Jody, Thanks for the nudge to join the JAI-Ext lists.
For reference... GeoTools 14-M1 w/ JAI-Ext is giving me a >12x speed increase w/ warp transforms and bicubic interpolation against GridCoverages with floating point data types. The timing of this work is great. I had to make some changes to Resampler2D in GeoTools. I'll defer a PR until I consult with the JAI-Ext folk to understand why the workarounds I reverted were initially in place. I am not seeing any artifacts with my resampling thus far (but maybe I don't know what to look for or my use-case doesn't expose the issue). Tom On Wed, Sep 2, 2015 at 1:58 PM, Jody Garnett <jody.garn...@gmail.com> wrote: > At the very least you could stage your own replacement operation in that > project (and remove the need for a workaround). > > Tom you may seriously wish to join that project's mailing list and ask > your question there (since that project is much closer to the metal so to > speak). > > -- > Jody Garnett > > On 2 September 2015 at 11:00, Tom Kunicki <tom.kuni...@weather.com> wrote: > >> Jody, >> >> Thanks, I am still looking into this. Any idea if the new JAI-Ext work >> will remove the need for that Warp workaround w/ floating point data types? >> >> On Tue, Sep 1, 2015 at 4:18 PM, Jody Garnett <jody.garn...@gmail.com> >> wrote: >> >>> No great insight here Tom, do you want to create a bug report? >>> >>> -- >>> Jody Garnett >>> >>> On 31 August 2015 at 16:41, Tom Kunicki <tom.kuni...@weather.com> wrote: >>> >>>> >>>> Hi, >>>> >>>> I have a floating point coverage that I would like to resample with >>>> BiCubic interpolation that requires a warp transform using GeoTools 13.2. >>>> >>>> I am running into an issue attempting to resample a floating point >>>> coverage and have isolated the issue to the following workaround here >>>> <https://github.com/geotools/geotools/blob/master/modules/library/coverage/src/main/java/org/geotools/coverage/processing/operation/Resampler2D.java#L705> >>>> . >>>> >>>> I seems the code is trying to reenter the reproject method by nulling >>>> out the interpolation argument (enters with BiCubic). It seems the >>>> original interpolation is set in the hints parameters so on each reentry >>>> BiCubic is pulled out of the hints and then the reproject method called >>>> again (until a StackOverflow is finally thrown). >>>> >>>> Is there any workaround? Happy to code with guidance. >>>> >>>> -- >>>> * Tom Kunicki* | Software Engineer >>>> *w:* 608 695 4251 *e:* tom.kuni...@weather.com >>>> >>>> >>>> >>>> >>>> -- >>>> * Tom Kunicki* | Software Engineer >>>> *w:* 608 695 4251 *e:* tom.kuni...@weather.com >>>> >>>> >>>> >>>> ------------------------------------------------------------------------------ >>>> >>>> _______________________________________________ >>>> GeoTools-GT2-Users mailing list >>>> GeoTools-GT2-Users@lists.sourceforge.net >>>> https://lists.sourceforge.net/lists/listinfo/geotools-gt2-users >>>> >>>> >>> >> >> >> -- >> * Tom Kunicki* | Software Engineer >> *w:* 608 695 4251 *e:* tom.kuni...@weather.com >> >> > -- * Tom Kunicki* | Software Engineer *w:* 608 695 4251 *e:* tom.kuni...@weather.com
------------------------------------------------------------------------------
_______________________________________________ GeoTools-GT2-Users mailing list GeoTools-GT2-Users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/geotools-gt2-users
