ShuriZma
/
desmume
mirror of https://github.com/TASEmulators/desmume.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

GPU: Fix the clear color on big-endian systems when running RGB666 and RGB888. 

- Also endian swap the BGnX and BGnY values on big-endian systems. This is a non-functional change, and is only meant to show that the endian swaps are indeed the correct choice for big-endian.
This commit is contained in:
rogerman 2017-10-17 18:49:05 -07:00
parent f5ead86d89
commit 43e740c661
1 changed files with 15 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

26
desmume/src/GPU.cpp
View File

@ -1617,11 +1617,11 @@ void GPUEngineBase::_RenderLine_Clear(GPUEngineCompositorInfo &compInfo)
break;
case NDSColorFormat_BGR666_Rev:
memset_u32_fast<GPU_FRAMEBUFFER_NATIVE_WIDTH>(*compInfo.target.lineColor, COLOR555TO666(dstClearColor16));
memset_u32_fast<GPU_FRAMEBUFFER_NATIVE_WIDTH>(*compInfo.target.lineColor, COLOR555TO666(LOCAL_TO_LE_16(dstClearColor16)));
break;
case NDSColorFormat_BGR888_Rev:
memset_u32_fast<GPU_FRAMEBUFFER_NATIVE_WIDTH>(*compInfo.target.lineColor, COLOR555TO888(dstClearColor16));
memset_u32_fast<GPU_FRAMEBUFFER_NATIVE_WIDTH>(*compInfo.target.lineColor, COLOR555TO888(LOCAL_TO_LE_16(dstClearColor16)));
break;
}
@ -2927,17 +2927,21 @@ void GPUEngineBase::_RenderPixelIterate_Final(GPUEngineCompositorInfo &compInfo,
const s32 wmask = wh - 1;
const s32 hmask = ht - 1;
IOREG_BGnX x = param.BGnX;
IOREG_BGnY y = param.BGnY;
IOREG_BGnX x;
IOREG_BGnY y;
x.value = LOCAL_TO_LE_32(param.BGnX.value);
y.value = LOCAL_TO_LE_32(param.BGnY.value);
#ifdef MSB_FIRST
// This only seems to work in the unrotated/unscaled case. I'm not too sure
// about how these bits should really be arranged on big-endian, but at
// least this arrangement fixes a bunch of games that use affine or extended
// layers, just as long as they don't perform any rotation/scaling.
// - rogerman, 2016-07-05
x.value = ((x.value & 0x00FFFFFF) << 8) | ((x.value & 0xFF000000) >> 24);
y.value = ((y.value & 0x00FFFFFF) << 8) | ((y.value & 0xFF000000) >> 24);
// This only seems to work in the unrotated/unscaled case.
//
// All the working values should be correct on big-endian, but there is something else wrong going
// on somewhere else, but that remains a mystery at this time. In the meantime, this hack will have
// to remain in order to fix a bunch of games that use affine or extended layers, just as long as
// they don't perform any rotation/scaling.
// - rogerman, 2017-10-17
x.value = ((x.value & 0xFF000000) >> 16) | (x.value & 0x00FF00FF) | ((x.value & 0x0000FF00) << 16);
y.value = ((y.value & 0xFF000000) >> 16) | (y.value & 0x00FF00FF) | ((y.value & 0x0000FF00) << 16);
#endif
u8 index;