Regarding the ImageMagick memory handlers. The MagickCore API provides a method to assign your own memory handlers if desired. In the mean-time to reduce memory consumption you can set the memory and memory-mapped resource limits to force pixels for large images to disk. This is of course slower but it stops any one process from consuming too much of the limited memory resource. We do this for the ImageMagick web interface, for example, at http://www.imagemagick.org/MagickStudio/scripts/MagickStudio.cgi.
In the mean-time we have investigated returning memory from the system with anonymous memory mapping. It's easy enough to do but it does slow down ImageMagick a bit because anonymous memory mapping is zeroed whereas heap memory is not. The embeddable memory methods are capable of returning memory to the system when a free block is coalesced and all segments in the coalesced free block are unused. Unfortuately we have not implemented on this option just yet, although it would be just a few lines of code. _______________________________________________ Magick-developers mailing list Magick-developers@imagemagick.org http://studio.imagemagick.org/mailman/listinfo/magick-developers
