sgilissen/llamaplayer
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

GPU: Implement weave deinterlacing

Browse Source
This commit is contained in:
Connor McLaughlin
2019-10-23 15:36:37 +10:00
parent 87f9f99938
commit 2d0dd03705
5 changed files with 101 additions and 74 deletions
Download Patch File Download Diff File Expand all files Collapse all files

99
src/core/gpu_hw.cpp
View File

@ -406,61 +406,80 @@ void main()
return ss.str();
}
std::string GPU_HW::GenerateRGB24DecodeFragmentShader()
std::string GPU_HW::GenerateDisplayFragmentShader(bool depth_24bit, bool interlaced)
{
std::stringstream ss;
GenerateShaderHeader(ss);
DefineMacro(ss, "DEPTH_24BIT", depth_24bit);
DefineMacro(ss, "INTERLACED", interlaced);
ss << R"(
in vec2 v_tex0;
out vec4 o_col0;
uniform sampler2D samp0;
uniform ivec2 u_base_coords;
uniform ivec3 u_base_coords;
ivec2 GetCoords(vec2 fragcoord)
{
ivec2 icoords = ivec2(fragcoord);
#if INTERLACED
if (((icoords.y - u_base_coords.z) & 1) != 0)
discard;
#endif
return icoords;
}
void main()
{
// compute offset in dwords from the start of the 24-bit values
ivec2 base = ivec2(u_base_coords.x, u_base_coords.y + int(gl_FragCoord.y));
int xoff = int(gl_FragCoord.x);
int dword_index = (xoff / 2) + (xoff / 4);
ivec2 icoords = GetCoords(gl_FragCoord.xy);
// sample two adjacent dwords, or four 16-bit values as the 24-bit value will lie somewhere between these
uint s0 = RGBA8ToRGBA5551(texelFetch(samp0, ivec2(base.x + dword_index * 2 + 0, base.y), 0));
uint s1 = RGBA8ToRGBA5551(texelFetch(samp0, ivec2(base.x + dword_index * 2 + 1, base.y), 0));
uint s2 = RGBA8ToRGBA5551(texelFetch(samp0, ivec2(base.x + (dword_index + 1) * 2 + 0, base.y), 0));
uint s3 = RGBA8ToRGBA5551(texelFetch(samp0, ivec2(base.x + (dword_index + 1) * 2 + 1, base.y), 0));
#if DEPTH_24BIT
// compute offset in dwords from the start of the 24-bit values
ivec2 base = ivec2(u_base_coords.x, u_base_coords.y + icoords.y);
int xoff = int(icoords.x);
int dword_index = (xoff / 2) + (xoff / 4);
// select the bit for this pixel depending on its offset in the 4-pixel block
uint r, g, b;
int block_offset = xoff & 3;
if (block_offset == 0)
{
r = s0 & 0xFFu;
g = s0 >> 8;
b = s1 & 0xFFu;
}
else if (block_offset == 1)
{
r = s1 >> 8;
g = s2 & 0xFFu;
b = s2 >> 8;
}
else if (block_offset == 2)
{
r = s1 & 0xFFu;
g = s1 >> 8;
b = s2 & 0xFFu;
}
else
{
r = s2 >> 8;
g = s3 & 0xFFu;
b = s3 >> 8;
}
// sample two adjacent dwords, or four 16-bit values as the 24-bit value will lie somewhere between these
uint s0 = RGBA8ToRGBA5551(texelFetch(samp0, ivec2(base.x + dword_index * 2 + 0, base.y), 0));
uint s1 = RGBA8ToRGBA5551(texelFetch(samp0, ivec2(base.x + dword_index * 2 + 1, base.y), 0));
uint s2 = RGBA8ToRGBA5551(texelFetch(samp0, ivec2(base.x + (dword_index + 1) * 2 + 0, base.y), 0));
uint s3 = RGBA8ToRGBA5551(texelFetch(samp0, ivec2(base.x + (dword_index + 1) * 2 + 1, base.y), 0));
// and normalize
o_col0 = vec4(float(r) / 255.0, float(g) / 255.0, float(b) / 255, 1.0);
// select the bit for this pixel depending on its offset in the 4-pixel block
uint r, g, b;
int block_offset = xoff & 3;
if (block_offset == 0)
{
r = s0 & 0xFFu;
g = s0 >> 8;
b = s1 & 0xFFu;
}
else if (block_offset == 1)
{
r = s1 >> 8;
g = s2 & 0xFFu;
b = s2 >> 8;
}
else if (block_offset == 2)
{
r = s1 & 0xFFu;
g = s1 >> 8;
b = s2 & 0xFFu;
}
else
{
r = s2 >> 8;
g = s3 & 0xFFu;
b = s3 >> 8;
}
// and normalize
o_col0 = vec4(float(r) / 255.0, float(g) / 255.0, float(b) / 255, 1.0);
#else
// load and return
o_col0 = texelFetch(samp0, u_base_coords.xy + icoords, 0);
#endif
}
)";