ShaderGen: Split HW shadergen from base

This commit is contained in:
Connor McLaughlin 2020-09-12 14:52:18 +10:00
parent ccaff2e7cd
commit 2b4568df6a
7 changed files with 544 additions and 514 deletions

View file

@ -71,6 +71,8 @@ add_library(core
save_state_version.h
settings.cpp
settings.h
shadergen.cpp
shadergen.h
sio.cpp
sio.h
spu.cpp

View file

@ -85,6 +85,7 @@
<ClCompile Include="psf_loader.cpp" />
<ClCompile Include="resources.cpp" />
<ClCompile Include="settings.cpp" />
<ClCompile Include="shadergen.cpp" />
<ClCompile Include="sio.cpp" />
<ClCompile Include="spu.cpp" />
<ClCompile Include="system.cpp" />
@ -134,6 +135,7 @@
<ClInclude Include="resources.h" />
<ClInclude Include="save_state_version.h" />
<ClInclude Include="settings.h" />
<ClInclude Include="shadergen.h" />
<ClInclude Include="sio.h" />
<ClInclude Include="spu.h" />
<ClInclude Include="system.h" />

View file

@ -47,6 +47,7 @@
<ClCompile Include="host_interface_progress_callback.cpp" />
<ClCompile Include="pgxp.cpp" />
<ClCompile Include="cheats.cpp" />
<ClCompile Include="shadergen.cpp" />
</ItemGroup>
<ItemGroup>
<ClInclude Include="types.h" />
@ -97,5 +98,6 @@
<ClInclude Include="pgxp.h" />
<ClInclude Include="cpu_core_private.h" />
<ClInclude Include="cheats.h" />
<ClInclude Include="shadergen.h" />
</ItemGroup>
</Project>

View file

@ -5,174 +5,15 @@
#include <glad.h>
Log_SetChannel(GPU_HW_ShaderGen);
GPU_HW_ShaderGen::GPU_HW_ShaderGen(HostDisplay::RenderAPI render_api, u32 resolution_scale, bool true_color,
bool scaled_dithering, GPUTextureFilter texture_filtering, bool uv_limits,
bool supports_dual_source_blend)
: m_render_api(render_api), m_resolution_scale(resolution_scale), m_true_color(true_color),
m_scaled_dithering(scaled_dithering), m_texture_filter(texture_filtering), m_uv_limits(uv_limits),
m_glsl(render_api != HostDisplay::RenderAPI::D3D11), m_supports_dual_source_blend(supports_dual_source_blend),
m_use_glsl_interface_blocks(false)
GPU_HW_ShaderGen::GPU_HW_ShaderGen(HostDisplay::RenderAPI render_api, u32 resolution_scale, bool true_color, bool scaled_dithering, GPUTextureFilter texture_filtering, bool uv_limits, bool supports_dual_source_blend) :
ShaderGen(render_api, supports_dual_source_blend),
m_resolution_scale(resolution_scale), m_true_color(true_color),
m_scaled_dithering(scaled_dithering), m_texture_filter(texture_filtering), m_uv_limits(uv_limits)
{
if (m_glsl)
{
if (m_render_api == HostDisplay::RenderAPI::OpenGL || m_render_api == HostDisplay::RenderAPI::OpenGLES)
SetGLSLVersionString();
m_use_glsl_interface_blocks = (IsVulkan() || GLAD_GL_ES_VERSION_3_2 || GLAD_GL_VERSION_3_2);
m_use_glsl_binding_layout = (IsVulkan() || UseGLSLBindingLayout());
}
}
GPU_HW_ShaderGen::~GPU_HW_ShaderGen() = default;
bool GPU_HW_ShaderGen::UseGLSLBindingLayout()
{
return (GLAD_GL_ES_VERSION_3_1 || GLAD_GL_VERSION_4_2 ||
(GLAD_GL_ARB_explicit_attrib_location && GLAD_GL_ARB_explicit_uniform_location &&
GLAD_GL_ARB_shading_language_420pack));
}
static void DefineMacro(std::stringstream& ss, const char* name, bool enabled)
{
ss << "#define " << name << " " << BoolToUInt32(enabled) << "\n";
}
void GPU_HW_ShaderGen::SetGLSLVersionString()
{
const char* glsl_version = reinterpret_cast<const char*>(glGetString(GL_SHADING_LANGUAGE_VERSION));
const bool glsl_es = (m_render_api == HostDisplay::RenderAPI::OpenGLES);
Assert(glsl_version != nullptr);
// Skip any strings in front of the version code.
const char* glsl_version_start = glsl_version;
while (*glsl_version_start != '\0' && (*glsl_version_start < '0' || *glsl_version_start > '9'))
glsl_version_start++;
int major_version = 0, minor_version = 0;
if (std::sscanf(glsl_version_start, "%d.%d", &major_version, &minor_version) == 2)
{
// Cap at GLSL 4.3, we're not using anything newer for now.
if (!glsl_es && (major_version > 4 || (major_version == 4 && minor_version > 30)))
{
major_version = 4;
minor_version = 30;
}
else if (glsl_es && (major_version > 3 || (major_version == 3 && minor_version > 20)))
{
major_version = 3;
minor_version = 20;
}
}
else
{
Log_ErrorPrintf("Invalid GLSL version string: '%s' ('%s')", glsl_version, glsl_version_start);
if (glsl_es)
{
major_version = 3;
minor_version = 0;
}
m_glsl_version_string = glsl_es ? "300" : "130";
}
char buf[128];
std::snprintf(buf, sizeof(buf), "#version %d%02d%s", major_version, minor_version,
(glsl_es && major_version >= 3) ? " es" : "");
m_glsl_version_string = buf;
}
void GPU_HW_ShaderGen::WriteHeader(std::stringstream& ss)
{
if (m_render_api == HostDisplay::RenderAPI::OpenGL || m_render_api == HostDisplay::RenderAPI::OpenGLES)
ss << m_glsl_version_string << "\n\n";
else if (m_render_api == HostDisplay::RenderAPI::Vulkan)
ss << "#version 450 core\n\n";
// Extension enabling for OpenGL.
if (m_render_api == HostDisplay::RenderAPI::OpenGLES)
{
// Enable EXT_blend_func_extended for dual-source blend on OpenGL ES.
if (GLAD_GL_EXT_blend_func_extended)
ss << "#extension GL_EXT_blend_func_extended : require\n";
}
else if (m_render_api == HostDisplay::RenderAPI::OpenGL)
{
// Need extensions for binding layout if GL<4.3.
if (m_use_glsl_binding_layout && !GLAD_GL_VERSION_4_3)
{
ss << "#extension GL_ARB_explicit_attrib_location : require\n";
ss << "#extension GL_ARB_explicit_uniform_location : require\n";
ss << "#extension GL_ARB_shading_language_420pack : require\n";
}
if (!GLAD_GL_VERSION_3_2)
ss << "#extension GL_ARB_uniform_buffer_object : require\n";
// Enable SSBOs if it's not required by the version.
if (!GLAD_GL_VERSION_4_3 && !GLAD_GL_ES_VERSION_3_1 && GLAD_GL_ARB_shader_storage_buffer_object)
ss << "#extension GL_ARB_shader_storage_buffer_object : require\n";
}
DefineMacro(ss, "API_OPENGL", m_render_api == HostDisplay::RenderAPI::OpenGL);
DefineMacro(ss, "API_OPENGL_ES", m_render_api == HostDisplay::RenderAPI::OpenGLES);
DefineMacro(ss, "API_D3D11", m_render_api == HostDisplay::RenderAPI::D3D11);
DefineMacro(ss, "API_VULKAN", m_render_api == HostDisplay::RenderAPI::Vulkan);
if (m_render_api == HostDisplay::RenderAPI::OpenGLES)
{
ss << "precision highp float;\n";
ss << "precision highp int;\n";
ss << "precision highp sampler2D;\n";
if (GLAD_GL_ES_VERSION_3_2)
ss << "precision highp usamplerBuffer;\n";
ss << "\n";
}
if (m_glsl)
{
ss << "#define GLSL 1\n";
ss << "#define float2 vec2\n";
ss << "#define float3 vec3\n";
ss << "#define float4 vec4\n";
ss << "#define int2 ivec2\n";
ss << "#define int3 ivec3\n";
ss << "#define int4 ivec4\n";
ss << "#define uint2 uvec2\n";
ss << "#define uint3 uvec3\n";
ss << "#define uint4 uvec4\n";
ss << "#define nointerpolation flat\n";
ss << "#define frac fract\n";
ss << "#define lerp mix\n";
ss << "#define CONSTANT const\n";
ss << "#define VECTOR_EQ(a, b) ((a) == (b))\n";
ss << "#define VECTOR_NEQ(a, b) ((a) != (b))\n";
ss << "#define VECTOR_COMP_EQ(a, b) equal((a), (b))\n";
ss << "#define VECTOR_COMP_NEQ(a, b) notEqual((a), (b))\n";
ss << "#define SAMPLE_TEXTURE(name, coords) texture(name, coords)\n";
ss << "#define LOAD_TEXTURE(name, coords, mip) texelFetch(name, coords, mip)\n";
ss << "#define LOAD_TEXTURE_OFFSET(name, coords, mip, offset) texelFetchOffset(name, coords, mip, offset)\n";
ss << "#define LOAD_TEXTURE_BUFFER(name, index) texelFetch(name, index)\n";
}
else
{
ss << "#define HLSL 1\n";
ss << "#define roundEven round\n";
ss << "#define CONSTANT static const\n";
ss << "#define VECTOR_EQ(a, b) (all((a) == (b)))\n";
ss << "#define VECTOR_NEQ(a, b) (any((a) != (b)))\n";
ss << "#define VECTOR_COMP_EQ(a, b) ((a) == (b))\n";
ss << "#define VECTOR_COMP_NEQ(a, b) ((a) != (b))\n";
ss << "#define SAMPLE_TEXTURE(name, coords) name.Sample(name##_ss, coords)\n";
ss << "#define LOAD_TEXTURE(name, coords, mip) name.Load(int3(coords, mip))\n";
ss << "#define LOAD_TEXTURE_OFFSET(name, coords, mip, offset) name.Load(int3(coords, mip), offset)\n";
ss << "#define LOAD_TEXTURE_BUFFER(name, index) name.Load(index)\n";
}
ss << "\n";
}
void GPU_HW_ShaderGen::WriteCommonFunctions(std::stringstream& ss)
{
ss << "CONSTANT uint RESOLUTION_SCALE = " << m_resolution_scale << "u;\n";
@ -224,272 +65,6 @@ float4 RGBA5551ToRGBA8(uint v)
)";
}
void GPU_HW_ShaderGen::DeclareUniformBuffer(std::stringstream& ss, const std::initializer_list<const char*>& members,
bool push_constant_on_vulkan)
{
if (IsVulkan())
{
if (push_constant_on_vulkan)
ss << "layout(push_constant) uniform PushConstants\n";
else
ss << "layout(std140, set = 0, binding = 0) uniform UBOBlock\n";
}
else if (m_glsl)
{
if (m_use_glsl_binding_layout)
ss << "layout(std140, binding = 1) uniform UBOBlock\n";
else
ss << "layout(std140) uniform UBOBlock\n";
}
else
{
ss << "cbuffer UBOBlock : register(b0)\n";
}
ss << "{\n";
for (const char* member : members)
ss << member << ";\n";
ss << "};\n\n";
}
void GPU_HW_ShaderGen::DeclareTexture(std::stringstream& ss, const char* name, u32 index)
{
if (m_glsl)
{
if (IsVulkan())
ss << "layout(set = 0, binding = " << (index + 1u) << ") ";
else if (m_use_glsl_binding_layout)
ss << "layout(binding = " << index << ") ";
ss << "uniform sampler2D " << name << ";\n";
}
else
{
ss << "Texture2D " << name << " : register(t" << index << ");\n";
ss << "SamplerState " << name << "_ss : register(s" << index << ");\n";
}
}
void GPU_HW_ShaderGen::DeclareTextureBuffer(std::stringstream& ss, const char* name, u32 index, bool is_int,
bool is_unsigned)
{
if (m_glsl)
{
if (IsVulkan())
ss << "layout(set = 0, binding = " << index << ") ";
else if (m_use_glsl_binding_layout)
ss << "layout(binding = " << index << ") ";
ss << "uniform " << (is_int ? (is_unsigned ? "u" : "i") : "") << "samplerBuffer " << name << ";\n";
}
else
{
ss << "Buffer<" << (is_int ? (is_unsigned ? "uint4" : "int4") : "float4") << "> " << name << " : register(t"
<< index << ");\n";
}
}
void GPU_HW_ShaderGen::DeclareVertexEntryPoint(
std::stringstream& ss, const std::initializer_list<const char*>& attributes, u32 num_color_outputs,
u32 num_texcoord_outputs, const std::initializer_list<std::pair<const char*, const char*>>& additional_outputs,
bool declare_vertex_id, const char* output_block_suffix)
{
if (m_glsl)
{
if (m_use_glsl_binding_layout)
{
u32 attribute_counter = 0;
for (const char* attribute : attributes)
{
ss << "layout(location = " << attribute_counter << ") in " << attribute << ";\n";
attribute_counter++;
}
}
else
{
for (const char* attribute : attributes)
ss << "in " << attribute << ";\n";
}
if (m_use_glsl_interface_blocks)
{
if (IsVulkan())
ss << "layout(location = 0) ";
ss << "out VertexData" << output_block_suffix << " {\n";
for (u32 i = 0; i < num_color_outputs; i++)
ss << " float4 v_col" << i << ";\n";
for (u32 i = 0; i < num_texcoord_outputs; i++)
ss << " float2 v_tex" << i << ";\n";
for (const auto [qualifiers, name] : additional_outputs)
ss << " " << qualifiers << " " << name << ";\n";
ss << "};\n";
}
else
{
for (u32 i = 0; i < num_color_outputs; i++)
ss << "out float4 v_col" << i << ";\n";
for (u32 i = 0; i < num_texcoord_outputs; i++)
ss << "out float2 v_tex" << i << ";\n";
for (const auto [qualifiers, name] : additional_outputs)
ss << qualifiers << " out " << name << ";\n";
}
ss << "#define v_pos gl_Position\n\n";
if (declare_vertex_id)
{
if (IsVulkan())
ss << "#define v_id uint(gl_VertexIndex)\n";
else
ss << "#define v_id uint(gl_VertexID)\n";
}
ss << "\n";
ss << "void main()\n";
}
else
{
ss << "void main(\n";
if (declare_vertex_id)
ss << " in uint v_id : SV_VertexID,\n";
u32 attribute_counter = 0;
for (const char* attribute : attributes)
{
ss << " in " << attribute << " : ATTR" << attribute_counter << ",\n";
attribute_counter++;
}
for (u32 i = 0; i < num_color_outputs; i++)
ss << " out float4 v_col" << i << " : COLOR" << i << ",\n";
for (u32 i = 0; i < num_texcoord_outputs; i++)
ss << " out float2 v_tex" << i << " : TEXCOORD" << i << ",\n";
u32 additional_counter = num_texcoord_outputs;
for (const auto [qualifiers, name] : additional_outputs)
{
ss << " " << qualifiers << " out " << name << " : TEXCOORD" << additional_counter << ",\n";
additional_counter++;
}
ss << " out float4 v_pos : SV_Position)\n";
}
}
void GPU_HW_ShaderGen::DeclareFragmentEntryPoint(
std::stringstream& ss, u32 num_color_inputs, u32 num_texcoord_inputs,
const std::initializer_list<std::pair<const char*, const char*>>& additional_inputs,
bool declare_fragcoord /* = false */, u32 num_color_outputs /* = 1 */, bool depth_output /* = false */)
{
if (m_glsl)
{
if (m_use_glsl_interface_blocks)
{
if (IsVulkan())
ss << "layout(location = 0) ";
ss << "in VertexData {\n";
for (u32 i = 0; i < num_color_inputs; i++)
ss << " float4 v_col" << i << ";\n";
for (u32 i = 0; i < num_texcoord_inputs; i++)
ss << " float2 v_tex" << i << ";\n";
for (const auto [qualifiers, name] : additional_inputs)
ss << " " << qualifiers << " " << name << ";\n";
ss << "};\n";
}
else
{
for (u32 i = 0; i < num_color_inputs; i++)
ss << "in float4 v_col" << i << ";\n";
for (u32 i = 0; i < num_texcoord_inputs; i++)
ss << "in float2 v_tex" << i << ";\n";
for (const auto [qualifiers, name] : additional_inputs)
ss << qualifiers << " in " << name << ";\n";
}
if (declare_fragcoord)
ss << "#define v_pos gl_FragCoord\n";
if (depth_output)
ss << "#define o_depth gl_FragDepth\n";
if (m_use_glsl_binding_layout)
{
if (m_supports_dual_source_blend)
{
for (u32 i = 0; i < num_color_outputs; i++)
ss << "layout(location = 0, index = " << i << ") out float4 o_col" << i << ";\n";
}
else
{
Assert(num_color_outputs <= 1);
for (u32 i = 0; i < num_color_outputs; i++)
ss << "layout(location = 0" << i << ") out float4 o_col" << i << ";\n";
}
}
else
{
for (u32 i = 0; i < num_color_outputs; i++)
ss << "out float4 o_col" << i << ";\n";
}
ss << "\n";
ss << "void main()\n";
}
else
{
{
ss << "void main(\n";
for (u32 i = 0; i < num_color_inputs; i++)
ss << " in float4 v_col" << i << " : COLOR" << i << ",\n";
for (u32 i = 0; i < num_texcoord_inputs; i++)
ss << " in float2 v_tex" << i << " : TEXCOORD" << i << ",\n";
u32 additional_counter = num_texcoord_inputs;
for (const auto [qualifiers, name] : additional_inputs)
{
ss << " " << qualifiers << " in " << name << " : TEXCOORD" << additional_counter << ",\n";
additional_counter++;
}
if (declare_fragcoord)
ss << " in float4 v_pos : SV_Position,\n";
if (depth_output)
{
ss << " out float o_depth : SV_Depth";
if (num_color_outputs > 0)
ss << ",\n";
else
ss << ")\n";
}
for (u32 i = 0; i < num_color_outputs; i++)
{
ss << " out float4 o_col" << i << " : SV_Target" << i;
if (i == (num_color_outputs - 1))
ss << ")\n";
else
ss << ",\n";
}
}
}
}
void GPU_HW_ShaderGen::WriteBatchUniformBuffer(std::stringstream& ss)
{
DeclareUniformBuffer(ss,
@ -1371,41 +946,6 @@ float4 SampleFromVRAM(uint4 texpage, float2 coords)
return ss.str();
}
std::string GPU_HW_ShaderGen::GenerateScreenQuadVertexShader()
{
std::stringstream ss;
WriteHeader(ss);
DeclareVertexEntryPoint(ss, {}, 0, 1, {}, true);
ss << R"(
{
v_tex0 = float2(float((v_id << 1) & 2u), float(v_id & 2u));
v_pos = float4(v_tex0 * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), 0.0f, 1.0f);
#if API_OPENGL || API_OPENGL_ES || API_VULKAN
v_pos.y = -v_pos.y;
#endif
}
)";
return ss.str();
}
std::string GPU_HW_ShaderGen::GenerateFillFragmentShader()
{
std::stringstream ss;
WriteHeader(ss);
DeclareUniformBuffer(ss, {"float4 u_fill_color"}, true);
DeclareFragmentEntryPoint(ss, 0, 1, {}, false, 1, true);
ss << R"(
{
o_col0 = u_fill_color;
o_depth = u_fill_color.a;
}
)";
return ss.str();
}
std::string GPU_HW_ShaderGen::GenerateInterlacedFillFragmentShader()
{
std::stringstream ss;
@ -1427,24 +967,6 @@ std::string GPU_HW_ShaderGen::GenerateInterlacedFillFragmentShader()
return ss.str();
}
std::string GPU_HW_ShaderGen::GenerateCopyFragmentShader()
{
std::stringstream ss;
WriteHeader(ss);
DeclareUniformBuffer(ss, {"float4 u_src_rect"}, true);
DeclareTexture(ss, "samp0", 0);
DeclareFragmentEntryPoint(ss, 0, 1, {}, false, 1);
ss << R"(
{
float2 coords = u_src_rect.xy + v_tex0 * u_src_rect.zw;
o_col0 = SAMPLE_TEXTURE(samp0, coords);
}
)";
return ss.str();
}
std::string GPU_HW_ShaderGen::GenerateDisplayFragmentShader(bool depth_24bit,
GPU_HW::InterlacedRenderMode interlace_mode)
{

View file

@ -1,25 +1,18 @@
#pragma once
#include "gpu_hw.h"
#include "host_display.h"
#include <sstream>
#include <string>
#include "shadergen.h"
class GPU_HW_ShaderGen
class GPU_HW_ShaderGen : public ShaderGen
{
public:
GPU_HW_ShaderGen(HostDisplay::RenderAPI render_api, u32 resolution_scale, bool true_color, bool scaled_dithering,
GPUTextureFilter texture_filtering, bool uv_limits, bool supports_dual_source_blend);
~GPU_HW_ShaderGen();
static bool UseGLSLBindingLayout();
std::string GenerateBatchVertexShader(bool textured);
std::string GenerateBatchFragmentShader(GPU_HW::BatchRenderMode transparency, GPU::TextureMode texture_mode,
bool dithering, bool interlacing);
std::string GenerateScreenQuadVertexShader();
std::string GenerateFillFragmentShader();
std::string GenerateInterlacedFillFragmentShader();
std::string GenerateCopyFragmentShader();
std::string GenerateDisplayFragmentShader(bool depth_24bit, GPU_HW::InterlacedRenderMode interlace_mode);
std::string GenerateVRAMReadFragmentShader();
std::string GenerateVRAMWriteFragmentShader(bool use_ssbo);
@ -27,36 +20,13 @@ public:
std::string GenerateVRAMUpdateDepthFragmentShader();
private:
ALWAYS_INLINE bool IsVulkan() const { return (m_render_api == HostDisplay::RenderAPI::Vulkan); }
void SetGLSLVersionString();
void WriteHeader(std::stringstream& ss);
void DeclareUniformBuffer(std::stringstream& ss, const std::initializer_list<const char*>& members,
bool push_constant_on_vulkan);
void DeclareTexture(std::stringstream& ss, const char* name, u32 index);
void DeclareTextureBuffer(std::stringstream& ss, const char* name, u32 index, bool is_int, bool is_unsigned);
void DeclareVertexEntryPoint(std::stringstream& ss, const std::initializer_list<const char*>& attributes,
u32 num_color_outputs, u32 num_texcoord_outputs,
const std::initializer_list<std::pair<const char*, const char*>>& additional_outputs,
bool declare_vertex_id = false, const char* output_block_suffix = "");
void DeclareFragmentEntryPoint(std::stringstream& ss, u32 num_color_inputs, u32 num_texcoord_inputs,
const std::initializer_list<std::pair<const char*, const char*>>& additional_inputs,
bool declare_fragcoord = false, u32 num_color_outputs = 1, bool depth_output = false);
void WriteCommonFunctions(std::stringstream& ss);
void WriteBatchUniformBuffer(std::stringstream& ss);
void WriteBatchTextureFilter(std::stringstream& ss, GPUTextureFilter texture_filter);
HostDisplay::RenderAPI m_render_api;
u32 m_resolution_scale;
bool m_true_color;
bool m_scaled_dithering;
GPUTextureFilter m_texture_filter;
bool m_uv_limits;
bool m_glsl;
bool m_supports_dual_source_blend;
bool m_use_glsl_interface_blocks;
bool m_use_glsl_binding_layout;
std::string m_glsl_version_string;
};

488
src/core/shadergen.cpp Normal file
View file

@ -0,0 +1,488 @@
#include "shadergen.h"
#include "common/assert.h"
#include "common/log.h"
#include <cstdio>
#include <glad.h>
Log_SetChannel(ShaderGen);
ShaderGen::ShaderGen(HostDisplay::RenderAPI render_api, bool supports_dual_source_blend)
: m_render_api(render_api), m_glsl(render_api != HostDisplay::RenderAPI::D3D11),
m_supports_dual_source_blend(supports_dual_source_blend), m_use_glsl_interface_blocks(false)
{
if (m_glsl)
{
if (m_render_api == HostDisplay::RenderAPI::OpenGL || m_render_api == HostDisplay::RenderAPI::OpenGLES)
SetGLSLVersionString();
m_use_glsl_interface_blocks = (IsVulkan() || GLAD_GL_ES_VERSION_3_2 || GLAD_GL_VERSION_3_2);
m_use_glsl_binding_layout = (IsVulkan() || UseGLSLBindingLayout());
}
}
ShaderGen::~ShaderGen() = default;
bool ShaderGen::UseGLSLBindingLayout()
{
return (GLAD_GL_ES_VERSION_3_1 || GLAD_GL_VERSION_4_2 ||
(GLAD_GL_ARB_explicit_attrib_location && GLAD_GL_ARB_explicit_uniform_location &&
GLAD_GL_ARB_shading_language_420pack));
}
void ShaderGen::DefineMacro(std::stringstream& ss, const char* name, bool enabled)
{
ss << "#define " << name << " " << BoolToUInt32(enabled) << "\n";
}
void ShaderGen::SetGLSLVersionString()
{
const char* glsl_version = reinterpret_cast<const char*>(glGetString(GL_SHADING_LANGUAGE_VERSION));
const bool glsl_es = (m_render_api == HostDisplay::RenderAPI::OpenGLES);
Assert(glsl_version != nullptr);
// Skip any strings in front of the version code.
const char* glsl_version_start = glsl_version;
while (*glsl_version_start != '\0' && (*glsl_version_start < '0' || *glsl_version_start > '9'))
glsl_version_start++;
int major_version = 0, minor_version = 0;
if (std::sscanf(glsl_version_start, "%d.%d", &major_version, &minor_version) == 2)
{
// Cap at GLSL 4.3, we're not using anything newer for now.
if (!glsl_es && (major_version > 4 || (major_version == 4 && minor_version > 30)))
{
major_version = 4;
minor_version = 30;
}
else if (glsl_es && (major_version > 3 || (major_version == 3 && minor_version > 20)))
{
major_version = 3;
minor_version = 20;
}
}
else
{
Log_ErrorPrintf("Invalid GLSL version string: '%s' ('%s')", glsl_version, glsl_version_start);
if (glsl_es)
{
major_version = 3;
minor_version = 0;
}
m_glsl_version_string = glsl_es ? "300" : "130";
}
char buf[128];
std::snprintf(buf, sizeof(buf), "#version %d%02d%s", major_version, minor_version,
(glsl_es && major_version >= 3) ? " es" : "");
m_glsl_version_string = buf;
}
void ShaderGen::WriteHeader(std::stringstream& ss)
{
if (m_render_api == HostDisplay::RenderAPI::OpenGL || m_render_api == HostDisplay::RenderAPI::OpenGLES)
ss << m_glsl_version_string << "\n\n";
else if (m_render_api == HostDisplay::RenderAPI::Vulkan)
ss << "#version 450 core\n\n";
// Extension enabling for OpenGL.
if (m_render_api == HostDisplay::RenderAPI::OpenGLES)
{
// Enable EXT_blend_func_extended for dual-source blend on OpenGL ES.
if (GLAD_GL_EXT_blend_func_extended)
ss << "#extension GL_EXT_blend_func_extended : require\n";
}
else if (m_render_api == HostDisplay::RenderAPI::OpenGL)
{
// Need extensions for binding layout if GL<4.3.
if (m_use_glsl_binding_layout && !GLAD_GL_VERSION_4_3)
{
ss << "#extension GL_ARB_explicit_attrib_location : require\n";
ss << "#extension GL_ARB_explicit_uniform_location : require\n";
ss << "#extension GL_ARB_shading_language_420pack : require\n";
}
if (!GLAD_GL_VERSION_3_2)
ss << "#extension GL_ARB_uniform_buffer_object : require\n";
// Enable SSBOs if it's not required by the version.
if (!GLAD_GL_VERSION_4_3 && !GLAD_GL_ES_VERSION_3_1 && GLAD_GL_ARB_shader_storage_buffer_object)
ss << "#extension GL_ARB_shader_storage_buffer_object : require\n";
}
DefineMacro(ss, "API_OPENGL", m_render_api == HostDisplay::RenderAPI::OpenGL);
DefineMacro(ss, "API_OPENGL_ES", m_render_api == HostDisplay::RenderAPI::OpenGLES);
DefineMacro(ss, "API_D3D11", m_render_api == HostDisplay::RenderAPI::D3D11);
DefineMacro(ss, "API_VULKAN", m_render_api == HostDisplay::RenderAPI::Vulkan);
if (m_render_api == HostDisplay::RenderAPI::OpenGLES)
{
ss << "precision highp float;\n";
ss << "precision highp int;\n";
ss << "precision highp sampler2D;\n";
if (GLAD_GL_ES_VERSION_3_2)
ss << "precision highp usamplerBuffer;\n";
ss << "\n";
}
if (m_glsl)
{
ss << "#define GLSL 1\n";
ss << "#define float2 vec2\n";
ss << "#define float3 vec3\n";
ss << "#define float4 vec4\n";
ss << "#define int2 ivec2\n";
ss << "#define int3 ivec3\n";
ss << "#define int4 ivec4\n";
ss << "#define uint2 uvec2\n";
ss << "#define uint3 uvec3\n";
ss << "#define uint4 uvec4\n";
ss << "#define nointerpolation flat\n";
ss << "#define frac fract\n";
ss << "#define lerp mix\n";
ss << "#define CONSTANT const\n";
ss << "#define VECTOR_EQ(a, b) ((a) == (b))\n";
ss << "#define VECTOR_NEQ(a, b) ((a) != (b))\n";
ss << "#define VECTOR_COMP_EQ(a, b) equal((a), (b))\n";
ss << "#define VECTOR_COMP_NEQ(a, b) notEqual((a), (b))\n";
ss << "#define SAMPLE_TEXTURE(name, coords) texture(name, coords)\n";
ss << "#define LOAD_TEXTURE(name, coords, mip) texelFetch(name, coords, mip)\n";
ss << "#define LOAD_TEXTURE_OFFSET(name, coords, mip, offset) texelFetchOffset(name, coords, mip, offset)\n";
ss << "#define LOAD_TEXTURE_BUFFER(name, index) texelFetch(name, index)\n";
}
else
{
ss << "#define HLSL 1\n";
ss << "#define roundEven round\n";
ss << "#define CONSTANT static const\n";
ss << "#define VECTOR_EQ(a, b) (all((a) == (b)))\n";
ss << "#define VECTOR_NEQ(a, b) (any((a) != (b)))\n";
ss << "#define VECTOR_COMP_EQ(a, b) ((a) == (b))\n";
ss << "#define VECTOR_COMP_NEQ(a, b) ((a) != (b))\n";
ss << "#define SAMPLE_TEXTURE(name, coords) name.Sample(name##_ss, coords)\n";
ss << "#define LOAD_TEXTURE(name, coords, mip) name.Load(int3(coords, mip))\n";
ss << "#define LOAD_TEXTURE_OFFSET(name, coords, mip, offset) name.Load(int3(coords, mip), offset)\n";
ss << "#define LOAD_TEXTURE_BUFFER(name, index) name.Load(index)\n";
}
ss << "\n";
}
void ShaderGen::DeclareUniformBuffer(std::stringstream& ss, const std::initializer_list<const char*>& members,
bool push_constant_on_vulkan)
{
if (IsVulkan())
{
if (push_constant_on_vulkan)
ss << "layout(push_constant) uniform PushConstants\n";
else
ss << "layout(std140, set = 0, binding = 0) uniform UBOBlock\n";
}
else if (m_glsl)
{
if (m_use_glsl_binding_layout)
ss << "layout(std140, binding = 1) uniform UBOBlock\n";
else
ss << "layout(std140) uniform UBOBlock\n";
}
else
{
ss << "cbuffer UBOBlock : register(b0)\n";
}
ss << "{\n";
for (const char* member : members)
ss << member << ";\n";
ss << "};\n\n";
}
void ShaderGen::DeclareTexture(std::stringstream& ss, const char* name, u32 index)
{
if (m_glsl)
{
if (IsVulkan())
ss << "layout(set = 0, binding = " << (index + 1u) << ") ";
else if (m_use_glsl_binding_layout)
ss << "layout(binding = " << index << ") ";
ss << "uniform sampler2D " << name << ";\n";
}
else
{
ss << "Texture2D " << name << " : register(t" << index << ");\n";
ss << "SamplerState " << name << "_ss : register(s" << index << ");\n";
}
}
void ShaderGen::DeclareTextureBuffer(std::stringstream& ss, const char* name, u32 index, bool is_int, bool is_unsigned)
{
if (m_glsl)
{
if (IsVulkan())
ss << "layout(set = 0, binding = " << index << ") ";
else if (m_use_glsl_binding_layout)
ss << "layout(binding = " << index << ") ";
ss << "uniform " << (is_int ? (is_unsigned ? "u" : "i") : "") << "samplerBuffer " << name << ";\n";
}
else
{
ss << "Buffer<" << (is_int ? (is_unsigned ? "uint4" : "int4") : "float4") << "> " << name << " : register(t"
<< index << ");\n";
}
}
void ShaderGen::DeclareVertexEntryPoint(
std::stringstream& ss, const std::initializer_list<const char*>& attributes, u32 num_color_outputs,
u32 num_texcoord_outputs, const std::initializer_list<std::pair<const char*, const char*>>& additional_outputs,
bool declare_vertex_id, const char* output_block_suffix)
{
if (m_glsl)
{
if (m_use_glsl_binding_layout)
{
u32 attribute_counter = 0;
for (const char* attribute : attributes)
{
ss << "layout(location = " << attribute_counter << ") in " << attribute << ";\n";
attribute_counter++;
}
}
else
{
for (const char* attribute : attributes)
ss << "in " << attribute << ";\n";
}
if (m_use_glsl_interface_blocks)
{
if (IsVulkan())
ss << "layout(location = 0) ";
ss << "out VertexData" << output_block_suffix << " {\n";
for (u32 i = 0; i < num_color_outputs; i++)
ss << " float4 v_col" << i << ";\n";
for (u32 i = 0; i < num_texcoord_outputs; i++)
ss << " float2 v_tex" << i << ";\n";
for (const auto [qualifiers, name] : additional_outputs)
ss << " " << qualifiers << " " << name << ";\n";
ss << "};\n";
}
else
{
for (u32 i = 0; i < num_color_outputs; i++)
ss << "out float4 v_col" << i << ";\n";
for (u32 i = 0; i < num_texcoord_outputs; i++)
ss << "out float2 v_tex" << i << ";\n";
for (const auto [qualifiers, name] : additional_outputs)
ss << qualifiers << " out " << name << ";\n";
}
ss << "#define v_pos gl_Position\n\n";
if (declare_vertex_id)
{
if (IsVulkan())
ss << "#define v_id uint(gl_VertexIndex)\n";
else
ss << "#define v_id uint(gl_VertexID)\n";
}
ss << "\n";
ss << "void main()\n";
}
else
{
ss << "void main(\n";
if (declare_vertex_id)
ss << " in uint v_id : SV_VertexID,\n";
u32 attribute_counter = 0;
for (const char* attribute : attributes)
{
ss << " in " << attribute << " : ATTR" << attribute_counter << ",\n";
attribute_counter++;
}
for (u32 i = 0; i < num_color_outputs; i++)
ss << " out float4 v_col" << i << " : COLOR" << i << ",\n";
for (u32 i = 0; i < num_texcoord_outputs; i++)
ss << " out float2 v_tex" << i << " : TEXCOORD" << i << ",\n";
u32 additional_counter = num_texcoord_outputs;
for (const auto [qualifiers, name] : additional_outputs)
{
ss << " " << qualifiers << " out " << name << " : TEXCOORD" << additional_counter << ",\n";
additional_counter++;
}
ss << " out float4 v_pos : SV_Position)\n";
}
}
void ShaderGen::DeclareFragmentEntryPoint(
std::stringstream& ss, u32 num_color_inputs, u32 num_texcoord_inputs,
const std::initializer_list<std::pair<const char*, const char*>>& additional_inputs,
bool declare_fragcoord /* = false */, u32 num_color_outputs /* = 1 */, bool depth_output /* = false */)
{
if (m_glsl)
{
if (m_use_glsl_interface_blocks)
{
if (IsVulkan())
ss << "layout(location = 0) ";
ss << "in VertexData {\n";
for (u32 i = 0; i < num_color_inputs; i++)
ss << " float4 v_col" << i << ";\n";
for (u32 i = 0; i < num_texcoord_inputs; i++)
ss << " float2 v_tex" << i << ";\n";
for (const auto [qualifiers, name] : additional_inputs)
ss << " " << qualifiers << " " << name << ";\n";
ss << "};\n";
}
else
{
for (u32 i = 0; i < num_color_inputs; i++)
ss << "in float4 v_col" << i << ";\n";
for (u32 i = 0; i < num_texcoord_inputs; i++)
ss << "in float2 v_tex" << i << ";\n";
for (const auto [qualifiers, name] : additional_inputs)
ss << qualifiers << " in " << name << ";\n";
}
if (declare_fragcoord)
ss << "#define v_pos gl_FragCoord\n";
if (depth_output)
ss << "#define o_depth gl_FragDepth\n";
if (m_use_glsl_binding_layout)
{
if (m_supports_dual_source_blend)
{
for (u32 i = 0; i < num_color_outputs; i++)
ss << "layout(location = 0, index = " << i << ") out float4 o_col" << i << ";\n";
}
else
{
Assert(num_color_outputs <= 1);
for (u32 i = 0; i < num_color_outputs; i++)
ss << "layout(location = 0" << i << ") out float4 o_col" << i << ";\n";
}
}
else
{
for (u32 i = 0; i < num_color_outputs; i++)
ss << "out float4 o_col" << i << ";\n";
}
ss << "\n";
ss << "void main()\n";
}
else
{
{
ss << "void main(\n";
for (u32 i = 0; i < num_color_inputs; i++)
ss << " in float4 v_col" << i << " : COLOR" << i << ",\n";
for (u32 i = 0; i < num_texcoord_inputs; i++)
ss << " in float2 v_tex" << i << " : TEXCOORD" << i << ",\n";
u32 additional_counter = num_texcoord_inputs;
for (const auto [qualifiers, name] : additional_inputs)
{
ss << " " << qualifiers << " in " << name << " : TEXCOORD" << additional_counter << ",\n";
additional_counter++;
}
if (declare_fragcoord)
ss << " in float4 v_pos : SV_Position,\n";
if (depth_output)
{
ss << " out float o_depth : SV_Depth";
if (num_color_outputs > 0)
ss << ",\n";
else
ss << ")\n";
}
for (u32 i = 0; i < num_color_outputs; i++)
{
ss << " out float4 o_col" << i << " : SV_Target" << i;
if (i == (num_color_outputs - 1))
ss << ")\n";
else
ss << ",\n";
}
}
}
}
std::string ShaderGen::GenerateScreenQuadVertexShader()
{
std::stringstream ss;
WriteHeader(ss);
DeclareVertexEntryPoint(ss, {}, 0, 1, {}, true);
ss << R"(
{
v_tex0 = float2(float((v_id << 1) & 2u), float(v_id & 2u));
v_pos = float4(v_tex0 * float2(2.0f, -2.0f) + float2(-1.0f, 1.0f), 0.0f, 1.0f);
#if API_OPENGL || API_OPENGL_ES || API_VULKAN
v_pos.y = -v_pos.y;
#endif
}
)";
return ss.str();
}
std::string ShaderGen::GenerateFillFragmentShader()
{
std::stringstream ss;
WriteHeader(ss);
DeclareUniformBuffer(ss, {"float4 u_fill_color"}, true);
DeclareFragmentEntryPoint(ss, 0, 1, {}, false, 1, true);
ss << R"(
{
o_col0 = u_fill_color;
o_depth = u_fill_color.a;
}
)";
return ss.str();
}
std::string ShaderGen::GenerateCopyFragmentShader()
{
std::stringstream ss;
WriteHeader(ss);
DeclareUniformBuffer(ss, {"float4 u_src_rect"}, true);
DeclareTexture(ss, "samp0", 0);
DeclareFragmentEntryPoint(ss, 0, 1, {}, false, 1);
ss << R"(
{
float2 coords = u_src_rect.xy + v_tex0 * u_src_rect.zw;
o_col0 = SAMPLE_TEXTURE(samp0, coords);
}
)";
return ss.str();
}

44
src/core/shadergen.h Normal file
View file

@ -0,0 +1,44 @@
#pragma once
#include "gpu_hw.h"
#include "host_display.h"
#include <sstream>
#include <string>
class ShaderGen
{
public:
ShaderGen(HostDisplay::RenderAPI render_api, bool supports_dual_source_blend);
~ShaderGen();
static bool UseGLSLBindingLayout();
std::string GenerateScreenQuadVertexShader();
std::string GenerateFillFragmentShader();
std::string GenerateCopyFragmentShader();
protected:
ALWAYS_INLINE bool IsVulkan() const { return (m_render_api == HostDisplay::RenderAPI::Vulkan); }
void SetGLSLVersionString();
void DefineMacro(std::stringstream& ss, const char* name, bool enabled);
void WriteHeader(std::stringstream& ss);
void DeclareUniformBuffer(std::stringstream& ss, const std::initializer_list<const char*>& members,
bool push_constant_on_vulkan);
void DeclareTexture(std::stringstream& ss, const char* name, u32 index);
void DeclareTextureBuffer(std::stringstream& ss, const char* name, u32 index, bool is_int, bool is_unsigned);
void DeclareVertexEntryPoint(std::stringstream& ss, const std::initializer_list<const char*>& attributes,
u32 num_color_outputs, u32 num_texcoord_outputs,
const std::initializer_list<std::pair<const char*, const char*>>& additional_outputs,
bool declare_vertex_id = false, const char* output_block_suffix = "");
void DeclareFragmentEntryPoint(std::stringstream& ss, u32 num_color_inputs, u32 num_texcoord_inputs,
const std::initializer_list<std::pair<const char*, const char*>>& additional_inputs,
bool declare_fragcoord = false, u32 num_color_outputs = 1, bool depth_output = false);
HostDisplay::RenderAPI m_render_api;
bool m_glsl;
bool m_supports_dual_source_blend;
bool m_use_glsl_interface_blocks;
bool m_use_glsl_binding_layout;
std::string m_glsl_version_string;
};