[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Gnash-dev] Any ideas about drawing character or shape in a bitmap cache

From: Wei Cao
Subject: [Gnash-dev] Any ideas about drawing character or shape in a bitmap cache instead of rendering in agg directly?
Date: Wed, 6 May 2009 01:55:13 +0800

Hi, everyone
Recently I am porting gnash to a 600MHz ARM like machine based on SDL+AGG. The resulting sdl-gnash palyer could run correctly, but with poor performance( cpu 100% even playing a simple swf file, e.g. an 400x200 advertisement ).
I read gnash's source, and added a setInvalidatedRegions call in SDLGui, to set invalidated regions in the renderer after movie_root::advance(), before movie_root::display(). That helps a lot, but performance is still not satisfied.
Compiling sdl-gnash staticly with -pg shows: generate_span and agg::render_scanlines_compound_layered cost most of time. So I wonder whether there is way to forbid rendering via agg pipelines (so costly , vertex generating, coordinate conversions, rasterizing, span generatering, and then span blending ) every time when drawing a bitmap/gradient filled shape that has never been changed during the whole lifetime. I guess the situation is common, for example, if the flash uses a picture of snow field as its background, and there are snowflakes fly in the sky, the bitmap background would be rendered from time to time because it always has intersection with the invalidated regions.
So, my idea is add a bitmap cache for each character( or bitmap/gradient filled shape only ), when a character is displayed for the first time, agg renderer will draw in its bitmap cache, then next time the character is showed, agg renderer could simply read from cache and do a blit. The cache is cleaned only when the character is rotated or scaled,
I notice one of  the developpers mensioned nearly the same idea in comments of render_handler_agg.cpp . Is the method feasible? Anyone did such a work in a branch?
BTW, I am also considering using SIMD instructions to accelerate agg's blending operation, any suggestion?
Wei Cao

reply via email to

[Prev in Thread] Current Thread [Next in Thread]