Fusing with down-sampling (bug?)

[adamkglaser]

I'm not sure if it is a bug, or this is a underlying reason why this should be the case, but have noticed something strange when I fuse with down-sampling. For my test, I have been saving a single HDF5 / XML file for BigStitcher in two ways:

1) At full resolution. 2) With 2x down-sampling in the HDF5 file itself, i.e. resolution level 0 is actually 2x down-sampled already.

Then I have been fusing these two files in BigStitcher, the first with 2x down-sampling, the second with 1x down-sampling (since I already 2x down-sampled the data in the input HDF5 iteslf). What is interesting is that the fusing times are the following:

1) ~5.5 hours 2) ~2.0 hours

Is there a reason the second case is almost twice as fast? In the end, the fused datasets are the same size. And in the first case, the HDF5 file contains a pyramid of resolutions, including resolution level 1 which is 2x down-sampled, so it shouldn't be doing any additional computations?

Thanks! Adam

[StephanPreibisch]

That is a very good question ... I never tried that. Could you send me some details (IJ-log & console output of both?). Thanks so much!

[adamkglaser]

Of course. Attached are two log files for case #1 and #2 above. Thanks! Adam case2.txt case1.txt