17 Commits

Author SHA1 Message Date
Max Horn
7db5cc0a9c While moving things down by one line 'looks better' (IMO), it also means that the last line is blitted outside the screen -> bad. So for now I'll undo my changes to the 2xSai filter family
svn-id: r13257
2004-03-13 15:45:39 +00:00
Max Horn
87e0e57567 the 2xsai class scalers all move the resulting image one up; work around that
svn-id: r13071
2004-02-27 19:36:52 +00:00
Max Horn
033ff78ae9 updated copyright notice
svn-id: r12176
2004-01-06 12:45:34 +00:00
Max Horn
a12b6606cc rewrote hq2x/hq3x code so that we can compile both the altivec and plain C version; on OSX, this will only use the altivec code if altivec is actually present
svn-id: r11504
2003-12-05 23:27:01 +00:00
Paweł Kołodziejski
b5a7ef0822 cleanup whitespaces
svn-id: r11219
2003-11-08 22:43:46 +00:00
Max Horn
ba1cf0e957 cut down dependencies a bit (so that I don't have to recompile hq2x/hq3x whenever common/scaler.h is changed :-)
svn-id: r11168
2003-11-06 23:54:12 +00:00
Max Horn
a02c13a52e fix scaler compilation for MSVC6. However, the produced code stll seems to be incorrect... Jamieson, maybe the MSVC6_COMPAT should be re-added then after all?
svn-id: r10552
2003-10-02 23:29:58 +00:00
Max Horn
bb5351f0a8 according to tsuteiuQ, this helps MSVC6 (seems it can cope with the enums but not with static consts)
svn-id: r10551
2003-10-02 23:09:54 +00:00
Jamieson Christian
8fbcb4ac15 Added MSVC6_COMPAT in order to be able to disable
new additions/optimizations that use constructs with
which MSVC6 is incompatible, such as the revamped
template-based super/hq/tv scalers.

svn-id: r10540
2003-10-02 15:00:36 +00:00
Max Horn
80eb1f8cb9 sync with hq2x (in particular, use my Altivec code here, too
svn-id: r10529
2003-10-01 20:59:05 +00:00
Max Horn
57eb9654b7 make MSVC7 happy
svn-id: r10524
2003-10-01 17:06:36 +00:00
Max Horn
628f65b639 move INTERPOLATE / Q_INTERPOLATE to intern.h; remove some jumps (pipelin trashers I call 'em :-) from diffYUV
svn-id: r10523
2003-10-01 16:47:49 +00:00
Max Horn
b6f85de558 instead of using a stack array, use regular variables (marked with the register keyword to hint the optimizer) for the color data. At least over here this helps the optimizer *A LOT* - instead of keeping w[] on the stack, on the PowerPC all values stay in registers, for a nice speed boost. Shouldn't hurt x86 either
svn-id: r10522
2003-10-01 16:39:31 +00:00
Max Horn
2aeb5c2414 added some AltiVec code. Still under development, and notice that this is my first time writing AltiVec code, so I am sure it could be done better :-). I am working on the interpolation function now.
svn-id: r10508
2003-09-30 16:59:01 +00:00
Max Horn
7ccafa467c synced hq2x with upstream changes; some performance tweaks to hq2x/hq3x (I think for bigger speed improvement we'd have to use ASM (with C++ fallback of course)...
svn-id: r10494
2003-09-29 23:34:43 +00:00
Max Horn
c40e7d68a5 templatized more of the scalers; also introduced template struct ColorMasks (shortens/simplifies other code a bit)
svn-id: r10489
2003-09-29 18:38:51 +00:00
Max Horn
4a7385b4e7 split out scalers into separate source files in common/scaler; some optimizations (in particular, the hq2x/hq3x scalers now are compiled into two versions, one for 555 and one for 565 mode)
svn-id: r10488
2003-09-29 16:02:47 +00:00