On 2015-01-11 09:27-0000 Phil Rosenberg wrote: > Regarding the "special values" where small errors in floating points lead > to jumps in integer representation - I think this is probably dealt with > now by anti-aliasing in most modern drivers.
No, it is worse than that. Generally, we have PLFLT input which is transformed to internal physical coordinates which are then transformed again (usually by simple scaling) by device drivers into the actual physical coordinates (pixels or whatever) that driver understands. So for certain magical combinations of input (which our standard examples used to hit a lot), the internal physical coordinates which are currently represented as 16-bit integers can flip to a different integer causing a section of the whole plot to shift visibly relative to the rest depending on platform. (Here we are defining platform as the combination of computer language, optimization for that language, OS, and hardware). In other words, the devices are following the orders they are given (with or without antialiasing applied) by the internal physical coordinates, but for magic input values those orders get changed by +/- 1 integer out of the limited total range of 16-bit integers depending on platform. We have historically adjusted our standard examples to avoid these magic situations so that we get more consistent test results comparing one platform with another. But that is just a cover up, and I would far prefer our internal physical coordinates simply had a lot more resolution so that in full generality regardless of input values the PLplot results always look the same (with small rounding errors consistent with 64-bit doubles rather than much larger and visible rounding errors consistent with 16-bit integers) from one platform to the next. Alan __________________________ Alan W. Irwin Astronomical research affiliation with Department of Physics and Astronomy, University of Victoria (astrowww.phys.uvic.ca). Programming affiliations with the FreeEOS equation-of-state implementation for stellar interiors (freeeos.sf.net); the Time Ephemerides project (timeephem.sf.net); PLplot scientific plotting software package (plplot.sf.net); the libLASi project (unifont.org/lasi); the Loads of Linux Links project (loll.sf.net); and the Linux Brochure Project (lbproject.sf.net). __________________________ Linux-powered Science __________________________ ------------------------------------------------------------------------------ Dive into the World of Parallel Programming! The Go Parallel Website, sponsored by Intel and developed in partnership with Slashdot Media, is your hub for all things parallel software development, from weekly thought leadership blogs to news, videos, case studies, tutorials and more. Take a look and join the conversation now. http://goparallel.sourceforge.net _______________________________________________ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
