static inline void GenerateVertexShader(T& out, u32 components, API_TYPE api_type)
{
// Non-uid template parameters will write to the dummy data (=> gets optimized out)
vertex_shader_uid_data dummy_data;
vertex_shader_uid_data* uid_data = out.template GetUidData<vertex_shader_uid_data>();
if (uid_data == nullptr)
uid_data = &dummy_data;
out.SetBuffer(text);
const bool is_writing_shadercode = (out.GetBuffer() != nullptr);
if (is_writing_shadercode)
text[sizeof(text) - 1] = 0x7C; // canary
_assert_(bpmem.genMode.numtexgens == xfmem.numTexGen.numTexGens);
_assert_(bpmem.genMode.numcolchans == xfmem.numChan.numColorChans);
if (DriverDetails::HasBug(DriverDetails::BUG_BROKENIVECSHIFTS))
{
// Add functions to do shifts on scalars and ivecs.
// This is included in the vertex shader for lighting shader generation.
out.Write("int ilshift(int a, int b) { return a << b; }\n"
"int irshift(int a, int b) { return a >> b; }\n"
"int2 ilshift(int2 a, int2 b) { return int2(a.x << b.x, a.y << b.y); }\n"
"int2 ilshift(int2 a, int b) { return int2(a.x << b, a.y << b); }\n"
"int2 irshift(int2 a, int2 b) { return int2(a.x >> b.x, a.y >> b.y); }\n"
"int2 irshift(int2 a, int b) { return int2(a.x >> b, a.y >> b); }\n"
"int3 ilshift(int3 a, int3 b) { return int3(a.x << b.x, a.y << b.y, a.z << b.z); }\n"
"int3 ilshift(int3 a, int b) { return int3(a.x << b, a.y << b, a.z << b); }\n"
"int3 irshift(int3 a, int3 b) { return int3(a.x >> b.x, a.y >> b.y, a.z >> b.z); }\n"
"int3 irshift(int3 a, int b) { return int3(a.x >> b, a.y >> b, a.z >> b); }\n"
"int4 ilshift(int4 a, int4 b) { return int4(a.x << b.x, a.y << b.y, a.z << b.z, a.w << b.w); }\n"
"int4 ilshift(int4 a, int b) { return int4(a.x << b, a.y << b, a.z << b, a.w << b); }\n"
"int4 irshift(int4 a, int4 b) { return int4(a.x >> b.x, a.y >> b.y, a.z >> b.z, a.w >> b.w); }\n"
"int4 irshift(int4 a, int b) { return int4(a.x >> b, a.y >> b, a.z >> b, a.w >> b); }\n\n");
}
out.Write("%s", s_lighting_struct);
// uniforms
if (api_type == API_OPENGL)
out.Write("layout(std140%s) uniform VSBlock {\n", g_ActiveConfig.backend_info.bSupportsBindingLayout ? ", binding = 2" : "");
else
out.Write("cbuffer VSBlock {\n");
out.Write(s_shader_uniforms);
out.Write("};\n");
out.Write("struct VS_OUTPUT {\n");
GenerateVSOutputMembers<T>(out, api_type);
out.Write("};\n");
uid_data->numTexGens = xfmem.numTexGen.numTexGens;
uid_data->components = components;
uid_data->pixel_lighting = g_ActiveConfig.bEnablePixelLighting;
if (api_type == API_OPENGL)
{
out.Write("in float4 rawpos; // ATTR%d,\n", SHADER_POSITION_ATTRIB);
if (components & VB_HAS_POSMTXIDX)
out.Write("in int posmtx; // ATTR%d,\n", SHADER_POSMTX_ATTRIB);
if (components & VB_HAS_NRM0)
out.Write("in float3 rawnorm0; // ATTR%d,\n", SHADER_NORM0_ATTRIB);
if (components & VB_HAS_NRM1)
out.Write("in float3 rawnorm1; // ATTR%d,\n", SHADER_NORM1_ATTRIB);
if (components & VB_HAS_NRM2)
out.Write("in float3 rawnorm2; // ATTR%d,\n", SHADER_NORM2_ATTRIB);
if (components & VB_HAS_COL0)
out.Write("in float4 color0; // ATTR%d,\n", SHADER_COLOR0_ATTRIB);
if (components & VB_HAS_COL1)
out.Write("in float4 color1; // ATTR%d,\n", SHADER_COLOR1_ATTRIB);
for (int i = 0; i < 8; ++i)
{
u32 hastexmtx = (components & (VB_HAS_TEXMTXIDX0<<i));
if ((components & (VB_HAS_UV0<<i)) || hastexmtx)
out.Write("in float%d tex%d; // ATTR%d,\n", hastexmtx ? 3 : 2, i, SHADER_TEXTURE0_ATTRIB + i);
}
if (g_ActiveConfig.backend_info.bSupportsGeometryShaders)
{
out.Write("out VertexData {\n");
GenerateVSOutputMembers<T>(out, api_type, g_ActiveConfig.backend_info.bSupportsBindingLayout ? "centroid" : "centroid out");
out.Write("} vs;\n");
}
else
{
// Let's set up attributes
for (size_t i = 0; i < 8; ++i)
{
if (i < xfmem.numTexGen.numTexGens)
{
out.Write("centroid out float3 uv%d;\n", i);
}
}
out.Write("centroid out float4 clipPos;\n");
if (g_ActiveConfig.bEnablePixelLighting)
{
out.Write("centroid out float3 Normal;\n");
out.Write("centroid out float3 WorldPos;\n");
}
out.Write("centroid out float4 colors_0;\n");
out.Write("centroid out float4 colors_1;\n");
}
out.Write("void main()\n{\n");
}
else // D3D
{
out.Write("VS_OUTPUT main(\n");
// inputs
if (components & VB_HAS_NRM0)
out.Write(" float3 rawnorm0 : NORMAL0,\n");
if (components & VB_HAS_NRM1)
out.Write(" float3 rawnorm1 : NORMAL1,\n");
if (components & VB_HAS_NRM2)
out.Write(" float3 rawnorm2 : NORMAL2,\n");
if (components & VB_HAS_COL0)
out.Write(" float4 color0 : COLOR0,\n");
if (components & VB_HAS_COL1)
out.Write(" float4 color1 : COLOR1,\n");
for (int i = 0; i < 8; ++i)
{
u32 hastexmtx = (components & (VB_HAS_TEXMTXIDX0<<i));
if ((components & (VB_HAS_UV0<<i)) || hastexmtx)
out.Write(" float%d tex%d : TEXCOORD%d,\n", hastexmtx ? 3 : 2, i, i);
}
if (components & VB_HAS_POSMTXIDX)
out.Write(" int posmtx : BLENDINDICES,\n");
out.Write(" float4 rawpos : POSITION) {\n");
}
out.Write("VS_OUTPUT o;\n");
// transforms
if (components & VB_HAS_POSMTXIDX)
{
if (is_writing_shadercode && (DriverDetails::HasBug(DriverDetails::BUG_NODYNUBOACCESS) && !DriverDetails::HasBug(DriverDetails::BUG_ANNIHILATEDUBOS)))
{
// This'll cause issues, but it can't be helped
out.Write("float4 pos = float4(dot(" I_TRANSFORMMATRICES"[0], rawpos), dot(" I_TRANSFORMMATRICES"[1], rawpos), dot(" I_TRANSFORMMATRICES"[2], rawpos), 1);\n");
if (components & VB_HAS_NRMALL)
out.Write("float3 N0 = " I_NORMALMATRICES"[0].xyz, N1 = " I_NORMALMATRICES"[1].xyz, N2 = " I_NORMALMATRICES"[2].xyz;\n");
}
else
{
out.Write("float4 pos = float4(dot(" I_TRANSFORMMATRICES"[posmtx], rawpos), dot(" I_TRANSFORMMATRICES"[posmtx+1], rawpos), dot(" I_TRANSFORMMATRICES"[posmtx+2], rawpos), 1);\n");
if (components & VB_HAS_NRMALL)
{
out.Write("int normidx = posmtx >= 32 ? (posmtx-32) : posmtx;\n");
out.Write("float3 N0 = " I_NORMALMATRICES"[normidx].xyz, N1 = " I_NORMALMATRICES"[normidx+1].xyz, N2 = " I_NORMALMATRICES"[normidx+2].xyz;\n");
}
}
if (components & VB_HAS_NRM0)
out.Write("float3 _norm0 = normalize(float3(dot(N0, rawnorm0), dot(N1, rawnorm0), dot(N2, rawnorm0)));\n");
if (components & VB_HAS_NRM1)
out.Write("float3 _norm1 = float3(dot(N0, rawnorm1), dot(N1, rawnorm1), dot(N2, rawnorm1));\n");
if (components & VB_HAS_NRM2)
out.Write("float3 _norm2 = float3(dot(N0, rawnorm2), dot(N1, rawnorm2), dot(N2, rawnorm2));\n");
}
else
{
out.Write("float4 pos = float4(dot(" I_POSNORMALMATRIX"[0], rawpos), dot(" I_POSNORMALMATRIX"[1], rawpos), dot(" I_POSNORMALMATRIX"[2], rawpos), 1.0);\n");
if (components & VB_HAS_NRM0)
out.Write("float3 _norm0 = normalize(float3(dot(" I_POSNORMALMATRIX"[3].xyz, rawnorm0), dot(" I_POSNORMALMATRIX"[4].xyz, rawnorm0), dot(" I_POSNORMALMATRIX"[5].xyz, rawnorm0)));\n");
if (components & VB_HAS_NRM1)
out.Write("float3 _norm1 = float3(dot(" I_POSNORMALMATRIX"[3].xyz, rawnorm1), dot(" I_POSNORMALMATRIX"[4].xyz, rawnorm1), dot(" I_POSNORMALMATRIX"[5].xyz, rawnorm1));\n");
if (components & VB_HAS_NRM2)
out.Write("float3 _norm2 = float3(dot(" I_POSNORMALMATRIX"[3].xyz, rawnorm2), dot(" I_POSNORMALMATRIX"[4].xyz, rawnorm2), dot(" I_POSNORMALMATRIX"[5].xyz, rawnorm2));\n");
}
if (!(components & VB_HAS_NRM0))
out.Write("float3 _norm0 = float3(0.0, 0.0, 0.0);\n");
out.Write("o.pos = float4(dot(" I_PROJECTION"[0], pos), dot(" I_PROJECTION"[1], pos), dot(" I_PROJECTION"[2], pos), dot(" I_PROJECTION"[3], pos));\n");
out.Write("int4 lacc;\n"
"float3 ldir, h, cosAttn, distAttn;\n"
"float dist, dist2, attn;\n");
uid_data->numColorChans = xfmem.numChan.numColorChans;
if (xfmem.numChan.numColorChans == 0)
{
if (components & VB_HAS_COL0)
out.Write("o.colors_0 = color0;\n");
else
out.Write("o.colors_0 = float4(1.0, 1.0, 1.0, 1.0);\n");
}
GenerateLightingShader<T>(out, uid_data->lighting, components, "color", "o.colors_");
if (xfmem.numChan.numColorChans < 2)
{
if (components & VB_HAS_COL1)
out.Write("o.colors_1 = color1;\n");
else
out.Write("o.colors_1 = o.colors_0;\n");
}
// special case if only pos and tex coord 0 and tex coord input is AB11
// donko - this has caused problems in some games. removed for now.
bool texGenSpecialCase = false;
/*bool texGenSpecialCase =
((g_main_cp_state.vtx_desc.Hex & 0x60600L) == g_main_cp_state.vtx_desc.Hex) && // only pos and tex coord 0
(g_main_cp_state.vtx_desc.Tex0Coord != NOT_PRESENT) &&
(xfmem.texcoords[0].texmtxinfo.inputform == XF_TEXINPUT_AB11);
*/
// transform texcoords
out.Write("float4 coord = float4(0.0, 0.0, 1.0, 1.0);\n");
for (unsigned int i = 0; i < xfmem.numTexGen.numTexGens; ++i)
{
TexMtxInfo& texinfo = xfmem.texMtxInfo[i];
out.Write("{\n");
out.Write("coord = float4(0.0, 0.0, 1.0, 1.0);\n");
uid_data->texMtxInfo[i].sourcerow = xfmem.texMtxInfo[i].sourcerow;
switch (texinfo.sourcerow)
{
case XF_SRCGEOM_INROW:
_assert_(texinfo.inputform == XF_TEXINPUT_ABC1);
out.Write("coord = rawpos;\n"); // pos.w is 1
break;
case XF_SRCNORMAL_INROW:
if (components & VB_HAS_NRM0)
{
_assert_(texinfo.inputform == XF_TEXINPUT_ABC1);
out.Write("coord = float4(rawnorm0.xyz, 1.0);\n");
}
break;
case XF_SRCCOLORS_INROW:
_assert_(texinfo.texgentype == XF_TEXGEN_COLOR_STRGBC0 || texinfo.texgentype == XF_TEXGEN_COLOR_STRGBC1);
break;
case XF_SRCBINORMAL_T_INROW:
if (components & VB_HAS_NRM1)
{
_assert_(texinfo.inputform == XF_TEXINPUT_ABC1);
out.Write("coord = float4(rawnorm1.xyz, 1.0);\n");
}
break;
case XF_SRCBINORMAL_B_INROW:
if (components & VB_HAS_NRM2)
{
_assert_(texinfo.inputform == XF_TEXINPUT_ABC1);
out.Write("coord = float4(rawnorm2.xyz, 1.0);\n");
}
break;
default:
_assert_(texinfo.sourcerow <= XF_SRCTEX7_INROW);
if (components & (VB_HAS_UV0 << (texinfo.sourcerow - XF_SRCTEX0_INROW)))
out.Write("coord = float4(tex%d.x, tex%d.y, 1.0, 1.0);\n", texinfo.sourcerow - XF_SRCTEX0_INROW, texinfo.sourcerow - XF_SRCTEX0_INROW);
break;
}
// first transformation
uid_data->texMtxInfo[i].texgentype = xfmem.texMtxInfo[i].texgentype;
switch (texinfo.texgentype)
{
case XF_TEXGEN_EMBOSS_MAP: // calculate tex coords into bump map
if (components & (VB_HAS_NRM1|VB_HAS_NRM2))
{
// transform the light dir into tangent space
uid_data->texMtxInfo[i].embosslightshift = xfmem.texMtxInfo[i].embosslightshift;
uid_data->texMtxInfo[i].embosssourceshift = xfmem.texMtxInfo[i].embosssourceshift;
out.Write("ldir = normalize(" LIGHT_POS".xyz - pos.xyz);\n", LIGHT_POS_PARAMS(texinfo.embosslightshift));
out.Write("o.tex%d.xyz = o.tex%d.xyz + float3(dot(ldir, _norm1), dot(ldir, _norm2), 0.0);\n", i, texinfo.embosssourceshift);
}
else
{
_assert_(0); // should have normals
uid_data->texMtxInfo[i].embosssourceshift = xfmem.texMtxInfo[i].embosssourceshift;
out.Write("o.tex%d.xyz = o.tex%d.xyz;\n", i, texinfo.embosssourceshift);
}
break;
case XF_TEXGEN_COLOR_STRGBC0:
_assert_(texinfo.sourcerow == XF_SRCCOLORS_INROW);
out.Write("o.tex%d.xyz = float3(o.colors_0.x, o.colors_0.y, 1);\n", i);
break;
case XF_TEXGEN_COLOR_STRGBC1:
_assert_(texinfo.sourcerow == XF_SRCCOLORS_INROW);
out.Write("o.tex%d.xyz = float3(o.colors_1.x, o.colors_1.y, 1);\n", i);
break;
case XF_TEXGEN_REGULAR:
default:
uid_data->texMtxInfo_n_projection |= xfmem.texMtxInfo[i].projection << i;
if (components & (VB_HAS_TEXMTXIDX0<<i))
{
out.Write("int tmp = int(tex%d.z);\n", i);
if (texinfo.projection == XF_TEXPROJ_STQ)
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TRANSFORMMATRICES"[tmp]), dot(coord, " I_TRANSFORMMATRICES"[tmp+1]), dot(coord, " I_TRANSFORMMATRICES"[tmp+2]));\n", i);
else
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TRANSFORMMATRICES"[tmp]), dot(coord, " I_TRANSFORMMATRICES"[tmp+1]), 1);\n", i);
}
else
{
if (texinfo.projection == XF_TEXPROJ_STQ)
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TEXMATRICES"[%d]), dot(coord, " I_TEXMATRICES"[%d]), dot(coord, " I_TEXMATRICES"[%d]));\n", i, 3*i, 3*i+1, 3*i+2);
else
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TEXMATRICES"[%d]), dot(coord, " I_TEXMATRICES"[%d]), 1);\n", i, 3*i, 3*i+1);
}
break;
}
uid_data->dualTexTrans_enabled = xfmem.dualTexTrans.enabled;
// CHECKME: does this only work for regular tex gen types?
if (xfmem.dualTexTrans.enabled && texinfo.texgentype == XF_TEXGEN_REGULAR)
{
const PostMtxInfo& postInfo = xfmem.postMtxInfo[i];
uid_data->postMtxInfo[i].index = xfmem.postMtxInfo[i].index;
int postidx = postInfo.index;
out.Write("float4 P0 = " I_POSTTRANSFORMMATRICES"[%d];\n"
"float4 P1 = " I_POSTTRANSFORMMATRICES"[%d];\n"
"float4 P2 = " I_POSTTRANSFORMMATRICES"[%d];\n",
postidx & 0x3f, (postidx + 1) & 0x3f, (postidx + 2) & 0x3f);
if (texGenSpecialCase)
{
// no normalization
// q of input is 1
// q of output is unknown
// multiply by postmatrix
out.Write("o.tex%d.xyz = float3(dot(P0.xy, o.tex%d.xy) + P0.z + P0.w, dot(P1.xy, o.tex%d.xy) + P1.z + P1.w, 0.0);\n", i, i, i);
}
else
{
uid_data->postMtxInfo[i].normalize = xfmem.postMtxInfo[i].normalize;
if (postInfo.normalize)
out.Write("o.tex%d.xyz = normalize(o.tex%d.xyz);\n", i, i);
// multiply by postmatrix
out.Write("o.tex%d.xyz = float3(dot(P0.xyz, o.tex%d.xyz) + P0.w, dot(P1.xyz, o.tex%d.xyz) + P1.w, dot(P2.xyz, o.tex%d.xyz) + P2.w);\n", i, i, i, i);
}
}
out.Write("}\n");
}
// clipPos/w needs to be done in pixel shader, not here
out.Write("o.clipPos = o.pos;\n");
if (g_ActiveConfig.bEnablePixelLighting)
{
out.Write("o.Normal = _norm0;\n");
out.Write("o.WorldPos = pos.xyz;\n");
if (components & VB_HAS_COL0)
out.Write("o.colors_0 = color0;\n");
if (components & VB_HAS_COL1)
out.Write("o.colors_1 = color1;\n");
}
//write the true depth value, if the game uses depth textures pixel shaders will override with the correct values
//if not early z culling will improve speed
if (g_ActiveConfig.backend_info.bSupportsClipControl)
{
out.Write("o.pos.z = -o.pos.z;\n");
}
else if (api_type == API_D3D)
{
out.Write("o.pos.z = -o.pos.z;\n");
}
else // OGL
{
// this results in a scale from -1..0 to -1..1 after perspective
// divide
out.Write("o.pos.z = o.pos.z * -2.0 - o.pos.w;\n");
// the next steps of the OGL pipeline are:
// (x_c,y_c,z_c,w_c) = o.pos //switch to OGL spec terminology
// clipping to -w_c <= (x_c,y_c,z_c) <= w_c
// (x_d,y_d,z_d) = (x_c,y_c,z_c)/w_c//perspective divide
// z_w = (f-n)/2*z_d + (n+f)/2
// z_w now contains the value to go to the 0..1 depth buffer
//trying to get the correct semantic while not using glDepthRange
//seems to get rather complicated
}
// 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
// in some primitives being placed one pixel too far to the bottom-right,
// 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");
if (api_type == API_OPENGL)
{
if (g_ActiveConfig.backend_info.bSupportsGeometryShaders)
{
AssignVSOutputMembers(out, "vs", "o");
}
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 (unsigned int i = 0; i < xfmem.numTexGen.numTexGens; ++i)
out.Write("uv%d.xyz = o.tex%d;\n", i, i);
out.Write("clipPos = o.clipPos;\n");
if (g_ActiveConfig.bEnablePixelLighting)
{
out.Write("Normal = o.Normal;\n");
out.Write("WorldPos = o.WorldPos;\n");
}
out.Write("colors_0 = o.colors_0;\n");
out.Write("colors_1 = o.colors_1;\n");
}
out.Write("gl_Position = o.pos;\n");
}
else // D3D
{
out.Write("return o;\n");
}
out.Write("}\n");
if (is_writing_shadercode)
{
if (text[sizeof(text) - 1] != 0x7C)
PanicAlert("VertexShader generator - buffer too small, canary has been eaten!");
}
}
static void GenerateLightShader(ShaderCode& object, const LightingUidData& uid_data, int index,
int litchan_index, bool alpha)
{
const char* swizzle = alpha ? "a" : "rgb";
const char* swizzle_components = (alpha) ? "" : "3";
int attnfunc = (uid_data.attnfunc >> (2 * litchan_index)) & 0x3;
int diffusefunc = (uid_data.diffusefunc >> (2 * litchan_index)) & 0x3;
switch (attnfunc)
{
case LIGHTATTN_NONE:
case LIGHTATTN_DIR:
object.Write("ldir = normalize(" LIGHT_POS ".xyz - pos.xyz);\n", LIGHT_POS_PARAMS(index));
object.Write("attn = 1.0;\n");
object.Write("if (length(ldir) == 0.0)\n\t ldir = _norm0;\n");
break;
case LIGHTATTN_SPEC:
object.Write("ldir = normalize(" LIGHT_POS ".xyz - pos.xyz);\n", LIGHT_POS_PARAMS(index));
object.Write("attn = (dot(_norm0, ldir) >= 0.0) ? max(0.0, dot(_norm0, " LIGHT_DIR
".xyz)) : 0.0;\n",
LIGHT_DIR_PARAMS(index));
object.Write("cosAttn = " LIGHT_COSATT ".xyz;\n", LIGHT_COSATT_PARAMS(index));
object.Write("distAttn = %s(" LIGHT_DISTATT ".xyz);\n",
(diffusefunc == LIGHTDIF_NONE) ? "" : "normalize", LIGHT_DISTATT_PARAMS(index));
object.Write("attn = max(0.0f, dot(cosAttn, float3(1.0, attn, attn*attn))) / dot(distAttn, "
"float3(1.0, attn, attn*attn));\n");
break;
case LIGHTATTN_SPOT:
object.Write("ldir = " LIGHT_POS ".xyz - pos.xyz;\n", LIGHT_POS_PARAMS(index));
object.Write("dist2 = dot(ldir, ldir);\n"
"dist = sqrt(dist2);\n"
"ldir = ldir / dist;\n"
"attn = max(0.0, dot(ldir, " LIGHT_DIR ".xyz));\n",
LIGHT_DIR_PARAMS(index));
// attn*attn may overflow
object.Write("attn = max(0.0, " LIGHT_COSATT ".x + " LIGHT_COSATT ".y*attn + " LIGHT_COSATT
".z*attn*attn) / dot(" LIGHT_DISTATT ".xyz, float3(1.0,dist,dist2));\n",
LIGHT_COSATT_PARAMS(index), LIGHT_COSATT_PARAMS(index), LIGHT_COSATT_PARAMS(index),
LIGHT_DISTATT_PARAMS(index));
break;
}
switch (diffusefunc)
{
case LIGHTDIF_NONE:
object.Write("lacc.%s += int%s(round(attn * float%s(" LIGHT_COL ")));\n", swizzle,
swizzle_components, swizzle_components, LIGHT_COL_PARAMS(index, swizzle));
break;
case LIGHTDIF_SIGN:
case LIGHTDIF_CLAMP:
object.Write("lacc.%s += int%s(round(attn * %sdot(ldir, _norm0)) * float%s(" LIGHT_COL ")));\n",
swizzle, swizzle_components, diffusefunc != LIGHTDIF_SIGN ? "max(0.0," : "(",
swizzle_components, LIGHT_COL_PARAMS(index, swizzle));
break;
default:
_assert_(0);
}
object.Write("\n");
}
inline void GenerateVertexShader(T& out, u32 components, const XFMemory &xfr, const BPMemory &bpm, bool use_integer_math)
{
// Non-uid template parameters will write to the dummy data (=> gets optimized out)
bool uidPresent = (&out.template GetUidData<vertex_shader_uid_data>() != NULL);
vertex_shader_uid_data dummy_data;
vertex_shader_uid_data& uid_data = uidPresent ? out.template GetUidData<vertex_shader_uid_data>() : dummy_data;
if (uidPresent)
{
out.ClearUID();
}
if (Write_Code)
{
_dbg_assert_log_(VIDEO, bpm.genMode.numtexgens == xfr.numTexGen.numTexGens, "numTexGens mismatch bpmem: %u xfmem: %u", bpm.genMode.numtexgens.Value(), xfr.numTexGen.numTexGens);
_dbg_assert_log_(VIDEO, bpm.genMode.numcolchans == xfr.numChan.numColorChans, "numColorChans mismatch bpmem: %u xfmem: %u", bpm.genMode.numcolchans.Value(), xfr.numChan.numColorChans);
}
uid_data.numTexGens = xfr.numTexGen.numTexGens;
uid_data.components = components;
bool lightingEnabled = xfr.numChan.numColorChans > 0;
bool enable_pl = g_ActiveConfig.PixelLightingEnabled(xfr, components);
bool needLightShader = lightingEnabled && !enable_pl;
for (unsigned int i = 0; i < xfr.numTexGen.numTexGens; ++i)
{
const TexMtxInfo& texinfo = xfr.texMtxInfo[i];
needLightShader = needLightShader || texinfo.texgentype == XF_TEXGEN_COLOR_STRGBC0 || texinfo.texgentype == XF_TEXGEN_COLOR_STRGBC1;
}
uid_data.pixel_lighting = enable_pl;
uid_data.numColorChans = xfr.numChan.numColorChans;
if (!(api_type & API_D3D9))
{
uid_data.msaa = g_ActiveConfig.iMultisamples > 1;
uid_data.ssaa = g_ActiveConfig.iMultisamples > 1 && g_ActiveConfig.bSSAA;
}
char * buffer = nullptr;
if (Write_Code)
{
buffer = out.GetBuffer();
if (buffer == nullptr)
{
buffer = text;
out.SetBuffer(text);
}
buffer[VERTEXSHADERGEN_BUFFERSIZE - 1] = 0x7C; // canary
// uniforms
if (api_type == API_OPENGL)
out.Write("layout(std140%s) uniform VSBlock {\n", g_ActiveConfig.backend_info.bSupportsBindingLayout ? ", binding = 2" : "");
else if (api_type == API_D3D11)
out.Write("cbuffer VSBlock : register(b0) {\n");
DeclareUniform<T, api_type>(out, C_PROJECTION, "float4", I_PROJECTION"[4]");
DeclareUniform<T, api_type>(out, C_DEPTHPARAMS, "float4", I_DEPTHPARAMS);
DeclareUniform<T, api_type>(out, C_MATERIALS, "float4", I_MATERIALS"[4]");
DeclareUniform<T, api_type>(out, C_LIGHTS, "float4", I_LIGHTS"[40]");
DeclareUniform<T, api_type>(out, C_TEXMATRICES, "float4", I_TEXMATRICES"[24]");
DeclareUniform<T, api_type>(out, C_TRANSFORMMATRICES, "float4", I_TRANSFORMMATRICES"[64]");
DeclareUniform<T, api_type>(out, C_NORMALMATRICES, "float4", I_NORMALMATRICES"[32]");
DeclareUniform<T, api_type>(out, C_POSTTRANSFORMMATRICES, "float4", I_POSTTRANSFORMMATRICES"[64]");
DeclareUniform<T, api_type>(out, C_PLOFFSETPARAMS, "float4", I_PLOFFSETPARAMS"[13]");
if (api_type == API_OPENGL || api_type == API_D3D11)
out.Write("};\n");
out.Write("struct VS_OUTPUT {\n");
GenerateVSOutputMembers<T, api_type>(out, enable_pl, xfr);
out.Write("};\n");
if (api_type == API_OPENGL)
{
out.Write("in float4 rawpos; // ATTR%d,\n", SHADER_POSITION_ATTRIB);
out.Write("in float fposmtx; // ATTR%d,\n", SHADER_POSMTX_ATTRIB);
if (components & VB_HAS_NRM0)
out.Write("in float3 rawnorm0; // ATTR%d,\n", SHADER_NORM0_ATTRIB);
if (components & VB_HAS_NRM1)
out.Write("in float3 rawnorm1; // ATTR%d,\n", SHADER_NORM1_ATTRIB);
if (components & VB_HAS_NRM2)
out.Write("in float3 rawnorm2; // ATTR%d,\n", SHADER_NORM2_ATTRIB);
if (components & VB_HAS_COL0)
out.Write("in float4 color0; // ATTR%d,\n", SHADER_COLOR0_ATTRIB);
if (components & VB_HAS_COL1)
out.Write("in float4 color1; // ATTR%d,\n", SHADER_COLOR1_ATTRIB);
for (int i = 0; i < 8; ++i)
{
u32 hastexmtx = (components & (VB_HAS_TEXMTXIDX0 << i));
if ((components & (VB_HAS_UV0 << i)) || hastexmtx)
out.Write("in float%d tex%d; // ATTR%d,\n", hastexmtx ? 3 : 2, i, SHADER_TEXTURE0_ATTRIB + i);
}
if (g_ActiveConfig.backend_info.bSupportsGeometryShaders)
{
out.Write("out VertexData {\n");
GenerateVSOutputMembers<T, api_type>(out, enable_pl, xfr, GetInterpolationQualifier(api_type, false, true));
out.Write("} vs;\n");
}
else
{
const char* optCentroid = GetInterpolationQualifier(api_type);
// Let's set up attributes
if (xfr.numTexGen.numTexGens < 7)
{
for (int i = 0; i < 8; ++i)
out.Write("%s out float3 uv%d_2;\n", optCentroid, i);
out.Write("%s out float4 clipPos_2;\n", optCentroid);
if (enable_pl)
out.Write("%s out float4 Normal_2;\n", optCentroid);
}
else
{
// wpos is in w of first 4 texcoords
if (enable_pl)
{
for (int i = 0; i < 8; ++i)
out.Write("%s out float4 uv%d_2;\n", optCentroid, i);
}
else
{
for (unsigned int i = 0; i < xfr.numTexGen.numTexGens; ++i)
out.Write("%s out float%d uv%d_2;\n", optCentroid, i < 4 ? 4 : 3, i);
}
}
out.Write("%s out float4 colors_0;\n", optCentroid);
out.Write("%s out float4 colors_1;\n", optCentroid);
}
out.Write("void main()\n{\n");
}
else
{
out.Write("VS_OUTPUT main(\n");
// inputs
if (components & VB_HAS_NRM0)
out.Write(" float3 rawnorm0 : NORMAL0,\n");
if (components & VB_HAS_NRM1)
out.Write(" float3 rawnorm1 : NORMAL1,\n");
if (components & VB_HAS_NRM2)
out.Write(" float3 rawnorm2 : NORMAL2,\n");
if (components & VB_HAS_COL0)
out.Write(" float4 color0 : COLOR0,\n");
if (components & VB_HAS_COL1)
out.Write(" float4 color1 : COLOR1,\n");
for (int i = 0; i < 8; ++i)
{
u32 hastexmtx = (components & (VB_HAS_TEXMTXIDX0 << i));
if ((components & (VB_HAS_UV0 << i)) || hastexmtx)
out.Write(" float%d tex%d : TEXCOORD%d,\n", hastexmtx ? 3 : 2, i, i);
}
out.Write(" float4 blend_indices : BLENDINDICES,\n");
out.Write(" float4 rawpos : POSITION) {\n");
}
out.Write("VS_OUTPUT o;\n");
if (api_type & API_D3D9)
{
out.Write("int4 indices = D3DCOLORtoUBYTE4(blend_indices);\n");
}
// transforms
if (api_type & API_D3D9)
{
out.Write("int posmtx = indices.x;\n");
}
else if (api_type == API_D3D11)
{
out.Write("int posmtx = blend_indices.x * 255.0;\n");
}
else
{
out.Write("int posmtx = int(fposmtx);\n");
}
out.Write("float4 pos = float4(dot(" I_TRANSFORMMATRICES"[posmtx], rawpos), dot(" I_TRANSFORMMATRICES"[posmtx+1], rawpos), dot(" I_TRANSFORMMATRICES"[posmtx+2], rawpos), 1);\n");
if (components & VB_HAS_NRMALL) {
out.Write("int normidx = posmtx >= 32 ? (posmtx-32) : posmtx;\n");
out.Write("float3 N0 = " I_NORMALMATRICES"[normidx].xyz, N1 = " I_NORMALMATRICES"[normidx+1].xyz, N2 = " I_NORMALMATRICES"[normidx+2].xyz;\n");
}
if (components & VB_HAS_NRM0)
out.Write("float3 _norm0 = normalize(float3(dot(N0, rawnorm0), dot(N1, rawnorm0), dot(N2, rawnorm0)));\n");
if (components & VB_HAS_NRM1)
out.Write("float3 _norm1 = float3(dot(N0, rawnorm1), dot(N1, rawnorm1), dot(N2, rawnorm1));\n");
if (components & VB_HAS_NRM2)
out.Write("float3 _norm2 = float3(dot(N0, rawnorm2), dot(N1, rawnorm2), dot(N2, rawnorm2));\n");
if (!(components & VB_HAS_NRM0))
out.Write("float3 _norm0 = float3(0.0, 0.0, 0.0);\n");
out.Write("o.pos = float4(dot(" I_PROJECTION"[0], pos), dot(" I_PROJECTION"[1], pos), dot(" I_PROJECTION"[2], pos), dot(" I_PROJECTION"[3], pos));\n");
if (api_type & API_D3D9)
{
//Write Pos offset for Point/Line Rendering
out.Write("o.pos.xy = o.pos.xy + " I_PLOFFSETPARAMS"[indices.z].xy * o.pos.w;\n");
}
if (needLightShader)
{
out.Write("float4 mat, lacc;\n"
"float3 ldir, h;\n"
"float dist, dist2, attn;\n");
if (use_integer_math)
{
out.Write("int4 ilacc;\n");
}
}
if (!lightingEnabled)
{
if (components & VB_HAS_COL0)
out.Write("o.colors_0 = color0;\n");
else
out.Write("o.colors_0 = float4(1.0, 1.0, 1.0, 1.0);\n");
if (components & VB_HAS_COL1)
out.Write("o.colors_1 = color1;\n");
else
out.Write("o.colors_1 = o.colors_0;\n");
}
}
if (needLightShader)
GenerateLightingShader<T, Write_Code>(out, uid_data.lighting, components, I_MATERIALS, I_LIGHTS, "color", "o.colors_", xfr, use_integer_math);
// special case if only pos and tex coord 0 and tex coord input is AB11
// donko - this has caused problems in some games. removed for now.
bool texGenSpecialCase = false;
/*bool texGenSpecialCase =
((g_main_cp_state.vtx_desc.Hex & 0x60600L) == g_main_cp_state.vtx_desc.Hex) && // only pos and tex coord 0
(g_main_cp_state.vtx_desc.Tex0Coord != NOT_PRESENT) &&
(xfr.texcoords[0].texmtxinfo.inputform == XF_TEXINPUT_AB11);
*/
if (Write_Code)
{
if (xfr.numChan.numColorChans < 2 && needLightShader)
{
if (components & VB_HAS_COL1)
out.Write("o.colors_1 = color1;\n");
else
out.Write("o.colors_1 = o.colors_0;\n");
}
// transform texcoords
out.Write("float4 coord = float4(0.0, 0.0, 1.0, 1.0);\n");
}
for (unsigned int i = 0; i < xfr.numTexGen.numTexGens; ++i)
{
const TexMtxInfo& texinfo = xfr.texMtxInfo[i];
uid_data.texMtxInfo[i].sourcerow = xfr.texMtxInfo[i].sourcerow;
if (Write_Code)
{
out.Write("{\n");
out.Write("coord = float4(0.0, 0.0, 1.0, 1.0);\n");
switch (texinfo.sourcerow)
{
case XF_SRCGEOM_INROW:
_dbg_assert_log_(VIDEO, texinfo.inputform == XF_TEXINPUT_ABC1, "Incorrect inputform sourcerow: XF_SRCGEOM_INROW inputform: %u", texinfo.inputform);
out.Write("coord = rawpos;\n"); // pos.w is 1
break;
case XF_SRCNORMAL_INROW:
if (components & VB_HAS_NRM0)
{
_dbg_assert_log_(VIDEO, texinfo.inputform == XF_TEXINPUT_ABC1, "Incorrect inputform sourcerow: XF_SRCNORMAL_INROW inputform: %u", texinfo.inputform);
out.Write("coord = float4(rawnorm0.xyz, 1.0);\n");
}
break;
case XF_SRCCOLORS_INROW:
_dbg_assert_log_(VIDEO, texinfo.texgentype == XF_TEXGEN_COLOR_STRGBC0 || texinfo.texgentype == XF_TEXGEN_COLOR_STRGBC1, "texgentype missmatch: %u", texinfo.texgentype);
break;
case XF_SRCBINORMAL_T_INROW:
if (components & VB_HAS_NRM1)
{
_dbg_assert_log_(VIDEO, texinfo.inputform == XF_TEXINPUT_ABC1, "Incorrect inputform sourcerow: XF_SRCBINORMAL_T_INROW inputform: %u", texinfo.inputform);
out.Write("coord = float4(rawnorm1.xyz, 1.0);\n");
}
break;
case XF_SRCBINORMAL_B_INROW:
if (components & VB_HAS_NRM2)
{
_dbg_assert_log_(VIDEO, texinfo.inputform == XF_TEXINPUT_ABC1, "Incorrect inputform sourcerow: XF_SRCBINORMAL_B_INROW inputform: %u", texinfo.inputform);
out.Write("coord = float4(rawnorm2.xyz, 1.0);\n");
}
break;
default:
_dbg_assert_log_(VIDEO, texinfo.sourcerow <= XF_SRCTEX7_INROW, "sourcerow missmatch: %u", texinfo.sourcerow);
if (components & (VB_HAS_UV0 << (texinfo.sourcerow - XF_SRCTEX0_INROW)))
out.Write("coord = float4(tex%d.x, tex%d.y, 1.0, 1.0);\n", texinfo.sourcerow - XF_SRCTEX0_INROW, texinfo.sourcerow - XF_SRCTEX0_INROW);
break;
}
}
// first transformation
uid_data.texMtxInfo[i].texgentype = xfr.texMtxInfo[i].texgentype;
switch (texinfo.texgentype)
{
case XF_TEXGEN_EMBOSS_MAP: // calculate tex coords into bump map
if (components & (VB_HAS_NRM1 | VB_HAS_NRM2))
{
// transform the light dir into tangent space
uid_data.texMtxInfo[i].embosslightshift = xfr.texMtxInfo[i].embosslightshift;
uid_data.texMtxInfo[i].embosssourceshift = xfr.texMtxInfo[i].embosssourceshift;
if (Write_Code)
{
out.Write("float3 eldir = normalize(" LIGHT_POS".xyz - pos.xyz);\n", LIGHT_POS_PARAMS(I_LIGHTS, texinfo.embosslightshift));
out.Write("o.tex%d.xyz = o.tex%d.xyz + float3(dot(eldir, _norm1), dot(eldir, _norm2), 0.0);\n", i, texinfo.embosssourceshift);
}
}
else
{
// The following assert was triggered in House of the Dead Overkill and Star Wars Rogue Squadron 2
//
uid_data.texMtxInfo[i].embosssourceshift = xfr.texMtxInfo[i].embosssourceshift;
if (Write_Code)
{
_dbg_assert_log_(VIDEO, 0, "vertex normals spected");
out.Write("o.tex%d.xyz = o.tex%d.xyz;\n", i, texinfo.embosssourceshift);
}
}
break;
case XF_TEXGEN_COLOR_STRGBC0:
if (Write_Code)
{
_dbg_assert_log_(VIDEO, texinfo.sourcerow == XF_SRCCOLORS_INROW, "sourcerow missmatch spected: XF_SRCCOLORS_INROW found: %u", texinfo.sourcerow);
out.Write("o.tex%d.xyz = float3(o.colors_0.x, o.colors_0.y, 1);\n", i);
}
break;
case XF_TEXGEN_COLOR_STRGBC1:
if (Write_Code)
{
_dbg_assert_log_(VIDEO, texinfo.sourcerow == XF_SRCCOLORS_INROW, "sourcerow missmatch spected: XF_SRCCOLORS_INROW found: %u", texinfo.sourcerow);
out.Write("o.tex%d.xyz = float3(o.colors_1.x, o.colors_1.y, 1);\n", i);
}
break;
case XF_TEXGEN_REGULAR:
default:
uid_data.texMtxInfo_n_projection |= xfr.texMtxInfo[i].projection << i;
if (Write_Code)
{
if (components & (VB_HAS_TEXMTXIDX0 << i))
{
out.Write("int tmp = int(tex%d.z);\n", i);
if (texinfo.projection == XF_TEXPROJ_STQ)
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TRANSFORMMATRICES"[tmp]), dot(coord, " I_TRANSFORMMATRICES"[tmp+1]), dot(coord, " I_TRANSFORMMATRICES"[tmp+2]));\n", i);
else
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TRANSFORMMATRICES"[tmp]), dot(coord, " I_TRANSFORMMATRICES"[tmp+1]), 1);\n", i);
}
else
{
if (texinfo.projection == XF_TEXPROJ_STQ)
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TEXMATRICES"[%d]), dot(coord, " I_TEXMATRICES"[%d]), dot(coord, " I_TEXMATRICES"[%d]));\n", i, 3 * i, 3 * i + 1, 3 * i + 2);
else
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TEXMATRICES"[%d]), dot(coord, " I_TEXMATRICES"[%d]), 1);\n", i, 3 * i, 3 * i + 1);
}
}
break;
}
uid_data.dualTexTrans_enabled = xfr.dualTexTrans.enabled;
// CHECKME: does this only work for regular tex gen types?
if (xfr.dualTexTrans.enabled && texinfo.texgentype == XF_TEXGEN_REGULAR)
{
const PostMtxInfo& postInfo = xfr.postMtxInfo[i];
uid_data.postMtxInfo[i].index = xfr.postMtxInfo[i].index;
int postidx = postInfo.index;
if (Write_Code)
{
out.Write("float4 P0 = " I_POSTTRANSFORMMATRICES"[%d];\n"
"float4 P1 = " I_POSTTRANSFORMMATRICES"[%d];\n"
"float4 P2 = " I_POSTTRANSFORMMATRICES"[%d];\n",
postidx & 0x3f, (postidx + 1) & 0x3f, (postidx + 2) & 0x3f);
}
if (texGenSpecialCase)
{
// no normalization
// q of input is 1
// q of output is unknown
// multiply by postmatrix
if (Write_Code)
out.Write("o.tex%d.xyz = float3(dot(P0.xy, o.tex%d.xy) + P0.z + P0.w, dot(P1.xy, o.tex%d.xy) + P1.z + P1.w, 0.0);\n", i, i, i);
}
else
{
uid_data.postMtxInfo[i].normalize = xfr.postMtxInfo[i].normalize;
if (Write_Code)
{
if (postInfo.normalize)
out.Write("o.tex%d.xyz = normalize(o.tex%d.xyz);\n", i, i);
// multiply by postmatrix
out.Write("o.tex%d.xyz = float3(dot(P0.xyz, o.tex%d.xyz) + P0.w, dot(P1.xyz, o.tex%d.xyz) + P1.w, dot(P2.xyz, o.tex%d.xyz) + P2.w);\n", i, i, i, i);
}
}
}
if (Write_Code)
out.Write("}\n");
}
if (Write_Code)
{
// clipPos/w needs to be done in pixel shader, not here
if (xfr.numTexGen.numTexGens < 7)
{
out.Write("o.clipPos%s = float4(pos.x,pos.y,o.pos.z,o.pos.w);\n", (api_type == API_OPENGL) ? "_2" : "");
}
else
{
out.Write("o.tex0.w = pos.x;\n");
out.Write("o.tex1.w = pos.y;\n");
out.Write("o.tex2.w = o.pos.z;\n");
out.Write("o.tex3.w = o.pos.w;\n");
}
if (enable_pl)
{
if (xfr.numTexGen.numTexGens < 7)
{
out.Write("o.Normal%s = float4(_norm0.x,_norm0.y,_norm0.z,pos.z);\n", (api_type == API_OPENGL) ? "_2" : "");
}
else
{
out.Write("o.tex4.w = _norm0.x;\n");
out.Write("o.tex5.w = _norm0.y;\n");
out.Write("o.tex6.w = _norm0.z;\n");
if (xfr.numTexGen.numTexGens < 8)
out.Write("o.tex7 = pos.xyzz;\n");
else
out.Write("o.tex7.w = pos.z;\n");
}
if (components & VB_HAS_COL0)
out.Write("o.colors_0 = color0;\n");
else
out.Write("o.colors_0 = float4(1.0, 1.0, 1.0, 1.0);\n");
if (components & VB_HAS_COL1)
out.Write("o.colors_1 = color1;\n");
else
out.Write("o.colors_1 = o.colors_0;\n");
}
//write the true depth value, if the game uses depth textures pixel shaders will override with the correct values
//if not early z culling will improve speed
if (g_ActiveConfig.backend_info.bSupportsClipControl)
{
out.Write("o.pos.z = -o.pos.z;\n");
}
else if (api_type & API_D3D9 || api_type == API_D3D11)
{
out.Write("o.pos.z = -((" I_DEPTHPARAMS".x - 1.0) * o.pos.w + o.pos.z * " I_DEPTHPARAMS".y);\n");
}
else
{
// this results in a scale from -1..0 to -1..1 after perspective
// divide
out.Write("o.pos.z = o.pos.z * -2.0 - o.pos.w;\n");
// the next steps of the OGL pipeline are:
// (x_c,y_c,z_c,w_c) = o.pos //switch to OGL spec terminology
// clipping to -w_c <= (x_c,y_c,z_c) <= w_c
// (x_d,y_d,z_d) = (x_c,y_c,z_c)/w_c//perspective divide
// z_w = (f-n)/2*z_d + (n+f)/2
// z_w now contains the value to go to the 0..1 depth buffer
//trying to get the correct semantic while not using glDepthRange
//seems to get rather complicated
}
// The console GPU places the pixel center at 7/12 in screen space unless
// antialiasing is enabled, while D3D11 and OpenGL place it at 0.5, and D3D9 at 0. This results
// in some primitives being placed one pixel too far to the bottom-right,
// 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_DEPTHPARAMS".zw;\n");
if (api_type & API_D3D9)
{
// Write Texture Offsets for Point/Line Rendering
for (unsigned int i = 0; i < xfr.numTexGen.numTexGens; ++i)
{
out.Write("o.tex%d.xy = o.tex%d.xy + (" I_PLOFFSETPARAMS"[indices.w].zw * " I_PLOFFSETPARAMS"[indices.y + %d].%s );\n", i, i, ((i / 4) + 1), texOffsetMemberSelector[i % 4]);
}
}
if (api_type == API_OPENGL)
{
if (g_ActiveConfig.backend_info.bSupportsGeometryShaders)
{
AssignVSOutputMembers<T, api_type>(out, "vs", "o", enable_pl, xfr);
}
else
{
if (xfr.numTexGen.numTexGens < 7)
{
for (unsigned int i = 0; i < 8; ++i)
{
if (i < xfr.numTexGen.numTexGens)
out.Write(" uv%d_2.xyz = o.tex%d;\n", i, i);
else
out.Write(" uv%d_2.xyz = float3(0.0, 0.0, 0.0);\n", i);
}
out.Write(" clipPos_2 = o.clipPos;\n");
if (enable_pl)
out.Write(" Normal_2 = o.Normal;\n");
}
else
{
// clip position is in w of first 4 texcoords
if (enable_pl)
{
for (int i = 0; i < 8; ++i)
out.Write(" uv%d_2 = o.tex%d;\n", i, i);
}
else
{
for (unsigned int i = 0; i < xfr.numTexGen.numTexGens; ++i)
out.Write(" uv%d_2%s = o.tex%d;\n", i, i < 4 ? ".xyzw" : ".xyz", i);
}
}
out.Write("colors_0 = o.colors_0;\n");
out.Write("colors_1 = o.colors_1;\n");
}
out.Write("gl_Position = o.pos;\n");
out.Write("}\n");
}
else
{
out.Write("return o;\n}\n");
}
if (buffer[VERTEXSHADERGEN_BUFFERSIZE - 1] != 0x7C)
PanicAlert("VertexShader generator - buffer too small, canary has been eaten!");
}
if (uidPresent)
{
out.CalculateUIDHash();
}
}
static T GenerateVertexShader(API_TYPE api_type)
{
T out;
const u32 components = VertexLoaderManager::g_current_components;
// Non-uid template parameters will write to the dummy data (=> gets optimized out)
vertex_shader_uid_data dummy_data;
vertex_shader_uid_data* uid_data = out.template GetUidData<vertex_shader_uid_data>();
if (uid_data != nullptr)
memset(uid_data, 0, sizeof(*uid_data));
else
uid_data = &dummy_data;
_assert_(bpmem.genMode.numtexgens == xfmem.numTexGen.numTexGens);
_assert_(bpmem.genMode.numcolchans == xfmem.numChan.numColorChans);
out.Write("%s", s_lighting_struct);
// uniforms
if (api_type == API_OPENGL)
out.Write("layout(std140%s) uniform VSBlock {\n",
g_ActiveConfig.backend_info.bSupportsBindingLayout ? ", binding = 2" : "");
else
out.Write("cbuffer VSBlock {\n");
out.Write(s_shader_uniforms);
out.Write("};\n");
out.Write("struct VS_OUTPUT {\n");
GenerateVSOutputMembers<T>(out, api_type, "");
out.Write("};\n");
uid_data->numTexGens = xfmem.numTexGen.numTexGens;
uid_data->components = components;
uid_data->pixel_lighting = g_ActiveConfig.bEnablePixelLighting;
if (api_type == API_OPENGL)
{
out.Write("in float4 rawpos; // ATTR%d,\n", SHADER_POSITION_ATTRIB);
if (components & VB_HAS_POSMTXIDX)
out.Write("in int posmtx; // ATTR%d,\n", SHADER_POSMTX_ATTRIB);
if (components & VB_HAS_NRM0)
out.Write("in float3 rawnorm0; // ATTR%d,\n", SHADER_NORM0_ATTRIB);
if (components & VB_HAS_NRM1)
out.Write("in float3 rawnorm1; // ATTR%d,\n", SHADER_NORM1_ATTRIB);
if (components & VB_HAS_NRM2)
out.Write("in float3 rawnorm2; // ATTR%d,\n", SHADER_NORM2_ATTRIB);
if (components & VB_HAS_COL0)
out.Write("in float4 color0; // ATTR%d,\n", SHADER_COLOR0_ATTRIB);
if (components & VB_HAS_COL1)
out.Write("in float4 color1; // ATTR%d,\n", SHADER_COLOR1_ATTRIB);
for (int i = 0; i < 8; ++i)
{
u32 hastexmtx = (components & (VB_HAS_TEXMTXIDX0 << i));
if ((components & (VB_HAS_UV0 << i)) || hastexmtx)
out.Write("in float%d tex%d; // ATTR%d,\n", hastexmtx ? 3 : 2, i,
SHADER_TEXTURE0_ATTRIB + i);
}
if (g_ActiveConfig.backend_info.bSupportsGeometryShaders)
{
out.Write("out VertexData {\n");
GenerateVSOutputMembers<T>(out, api_type, GetInterpolationQualifier(true, false));
out.Write("} vs;\n");
}
else
{
// Let's set up attributes
for (u32 i = 0; i < 8; ++i)
{
if (i < xfmem.numTexGen.numTexGens)
{
out.Write("%s out float3 uv%u;\n", GetInterpolationQualifier(), i);
}
}
out.Write("%s out float4 clipPos;\n", GetInterpolationQualifier());
if (g_ActiveConfig.bEnablePixelLighting)
{
out.Write("%s out float3 Normal;\n", GetInterpolationQualifier());
out.Write("%s out float3 WorldPos;\n", GetInterpolationQualifier());
}
out.Write("%s out float4 colors_0;\n", GetInterpolationQualifier());
out.Write("%s out float4 colors_1;\n", GetInterpolationQualifier());
}
out.Write("void main()\n{\n");
}
else // D3D
{
out.Write("VS_OUTPUT main(\n");
// inputs
if (components & VB_HAS_NRM0)
out.Write(" float3 rawnorm0 : NORMAL0,\n");
if (components & VB_HAS_NRM1)
out.Write(" float3 rawnorm1 : NORMAL1,\n");
if (components & VB_HAS_NRM2)
out.Write(" float3 rawnorm2 : NORMAL2,\n");
if (components & VB_HAS_COL0)
out.Write(" float4 color0 : COLOR0,\n");
if (components & VB_HAS_COL1)
out.Write(" float4 color1 : COLOR1,\n");
for (int i = 0; i < 8; ++i)
{
u32 hastexmtx = (components & (VB_HAS_TEXMTXIDX0 << i));
if ((components & (VB_HAS_UV0 << i)) || hastexmtx)
out.Write(" float%d tex%d : TEXCOORD%d,\n", hastexmtx ? 3 : 2, i, i);
}
if (components & VB_HAS_POSMTXIDX)
out.Write(" int posmtx : BLENDINDICES,\n");
out.Write(" float4 rawpos : POSITION) {\n");
}
out.Write("VS_OUTPUT o;\n");
// transforms
if (components & VB_HAS_POSMTXIDX)
{
out.Write("float4 pos = float4(dot(" I_TRANSFORMMATRICES
"[posmtx], rawpos), dot(" I_TRANSFORMMATRICES
"[posmtx+1], rawpos), dot(" I_TRANSFORMMATRICES "[posmtx+2], rawpos), 1);\n");
if (components & VB_HAS_NRMALL)
{
out.Write("int normidx = posmtx & 31;\n");
out.Write("float3 N0 = " I_NORMALMATRICES "[normidx].xyz, N1 = " I_NORMALMATRICES
"[normidx+1].xyz, N2 = " I_NORMALMATRICES "[normidx+2].xyz;\n");
}
if (components & VB_HAS_NRM0)
out.Write("float3 _norm0 = normalize(float3(dot(N0, rawnorm0), dot(N1, rawnorm0), dot(N2, "
"rawnorm0)));\n");
if (components & VB_HAS_NRM1)
out.Write(
"float3 _norm1 = float3(dot(N0, rawnorm1), dot(N1, rawnorm1), dot(N2, rawnorm1));\n");
if (components & VB_HAS_NRM2)
out.Write(
"float3 _norm2 = float3(dot(N0, rawnorm2), dot(N1, rawnorm2), dot(N2, rawnorm2));\n");
}
else
{
out.Write("float4 pos = float4(dot(" I_POSNORMALMATRIX "[0], rawpos), dot(" I_POSNORMALMATRIX
"[1], rawpos), dot(" I_POSNORMALMATRIX "[2], rawpos), 1.0);\n");
if (components & VB_HAS_NRM0)
out.Write("float3 _norm0 = normalize(float3(dot(" I_POSNORMALMATRIX
"[3].xyz, rawnorm0), dot(" I_POSNORMALMATRIX
"[4].xyz, rawnorm0), dot(" I_POSNORMALMATRIX "[5].xyz, rawnorm0)));\n");
if (components & VB_HAS_NRM1)
out.Write("float3 _norm1 = float3(dot(" I_POSNORMALMATRIX
"[3].xyz, rawnorm1), dot(" I_POSNORMALMATRIX
"[4].xyz, rawnorm1), dot(" I_POSNORMALMATRIX "[5].xyz, rawnorm1));\n");
if (components & VB_HAS_NRM2)
out.Write("float3 _norm2 = float3(dot(" I_POSNORMALMATRIX
"[3].xyz, rawnorm2), dot(" I_POSNORMALMATRIX
"[4].xyz, rawnorm2), dot(" I_POSNORMALMATRIX "[5].xyz, rawnorm2));\n");
}
if (!(components & VB_HAS_NRM0))
out.Write("float3 _norm0 = float3(0.0, 0.0, 0.0);\n");
out.Write("o.pos = float4(dot(" I_PROJECTION "[0], pos), dot(" I_PROJECTION
"[1], pos), dot(" I_PROJECTION "[2], pos), dot(" I_PROJECTION "[3], pos));\n");
out.Write("int4 lacc;\n"
"float3 ldir, h, cosAttn, distAttn;\n"
"float dist, dist2, attn;\n");
uid_data->numColorChans = xfmem.numChan.numColorChans;
if (xfmem.numChan.numColorChans == 0)
{
if (components & VB_HAS_COL0)
out.Write("o.colors_0 = color0;\n");
else
out.Write("o.colors_0 = float4(1.0, 1.0, 1.0, 1.0);\n");
}
GenerateLightingShader<T>(out, uid_data->lighting, components, "color", "o.colors_");
if (xfmem.numChan.numColorChans < 2)
{
if (components & VB_HAS_COL1)
out.Write("o.colors_1 = color1;\n");
else
out.Write("o.colors_1 = o.colors_0;\n");
}
// transform texcoords
out.Write("float4 coord = float4(0.0, 0.0, 1.0, 1.0);\n");
for (unsigned int i = 0; i < xfmem.numTexGen.numTexGens; ++i)
{
TexMtxInfo& texinfo = xfmem.texMtxInfo[i];
out.Write("{\n");
out.Write("coord = float4(0.0, 0.0, 1.0, 1.0);\n");
uid_data->texMtxInfo[i].sourcerow = xfmem.texMtxInfo[i].sourcerow;
switch (texinfo.sourcerow)
{
case XF_SRCGEOM_INROW:
out.Write("coord.xyz = rawpos.xyz;\n");
break;
case XF_SRCNORMAL_INROW:
if (components & VB_HAS_NRM0)
{
out.Write("coord.xyz = rawnorm0.xyz;\n");
}
break;
case XF_SRCCOLORS_INROW:
_assert_(texinfo.texgentype == XF_TEXGEN_COLOR_STRGBC0 ||
texinfo.texgentype == XF_TEXGEN_COLOR_STRGBC1);
break;
case XF_SRCBINORMAL_T_INROW:
if (components & VB_HAS_NRM1)
{
out.Write("coord.xyz = rawnorm1.xyz;\n");
}
break;
case XF_SRCBINORMAL_B_INROW:
if (components & VB_HAS_NRM2)
{
out.Write("coord.xyz = rawnorm2.xyz;\n");
}
break;
default:
_assert_(texinfo.sourcerow <= XF_SRCTEX7_INROW);
if (components & (VB_HAS_UV0 << (texinfo.sourcerow - XF_SRCTEX0_INROW)))
out.Write("coord = float4(tex%d.x, tex%d.y, 1.0, 1.0);\n",
texinfo.sourcerow - XF_SRCTEX0_INROW, texinfo.sourcerow - XF_SRCTEX0_INROW);
break;
}
// Input form of AB11 sets z element to 1.0
uid_data->texMtxInfo[i].inputform = xfmem.texMtxInfo[i].inputform;
if (texinfo.inputform == XF_TEXINPUT_AB11)
out.Write("coord.z = 1.0;\n");
// first transformation
uid_data->texMtxInfo[i].texgentype = xfmem.texMtxInfo[i].texgentype;
switch (texinfo.texgentype)
{
case XF_TEXGEN_EMBOSS_MAP: // calculate tex coords into bump map
if (components & (VB_HAS_NRM1 | VB_HAS_NRM2))
{
// transform the light dir into tangent space
uid_data->texMtxInfo[i].embosslightshift = xfmem.texMtxInfo[i].embosslightshift;
uid_data->texMtxInfo[i].embosssourceshift = xfmem.texMtxInfo[i].embosssourceshift;
out.Write("ldir = normalize(" LIGHT_POS ".xyz - pos.xyz);\n",
LIGHT_POS_PARAMS(texinfo.embosslightshift));
out.Write(
"o.tex%d.xyz = o.tex%d.xyz + float3(dot(ldir, _norm1), dot(ldir, _norm2), 0.0);\n", i,
texinfo.embosssourceshift);
}
else
{
// The following assert was triggered in House of the Dead Overkill and Star Wars Rogue
// Squadron 2
//_assert_(0); // should have normals
uid_data->texMtxInfo[i].embosssourceshift = xfmem.texMtxInfo[i].embosssourceshift;
out.Write("o.tex%d.xyz = o.tex%d.xyz;\n", i, texinfo.embosssourceshift);
}
break;
case XF_TEXGEN_COLOR_STRGBC0:
out.Write("o.tex%d.xyz = float3(o.colors_0.x, o.colors_0.y, 1);\n", i);
break;
case XF_TEXGEN_COLOR_STRGBC1:
out.Write("o.tex%d.xyz = float3(o.colors_1.x, o.colors_1.y, 1);\n", i);
break;
case XF_TEXGEN_REGULAR:
default:
uid_data->texMtxInfo_n_projection |= xfmem.texMtxInfo[i].projection << i;
if (components & (VB_HAS_TEXMTXIDX0 << i))
{
out.Write("int tmp = int(tex%d.z);\n", i);
if (texinfo.projection == XF_TEXPROJ_STQ)
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TRANSFORMMATRICES
"[tmp]), dot(coord, " I_TRANSFORMMATRICES
"[tmp+1]), dot(coord, " I_TRANSFORMMATRICES "[tmp+2]));\n",
i);
else
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TRANSFORMMATRICES
"[tmp]), dot(coord, " I_TRANSFORMMATRICES "[tmp+1]), 1);\n",
i);
}
else
{
if (texinfo.projection == XF_TEXPROJ_STQ)
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TEXMATRICES
"[%d]), dot(coord, " I_TEXMATRICES "[%d]), dot(coord, " I_TEXMATRICES
"[%d]));\n",
i, 3 * i, 3 * i + 1, 3 * i + 2);
else
out.Write("o.tex%d.xyz = float3(dot(coord, " I_TEXMATRICES
"[%d]), dot(coord, " I_TEXMATRICES "[%d]), 1);\n",
i, 3 * i, 3 * i + 1);
}
break;
}
uid_data->dualTexTrans_enabled = xfmem.dualTexTrans.enabled;
// CHECKME: does this only work for regular tex gen types?
if (xfmem.dualTexTrans.enabled && texinfo.texgentype == XF_TEXGEN_REGULAR)
{
const PostMtxInfo& postInfo = xfmem.postMtxInfo[i];
uid_data->postMtxInfo[i].index = xfmem.postMtxInfo[i].index;
int postidx = postInfo.index;
out.Write("float4 P0 = " I_POSTTRANSFORMMATRICES "[%d];\n"
"float4 P1 = " I_POSTTRANSFORMMATRICES "[%d];\n"
"float4 P2 = " I_POSTTRANSFORMMATRICES "[%d];\n",
postidx & 0x3f, (postidx + 1) & 0x3f, (postidx + 2) & 0x3f);
uid_data->postMtxInfo[i].normalize = xfmem.postMtxInfo[i].normalize;
if (postInfo.normalize)
out.Write("o.tex%d.xyz = normalize(o.tex%d.xyz);\n", i, i);
// multiply by postmatrix
out.Write("o.tex%d.xyz = float3(dot(P0.xyz, o.tex%d.xyz) + P0.w, dot(P1.xyz, o.tex%d.xyz) + "
"P1.w, dot(P2.xyz, o.tex%d.xyz) + P2.w);\n",
i, i, i, i);
}
out.Write("}\n");
}
// clipPos/w needs to be done in pixel shader, not here
out.Write("o.clipPos = o.pos;\n");
if (g_ActiveConfig.bEnablePixelLighting)
{
out.Write("o.Normal = _norm0;\n");
out.Write("o.WorldPos = pos.xyz;\n");
if (components & VB_HAS_COL0)
out.Write("o.colors_0 = color0;\n");
if (components & VB_HAS_COL1)
out.Write("o.colors_1 = color1;\n");
}
// write the true depth value, if the game uses depth textures pixel shaders will override with
// the correct values
// if not early z culling will improve speed
if (g_ActiveConfig.backend_info.bSupportsClipControl)
{
out.Write("o.pos.z = -o.pos.z;\n");
}
else // OGL
{
// this results in a scale from -1..0 to -1..1 after perspective
// divide
out.Write("o.pos.z = o.pos.z * -2.0 - o.pos.w;\n");
// the next steps of the OGL pipeline are:
// (x_c,y_c,z_c,w_c) = o.pos //switch to OGL spec terminology
// clipping to -w_c <= (x_c,y_c,z_c) <= w_c
// (x_d,y_d,z_d) = (x_c,y_c,z_c)/w_c//perspective divide
// z_w = (f-n)/2*z_d + (n+f)/2
// z_w now contains the value to go to the 0..1 depth buffer
// trying to get the correct semantic while not using glDepthRange
// seems to get rather complicated
}
// 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
// in some primitives being placed one pixel too far to the bottom-right,
// 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");
if (api_type == API_OPENGL)
{
if (g_ActiveConfig.backend_info.bSupportsGeometryShaders)
{
AssignVSOutputMembers(out, "vs", "o");
}
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 (unsigned int i = 0; i < xfmem.numTexGen.numTexGens; ++i)
out.Write("uv%d.xyz = o.tex%d;\n", i, i);
out.Write("clipPos = o.clipPos;\n");
if (g_ActiveConfig.bEnablePixelLighting)
{
out.Write("Normal = o.Normal;\n");
out.Write("WorldPos = o.WorldPos;\n");
}
out.Write("colors_0 = o.colors_0;\n");
out.Write("colors_1 = o.colors_1;\n");
}
out.Write("gl_Position = o.pos;\n");
}
else // D3D
{
out.Write("return o;\n");
}
out.Write("}\n");
return out;
}
