Merge pull request #9169 from lioncash/uber-vert

UberShaderVertex: Migrate over to fmt
This commit is contained in:
Léo Lam 2020-10-22 01:21:31 +02:00 committed by GitHub
commit 89b01cd6d3
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 296 additions and 288 deletions

View File

@ -23,105 +23,105 @@ VertexShaderUid GetVertexShaderUid()
return out;
}
static void GenVertexShaderTexGens(APIType ApiType, u32 numTexgen, ShaderCode& out);
static void GenVertexShaderTexGens(APIType api_type, u32 num_texgen, ShaderCode& out);
ShaderCode GenVertexShader(APIType ApiType, const ShaderHostConfig& host_config,
ShaderCode GenVertexShader(APIType api_type, const ShaderHostConfig& host_config,
const vertex_ubershader_uid_data* uid_data)
{
const bool msaa = host_config.msaa;
const bool ssaa = host_config.ssaa;
const bool per_pixel_lighting = host_config.per_pixel_lighting;
const bool vertex_rounding = host_config.vertex_rounding;
const u32 numTexgen = uid_data->num_texgens;
const u32 num_texgen = uid_data->num_texgens;
ShaderCode out;
out.Write("// Vertex UberShader\n\n");
out.Write("%s", s_lighting_struct);
out.WriteFmt("// Vertex UberShader\n\n");
out.WriteFmt("{}", s_lighting_struct);
// uniforms
if (ApiType == APIType::OpenGL || ApiType == APIType::Vulkan)
out.Write("UBO_BINDING(std140, 2) uniform VSBlock {\n");
if (api_type == APIType::OpenGL || api_type == APIType::Vulkan)
out.WriteFmt("UBO_BINDING(std140, 2) uniform VSBlock {{\n");
else
out.Write("cbuffer VSBlock {\n");
out.Write(s_shader_uniforms);
out.Write("};\n");
out.WriteFmt("cbuffer VSBlock {{\n");
out.WriteFmt("{}", s_shader_uniforms);
out.WriteFmt("}};\n");
out.Write("struct VS_OUTPUT {\n");
GenerateVSOutputMembers(out, ApiType, numTexgen, host_config, "");
out.Write("};\n\n");
out.WriteFmt("struct VS_OUTPUT {{\n");
GenerateVSOutputMembers(out, api_type, num_texgen, host_config, "");
out.WriteFmt("}};\n\n");
WriteUberShaderCommonHeader(out, ApiType, host_config);
WriteUberShaderCommonHeader(out, api_type, host_config);
WriteLightingFunction(out);
if (ApiType == APIType::OpenGL || ApiType == APIType::Vulkan)
if (api_type == APIType::OpenGL || api_type == APIType::Vulkan)
{
out.Write("ATTRIBUTE_LOCATION(%d) in float4 rawpos;\n", SHADER_POSITION_ATTRIB);
out.Write("ATTRIBUTE_LOCATION(%d) in uint4 posmtx;\n", SHADER_POSMTX_ATTRIB);
out.Write("ATTRIBUTE_LOCATION(%d) in float3 rawnorm0;\n", SHADER_NORM0_ATTRIB);
out.Write("ATTRIBUTE_LOCATION(%d) in float3 rawnorm1;\n", SHADER_NORM1_ATTRIB);
out.Write("ATTRIBUTE_LOCATION(%d) in float3 rawnorm2;\n", SHADER_NORM2_ATTRIB);
out.Write("ATTRIBUTE_LOCATION(%d) in float4 rawcolor0;\n", SHADER_COLOR0_ATTRIB);
out.Write("ATTRIBUTE_LOCATION(%d) in float4 rawcolor1;\n", SHADER_COLOR1_ATTRIB);
out.WriteFmt("ATTRIBUTE_LOCATION({}) in float4 rawpos;\n", SHADER_POSITION_ATTRIB);
out.WriteFmt("ATTRIBUTE_LOCATION({}) in uint4 posmtx;\n", SHADER_POSMTX_ATTRIB);
out.WriteFmt("ATTRIBUTE_LOCATION({}) in float3 rawnorm0;\n", SHADER_NORM0_ATTRIB);
out.WriteFmt("ATTRIBUTE_LOCATION({}) in float3 rawnorm1;\n", SHADER_NORM1_ATTRIB);
out.WriteFmt("ATTRIBUTE_LOCATION({}) in float3 rawnorm2;\n", SHADER_NORM2_ATTRIB);
out.WriteFmt("ATTRIBUTE_LOCATION({}) in float4 rawcolor0;\n", SHADER_COLOR0_ATTRIB);
out.WriteFmt("ATTRIBUTE_LOCATION({}) in float4 rawcolor1;\n", SHADER_COLOR1_ATTRIB);
for (int i = 0; i < 8; ++i)
out.Write("ATTRIBUTE_LOCATION(%d) in float3 rawtex%d;\n", SHADER_TEXTURE0_ATTRIB + i, i);
out.WriteFmt("ATTRIBUTE_LOCATION({}) in float3 rawtex{};\n", SHADER_TEXTURE0_ATTRIB + i, i);
if (host_config.backend_geometry_shaders)
{
out.Write("VARYING_LOCATION(0) out VertexData {\n");
GenerateVSOutputMembers(out, ApiType, numTexgen, host_config,
out.WriteFmt("VARYING_LOCATION(0) out VertexData {{\n");
GenerateVSOutputMembers(out, api_type, num_texgen, host_config,
GetInterpolationQualifier(msaa, ssaa, true, false));
out.Write("} vs;\n");
out.WriteFmt("}} vs;\n");
}
else
{
// Let's set up attributes
u32 counter = 0;
out.Write("VARYING_LOCATION(%u) %s out float4 colors_0;\n", counter++,
out.WriteFmt("VARYING_LOCATION({}) {} out float4 colors_0;\n", counter++,
GetInterpolationQualifier(msaa, ssaa));
out.Write("VARYING_LOCATION(%u) %s out float4 colors_1;\n", counter++,
out.WriteFmt("VARYING_LOCATION({}) {} out float4 colors_1;\n", counter++,
GetInterpolationQualifier(msaa, ssaa));
for (u32 i = 0; i < numTexgen; ++i)
for (u32 i = 0; i < num_texgen; ++i)
{
out.Write("VARYING_LOCATION(%u) %s out float3 tex%u;\n", counter++,
out.WriteFmt("VARYING_LOCATION({}) {} out float3 tex{};\n", counter++,
GetInterpolationQualifier(msaa, ssaa), i);
}
if (!host_config.fast_depth_calc)
{
out.Write("VARYING_LOCATION(%u) %s out float4 clipPos;\n", counter++,
out.WriteFmt("VARYING_LOCATION({}) {} out float4 clipPos;\n", counter++,
GetInterpolationQualifier(msaa, ssaa));
}
if (per_pixel_lighting)
{
out.Write("VARYING_LOCATION(%u) %s out float3 Normal;\n", counter++,
out.WriteFmt("VARYING_LOCATION({}) {} out float3 Normal;\n", counter++,
GetInterpolationQualifier(msaa, ssaa));
out.Write("VARYING_LOCATION(%u) %s out float3 WorldPos;\n", counter++,
out.WriteFmt("VARYING_LOCATION({}) {} out float3 WorldPos;\n", counter++,
GetInterpolationQualifier(msaa, ssaa));
}
}
out.Write("void main()\n{\n");
out.WriteFmt("void main()\n{{\n");
}
else // D3D
{
out.Write("VS_OUTPUT main(\n");
out.WriteFmt("VS_OUTPUT main(\n");
// inputs
out.Write(" float3 rawnorm0 : NORMAL0,\n");
out.Write(" float3 rawnorm1 : NORMAL1,\n");
out.Write(" float3 rawnorm2 : NORMAL2,\n");
out.Write(" float4 rawcolor0 : COLOR0,\n");
out.Write(" float4 rawcolor1 : COLOR1,\n");
out.WriteFmt(" float3 rawnorm0 : NORMAL0,\n"
" float3 rawnorm1 : NORMAL1,\n"
" float3 rawnorm2 : NORMAL2,\n"
" float4 rawcolor0 : COLOR0,\n"
" float4 rawcolor1 : COLOR1,\n");
for (int i = 0; i < 8; ++i)
out.Write(" float3 rawtex%d : TEXCOORD%d,\n", i, i);
out.Write(" uint posmtx : BLENDINDICES,\n");
out.Write(" float4 rawpos : POSITION) {\n");
out.WriteFmt(" float3 rawtex{} : TEXCOORD{},\n", i, i);
out.WriteFmt(" uint posmtx : BLENDINDICES,\n");
out.WriteFmt(" float4 rawpos : POSITION) {{\n");
}
out.Write("VS_OUTPUT o;\n"
out.WriteFmt("VS_OUTPUT o;\n"
"\n");
// Transforms
out.Write("// Position matrix\n"
out.WriteFmt("// Position matrix\n"
"float4 P0;\n"
"float4 P1;\n"
"float4 P2;\n"
@ -131,9 +131,9 @@ ShaderCode GenVertexShader(APIType ApiType, const ShaderHostConfig& host_config,
"float3 N1;\n"
"float3 N2;\n"
"\n"
"if ((components & %uu) != 0u) {// VB_HAS_POSMTXIDX\n",
"if ((components & {}u) != 0u) {{// VB_HAS_POSMTXIDX\n",
VB_HAS_POSMTXIDX);
out.Write(" // Vertex format has a per-vertex matrix\n"
out.WriteFmt(" // Vertex format has a per-vertex matrix\n"
" int posidx = int(posmtx.r);\n"
" P0 = " I_TRANSFORMMATRICES "[posidx];\n"
" P1 = " I_TRANSFORMMATRICES "[posidx+1];\n"
@ -143,7 +143,7 @@ ShaderCode GenVertexShader(APIType ApiType, const ShaderHostConfig& host_config,
" N0 = " I_NORMALMATRICES "[normidx].xyz;\n"
" N1 = " I_NORMALMATRICES "[normidx+1].xyz;\n"
" N2 = " I_NORMALMATRICES "[normidx+2].xyz;\n"
"} else {\n"
"}} else {{\n"
" // One shared matrix\n"
" P0 = " I_POSNORMALMATRIX "[0];\n"
" P1 = " I_POSNORMALMATRIX "[1];\n"
@ -151,7 +151,7 @@ ShaderCode GenVertexShader(APIType ApiType, const ShaderHostConfig& host_config,
" N0 = " I_POSNORMALMATRIX "[3].xyz;\n"
" N1 = " I_POSNORMALMATRIX "[4].xyz;\n"
" N2 = " I_POSNORMALMATRIX "[5].xyz;\n"
"}\n"
"}}\n"
"\n"
"float4 pos = float4(dot(P0, rawpos), dot(P1, rawpos), dot(P2, rawpos), 1.0);\n"
"o.pos = float4(dot(" I_PROJECTION "[0], pos), dot(" I_PROJECTION
@ -159,57 +159,61 @@ ShaderCode GenVertexShader(APIType ApiType, const ShaderHostConfig& host_config,
"\n"
"// Only the first normal gets normalized (TODO: why?)\n"
"float3 _norm0 = float3(0.0, 0.0, 0.0);\n"
"if ((components & %uu) != 0u) // VB_HAS_NRM0\n",
"if ((components & {}u) != 0u) // VB_HAS_NRM0\n",
VB_HAS_NRM0);
out.Write(
out.WriteFmt(
" _norm0 = normalize(float3(dot(N0, rawnorm0), dot(N1, rawnorm0), dot(N2, rawnorm0)));\n"
"\n"
"float3 _norm1 = float3(0.0, 0.0, 0.0);\n"
"if ((components & %uu) != 0u) // VB_HAS_NRM1\n",
"if ((components & {}u) != 0u) // VB_HAS_NRM1\n",
VB_HAS_NRM1);
out.Write(" _norm1 = float3(dot(N0, rawnorm1), dot(N1, rawnorm1), dot(N2, rawnorm1));\n"
out.WriteFmt(" _norm1 = float3(dot(N0, rawnorm1), dot(N1, rawnorm1), dot(N2, rawnorm1));\n"
"\n"
"float3 _norm2 = float3(0.0, 0.0, 0.0);\n"
"if ((components & %uu) != 0u) // VB_HAS_NRM2\n",
"if ((components & {}u) != 0u) // VB_HAS_NRM2\n",
VB_HAS_NRM2);
out.Write(" _norm2 = float3(dot(N0, rawnorm2), dot(N1, rawnorm2), dot(N2, rawnorm2));\n"
out.WriteFmt(" _norm2 = float3(dot(N0, rawnorm2), dot(N1, rawnorm2), dot(N2, rawnorm2));\n"
"\n");
// Hardware Lighting
WriteVertexLighting(out, ApiType, "pos.xyz", "_norm0", "rawcolor0", "rawcolor1", "o.colors_0",
WriteVertexLighting(out, api_type, "pos.xyz", "_norm0", "rawcolor0", "rawcolor1", "o.colors_0",
"o.colors_1");
// Texture Coordinates
if (numTexgen > 0)
GenVertexShaderTexGens(ApiType, numTexgen, out);
if (num_texgen > 0)
GenVertexShaderTexGens(api_type, num_texgen, out);
out.Write("if (xfmem_numColorChans == 0u) {\n");
out.Write(" if ((components & %uu) != 0u)\n", VB_HAS_COL0);
out.Write(" o.colors_0 = rawcolor0;\n");
out.Write(" else\n");
out.Write(" o.colors_1 = float4(1.0, 1.0, 1.0, 1.0);\n");
out.Write("}\n");
out.Write("if (xfmem_numColorChans < 2u) {\n");
out.Write(" if ((components & %uu) != 0u)\n", VB_HAS_COL1);
out.Write(" o.colors_0 = rawcolor1;\n");
out.Write(" else\n");
out.Write(" o.colors_1 = float4(1.0, 1.0, 1.0, 1.0);\n");
out.Write("}\n");
out.WriteFmt("if (xfmem_numColorChans == 0u) {{\n"
" if ((components & {}u) != 0u)\n"
" o.colors_0 = rawcolor0;\n"
" else\n"
" o.colors_1 = float4(1.0, 1.0, 1.0, 1.0);\n"
"}}\n",
VB_HAS_COL0);
out.WriteFmt("if (xfmem_numColorChans < 2u) {{\n"
" if ((components & {}u) != 0u)\n"
" o.colors_0 = rawcolor1;\n"
" else\n"
" o.colors_1 = float4(1.0, 1.0, 1.0, 1.0);\n"
"}}\n",
VB_HAS_COL1);
if (!host_config.fast_depth_calc)
{
// clipPos/w needs to be done in pixel shader, not here
out.Write("o.clipPos = o.pos;\n");
out.WriteFmt("o.clipPos = o.pos;\n");
}
if (per_pixel_lighting)
{
out.Write("o.Normal = _norm0;\n");
out.Write("o.WorldPos = pos.xyz;\n");
out.Write("if ((components & %uu) != 0u) // VB_HAS_COL0\n", VB_HAS_COL0);
out.Write(" o.colors_0 = rawcolor0;\n");
out.Write("if ((components & %uu) != 0u) // VB_HAS_COL1\n", VB_HAS_COL1);
out.Write(" o.colors_1 = rawcolor1;\n");
out.WriteFmt("o.Normal = _norm0;\n"
"o.WorldPos = pos.xyz;\n");
out.WriteFmt("if ((components & {}u) != 0u) // VB_HAS_COL0\n"
" o.colors_0 = rawcolor0;\n",
VB_HAS_COL0);
out.WriteFmt("if ((components & {}u) != 0u) // VB_HAS_COL1\n"
" o.colors_1 = rawcolor1;\n",
VB_HAS_COL1);
}
// If we can disable the incorrect depth clipping planes using depth clamping, then we can do
@ -221,13 +225,13 @@ ShaderCode GenVertexShader(APIType ApiType, const ShaderHostConfig& host_config,
// own clipping. We want to clip so that -w <= z <= 0, which matches the console -1..0 range.
// We adjust our depth value for clipping purposes to match the perspective projection in the
// software backend, which is a hack to fix Sonic Adventure and Unleashed games.
out.Write("float clipDepth = o.pos.z * (1.0 - 1e-7);\n");
out.Write("float clipDist0 = clipDepth + o.pos.w;\n"); // Near: z < -w
out.Write("float clipDist1 = -clipDepth;\n"); // Far: z > 0
out.WriteFmt("float clipDepth = o.pos.z * (1.0 - 1e-7);\n"
"float clipDist0 = clipDepth + o.pos.w;\n" // Near: z < -w
"float clipDist1 = -clipDepth;\n"); // Far: z > 0
if (host_config.backend_geometry_shaders)
{
out.Write("o.clipDist0 = clipDist0;\n");
out.Write("o.clipDist1 = clipDist1;\n");
out.WriteFmt("o.clipDist0 = clipDist0;\n"
"o.clipDist1 = clipDist1;\n");
}
}
@ -242,7 +246,7 @@ ShaderCode GenVertexShader(APIType ApiType, const ShaderHostConfig& host_config,
// divide, because some games will use a depth range larger than what is allowed by the
// graphics API. These large depth ranges will still be clipped to the 0..1 range, so these
// games effectively add a depth bias to the values written to the depth buffer.
out.Write("o.pos.z = o.pos.w * " I_PIXELCENTERCORRECTION ".w - "
out.WriteFmt("o.pos.z = o.pos.w * " I_PIXELCENTERCORRECTION ".w - "
"o.pos.z * " I_PIXELCENTERCORRECTION ".z;\n");
if (!host_config.backend_clip_control)
@ -250,12 +254,12 @@ ShaderCode GenVertexShader(APIType ApiType, const ShaderHostConfig& host_config,
// If the graphics API doesn't support a depth range of 0..1, then we need to map z to
// the -1..1 range. Unfortunately we have to use a substraction, which is a lossy floating-point
// operation that can introduce a round-trip error.
out.Write("o.pos.z = o.pos.z * 2.0 - o.pos.w;\n");
out.WriteFmt("o.pos.z = o.pos.z * 2.0 - o.pos.w;\n");
}
// Correct for negative viewports by mirroring all vertices. We need to negate the height here,
// since the viewport height is already negated by the render backend.
out.Write("o.pos.xy *= sign(" I_PIXELCENTERCORRECTION ".xy * float2(1.0, -1.0));\n");
out.WriteFmt("o.pos.xy *= sign(" I_PIXELCENTERCORRECTION ".xy * float2(1.0, -1.0));\n");
// The console GPU places the pixel center at 7/12 in screen space unless
// antialiasing is enabled, while D3D and OpenGL place it at 0.5. This results
@ -263,7 +267,7 @@ ShaderCode GenVertexShader(APIType ApiType, const ShaderHostConfig& host_config,
// which in turn can be critical if it happens for clear quads.
// Hence, we compensate for this pixel center difference so that primitives
// get rasterized correctly.
out.Write("o.pos.xy = o.pos.xy - o.pos.w * " I_PIXELCENTERCORRECTION ".xy;\n");
out.WriteFmt("o.pos.xy = o.pos.xy - o.pos.w * " I_PIXELCENTERCORRECTION ".xy;\n");
if (vertex_rounding)
{
@ -271,219 +275,223 @@ ShaderCode GenVertexShader(APIType ApiType, const ShaderHostConfig& host_config,
// cause an additional pixel offset. Due to a higher pixel density we need to correct this
// by converting our clip-space position into the Wii's screen-space.
// Acquire the right pixel and then convert it back.
out.Write("if (o.pos.w == 1.0f)\n");
out.Write("{\n");
out.WriteFmt("if (o.pos.w == 1.0f)\n"
"{{\n");
out.Write("\tfloat ss_pixel_x = ((o.pos.x + 1.0f) * (" I_VIEWPORT_SIZE ".x * 0.5f));\n");
out.Write("\tfloat ss_pixel_y = ((o.pos.y + 1.0f) * (" I_VIEWPORT_SIZE ".y * 0.5f));\n");
out.WriteFmt("\tfloat ss_pixel_x = ((o.pos.x + 1.0f) * (" I_VIEWPORT_SIZE ".x * 0.5f));\n"
"\tfloat ss_pixel_y = ((o.pos.y + 1.0f) * (" I_VIEWPORT_SIZE ".y * 0.5f));\n");
out.Write("\tss_pixel_x = round(ss_pixel_x);\n");
out.Write("\tss_pixel_y = round(ss_pixel_y);\n");
out.WriteFmt("\tss_pixel_x = round(ss_pixel_x);\n"
"\tss_pixel_y = round(ss_pixel_y);\n");
out.Write("\to.pos.x = ((ss_pixel_x / (" I_VIEWPORT_SIZE ".x * 0.5f)) - 1.0f);\n");
out.Write("\to.pos.y = ((ss_pixel_y / (" I_VIEWPORT_SIZE ".y * 0.5f)) - 1.0f);\n");
out.Write("}\n");
out.WriteFmt("\to.pos.x = ((ss_pixel_x / (" I_VIEWPORT_SIZE ".x * 0.5f)) - 1.0f);\n"
"\to.pos.y = ((ss_pixel_y / (" I_VIEWPORT_SIZE ".y * 0.5f)) - 1.0f);\n"
"}}\n");
}
if (ApiType == APIType::OpenGL || ApiType == APIType::Vulkan)
if (api_type == APIType::OpenGL || api_type == APIType::Vulkan)
{
if (host_config.backend_geometry_shaders)
{
AssignVSOutputMembers(out, "vs", "o", numTexgen, host_config);
AssignVSOutputMembers(out, "vs", "o", num_texgen, host_config);
}
else
{
// TODO: Pass interface blocks between shader stages even if geometry shaders
// are not supported, however that will require at least OpenGL 3.2 support.
for (u32 i = 0; i < numTexgen; ++i)
out.Write("tex%d.xyz = o.tex%d;\n", i, i);
for (u32 i = 0; i < num_texgen; ++i)
out.WriteFmt("tex{}.xyz = o.tex{};\n", i, i);
if (!host_config.fast_depth_calc)
out.Write("clipPos = o.clipPos;\n");
out.WriteFmt("clipPos = o.clipPos;\n");
if (per_pixel_lighting)
{
out.Write("Normal = o.Normal;\n");
out.Write("WorldPos = o.WorldPos;\n");
out.WriteFmt("Normal = o.Normal;\n"
"WorldPos = o.WorldPos;\n");
}
out.Write("colors_0 = o.colors_0;\n");
out.Write("colors_1 = o.colors_1;\n");
out.WriteFmt("colors_0 = o.colors_0;\n"
"colors_1 = o.colors_1;\n");
}
if (host_config.backend_depth_clamp)
{
out.Write("gl_ClipDistance[0] = clipDist0;\n");
out.Write("gl_ClipDistance[1] = clipDist1;\n");
out.WriteFmt("gl_ClipDistance[0] = clipDist0;\n"
"gl_ClipDistance[1] = clipDist1;\n");
}
// Vulkan NDC space has Y pointing down (right-handed NDC space).
if (ApiType == APIType::Vulkan)
out.Write("gl_Position = float4(o.pos.x, -o.pos.y, o.pos.z, o.pos.w);\n");
if (api_type == APIType::Vulkan)
out.WriteFmt("gl_Position = float4(o.pos.x, -o.pos.y, o.pos.z, o.pos.w);\n");
else
out.Write("gl_Position = o.pos;\n");
out.WriteFmt("gl_Position = o.pos;\n");
}
else // D3D
{
out.Write("return o;\n");
out.WriteFmt("return o;\n");
}
out.Write("}\n");
out.WriteFmt("}}\n");
return out;
}
void GenVertexShaderTexGens(APIType ApiType, u32 numTexgen, ShaderCode& out)
static void GenVertexShaderTexGens(APIType api_type, u32 num_texgen, ShaderCode& out)
{
// The HLSL compiler complains that the output texture coordinates are uninitialized when trying
// to dynamically index them.
for (u32 i = 0; i < numTexgen; i++)
out.Write("o.tex%u = float3(0.0, 0.0, 0.0);\n", i);
for (u32 i = 0; i < num_texgen; i++)
out.WriteFmt("o.tex{} = float3(0.0, 0.0, 0.0);\n", i);
out.Write("// Texture coordinate generation\n");
if (numTexgen == 1)
out.Write("{ const uint texgen = 0u;\n");
out.WriteFmt("// Texture coordinate generation\n");
if (num_texgen == 1)
{
out.WriteFmt("{{ const uint texgen = 0u;\n");
}
else
out.Write("%sfor (uint texgen = 0u; texgen < %uu; texgen++) {\n",
ApiType == APIType::D3D ? "[loop] " : "", numTexgen);
{
out.WriteFmt("{}for (uint texgen = 0u; texgen < {}u; texgen++) {{\n",
api_type == APIType::D3D ? "[loop] " : "", num_texgen);
}
out.Write(" // Texcoord transforms\n");
out.Write(" float4 coord = float4(0.0, 0.0, 1.0, 1.0);\n"
out.WriteFmt(" // Texcoord transforms\n");
out.WriteFmt(" float4 coord = float4(0.0, 0.0, 1.0, 1.0);\n"
" uint texMtxInfo = xfmem_texMtxInfo(texgen);\n");
out.Write(" switch (%s) {\n", BitfieldExtract("texMtxInfo", TexMtxInfo().sourcerow).c_str());
out.Write(" case %uu: // XF_SRCGEOM_INROW\n", XF_SRCGEOM_INROW);
out.Write(" coord.xyz = rawpos.xyz;\n");
out.Write(" break;\n\n");
out.Write(" case %uu: // XF_SRCNORMAL_INROW\n", XF_SRCNORMAL_INROW);
out.Write(
" coord.xyz = ((components & %uu /* VB_HAS_NRM0 */) != 0u) ? rawnorm0.xyz : coord.xyz;",
out.WriteFmt(" switch ({}) {{\n", BitfieldExtract("texMtxInfo", TexMtxInfo().sourcerow));
out.WriteFmt(" case {}u: // XF_SRCGEOM_INROW\n", XF_SRCGEOM_INROW);
out.WriteFmt(" coord.xyz = rawpos.xyz;\n");
out.WriteFmt(" break;\n\n");
out.WriteFmt(" case {}u: // XF_SRCNORMAL_INROW\n", XF_SRCNORMAL_INROW);
out.WriteFmt(
" coord.xyz = ((components & {}u /* VB_HAS_NRM0 */) != 0u) ? rawnorm0.xyz : coord.xyz;",
VB_HAS_NRM0);
out.Write(" break;\n\n");
out.Write(" case %uu: // XF_SRCBINORMAL_T_INROW\n", XF_SRCBINORMAL_T_INROW);
out.Write(
" coord.xyz = ((components & %uu /* VB_HAS_NRM1 */) != 0u) ? rawnorm1.xyz : coord.xyz;",
out.WriteFmt(" break;\n\n");
out.WriteFmt(" case {}u: // XF_SRCBINORMAL_T_INROW\n", XF_SRCBINORMAL_T_INROW);
out.WriteFmt(
" coord.xyz = ((components & {}u /* VB_HAS_NRM1 */) != 0u) ? rawnorm1.xyz : coord.xyz;",
VB_HAS_NRM1);
out.Write(" break;\n\n");
out.Write(" case %uu: // XF_SRCBINORMAL_B_INROW\n", XF_SRCBINORMAL_B_INROW);
out.Write(
" coord.xyz = ((components & %uu /* VB_HAS_NRM2 */) != 0u) ? rawnorm2.xyz : coord.xyz;",
out.WriteFmt(" break;\n\n");
out.WriteFmt(" case {}u: // XF_SRCBINORMAL_B_INROW\n", XF_SRCBINORMAL_B_INROW);
out.WriteFmt(
" coord.xyz = ((components & {}u /* VB_HAS_NRM2 */) != 0u) ? rawnorm2.xyz : coord.xyz;",
VB_HAS_NRM2);
out.Write(" break;\n\n");
out.WriteFmt(" break;\n\n");
for (u32 i = 0; i < 8; i++)
{
out.Write(" case %uu: // XF_SRCTEX%u_INROW\n", XF_SRCTEX0_INROW + i, i);
out.Write(
" coord = ((components & %uu /* VB_HAS_UV%u */) != 0u) ? float4(rawtex%u.x, rawtex%u.y, "
out.WriteFmt(" case {}u: // XF_SRCTEX{}_INROW\n", XF_SRCTEX0_INROW + i, i);
out.WriteFmt(
" coord = ((components & {}u /* VB_HAS_UV{} */) != 0u) ? float4(rawtex{}.x, rawtex{}.y, "
"1.0, 1.0) : coord;\n",
VB_HAS_UV0 << i, i, i, i);
out.Write(" break;\n\n");
out.WriteFmt(" break;\n\n");
}
out.Write(" }\n");
out.Write("\n");
out.WriteFmt(" }}\n"
"\n");
out.Write(" // Input form of AB11 sets z element to 1.0\n");
out.Write(" if (%s == %uu) // inputform == XF_TEXINPUT_AB11\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().inputform).c_str(), XF_TEXINPUT_AB11);
out.Write(" coord.z = 1.0f;\n");
out.Write("\n");
out.WriteFmt(" // Input form of AB11 sets z element to 1.0\n");
out.WriteFmt(" if ({} == {}u) // inputform == XF_TEXINPUT_AB11\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().inputform), XF_TEXINPUT_AB11);
out.WriteFmt(" coord.z = 1.0f;\n"
"\n");
out.Write(" // first transformation\n");
out.Write(" uint texgentype = %s;\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().texgentype).c_str());
out.Write(" float3 output_tex;\n"
out.WriteFmt(" // first transformation\n");
out.WriteFmt(" uint texgentype = {};\n", BitfieldExtract("texMtxInfo", TexMtxInfo().texgentype));
out.WriteFmt(" float3 output_tex;\n"
" switch (texgentype)\n"
" {\n");
out.Write(" case %uu: // XF_TEXGEN_EMBOSS_MAP\n", XF_TEXGEN_EMBOSS_MAP);
out.Write(" {\n");
out.Write(" uint light = %s;\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().embosslightshift).c_str());
out.Write(" uint source = %s;\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().embosssourceshift).c_str());
out.Write(" switch (source) {\n");
for (u32 i = 0; i < numTexgen; i++)
out.Write(" case %uu: output_tex.xyz = o.tex%u; break;\n", i, i);
out.Write(" default: output_tex.xyz = float3(0.0, 0.0, 0.0); break;\n"
" }\n");
out.Write(" if ((components & %uu) != 0u) { // VB_HAS_NRM1 | VB_HAS_NRM2\n",
" {{\n");
out.WriteFmt(" case {}u: // XF_TEXGEN_EMBOSS_MAP\n", XF_TEXGEN_EMBOSS_MAP);
out.WriteFmt(" {{\n");
out.WriteFmt(" uint light = {};\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().embosslightshift));
out.WriteFmt(" uint source = {};\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().embosssourceshift));
out.WriteFmt(" switch (source) {{\n");
for (u32 i = 0; i < num_texgen; i++)
out.WriteFmt(" case {}u: output_tex.xyz = o.tex{}; break;\n", i, i);
out.WriteFmt(" default: output_tex.xyz = float3(0.0, 0.0, 0.0); break;\n"
" }}\n");
out.WriteFmt(" if ((components & {}u) != 0u) {{ // VB_HAS_NRM1 | VB_HAS_NRM2\n",
VB_HAS_NRM1 | VB_HAS_NRM2); // Should this be VB_HAS_NRM1 | VB_HAS_NRM2
out.Write(" float3 ldir = normalize(" I_LIGHTS "[light].pos.xyz - pos.xyz);\n"
out.WriteFmt(" float3 ldir = normalize(" I_LIGHTS "[light].pos.xyz - pos.xyz);\n"
" output_tex.xyz += float3(dot(ldir, _norm1), dot(ldir, _norm2), 0.0);\n"
" }\n"
" }\n"
" }}\n"
" }}\n"
" break;\n\n");
out.Write(" case %uu: // XF_TEXGEN_COLOR_STRGBC0\n", XF_TEXGEN_COLOR_STRGBC0);
out.Write(" output_tex.xyz = float3(o.colors_0.x, o.colors_0.y, 1.0);\n"
out.WriteFmt(" case {}u: // XF_TEXGEN_COLOR_STRGBC0\n", XF_TEXGEN_COLOR_STRGBC0);
out.WriteFmt(" output_tex.xyz = float3(o.colors_0.x, o.colors_0.y, 1.0);\n"
" break;\n\n");
out.Write(" case %uu: // XF_TEXGEN_COLOR_STRGBC1\n", XF_TEXGEN_COLOR_STRGBC1);
out.Write(" output_tex.xyz = float3(o.colors_1.x, o.colors_1.y, 1.0);\n"
out.WriteFmt(" case {}u: // XF_TEXGEN_COLOR_STRGBC1\n", XF_TEXGEN_COLOR_STRGBC1);
out.WriteFmt(" output_tex.xyz = float3(o.colors_1.x, o.colors_1.y, 1.0);\n"
" break;\n\n");
out.Write(" default: // Also XF_TEXGEN_REGULAR\n"
" {\n");
out.Write(" if ((components & (%uu /* VB_HAS_TEXMTXIDX0 */ << texgen)) != 0u) {\n",
out.WriteFmt(" default: // Also XF_TEXGEN_REGULAR\n"
" {{\n");
out.WriteFmt(" if ((components & ({}u /* VB_HAS_TEXMTXIDX0 */ << texgen)) != 0u) {{\n",
VB_HAS_TEXMTXIDX0);
out.Write(" // This is messy, due to dynamic indexing of the input texture coordinates.\n"
out.WriteFmt(
" // This is messy, due to dynamic indexing of the input texture coordinates.\n"
" // Hopefully the compiler will unroll this whole loop anyway and the switch.\n"
" int tmp = 0;\n"
" switch (texgen) {\n");
for (u32 i = 0; i < numTexgen; i++)
out.Write(" case %uu: tmp = int(rawtex%u.z); break;\n", i, i);
out.Write(" }\n"
" switch (texgen) {{\n");
for (u32 i = 0; i < num_texgen; i++)
out.WriteFmt(" case {}u: tmp = int(rawtex{}.z); break;\n", i, i);
out.WriteFmt(" }}\n"
"\n");
out.Write(" if (%s == %uu) {\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().projection).c_str(), XF_TEXPROJ_STQ);
out.Write(" output_tex.xyz = float3(dot(coord, " I_TRANSFORMMATRICES "[tmp]),\n"
out.WriteFmt(" if ({} == {}u) {{\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().projection), XF_TEXPROJ_STQ);
out.WriteFmt(" output_tex.xyz = float3(dot(coord, " I_TRANSFORMMATRICES "[tmp]),\n"
" dot(coord, " I_TRANSFORMMATRICES "[tmp + 1]),\n"
" dot(coord, " I_TRANSFORMMATRICES "[tmp + 2]));\n"
" } else {\n"
" }} else {{\n"
" output_tex.xyz = float3(dot(coord, " I_TRANSFORMMATRICES "[tmp]),\n"
" dot(coord, " I_TRANSFORMMATRICES "[tmp + 1]),\n"
" 1.0);\n"
" }\n"
" } else {\n");
out.Write(" if (%s == %uu) {\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().projection).c_str(), XF_TEXPROJ_STQ);
out.Write(" output_tex.xyz = float3(dot(coord, " I_TEXMATRICES "[3u * texgen]),\n"
" }}\n"
" }} else {{\n");
out.WriteFmt(" if ({} == {}u) {{\n",
BitfieldExtract("texMtxInfo", TexMtxInfo().projection), XF_TEXPROJ_STQ);
out.WriteFmt(
" output_tex.xyz = float3(dot(coord, " I_TEXMATRICES "[3u * texgen]),\n"
" dot(coord, " I_TEXMATRICES "[3u * texgen + 1u]),\n"
" dot(coord, " I_TEXMATRICES "[3u * texgen + 2u]));\n"
" } else {\n"
" }} else {{\n"
" output_tex.xyz = float3(dot(coord, " I_TEXMATRICES "[3u * texgen]),\n"
" dot(coord, " I_TEXMATRICES "[3u * texgen + 1u]),\n"
" 1.0);\n"
" }\n"
" }\n"
" }\n"
" }}\n"
" }}\n"
" }}\n"
" break;\n\n"
" }\n"
" }}\n"
"\n");
out.Write(" if (xfmem_dualTexInfo != 0u) {\n");
out.Write(" uint postMtxInfo = xfmem_postMtxInfo(texgen);");
out.Write(" uint base_index = %s;\n",
BitfieldExtract("postMtxInfo", PostMtxInfo().index).c_str());
out.Write(" float4 P0 = " I_POSTTRANSFORMMATRICES "[base_index & 0x3fu];\n"
out.WriteFmt(" if (xfmem_dualTexInfo != 0u) {{\n");
out.WriteFmt(" uint postMtxInfo = xfmem_postMtxInfo(texgen);");
out.WriteFmt(" uint base_index = {};\n", BitfieldExtract("postMtxInfo", PostMtxInfo().index));
out.WriteFmt(" float4 P0 = " I_POSTTRANSFORMMATRICES "[base_index & 0x3fu];\n"
" float4 P1 = " I_POSTTRANSFORMMATRICES "[(base_index + 1u) & 0x3fu];\n"
" float4 P2 = " I_POSTTRANSFORMMATRICES "[(base_index + 2u) & 0x3fu];\n"
"\n");
out.Write(" if (%s != 0u)\n", BitfieldExtract("postMtxInfo", PostMtxInfo().normalize).c_str());
out.Write(" output_tex.xyz = normalize(output_tex.xyz);\n"
out.WriteFmt(" if ({} != 0u)\n", BitfieldExtract("postMtxInfo", PostMtxInfo().normalize));
out.WriteFmt(" output_tex.xyz = normalize(output_tex.xyz);\n"
"\n"
" // multiply by postmatrix\n"
" output_tex.xyz = float3(dot(P0.xyz, output_tex.xyz) + P0.w,\n"
" dot(P1.xyz, output_tex.xyz) + P1.w,\n"
" dot(P2.xyz, output_tex.xyz) + P2.w);\n"
" }\n\n");
" }}\n\n");
// When q is 0, the GameCube appears to have a special case
// This can be seen in devkitPro's neheGX Lesson08 example for Wii
// Makes differences in Rogue Squadron 3 (Hoth sky) and The Last Story (shadow culling)
out.Write(" if (texgentype == %uu && output_tex.z == 0.0) // XF_TEXGEN_REGULAR\n",
out.WriteFmt(" if (texgentype == {}u && output_tex.z == 0.0) // XF_TEXGEN_REGULAR\n",
XF_TEXGEN_REGULAR);
out.Write(
out.WriteFmt(
" output_tex.xy = clamp(output_tex.xy / 2.0f, float2(-1.0f,-1.0f), float2(1.0f,1.0f));\n"
"\n");
out.Write(" // Hopefully GPUs that can support dynamic indexing will optimize this.\n");
out.Write(" switch (texgen) {\n");
for (u32 i = 0; i < numTexgen; i++)
out.Write(" case %uu: o.tex%u = output_tex; break;\n", i, i);
out.Write(" }\n"
"}\n");
out.WriteFmt(" // Hopefully GPUs that can support dynamic indexing will optimize this.\n");
out.WriteFmt(" switch (texgen) {{\n");
for (u32 i = 0; i < num_texgen; i++)
out.WriteFmt(" case {}u: o.tex{} = output_tex; break;\n", i, i);
out.WriteFmt(" }}\n"
"}}\n");
}
void EnumerateVertexShaderUids(const std::function<void(const VertexShaderUid&)>& callback)