history-museum/winamp
1
0
Fork 0
mirror of https://github.com/WinampDesktop/winamp.git synced 2025-06-17 11:15:46 -04:00
Code Issues Actions Packages Projects Releases Wiki Activity

GPU/HW: Clamp interpolated UVs to polygon limits

Browse Source 
Fixes texture filtering and PGXP issues in some games.
This commit is contained in:
Connor McLaughlin
2020-08-10 22:37:07 +10:00
parent f14dc6de27
commit b95ce993e0
7 changed files with 130 additions and 58 deletions
Download Patch File Download Diff File Expand all files Collapse all files

91
src/core/gpu_hw_shadergen.cpp
View File

@ -508,8 +508,9 @@ std::string GPU_HW_ShaderGen::GenerateBatchVertexShader(bool textured, bool upsc
const char* output_block_suffix = upscaled_lines ? "VS" : "";
if (textured)
{
DeclareVertexEntryPoint(ss, {"float4 a_pos", "float4 a_col0", "uint a_texcoord", "uint a_texpage"}, 1, 1,
{{"nointerpolation", "uint4 v_texpage"}}, false, output_block_suffix);
DeclareVertexEntryPoint(
ss, {"float4 a_pos", "float4 a_col0", "uint a_texcoord", "uint a_texpage", "float4 a_uv_limits"}, 1, 1,
{{"nointerpolation", "uint4 v_texpage"}, {"nointerpolation", "float4 v_uv_limits"}}, false, output_block_suffix);
}
else
{
@ -557,6 +558,8 @@ std::string GPU_HW_ShaderGen::GenerateBatchVertexShader(bool textured, bool upsc
v_texpage.y = ((a_texpage >> 4) & 1u) * 256u * RESOLUTION_SCALE;
v_texpage.z = ((a_texpage >> 16) & 63u) * 16u * RESOLUTION_SCALE;
v_texpage.w = ((a_texpage >> 22) & 511u) * RESOLUTION_SCALE;
v_uv_limits = a_uv_limits * float4(255.0, 255.0, 255.0, 255.0);
#endif
}
)";
@ -658,13 +661,7 @@ uint2 FloatToIntegerCoords(float2 coords)
float4 SampleFromVRAM(uint4 texpage, float2 coords)
{
#if PALETTE
// We can't currently use upscaled coordinate for palettes because of how they're packed.
// Not that it would be any benefit anyway, render-to-texture effects don't use palettes.
#if !TEXTURE_FILTERING
coords /= float2(RESOLUTION_SCALE, RESOLUTION_SCALE);
#endif
uint2 icoord = ApplyTextureWindow(FloatToIntegerCoords(coords));
uint2 index_coord = icoord;
#if PALETTE_4_BIT
index_coord.x /= 4u;
@ -698,12 +695,43 @@ float4 SampleFromVRAM(uint4 texpage, float2 coords)
return SAMPLE_TEXTURE(samp0, float2(direct_icoord) * RCP_VRAM_SIZE);
#endif
}
void BilinearSampleFromVRAM(uint4 texpage, float2 coords, float4 uv_limits,
out float4 texcol, out float ialpha)
{
// Compute the coordinates of the four texels we will be interpolating between.
// Clamp this to the triangle texture coordinates.
float2 texel_top_left = frac(coords) - float2(0.5, 0.5);
float2 texel_offset = sign(texel_top_left);
float4 fcoords = max(coords.xyxy + float4(0.0, 0.0, texel_offset.x, texel_offset.y),
float4(0.0, 0.0, 0.0, 0.0));
// Load four texels.
float4 s00 = SampleFromVRAM(texpage, clamp(fcoords.xy, uv_limits.xy, uv_limits.zw));
float4 s10 = SampleFromVRAM(texpage, clamp(fcoords.zy, uv_limits.xy, uv_limits.zw));
float4 s01 = SampleFromVRAM(texpage, clamp(fcoords.xw, uv_limits.xy, uv_limits.zw));
float4 s11 = SampleFromVRAM(texpage, clamp(fcoords.zw, uv_limits.xy, uv_limits.zw));
// Compute alpha from how many texels aren't pixel color 0000h.
float a00 = float(VECTOR_NEQ(s00, TRANSPARENT_PIXEL_COLOR));
float a10 = float(VECTOR_NEQ(s10, TRANSPARENT_PIXEL_COLOR));
float a01 = float(VECTOR_NEQ(s01, TRANSPARENT_PIXEL_COLOR));
float a11 = float(VECTOR_NEQ(s11, TRANSPARENT_PIXEL_COLOR));
// Bilinearly interpolate.
float2 weights = abs(texel_top_left);
texcol = lerp(lerp(s00, s10, weights.x), lerp(s01, s11, weights.x), weights.y);
ialpha = lerp(lerp(a00, a10, weights.x), lerp(a01, a11, weights.x), weights.y);
}
#endif
)";
if (textured)
{
DeclareFragmentEntryPoint(ss, 1, 1, {{"nointerpolation", "uint4 v_texpage"}}, true, use_dual_source ? 2 : 1, true);
DeclareFragmentEntryPoint(ss, 1, 1,
{{"nointerpolation", "uint4 v_texpage"}, {"nointerpolation", "float4 v_uv_limits"}}, true,
use_dual_source ? 2 : 1, true);
}
else
{
@ -725,48 +753,35 @@ float4 SampleFromVRAM(uint4 texpage, float2 coords)
#endif
#if TEXTURED
float2 coords = v_tex0;
float4 uv_limits = v_uv_limits;
float4 texcol;
// We can't currently use upscaled coordinate for palettes because of how they're packed.
// Not that it would be any benefit anyway, render-to-texture effects don't use palettes.
#if PALETTE
coords /= float2(RESOLUTION_SCALE, RESOLUTION_SCALE);
#else
uv_limits *= float4(RESOLUTION_SCALE, RESOLUTION_SCALE, RESOLUTION_SCALE, RESOLUTION_SCALE);
#endif
#if TEXTURE_FILTERING
// Compute the coordinates of the four texels we will be interpolating between.
// TODO: Find some way to clamp this to the triangle texture coordinates?
float2 downscaled_coords = v_tex0;
#if PALETTE
downscaled_coords /= float2(RESOLUTION_SCALE, RESOLUTION_SCALE);
#endif
float2 texel_top_left = frac(downscaled_coords) - float2(0.5, 0.5);
float2 texel_offset = sign(texel_top_left);
float4 fcoords = max(downscaled_coords.xyxy + float4(0.0, 0.0, texel_offset.x, texel_offset.y),
float4(0.0, 0.0, 0.0, 0.0));
// Load four texels.
float4 s00 = SampleFromVRAM(v_texpage, fcoords.xy);
float4 s10 = SampleFromVRAM(v_texpage, fcoords.zy);
float4 s01 = SampleFromVRAM(v_texpage, fcoords.xw);
float4 s11 = SampleFromVRAM(v_texpage, fcoords.zw);
// Compute alpha from how many texels aren't pixel color 0000h.
float a00 = float(VECTOR_NEQ(s00, TRANSPARENT_PIXEL_COLOR));
float a10 = float(VECTOR_NEQ(s10, TRANSPARENT_PIXEL_COLOR));
float a01 = float(VECTOR_NEQ(s01, TRANSPARENT_PIXEL_COLOR));
float a11 = float(VECTOR_NEQ(s11, TRANSPARENT_PIXEL_COLOR));
// Bilinearly interpolate.
float2 weights = abs(texel_top_left);
float4 texcol = lerp(lerp(s00, s10, weights.x), lerp(s01, s11, weights.x), weights.y);
ialpha = lerp(lerp(a00, a10, weights.x), lerp(a01, a11, weights.x), weights.y);
BilinearSampleFromVRAM(v_texpage, coords, uv_limits, texcol, ialpha);
if (ialpha < 0.5)
discard;
texcol.rgb /= float3(ialpha, ialpha, ialpha);
semitransparent = (texcol.a != 0.0);
#else
float4 texcol = SampleFromVRAM(v_texpage, v_tex0);
texcol = SampleFromVRAM(v_texpage, clamp(coords, uv_limits.xy, uv_limits.zw));
if (VECTOR_EQ(texcol, TRANSPARENT_PIXEL_COLOR))
discard;
semitransparent = (texcol.a != 0.0);
ialpha = 1.0;
#endif
semitransparent = (texcol.a != 0.0);
// If not using true color, truncate the framebuffer colors to 5-bit.
#if !TRUE_COLOR
icolor = uint3(texcol.rgb * float3(255.0, 255.0, 255.0)) >> 3;