Hi and welcome to gsl-dev, Clemens Eisserer wrote: > I am the student who submitted the proposal about implementing > subpixel-font-smoothing in OpenOffice's font rendering system.
though Philipp already wrote a good reply to your questions I have some additional comments. > I am not very familiar with OpenOffice's source-code, what could be > specific challenges? One important challenge for a new feature is not to become too dependent on specific system libraries or a specific platform environment. Regressions on other platforms should be avoided at all cost. E.g. though an X11 roundtrip may be acceptable for a local X11 display the latencies through network protocols, cards and switches can sum up... > I thought about it and this are the challenges I would see: > > - Problem with XOR painting (selection, cursor, ...). Currently even > AcrobatReader suffers from this, and selected text suffers from > colored pixels on the glyph outlines. > Now that XRender is widely supported (and hw implementations are very > slowly arriving now, maybe selection could be done using painting with > apha? The VCL infrastructure is already there (vcl/unx/source/gdi/salgdi2.cxx) and at least Calc used it. It has been reverted though, because not all relevant platforms support a round-trip free implementation and are thus slowed down by using transparant selection. > - Performance: > Subpixel-smoothing can't be done on hardware, however XRender has as > far as I know an API which would allow accaleration by the driver. No > idea a about windows, I am just afraid rendering to vram-surfaces and > reading back could be too inefficient :-/ On windows GDI already takes care of this if the option is enabled in the display settings. AFAIK almost all current Xrender implementations are not hardware accelerated. Even then it allows an implementation without X11 roundtrip though, which is quite nice. > - Maybe OpenGL accalerated rendering: > Maybe we could hook the complete drawing on top of some abstraction > layer (cairo or Java2d-native-backends when freed). My favourite would > be really Java2D, I hope it'll get some useful native C-bindings (not > going the native->jni->jvm->jni->native way). > As far as I know this stuff is already implemented there very > efficiently - and we could support several backends like OpenGL, D3D, > ... Have a look at the Canvas infrastructure, which is already used in Impress. > However I guess even the bindings would be enough for a seperate SoC > proposal ;) ;-) -- Herbert --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
