Hi, I send a series of patches for NEON optimizations of bilinear scaling. It's based on latest snapshot of pixman master repository.
First patch defines a new common header macro for bilinear a8 mask scanline function. In second patch, I implemented some bilinear scanline functions in a new file pixman-arm-neon-asm-bilinear.S with modified Makefile. Last patch just enables additional fast paths into pixman_implementation_t. I'm sorry about the test failure in previous patch. I used mask format symbol 'x' to indicate that there's no mask instead of 'n' which generally means 'solid mask'. So, in this convention, over_8888_x_8888 add_8888_x_8888 src_8888_8_8888 over_8888_8_8888 add_8888_8_8888 these functions are implemented in the new file. It passes pixman scaling test now. Any kind of review would be appreciated. -- Best Regards, Taekyun Kim
_______________________________________________ Pixman mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/pixman
