a wide
range of experiments (tomography, diffraction, spectroscopy, microscopy, etc...)
Also, the samples can vary widely.  I fear there may be too many variables to select a
single algorythm for a single type of experiment.  This may mean the type o copreion
willneed to be decided upon during an experiment.  We'll need more real world
experience before we can know this for certain, however.

Happy Holidays all!

Brian Tieman