On Sun, 19 Jul 2020 09:35:15 GMT, Laurent Bourgès <github.com+5607073+bourg...@openjdk.org> wrote:
>> I think you are conflating the rasterization step with the rendering step. >> In Prism the rasterization mainly includes >> clipping the shape and generating the mask data for the filled and/or >> stroked shape (the latter possibly being a wide >> and/or dashed stroke). The rasterized shape is then sent to the Prism >> GraphicsPipeline for drawing into the render >> target. This latter step is either done using shaders + D3D or OpenGL, or >> via the SW pipeline (or the J2D pipeline for >> printing). In the case of the SW pipeline, the renderer is based on code >> that was originally derived from Pisces, but >> it is distinct from the rasterization step that OpenPisces used to perform >> and Marlin now does perform. Take a look at >> the [SW >> pipeline](https://github.com/openjdk/jfx/tree/master/modules/javafx.graphics/src/main/java/com/sun/prism/sw) >> classes, the [Java >> PiscesRendering](https://github.com/openjdk/jfx/tree/master/modules/javafx.graphics/src/main/java/com/sun/pisces) >> classes, and the >> [prism-sw](https://github.com/openjdk/jfx/tree/master/modules/javafx.graphics/src/main/native-prism-sw) >> native code. > > @kevinrushforth I asked skara JBS to associate my github account with my > openjdk role... > I will approve your PR once I have proper credentials. > > Good job, 10k LOC removed. > > I wonder if I should also deprecate the Float-precision Marlin renderer to > remove 10k lines too. > The Double-precision Marlin renderer was enabled as the default renderer > since JDK10, so it is doing the job properly, > no reason to have 2 variants anymore and it will ease the maintenance. > Laurent That's a good question about removing the single-precision float Marlin rasterizer. For desktop platforms, I see no reason to keep it. Perhaps Johan could weigh in as to whether their might be value in keeping it for some embedded platforms. ------------- PR: https://git.openjdk.java.net/jfx/pull/268