650 Source/Core/VideoCommon/FramebufferShaderGen.cpp
View file
@@ -4,12 +4,12 @@

#include "VideoCommon/FramebufferShaderGen.h"

#include <sstream>
#include <string_view>

#include "Common/Logging/Log.h"

#include "VideoCommon/FramebufferManager.h"
#include "VideoCommon/ShaderGenCommon.h"
#include "VideoCommon/TextureDecoder.h"
#include "VideoCommon/VertexShaderGen.h"
#include "VideoCommon/VideoCommon.h"
@@ -24,38 +24,39 @@ APIType GetAPIType()
return g_ActiveConfig.backend_info.api_type;
}

void EmitUniformBufferDeclaration(std::ostringstream& ss)
void EmitUniformBufferDeclaration(ShaderCode& code)
{
if (GetAPIType() == APIType::D3D)
ss << "cbuffer PSBlock : register(b0)\n";
code.WriteFmt("cbuffer PSBlock : register(b0)\n");
else
ss << "UBO_BINDING(std140, 1) uniform PSBlock\n";
code.WriteFmt("UBO_BINDING(std140, 1) uniform PSBlock\n");
}

void EmitSamplerDeclarations(std::ostringstream& ss, u32 start = 0, u32 end = 1,
void EmitSamplerDeclarations(ShaderCode& code, u32 start = 0, u32 end = 1,
bool multisampled = false)
{
switch (GetAPIType())
{
case APIType::D3D:
{
const char* array_type = multisampled ? "Texture2DMSArray<float4>" : "Texture2DArray<float4>";

for (u32 i = start; i < end; i++)
{
ss << (multisampled ? "Texture2DMSArray<float4>" : "Texture2DArray<float4>") << " tex" << i
<< " : register(t" << i << ");\n";
ss << "SamplerState"
<< " samp" << i << " : register(s" << i << ");\n";
code.WriteFmt("{} tex{} : register(t{});\n", array_type, i, i);
code.WriteFmt("SamplerState samp{} : register(s{});\n", i, i);
}
}
break;

case APIType::OpenGL:
case APIType::Vulkan:
{
const char* array_type = multisampled ? "sampler2DMSArray" : "sampler2DArray";

for (u32 i = start; i < end; i++)
{
ss << "SAMPLER_BINDING(" << i << ") uniform "
<< (multisampled ? "sampler2DMSArray" : "sampler2DArray") << " samp" << i << ";\n";
code.WriteFmt("SAMPLER_BINDING({}) uniform {} samp{};\n", i, array_type, i);
}
}
break;
@@ -64,17 +65,17 @@ void EmitSamplerDeclarations(std::ostringstream& ss, u32 start = 0, u32 end = 1,
}
}

void EmitSampleTexture(std::ostringstream& ss, u32 n, std::string_view coords)
void EmitSampleTexture(ShaderCode& code, u32 n, std::string_view coords)
{
switch (GetAPIType())
{
case APIType::D3D:
ss << "tex" << n << ".Sample(samp" << n << ", " << coords << ')';
code.WriteFmt("tex{}.Sample(samp{}, {})", n, n, coords);
break;

case APIType::OpenGL:
case APIType::Vulkan:
ss << "texture(samp" << n << ", " << coords << ')';
code.WriteFmt("texture(samp{}, {})", n, coords);
break;

default:
@@ -84,45 +85,45 @@ void EmitSampleTexture(std::ostringstream& ss, u32 n, std::string_view coords)

// Emits a texel fetch/load instruction. Assumes that "coords" is a 4-element vector, with z
// containing the layer, and w containing the mipmap level.
void EmitTextureLoad(std::ostringstream& ss, u32 n, std::string_view coords)
void EmitTextureLoad(ShaderCode& code, u32 n, std::string_view coords)
{
switch (GetAPIType())
{
case APIType::D3D:
ss << "tex" << n << ".Load(" << coords << ')';
code.WriteFmt("tex{}.Load({})", n, coords);
break;

case APIType::OpenGL:
case APIType::Vulkan:
ss << "texelFetch(samp" << n << ", (" << coords << ").xyz, (" << coords << ").w)";
code.WriteFmt("texelFetch(samp{}, ({}).xyz, ({}).w)", n, coords, coords);
break;

default:
break;
}
}

void EmitVertexMainDeclaration(std::ostringstream& ss, u32 num_tex_inputs, u32 num_color_inputs,
void EmitVertexMainDeclaration(ShaderCode& code, u32 num_tex_inputs, u32 num_color_inputs,
bool position_input, u32 num_tex_outputs, u32 num_color_outputs,
std::string_view extra_inputs = {})
{
switch (GetAPIType())
{
case APIType::D3D:
{
ss << "void main(";
code.WriteFmt("void main(");
for (u32 i = 0; i < num_tex_inputs; i++)
ss << "in float3 rawtex" << i << " : TEXCOORD" << i << ", ";
code.WriteFmt("in float3 rawtex{} : TEXCOORD{}, ", i, i);
for (u32 i = 0; i < num_color_inputs; i++)
ss << "in float4 rawcolor" << i << " : COLOR" << i << ", ";
code.WriteFmt("in float4 rawcolor{} : COLOR{}, ", i, i);
if (position_input)
ss << "in float4 rawpos : POSITION, ";
ss << extra_inputs;
code.WriteFmt("in float4 rawpos : POSITION, ");
code.WriteFmt("{}", extra_inputs);
for (u32 i = 0; i < num_tex_outputs; i++)
ss << "out float3 v_tex" << i << " : TEXCOORD" << i << ", ";
code.WriteFmt("out float3 v_tex{} : TEXCOORD{}, ", i, i);
for (u32 i = 0; i < num_color_outputs; i++)
ss << "out float4 v_col" << i << " : COLOR" << i << ", ";
ss << "out float4 opos : SV_Position)\n";
code.WriteFmt("out float4 v_col{} : COLOR{}, ", i, i);
code.WriteFmt("out float4 opos : SV_Position)\n");
}
break;

@@ -131,59 +132,59 @@ void EmitVertexMainDeclaration(std::ostringstream& ss, u32 num_tex_inputs, u32 n
{
for (u32 i = 0; i < num_tex_inputs; i++)
{
ss << "ATTRIBUTE_LOCATION(" << (SHADER_TEXTURE0_ATTRIB + i) << ") in float3 rawtex" << i
<< ";\n";
const auto attribute = SHADER_TEXTURE0_ATTRIB + i;
code.WriteFmt("ATTRIBUTE_LOCATION({}) in float3 rawtex{};\n", attribute, i);
}
for (u32 i = 0; i < num_color_inputs; i++)
{
ss << "ATTRIBUTE_LOCATION(" << (SHADER_COLOR0_ATTRIB + i) << ") in float4 rawcolor" << i
<< ";\n";
const auto attribute = SHADER_COLOR0_ATTRIB + i;
code.WriteFmt("ATTRIBUTE_LOCATION({}) in float4 rawcolor{};\n", attribute, i);
}
if (position_input)
ss << "ATTRIBUTE_LOCATION(" << SHADER_POSITION_ATTRIB << ") in float4 rawpos;\n";
code.WriteFmt("ATTRIBUTE_LOCATION({}) in float4 rawpos;\n", SHADER_POSITION_ATTRIB);

if (g_ActiveConfig.backend_info.bSupportsGeometryShaders)
{
ss << "VARYING_LOCATION(0) out VertexData {\n";
code.WriteFmt("VARYING_LOCATION(0) out VertexData {{\n");
for (u32 i = 0; i < num_tex_outputs; i++)
ss << " float3 v_tex" << i << ";\n";
code.WriteFmt(" float3 v_tex{};\n", i);
for (u32 i = 0; i < num_color_outputs; i++)
ss << " float4 v_col" << i << ";\n";
ss << "};\n";
code.WriteFmt(" float4 v_col{};\n", i);
code.WriteFmt("}};\n");
}
else
{
for (u32 i = 0; i < num_tex_outputs; i++)
ss << "VARYING_LOCATION(" << i << ") out float3 v_tex" << i << ";\n";
code.WriteFmt("VARYING_LOCATION({}) out float3 v_tex{};\n", i, i);
for (u32 i = 0; i < num_color_outputs; i++)
ss << "VARYING_LOCATION(" << (num_tex_inputs + i) << ") out float4 v_col" << i << ";\n";
code.WriteFmt("VARYING_LOCATION({}) out float4 v_col{};\n", num_tex_inputs + i, i);
}
ss << "#define opos gl_Position\n";
ss << extra_inputs << '\n';
ss << "void main()\n";
code.WriteFmt("#define opos gl_Position\n");
code.WriteFmt("{}\n", extra_inputs);
code.WriteFmt("void main()\n");
}
break;
default:
break;
}
}

void EmitPixelMainDeclaration(std::ostringstream& ss, u32 num_tex_inputs, u32 num_color_inputs,
void EmitPixelMainDeclaration(ShaderCode& code, u32 num_tex_inputs, u32 num_color_inputs,
std::string_view output_type = "float4",
std::string_view extra_vars = {}, bool emit_frag_coord = false)
{
switch (GetAPIType())
{
case APIType::D3D:
{
ss << "void main(";
code.WriteFmt("void main(");
for (u32 i = 0; i < num_tex_inputs; i++)
ss << "in float3 v_tex" << i << " : TEXCOORD" << i << ", ";
code.WriteFmt("in float3 v_tex{} : TEXCOORD{}, ", i, i);
for (u32 i = 0; i < num_color_inputs; i++)
ss << "in float4 v_col" << i << " : COLOR" << i << ", ";
code.WriteFmt("in float4 v_col{} : COLOR{}, ", i, i);
if (emit_frag_coord)
ss << "in float4 frag_coord : SV_Position, ";
ss << extra_vars << "out " << output_type << " ocol0 : SV_Target)\n";
code.WriteFmt("in float4 frag_coord : SV_Position, ");
code.WriteFmt("{}out {} ocol0 : SV_Target)\n", extra_vars, output_type);
}
break;

@@ -192,26 +193,26 @@ void EmitPixelMainDeclaration(std::ostringstream& ss, u32 num_tex_inputs, u32 nu
{
if (g_ActiveConfig.backend_info.bSupportsGeometryShaders)
{
ss << "VARYING_LOCATION(0) in VertexData {\n";
code.WriteFmt("VARYING_LOCATION(0) in VertexData {{\n");
for (u32 i = 0; i < num_tex_inputs; i++)
ss << " in float3 v_tex" << i << ";\n";
code.WriteFmt(" in float3 v_tex{};\n", i);
for (u32 i = 0; i < num_color_inputs; i++)
ss << " in float4 v_col" << i << ";\n";
ss << "};\n";
code.WriteFmt(" in float4 v_col{};\n", i);
code.WriteFmt("}};\n");
}
else
{
for (u32 i = 0; i < num_tex_inputs; i++)
ss << "VARYING_LOCATION(" << i << ") in float3 v_tex" << i << ";\n";
code.WriteFmt("VARYING_LOCATION({}) in float3 v_tex{};\n", i, i);
for (u32 i = 0; i < num_color_inputs; i++)
ss << "VARYING_LOCATION(" << (num_tex_inputs + i) << ") in float4 v_col" << i << ";\n";
code.WriteFmt("VARYING_LOCATION({}) in float4 v_col{};\n", num_tex_inputs + i, i);
}

ss << "FRAGMENT_OUTPUT_LOCATION(0) out " << output_type << " ocol0;\n";
ss << extra_vars << "\n";
code.WriteFmt("FRAGMENT_OUTPUT_LOCATION(0) out {} ocol0;\n", output_type);
code.WriteFmt("{}\n", extra_vars);
if (emit_frag_coord)
ss << "#define frag_coord gl_FragCoord\n";
ss << "void main()\n";
code.WriteFmt("#define frag_coord gl_FragCoord\n");
code.WriteFmt("void main()\n");
}
break;

@@ -223,399 +224,406 @@ void EmitPixelMainDeclaration(std::ostringstream& ss, u32 num_tex_inputs, u32 nu

std::string GenerateScreenQuadVertexShader()
{
std::ostringstream ss;
EmitVertexMainDeclaration(ss, 0, 0, false, 1, 0,
ShaderCode code;
EmitVertexMainDeclaration(code, 0, 0, false, 1, 0,
GetAPIType() == APIType::D3D ? "in uint id : SV_VertexID, " :
"#define id gl_VertexID\n");
ss << "{\n"
" v_tex0 = float3(float((id << 1) & 2), float(id & 2), 0.0f);\n"
" opos = float4(v_tex0.xy * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), 0.0f, 1.0f);\n";
code.WriteFmt(
"{{\n"
" v_tex0 = float3(float((id << 1) & 2), float(id & 2), 0.0f);\n"
" opos = float4(v_tex0.xy * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), 0.0f, 1.0f);\n");

// NDC space is flipped in Vulkan. We also flip in GL so that (0,0) is in the lower-left.
if (GetAPIType() == APIType::Vulkan || GetAPIType() == APIType::OpenGL)
ss << " opos.y = -opos.y;\n";
code.WriteFmt(" opos.y = -opos.y;\n");

ss << "}\n";
code.WriteFmt("}}\n");

return ss.str();
return code.GetBuffer();
}

std::string GeneratePassthroughGeometryShader(u32 num_tex, u32 num_colors)
{
std::ostringstream ss;
ShaderCode code;
if (GetAPIType() == APIType::D3D)
{
ss << "struct VS_OUTPUT\n"
"{\n";
code.WriteFmt("struct VS_OUTPUT\n"
"{{\n");
for (u32 i = 0; i < num_tex; i++)
ss << " float3 tex" << i << " : TEXCOORD" << i << ";\n";
code.WriteFmt(" float3 tex{} : TEXCOORD{};\n", i, i);
for (u32 i = 0; i < num_colors; i++)
ss << " float4 color" << i << " : COLOR" << i << ";\n";
ss << " float4 position : SV_Position;\n"
"};\n";
code.WriteFmt(" float4 color{} : COLOR{};\n", i, i);
code.WriteFmt(" float4 position : SV_Position;\n"
"}};\n");

ss << "struct GS_OUTPUT\n"
"{";
code.WriteFmt("struct GS_OUTPUT\n"
"{{");
for (u32 i = 0; i < num_tex; i++)
ss << " float3 tex" << i << " : TEXCOORD" << i << ";\n";
code.WriteFmt(" float3 tex{} : TEXCOORD{};\n", i, i);
for (u32 i = 0; i < num_colors; i++)
ss << " float4 color" << i << " : COLOR" << i << ";\n";
ss << " float4 position : SV_Position;\n"
" uint slice : SV_RenderTargetArrayIndex;\n"
"};\n\n";

ss << "[maxvertexcount(6)]\n"
"void main(triangle VS_OUTPUT vso[3], inout TriangleStream<GS_OUTPUT> output)\n"
"{\n"
" for (uint slice = 0; slice < 2u; slice++)\n"
" {\n"
" for (int i = 0; i < 3; i++)\n"
" {\n"
" GS_OUTPUT gso;\n"
" gso.position = vso[i].position;\n";
code.WriteFmt(" float4 color{} : COLOR{};\n", i, i);
code.WriteFmt(" float4 position : SV_Position;\n"
" uint slice : SV_RenderTargetArrayIndex;\n"
"}};\n\n");

code.WriteFmt("[maxvertexcount(6)]\n"
"void main(triangle VS_OUTPUT vso[3], inout TriangleStream<GS_OUTPUT> output)\n"
"{{\n"
" for (uint slice = 0; slice < 2u; slice++)\n"
" {{\n"
" for (int i = 0; i < 3; i++)\n"
" {{\n"
" GS_OUTPUT gso;\n"
" gso.position = vso[i].position;\n");
for (u32 i = 0; i < num_tex; i++)
ss << " gso.tex" << i << " = float3(vso[i].tex" << i << ".xy, float(slice));\n";
code.WriteFmt(" gso.tex{} = float3(vso[i].tex{}.xy, float(slice));\n", i, i);
for (u32 i = 0; i < num_colors; i++)
ss << " gso.color" << i << " = vso[i].color" << i << ";\n";
ss << " gso.slice = slice;\n"
" output.Append(gso);\n"
" }\n"
" output.RestartStrip();\n"
" }\n"
"}\n";
code.WriteFmt(" gso.color{} = vso[i].color{};\n", i, i);
code.WriteFmt(" gso.slice = slice;\n"
" output.Append(gso);\n"
" }}\n"
" output.RestartStrip();\n"
" }}\n"
"}}\n");
}
else if (GetAPIType() == APIType::OpenGL || GetAPIType() == APIType::Vulkan)
{
ss << "layout(triangles) in;\n"
"layout(triangle_strip, max_vertices = 6) out;\n";
code.WriteFmt("layout(triangles) in;\n"
"layout(triangle_strip, max_vertices = 6) out;\n");

if (num_tex > 0 || num_colors > 0)
{
ss << "VARYING_LOCATION(0) in VertexData {\n";
code.WriteFmt("VARYING_LOCATION(0) in VertexData {{\n");
for (u32 i = 0; i < num_tex; i++)
ss << " float3 v_tex" << i << ";\n";
code.WriteFmt(" float3 v_tex{};\n", i);
for (u32 i = 0; i < num_colors; i++)
ss << " float4 v_col" << i << ";\n";
ss << "} v_in[];\n";
code.WriteFmt(" float4 v_col{};\n", i);
code.WriteFmt("}} v_in[];\n");

ss << "VARYING_LOCATION(0) out VertexData {\n";
code.WriteFmt("VARYING_LOCATION(0) out VertexData {{\n");
for (u32 i = 0; i < num_tex; i++)
ss << " float3 v_tex" << i << ";\n";
code.WriteFmt(" float3 v_tex{};\n", i);
for (u32 i = 0; i < num_colors; i++)
ss << " float4 v_col" << i << ";\n";
ss << "} v_out;\n";
code.WriteFmt(" float4 v_col{};\n", i);
code.WriteFmt("}} v_out;\n");
}
ss << "\n"
"void main()\n"
"{\n"
" for (int j = 0; j < 2; j++)\n"
" {\n"
" gl_Layer = j;\n";
code.WriteFmt("\n"
"void main()\n"
"{{\n"
" for (int j = 0; j < 2; j++)\n"
" {{\n"
" gl_Layer = j;\n");

// We have to explicitly unroll this loop otherwise the GL compiler gets cranky.
for (u32 v = 0; v < 3; v++)
{
ss << " gl_Position = gl_in[" << v << "].gl_Position;\n";
code.WriteFmt(" gl_Position = gl_in[{}].gl_Position;\n", v);
for (u32 i = 0; i < num_tex; i++)
ss << " v_out.v_tex" << i << " = float3(v_in[" << v << "].v_tex" << i
<< ".xy, float(j));\n";
{
code.WriteFmt(" v_out.v_tex{} = float3(v_in[{}].v_tex{}.xy, float(j));\n", i, v, i);
}
for (u32 i = 0; i < num_colors; i++)
ss << " v_out.v_col" << i << " = v_in[" << v << "].v_col" << i << ";\n";
ss << " EmitVertex();\n\n";
code.WriteFmt(" v_out.v_col{} = v_in[{}].v_col{};\n", i, v, i);
code.WriteFmt(" EmitVertex();\n\n");
}
ss << " EndPrimitive();\n"
" }\n"
"}\n";
code.WriteFmt(" EndPrimitive();\n"
" }}\n"
"}}\n");
}

return ss.str();
return code.GetBuffer();
}

std::string GenerateTextureCopyVertexShader()
{
std::ostringstream ss;
EmitUniformBufferDeclaration(ss);
ss << "{"
" float2 src_offset;\n"
" float2 src_size;\n"
"};\n\n";

EmitVertexMainDeclaration(ss, 0, 0, false, 1, 0,
ShaderCode code;
EmitUniformBufferDeclaration(code);
code.WriteFmt("{{"
" float2 src_offset;\n"
" float2 src_size;\n"
"}};\n\n");

EmitVertexMainDeclaration(code, 0, 0, false, 1, 0,
GetAPIType() == APIType::D3D ? "in uint id : SV_VertexID, " :
"#define id gl_VertexID");
ss << "{\n"
" v_tex0 = float3(float((id << 1) & 2), float(id & 2), 0.0f);\n"
" opos = float4(v_tex0.xy * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), 0.0f, 1.0f);\n"
" v_tex0 = float3(src_offset + (src_size * v_tex0.xy), 0.0f);\n";
code.WriteFmt(
"{{\n"
" v_tex0 = float3(float((id << 1) & 2), float(id & 2), 0.0f);\n"
" opos = float4(v_tex0.xy * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), 0.0f, 1.0f);\n"
" v_tex0 = float3(src_offset + (src_size * v_tex0.xy), 0.0f);\n");

// NDC space is flipped in Vulkan. We also flip in GL so that (0,0) is in the lower-left.
if (GetAPIType() == APIType::Vulkan || GetAPIType() == APIType::OpenGL)
ss << " opos.y = -opos.y;\n";
code.WriteFmt(" opos.y = -opos.y;\n");

ss << "}\n";
code.WriteFmt("}}\n");

return ss.str();
return code.GetBuffer();
}

std::string GenerateTextureCopyPixelShader()
{
std::ostringstream ss;
EmitSamplerDeclarations(ss, 0, 1, false);
EmitPixelMainDeclaration(ss, 1, 0);
ss << "{\n"
" ocol0 = ";
EmitSampleTexture(ss, 0, "v_tex0");
ss << ";\n"
"}\n";
return ss.str();
ShaderCode code;
EmitSamplerDeclarations(code, 0, 1, false);
EmitPixelMainDeclaration(code, 1, 0);
code.WriteFmt("{{\n"
" ocol0 = ");
EmitSampleTexture(code, 0, "v_tex0");
code.WriteFmt(";\n"
"}}\n");
return code.GetBuffer();
}

std::string GenerateColorPixelShader()
{
std::ostringstream ss;
EmitPixelMainDeclaration(ss, 0, 1);
ss << "{\n"
" ocol0 = v_col0;\n"
"}\n";
return ss.str();
ShaderCode code;
EmitPixelMainDeclaration(code, 0, 1);
code.WriteFmt("{{\n"
" ocol0 = v_col0;\n"
"}}\n");
return code.GetBuffer();
}

std::string GenerateResolveDepthPixelShader(u32 samples)
{
std::ostringstream ss;
EmitSamplerDeclarations(ss, 0, 1, true);
EmitPixelMainDeclaration(ss, 1, 0, "float",
ShaderCode code;
EmitSamplerDeclarations(code, 0, 1, true);
EmitPixelMainDeclaration(code, 1, 0, "float",
GetAPIType() == APIType::D3D ? "in float4 ipos : SV_Position, " : "");
ss << "{\n"
" int layer = int(v_tex0.z);\n";
code.WriteFmt("{{\n"
" int layer = int(v_tex0.z);\n");
if (GetAPIType() == APIType::D3D)
ss << " int3 coords = int3(int2(ipos.xy), layer);\n";
code.WriteFmt(" int3 coords = int3(int2(ipos.xy), layer);\n");
else
ss << " int3 coords = int3(int2(gl_FragCoord.xy), layer);\n";
code.WriteFmt(" int3 coords = int3(int2(gl_FragCoord.xy), layer);\n");

// Take the minimum of all depth samples.
if (GetAPIType() == APIType::D3D)
ss << " ocol0 = tex0.Load(coords, 0).r;\n";
code.WriteFmt(" ocol0 = tex0.Load(coords, 0).r;\n");
else
ss << " ocol0 = texelFetch(samp0, coords, 0).r;\n";
ss << " for (int i = 1; i < " << samples << "; i++)\n";
code.WriteFmt(" ocol0 = texelFetch(samp0, coords, 0).r;\n");
code.WriteFmt(" for (int i = 1; i < {}; i++)\n", samples);
if (GetAPIType() == APIType::D3D)
ss << " ocol0 = min(ocol0, tex0.Load(coords, i).r);\n";
code.WriteFmt(" ocol0 = min(ocol0, tex0.Load(coords, i).r);\n");
else
ss << " ocol0 = min(ocol0, texelFetch(samp0, coords, i).r);\n";
code.WriteFmt(" ocol0 = min(ocol0, texelFetch(samp0, coords, i).r);\n");

ss << "}\n";
return ss.str();
code.WriteFmt("}}\n");
return code.GetBuffer();
}

std::string GenerateClearVertexShader()
{
std::ostringstream ss;
EmitUniformBufferDeclaration(ss);
ss << "{\n"
" float4 clear_color;\n"
" float clear_depth;\n"
"};\n";

EmitVertexMainDeclaration(ss, 0, 0, false, 0, 1,
ShaderCode code;
EmitUniformBufferDeclaration(code);
code.WriteFmt("{{\n"
" float4 clear_color;\n"
" float clear_depth;\n"
"}};\n");

EmitVertexMainDeclaration(code, 0, 0, false, 0, 1,
GetAPIType() == APIType::D3D ? "in uint id : SV_VertexID, " :
"#define id gl_VertexID\n");
ss << "{\n"
" float2 coord = float2(float((id << 1) & 2), float(id & 2));\n"
" opos = float4(coord * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), clear_depth, 1.0f);\n"
" v_col0 = clear_color;\n";
code.WriteFmt(
"{{\n"
" float2 coord = float2(float((id << 1) & 2), float(id & 2));\n"
" opos = float4(coord * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), clear_depth, 1.0f);\n"
" v_col0 = clear_color;\n");

// NDC space is flipped in Vulkan
if (GetAPIType() == APIType::Vulkan)
ss << " opos.y = -opos.y;\n";
code.WriteFmt(" opos.y = -opos.y;\n");

ss << "}\n";
code.WriteFmt("}}\n");

return ss.str();
return code.GetBuffer();
}

std::string GenerateEFBPokeVertexShader()
{
std::ostringstream ss;
EmitVertexMainDeclaration(ss, 0, 1, true, 0, 1);
ss << "{\n"
" v_col0 = rawcolor0;\n"
" opos = float4(rawpos.xyz, 1.0f);\n";
ShaderCode code;
EmitVertexMainDeclaration(code, 0, 1, true, 0, 1);
code.WriteFmt("{{\n"
" v_col0 = rawcolor0;\n"
" opos = float4(rawpos.xyz, 1.0f);\n");
if (g_ActiveConfig.backend_info.bSupportsLargePoints)
ss << " gl_PointSize = rawpos.w;\n";
code.WriteFmt(" gl_PointSize = rawpos.w;\n");

// NDC space is flipped in Vulkan.
if (GetAPIType() == APIType::Vulkan)
ss << " opos.y = -opos.y;\n";
code.WriteFmt(" opos.y = -opos.y;\n");

ss << "}\n";
return ss.str();
code.WriteFmt("}}\n");
return code.GetBuffer();
}

std::string GenerateFormatConversionShader(EFBReinterpretType convtype, u32 samples)
{
std::ostringstream ss;
EmitSamplerDeclarations(ss, 0, 1, samples > 1);
ShaderCode code;
EmitSamplerDeclarations(code, 0, 1, samples > 1);
EmitPixelMainDeclaration(
ss, 1, 0, "float4",
code, 1, 0, "float4",
GetAPIType() == APIType::D3D ?
(g_ActiveConfig.bSSAA ?
"in float4 ipos : SV_Position, in uint isample : SV_SampleIndex, " :
"in float4 ipos : SV_Position, ") :
"");
ss << "{\n"
" int layer = int(v_tex0.z);\n";
code.WriteFmt("{{\n"
" int layer = int(v_tex0.z);\n");
if (GetAPIType() == APIType::D3D)
ss << " int3 coords = int3(int2(ipos.xy), layer);\n";
code.WriteFmt(" int3 coords = int3(int2(ipos.xy), layer);\n");
else
ss << " int3 coords = int3(int2(gl_FragCoord.xy), layer);\n";
code.WriteFmt(" int3 coords = int3(int2(gl_FragCoord.xy), layer);\n");

if (samples == 1)
{
// No MSAA at all.
if (GetAPIType() == APIType::D3D)
ss << " float4 val = tex0.Load(int4(coords, 0));\n";
code.WriteFmt(" float4 val = tex0.Load(int4(coords, 0));\n");
else
ss << " float4 val = texelFetch(samp0, coords, 0);\n";
code.WriteFmt(" float4 val = texelFetch(samp0, coords, 0);\n");
}
else if (g_ActiveConfig.bSSAA)
{
// Sample shading, shader runs once per sample
if (GetAPIType() == APIType::D3D)
ss << " float4 val = tex0.Load(coords, isample);";
code.WriteFmt(" float4 val = tex0.Load(coords, isample);");
else
ss << " float4 val = texelFetch(samp0, coords, gl_SampleID);";
code.WriteFmt(" float4 val = texelFetch(samp0, coords, gl_SampleID);");
}
else
{
// MSAA without sample shading, average out all samples.
ss << " float4 val = float4(0.0f, 0.0f, 0.0f, 0.0f);\n";
ss << " for (int i = 0; i < " << samples << "; i++)\n";
code.WriteFmt(" float4 val = float4(0.0f, 0.0f, 0.0f, 0.0f);\n");
code.WriteFmt(" for (int i = 0; i < {}; i++)\n", samples);
if (GetAPIType() == APIType::D3D)
ss << " val += tex0.Load(coords, i);\n";
code.WriteFmt(" val += tex0.Load(coords, i);\n");
else
ss << " val += texelFetch(samp0, coords, i);\n";
ss << " val /= float(" << samples << ");\n";
code.WriteFmt(" val += texelFetch(samp0, coords, i);\n");
code.WriteFmt(" val /= float({});\n", samples);
}

switch (convtype)
{
case EFBReinterpretType::RGB8ToRGBA6:
ss << " int4 src8 = int4(round(val * 255.f));\n"
" int4 dst6;\n"
" dst6.r = src8.r >> 2;\n"
" dst6.g = ((src8.r & 0x3) << 4) | (src8.g >> 4);\n"
" dst6.b = ((src8.g & 0xF) << 2) | (src8.b >> 6);\n"
" dst6.a = src8.b & 0x3F;\n"
" ocol0 = float4(dst6) / 63.f;\n";
code.WriteFmt(" int4 src8 = int4(round(val * 255.f));\n"
" int4 dst6;\n"
" dst6.r = src8.r >> 2;\n"
" dst6.g = ((src8.r & 0x3) << 4) | (src8.g >> 4);\n"
" dst6.b = ((src8.g & 0xF) << 2) | (src8.b >> 6);\n"
" dst6.a = src8.b & 0x3F;\n"
" ocol0 = float4(dst6) / 63.f;\n");
break;

case EFBReinterpretType::RGB8ToRGB565:
ss << " ocol0 = val;\n";
code.WriteFmt(" ocol0 = val;\n");
break;

case EFBReinterpretType::RGBA6ToRGB8:
ss << " int4 src6 = int4(round(val * 63.f));\n"
" int4 dst8;\n"
" dst8.r = (src6.r << 2) | (src6.g >> 4);\n"
" dst8.g = ((src6.g & 0xF) << 4) | (src6.b >> 2);\n"
" dst8.b = ((src6.b & 0x3) << 6) | src6.a;\n"
" dst8.a = 255;\n"
" ocol0 = float4(dst8) / 255.f;\n";
code.WriteFmt(" int4 src6 = int4(round(val * 63.f));\n"
" int4 dst8;\n"
" dst8.r = (src6.r << 2) | (src6.g >> 4);\n"
" dst8.g = ((src6.g & 0xF) << 4) | (src6.b >> 2);\n"
" dst8.b = ((src6.b & 0x3) << 6) | src6.a;\n"
" dst8.a = 255;\n"
" ocol0 = float4(dst8) / 255.f;\n");
break;

case EFBReinterpretType::RGBA6ToRGB565:
ss << " ocol0 = val;\n";
code.WriteFmt(" ocol0 = val;\n");
break;

case EFBReinterpretType::RGB565ToRGB8:
ss << " ocol0 = val;\n";
code.WriteFmt(" ocol0 = val;\n");
break;

case EFBReinterpretType::RGB565ToRGBA6:
//
ss << " ocol0 = val;\n";
code.WriteFmt(" ocol0 = val;\n");
break;
}

ss << "}\n";
return ss.str();
code.WriteFmt("}}\n");
return code.GetBuffer();
}

std::string GenerateTextureReinterpretShader(TextureFormat from_format, TextureFormat to_format)
{
std::ostringstream ss;
EmitSamplerDeclarations(ss, 0, 1, false);
EmitPixelMainDeclaration(ss, 1, 0, "float4", "", true);
ss << "{\n"
" int layer = int(v_tex0.z);\n"
" int4 coords = int4(int2(frag_coord.xy), layer, 0);\n";
ShaderCode code;
EmitSamplerDeclarations(code, 0, 1, false);
EmitPixelMainDeclaration(code, 1, 0, "float4", "", true);
code.WriteFmt("{{\n"
" int layer = int(v_tex0.z);\n"
" int4 coords = int4(int2(frag_coord.xy), layer, 0);\n");

// Convert to a 32-bit value encompassing all channels, filling the most significant bits with
// zeroes.
ss << " uint raw_value;\n";
code.WriteFmt(" uint raw_value;\n");
switch (from_format)
{
case TextureFormat::I8:
case TextureFormat::C8:
{
ss << " float4 temp_value = ";
EmitTextureLoad(ss, 0, "coords");
ss << ";\n"
" raw_value = uint(temp_value.r * 255.0);\n";
code.WriteFmt(" float4 temp_value = ");
EmitTextureLoad(code, 0, "coords");
code.WriteFmt(";\n"
" raw_value = uint(temp_value.r * 255.0);\n");
}
break;

case TextureFormat::IA8:
{
ss << " float4 temp_value = ";
EmitTextureLoad(ss, 0, "coords");
ss << ";\n"
" raw_value = uint(temp_value.r * 255.0) | (uint(temp_value.a * 255.0) << 8);\n";
code.WriteFmt(" float4 temp_value = ");
EmitTextureLoad(code, 0, "coords");
code.WriteFmt(
";\n"
" raw_value = uint(temp_value.r * 255.0) | (uint(temp_value.a * 255.0) << 8);\n");
}
break;

case TextureFormat::I4:
{
ss << " float4 temp_value = ";
EmitTextureLoad(ss, 0, "coords");
ss << ";\n"
" raw_value = uint(temp_value.r * 15.0);\n";
code.WriteFmt(" float4 temp_value = ");
EmitTextureLoad(code, 0, "coords");
code.WriteFmt(";\n"
" raw_value = uint(temp_value.r * 15.0);\n");
}
break;

case TextureFormat::IA4:
{
ss << " float4 temp_value = ";
EmitTextureLoad(ss, 0, "coords");
ss << ";\n"
" raw_value = uint(temp_value.r * 15.0) | (uint(temp_value.a * 15.0) << 4);\n";
code.WriteFmt(" float4 temp_value = ");
EmitTextureLoad(code, 0, "coords");
code.WriteFmt(";\n"
" raw_value = uint(temp_value.r * 15.0) | (uint(temp_value.a * 15.0) << 4);\n");
}
break;

case TextureFormat::RGB565:
{
ss << " float4 temp_value = ";
EmitTextureLoad(ss, 0, "coords");
ss << ";\n"
" raw_value = uint(temp_value.b * 31.0) | (uint(temp_value.g * 63.0) << 5) |\n"
" (uint(temp_value.r * 31.0) << 11);\n";
code.WriteFmt(" float4 temp_value = ");
EmitTextureLoad(code, 0, "coords");
code.WriteFmt(";\n"
" raw_value = uint(temp_value.b * 31.0) | (uint(temp_value.g * 63.0) << 5) |\n"
" (uint(temp_value.r * 31.0) << 11);\n");
}
break;

case TextureFormat::RGB5A3:
{
ss << " float4 temp_value = ";
EmitTextureLoad(ss, 0, "coords");
ss << ";\n";
code.WriteFmt(" float4 temp_value = ");
EmitTextureLoad(code, 0, "coords");
code.WriteFmt(";\n");

// 0.8784 = 224 / 255 which is the maximum alpha value that can be represented in 3 bits
ss << " if (temp_value.a > 0.878f) {\n"
" raw_value = (uint(temp_value.b * 31.0)) | (uint(temp_value.g * 31.0) << 5) |\n"
" (uint(temp_value.r * 31.0) << 10) | 0x8000u;\n"
" } else {\n"
" raw_value = (uint(temp_value.b * 15.0)) | (uint(temp_value.g * 15.0) << 4) |\n"
" (uint(temp_value.r * 15.0) << 8) | (uint(temp_value.a * 7.0) << 12);\n"
" }\n";
code.WriteFmt(
" if (temp_value.a > 0.878f) {{\n"
" raw_value = (uint(temp_value.b * 31.0)) | (uint(temp_value.g * 31.0) << 5) |\n"
" (uint(temp_value.r * 31.0) << 10) | 0x8000u;\n"
" }} else {{\n"
" raw_value = (uint(temp_value.b * 15.0)) | (uint(temp_value.g * 15.0) << 4) |\n"
" (uint(temp_value.r * 15.0) << 8) | (uint(temp_value.a * 7.0) << 12);\n"
" }}\n");
}
break;

@@ -630,110 +638,110 @@ std::string GenerateTextureReinterpretShader(TextureFormat from_format, TextureF
case TextureFormat::I8:
case TextureFormat::C8:
{
ss << " float orgba = float(raw_value & 0xFFu) / 255.0;\n"
" ocol0 = float4(orgba, orgba, orgba, orgba);\n";
code.WriteFmt(" float orgba = float(raw_value & 0xFFu) / 255.0;\n"
" ocol0 = float4(orgba, orgba, orgba, orgba);\n");
}
break;

case TextureFormat::IA8:
{
ss << " float orgb = float(raw_value & 0xFFu) / 255.0;\n"
" ocol0 = float4(orgb, orgb, orgb, float((raw_value >> 8) & 0xFFu) / 255.0);\n";
code.WriteFmt(" float orgb = float(raw_value & 0xFFu) / 255.0;\n"
" ocol0 = float4(orgb, orgb, orgb, float((raw_value >> 8) & 0xFFu) / 255.0);\n");
}
break;

case TextureFormat::IA4:
{
ss << " float orgb = float(raw_value & 0xFu) / 15.0;\n"
" ocol0 = float4(orgb, orgb, orgb, float((raw_value >> 4) & 0xFu) / 15.0);\n";
code.WriteFmt(" float orgb = float(raw_value & 0xFu) / 15.0;\n"
" ocol0 = float4(orgb, orgb, orgb, float((raw_value >> 4) & 0xFu) / 15.0);\n");
}
break;

case TextureFormat::RGB565:
{
ss << " ocol0 = float4(float((raw_value >> 10) & 0x1Fu) / 31.0,\n"
" float((raw_value >> 5) & 0x1Fu) / 31.0,\n"
" float(raw_value & 0x1Fu) / 31.0, 1.0);\n";
code.WriteFmt(" ocol0 = float4(float((raw_value >> 10) & 0x1Fu) / 31.0,\n"
" float((raw_value >> 5) & 0x1Fu) / 31.0,\n"
" float(raw_value & 0x1Fu) / 31.0, 1.0);\n");
}
break;

case TextureFormat::RGB5A3:
{
ss << " if ((raw_value & 0x8000u) != 0u) {\n"
" ocol0 = float4(float((raw_value >> 10) & 0x1Fu) / 31.0,\n"
" float((raw_value >> 5) & 0x1Fu) / 31.0,\n"
" float(raw_value & 0x1Fu) / 31.0, 1.0);\n"
" } else {\n"
" ocol0 = float4(float((raw_value >> 8) & 0x0Fu) / 15.0,\n"
" float((raw_value >> 4) & 0x0Fu) / 15.0,\n"
" float(raw_value & 0x0Fu) / 15.0,\n"
" float((raw_value >> 12) & 0x07u) / 7.0);\n"
" }\n";
code.WriteFmt(" if ((raw_value & 0x8000u) != 0u) {{\n"
" ocol0 = float4(float((raw_value >> 10) & 0x1Fu) / 31.0,\n"
" float((raw_value >> 5) & 0x1Fu) / 31.0,\n"
" float(raw_value & 0x1Fu) / 31.0, 1.0);\n"
" }} else {{\n"
" ocol0 = float4(float((raw_value >> 8) & 0x0Fu) / 15.0,\n"
" float((raw_value >> 4) & 0x0Fu) / 15.0,\n"
" float(raw_value & 0x0Fu) / 15.0,\n"
" float((raw_value >> 12) & 0x07u) / 7.0);\n"
" }}\n");
}
break;
default:
WARN_LOG(VIDEO, "To format %u is not supported", static_cast<u32>(to_format));
return "{}\n";
}

ss << "}\n";
return ss.str();
code.WriteFmt("}}\n");
return code.GetBuffer();
}

std::string GenerateEFBRestorePixelShader()
{
std::ostringstream ss;
EmitSamplerDeclarations(ss, 0, 2, false);
EmitPixelMainDeclaration(ss, 1, 0, "float4",
ShaderCode code;
EmitSamplerDeclarations(code, 0, 2, false);
EmitPixelMainDeclaration(code, 1, 0, "float4",
GetAPIType() == APIType::D3D ? "out float depth : SV_Depth, " : "");
ss << "{\n"
" ocol0 = ";
EmitSampleTexture(ss, 0, "v_tex0");
ss << ";\n";
ss << " " << (GetAPIType() == APIType::D3D ? "depth" : "gl_FragDepth") << " = ";
EmitSampleTexture(ss, 1, "v_tex0");
ss << ".r;\n"
"}\n";
return ss.str();
code.WriteFmt("{{\n"
" ocol0 = ");
EmitSampleTexture(code, 0, "v_tex0");
code.WriteFmt(";\n");
code.WriteFmt(" {} = ", GetAPIType() == APIType::D3D ? "depth" : "gl_FragDepth");
EmitSampleTexture(code, 1, "v_tex0");
code.WriteFmt(".r;\n"
"}}\n");
return code.GetBuffer();
}

std::string GenerateImGuiVertexShader()
{
std::ostringstream ss;
ShaderCode code;

// Uniform buffer contains the viewport size, and we transform in the vertex shader.
EmitUniformBufferDeclaration(ss);
ss << "{\n"
"float2 u_rcp_viewport_size_mul2;\n"
"};\n\n";

EmitVertexMainDeclaration(ss, 1, 1, true, 1, 1);
ss << "{\n"
" v_tex0 = float3(rawtex0.xy, 0.0);\n"
" v_col0 = rawcolor0;\n"
" opos = float4(rawpos.x * u_rcp_viewport_size_mul2.x - 1.0,"
" 1.0 - rawpos.y * u_rcp_viewport_size_mul2.y, 0.0, 1.0);\n";
EmitUniformBufferDeclaration(code);
code.WriteFmt("{{\n"
"float2 u_rcp_viewport_size_mul2;\n"
"}};\n\n");

EmitVertexMainDeclaration(code, 1, 1, true, 1, 1);
code.WriteFmt("{{\n"
" v_tex0 = float3(rawtex0.xy, 0.0);\n"
" v_col0 = rawcolor0;\n"
" opos = float4(rawpos.x * u_rcp_viewport_size_mul2.x - 1.0,"
" 1.0 - rawpos.y * u_rcp_viewport_size_mul2.y, 0.0, 1.0);\n");

// NDC space is flipped in Vulkan.
if (GetAPIType() == APIType::Vulkan)
ss << " opos.y = -opos.y;\n";
code.WriteFmt(" opos.y = -opos.y;\n");

ss << "}\n";
return ss.str();
code.WriteFmt("}}\n");
return code.GetBuffer();
}

std::string GenerateImGuiPixelShader()
{
std::ostringstream ss;
EmitSamplerDeclarations(ss, 0, 1, false);
EmitPixelMainDeclaration(ss, 1, 1);
ss << "{\n"
" ocol0 = ";
EmitSampleTexture(ss, 0, "float3(v_tex0.xy, 0.0)");
ss << " * v_col0;\n"
"}\n";

return ss.str();
ShaderCode code;
EmitSamplerDeclarations(code, 0, 1, false);
EmitPixelMainDeclaration(code, 1, 1);
code.WriteFmt("{{\n"
" ocol0 = ");
EmitSampleTexture(code, 0, "float3(v_tex0.xy, 0.0)");
code.WriteFmt(" * v_col0;\n"
"}}\n");

return code.GetBuffer();
}

} // namespace FramebufferShaderGen