"small" cells don't show cellname, "empty" cells always show it

[stefanottili]

How to optimize display spreed when drawing lots of "empty" cells.
This issue showed up debugging. This is not a "normal" use case.

Using "hide all layers", "save as" "visible layers only", one can create an oas with lot's of "empty cells".
E.g. the greyhound gds.

For "small cells" the cellname doesn't get displayed, but for the even smaller (bbox 0, 0, 0, 0) "empty cells" the cellname is always drawn.
This can lead to a slow display at "level 1" because a lot of cellnames to be drawn.

[klayoutmatthias]

For text drawing there is an optimization that quads are skipped when a pixel is set already, covering the whole quad. So effectively there is a density limit below which no texts are drawn. I will try the same on empty cells.

Matthias

[klayoutmatthias]

That wasn't as effective as thought. Greyhound has a stupid density of cells at top level which is too low for pixelwise compaction, but too high for good drawing performance. The best option I found is to introduce a new setting - the effective dimension of an empty cell. The empty cell is thought as a square cell of that dimension for the purpose of scaling the font. With the "drop small cells" option and with suppressing drawing of small glyphs for scalable font, the drawing becomes reasonably fast.

[klayoutmatthias]

The PR introduces the "empty cell dimension" parameter. By default it's set to 4µm which appears to be a good value for Greyhound. At level 1, all empty cells draw in less than a second.

[stefanottili]

Works, thanks.