Investigate performance of TIFF-based readers

[mlinkert]

See files in ome-tiff/johannes and zeiss-lsm/jerome/CONFOCAL*, and the attached ImageJ macro and BeanShell? script for testing.

For the OME-TIFF file and the ImageJ macro, the baseline timing is:

Plain IJ1: 531ms, Bio-Formats: 7669ms

For the OME-TIFF file and the BeanShell? script, the baseline timing is:

create reader: 677 ms
create metadata structure: 1577 ms
initialize file: 2004 ms
read all images: 494 ms

...which basically means that constructing a reader and an OMEXMLMetadata is more expensive than calling setId. So the first priority is to reduce the amount of time required to set up, and then we can work on reducing the setId execution time so that it is closer to what we see when reading files with ImageJ's built-in TIFF reader.

[mlinkert]

These commits:

​https://github.com/openmicroscopy/bioformats/commit/ccfdb78b4979112d87c3afb3c2d25c4d93a7f0aa
​https://github.com/openmicroscopy/bioformats/commit/62c862c4f51838ca63201115ec9c6f27bc179332

bring the output of timing.ijm down to:

Plain IJ1: 390ms, Bio-Formats: 2248ms

and the output of the BeanShell? script to:

create reader: 700 ms
create metadata structure: 144 ms
initialize file: 2877 ms
read all images: 229 ms

Still not as good as we'd like, but it's progress.

[jburel]

Moved from sprint 2012-08-28 (3)

[mlinkert]

See also this thread: ​http://lists.openmicroscopy.org.uk/pipermail/ome-devel/2012-November/002319.html