transforms so we can avoid doing the expensive
64-bit math. Unroll the bilinear interpolation loops for an extra boost. I
tested this with the up/downscaling cairo-benchmarks with Xvfb and saw
a 12% +/- 4% decrease in time taken to run them.
Add XSERV_t, TRANS_SERVER, TRANS_REOPEN to quash warnings.
Add #include <dix-config.h> or <xorg-config.h>, as appropriate, to all
source files in the xserver/xorg tree, predicated on defines of
HAVE_{DIX,XORG}_CONFIG_H. Change all Xfont includes to
<X11/fonts/foo.h>.
Changed the semantics of the Convolution filter a bit. It now doesn't try
to normalize the filter values but leaves this to the client. This
gives more reasonable behaviour in the limit where the filter
parameters sum up to 0.
Add MMX implementation of non-repeating source IN repeating mask, aka
"translucent window".
Add MMX implementation of CopyArea.
Use MMX implementation of CopyArea.
Use the new implementations.
- fbCombineSaturate was pointed at fbCombineDisjointOver, instead of
fbCombineDisjointOverReverse as it should. Instead, point
fbCombineDisjointOverReverse at fbCombineSaturate (which is likely to
be faster).
- fix previously-unused fbCombineSaturate implementation.
- fbCombineMaskAlphaC was just a copy of fbCombineMaskValueC. Make it do
what it's supposed to (return a cs.alpha).
- fbCombineAtopC didn't invert the source alpha value.
- fix copy'n'paste errors in fbCombine(Dis/Con)jointGeneralC, also source
alpha wasn't treated in a component fashion.
- fbCompositeSrc_8888* didn't handle when the source lacks an alpha
channel. Rather than adding that and possilby slowing down the (normal)
alpha case, don't let x8r8g8b8/x8b8g8r8 Pictures be used in
fbCompositeSrc_8888* because Over with one of these is just Src.