aboutsummaryrefslogtreecommitdiff
path: root/graphics/scaler
AgeCommit message (Collapse)Author
2009-05-28Tweak the hq2x/hq3x a little bit (interpolation code is little bit shorter, ↵Max Horn
faster and easier to read... but 70+% of the time is spent on computing diffYUV etc. anyway, so I guess this a bit pointless... whatever... ;) svn-id: r40964
2009-02-15Attempt to fix compilation problems on MSVC 7 (I wonder why it compiled ↵Max Horn
elsewhere, actually ;) svn-id: r38314
2009-01-27Fixed hq2x ASM scalerMax Horn
svn-id: r36115
2009-01-27Re-enabled ASM versions of HQx scalers (oops)Max Horn
svn-id: r36099
2009-01-27Added new interpolate16_1_1_1_1 func, got rid of interpolate32_1_1_1_1Max Horn
svn-id: r36090
2009-01-27Replaced small lookup table in 2xsai by direct computationMax Horn
svn-id: r36089
2009-01-27Removed interpolate16_2 and interpolate16_3Max Horn
svn-id: r36088
2009-01-27Speed up HQ2x/HQ3x C++ version a bit (about 20% faster with -O3 on my system)Max Horn
svn-id: r36087
2009-01-27New code for interpolate32_1_1 overflowed in 565 mode, so reverting to the ↵Max Horn
old code (with some variable name changes to match the other interpolators) svn-id: r36086
2009-01-26Removed use of LUT16to32 in HQx asm versions, replacing some MMX code with ↵Max Horn
'plain' x86 code. Advantage: got rid of a 256kb table (reduces cache load, so over here the code is about as fast as before; in particular, since the affected interpolators are not used that often, it seems). Moreover, the new code is more accurate than the old ASM code, which actually differed from what our C++ HQx did (sacrificing precision for speed, i.e., cheating ;-) svn-id: r36078
2009-01-26Slightly improved interpolate32_1_1Max Horn
svn-id: r36075
2009-01-25Renamed lowBits -> kLowBitsMask and highBits -> kHighBitsMaskMax Horn
svn-id: r36049
2009-01-24HQX: Removed some more branches (like in my previous commit) from ↵Max Horn
interpolation macros (they are meant to speed up things, but in reality cause a slowdown -- branches are expensive on modern CPUs) svn-id: r36047
2009-01-24Turned two vars of the HQ2x/HQ3x ASM implementation into global vars, to ↵Max Horn
make it possible to adjust them for 555 vs. 565 mode (555 mode is still a bit buggy, due to the interpolation code they use) svn-id: r36046
2009-01-24Moved 565 mode specific bits of HQ2x/HQ3x asm to 'variables'Max Horn
svn-id: r36034
2009-01-23Remark that the HQnx ASM code only works in 565 modeMax Horn
svn-id: r36025
2009-01-23Added some comments to hq2x/hq3x asm codeMax Horn
svn-id: r36024
2009-01-22Speed up HQ2x/HQ3x C versions by reordering tests in diffYUV: most likeliest ↵Max Horn
to be taken now is first, least likely is last svn-id: r35996
2009-01-22Got rid of gBitFormat in thumbnail scaler code (and even simplified it in ↵Max Horn
the process) svn-id: r35994
2008-11-29Remove warnings from NDS build.Robin Watts
svn-id: r35172
2008-11-03Committed my patch #2216641 "GRAPHICS: PixelFormat introduction".Johannes Schickel
svn-id: r34875
2008-08-20Renamed graphics/scaler/thumbnail.cpp to thumbnail_intern.cpp. Fixes ↵Kari Salminen
compiling under MSVC. svn-id: r34063
2008-08-20Committed patch #2050337 "KYRA/SCUMM: Thumbnail support/improvement". ↵Johannes Schickel
(Without Max' compressed backward seeking support for now) svn-id: r34053
2008-08-05Patch #2014663: Do not mark stack as executable in HQx scalerMax Horn
svn-id: r33645
2008-03-23Patch #1923547: enable nasm usage on OS/2Jordi Vilalta Prat
svn-id: r31226
2008-02-05Committed my own patch #1885901 to fix NASM warnings.Torbjörn Andersson
svn-id: r30804
2008-01-27Removed trailing spaces.Jordi Vilalta Prat
svn-id: r30664
2007-09-19Merged common/stdafx.h into common/scummsys.h. All referencing files have ↵Nicola Mettifogo
been updated. svn-id: r28966
2007-08-11make use of RGBToColor<ColorMasks> > template functionMax Horn
svn-id: r28516
2007-06-19Implemented the OSystem framebuffer API, as discussed on scummvm-devel. All ↵Max Horn
changes are just fine, and won't cause any compile problems or regressions, despite the fact that I can't test most of the non-SDL backend changes, at an improbability level of two to the power of two hundred and seventy-six thousand to one against - possibly much higher. Anything you still can't cope with is therefore your own problem. Please relax. svn-id: r27548
2007-05-30Updated legal headers in source files, based on what Pidgin (the IM client ↵Max Horn
formerly knowns as Gaim) does; added new (incomplete) COPYRIGHT file; updated copyright dates in a few spots svn-id: r27024
2006-04-21Implementation and declartion in intern.h differed. Changed so both is using ↵Lars Persson
the same datatype. svn-id: r22074
2006-04-19Fix ASM HQ routines, because RGBtoYUV and LUT16to32 are on heap now.Eugene Sandulenko
svn-id: r22041
2006-04-17Make use of triple-interpolation function in SuperEagel (should save a few ↵Max Horn
cycles... not that it really matters :) svn-id: r21985
2006-04-17Replace interpolate32 calls by interpolate16 calls (as we only work on ↵Max Horn
single pixels here currently) svn-id: r21983
2006-04-17Simple optimizationMax Horn
svn-id: r21982
2006-04-17Renamed and merge some of the pixel interpolation functions (the names still ↵Max Horn
could be better) svn-id: r21981
2006-04-17Moved ColorMasks to their own new header graphics/colormasks.h, so that both ↵Max Horn
the scalers and the GUI code can access them. Also added (untested) entries for 1555, 444, 888 and 8888 modes svn-id: r21966
2006-04-17Removed my Altivec version of the hqx blitters, since it (a) didn't work ↵Max Horn
with newer GCC versions, (b) speed was improved on some systems, but actually slower on others, and (c) even on my old 400 Mhz G4 hqx is fast enough w/o it svn-id: r21961
2006-04-04Don't use HAVE_CONFIG_H / config.h directly, use scummsys.h insteadMax Horn
svn-id: r21605
2006-03-15* 'make bundle' now copies the default theme files into the bundle.Oystein Eftevaag
* Common::File will now look inside the application bundle on MacOS X for a file if it can't find it anywhere else (so it'll find default-theme.ini and .zip) * Cleaned up some #ifs. svn-id: r21308
2006-03-14Adds support for building on Intel Macs, using a modifiedOystein Eftevaag
version of patch 1448395 by Aubin and some of my own changes. Notably: * Two elements of a struct in backends/midi/quicktime.cpp needed to be forced to BE. * NASM set to build Mach-O binaries * configure now checks for Altivec support on Darwin hosts, instead of assuming it. Altivec support can also be disabled (not 100% sure if t his is actually needed, but might be handy for crosscompiling). * The aliasing of the symbols contained in the .asm files done on Windows were also needed on MacOS. Note that while the asm code is properly compiled and linked in, HQ2/3x support will still not work properly on Intel macs if NASM is enabled. svn-id: r21280
2006-02-20Fixing up my previous commit: finished moving common/scaler* to graphicsMax Horn
svn-id: r20798
2006-02-20Moved common/scaler* to graphics/Max Horn
svn-id: r20797