Tile target README
==================

Tile is a target that allows you to lay a (large) display area over multiple
displays.  It is like those large screens on the wall that are actually
composed of rows of monitors each displaying a smaller part of the image,
but this target is less efficient because it's done by software rather than
hardware.  Howvever it is more flexible because the tiles are not restricted
to any grid; in fact, the "tiles" can be mapped to any area of the display,
and some areas can even be unmapped!



USAGE
=====

The target requires arguments in the form of 

"offx,offy,sizex,sizey,(subdisplay):..."

where offx and offy are the tile's offset from the main display,
      sizex and sizey are the size of the tile,
      subdisplay is the display name to open for the tile,

  and ... is more tiles following the same format as above.

The subdisplay name must be in parantheses.  A colon separates the
individual tiles. offx,offy,sizex,sizey are integers.

For example, this makes an equal tile display of 2x2 X-target visuals
with a size of 320x200 each:

       "display-tile:0,0,320,200,(display-x):320,0,320,200,(display-x):0,200,320,200,(display-x):320,200,320,200,(display-x)"



NOTES
=====

Visible.{x/y} passed to ggiSetMode() is ignored, because
the physical dimensions are already specified in the arguments.

All input comes from the first tile (child visual).

Long-standing bug where display garbage results from bltting to a tile
extending over a null area.

The default behavior of the tile target is to emulate a DirectBuffer
(framebuffer).  It is possible to disable it, which may result in speed ups.
Use the arguments "-nodb" or "-usedb" to select.

Non-DB mode still has some rough edges:

Optimization needed. e.g. if everything is in the same visual than just do
a ggiPutBox rather than lots of ggiPutHLines.  Another is ggiCopyBox.

An unavoidable limitation is that the ggiGet* calls will be very slow
because it has to poll every tile for the data rather than just using the
backstore.