Merge pull request #1519 from vlj/rsx

Rsx: Vertex attribute format fixes + ssa variable support

rpcs3-tests/ps3-rsx-common.cpp

@@ -0,0 +1,15 @@
+#include "stdafx.h"
+#include "Emu\RSX\Common\BufferUtils.h"
+
+
+TEST_CLASS(rsx_common)
+{
+	// Check UB256 size is correctly set in write_vertex_array_data_to_buffer
+	TEST_METHOD(ub256_upload)
+	{
+		std::vector<gsl::byte> dest_buffer(2200);
+		std::vector<gsl::byte> src_buffer(100000); // Big enough
+
+		write_vertex_array_data_to_buffer(gsl::span<gsl::byte>(dest_buffer), src_buffer.data(), 0, 550, rsx::vertex_base_type::ub256, 4, 20);
+	}
+};

rpcs3-tests/rpcs3-tests.vcxproj

@@ -56,7 +56,7 @@
     </ClCompile>
     <Link>
       <AdditionalLibraryDirectories>$(VCInstallDir)UnitTest\lib;..\OpenAL\libs\Win64;..\ffmpeg\Windows\x86_64\lib;%(AdditionalLibraryDirectories)</AdditionalLibraryDirectories>
-   </Link>
+    </Link>
   </ItemDefinitionGroup>
   <ItemDefinitionGroup Condition="'$(Configuration)|$(Platform)'=='Release|x64'">
     <ClCompile>
@@ -76,6 +76,7 @@
     <ClInclude Include="stdafx.h" />
   </ItemGroup>
   <ItemGroup>
+    <ClCompile Include="ps3-rsx-common.cpp" />
     <ClCompile Include="ps3_ppu_llvm.cpp" />
     <ClCompile Include="stdafx.cpp">
       <PrecompiledHeader Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">Create</PrecompiledHeader>

rpcs3-tests/rpcs3-tests.vcxproj.filters

@@ -17,6 +17,9 @@
     <ClCompile Include="ps3_ppu_llvm.cpp">
       <Filter>Source Files</Filter>
     </ClCompile>
+    <ClCompile Include="ps3-rsx-common.cpp">
+      <Filter>Source Files</Filter>
+    </ClCompile>
   </ItemGroup>
   <ItemGroup>
     <ClInclude Include="stdafx.h">

rpcs3/Emu/RSX/Common/BufferUtils.cpp

@@ -58,6 +58,7 @@ void write_vertex_array_data_to_buffer(gsl::span<gsl::byte> raw_dst_span, const
 	switch (type)
 	{
 	case rsx::vertex_base_type::ub:
+	case rsx::vertex_base_type::ub256:
 	{
 		gsl::span<u8> dst_span = as_span_workaround<u8>(raw_dst_span);
 		copy_whole_attribute_array<u8>(dst_span, src_ptr, vector_element_count, element_size, attribute_src_stride, first, count);
@@ -65,14 +66,13 @@ void write_vertex_array_data_to_buffer(gsl::span<gsl::byte> raw_dst_span, const
 	}
 	case rsx::vertex_base_type::s1:
 	case rsx::vertex_base_type::sf:
+	case rsx::vertex_base_type::s32k:
 	{
 		gsl::span<u16> dst_span = as_span_workaround<u16>(raw_dst_span);
 		copy_whole_attribute_array<be_t<u16>>(dst_span, src_ptr, vector_element_count, element_size, attribute_src_stride, first, count);
 		return;
 	}
 	case rsx::vertex_base_type::f:
-	case rsx::vertex_base_type::s32k:
-	case rsx::vertex_base_type::ub256:
 	{
 		gsl::span<u32> dst_span = as_span_workaround<u32>(raw_dst_span);
 		copy_whole_attribute_array<be_t<u32>>(dst_span, src_ptr, vector_element_count, element_size, attribute_src_stride, first, count);

rpcs3/Emu/RSX/Common/TextureUtils.cpp

@@ -295,6 +295,8 @@ std::vector<MipmapLevelInfo> upload_placed_texture(gsl::span<gsl::byte> mapped_b
 			return copy_texture_data<copy_unmodified_block_swizzled, false, 1>(as_span_workaround<u8>(mapped_buffer), reinterpret_cast<const u8*>(pixels), w, h, depth, layer, texture.mipmap(), texture.pitch());
 		else
 			return copy_texture_data<copy_unmodified_block, true, 1>(as_span_workaround<u8>(mapped_buffer), reinterpret_cast<const u8*>(pixels), w, h, depth, layer, texture.mipmap(), texture.pitch());
+	case CELL_GCM_TEXTURE_DEPTH24_D8: // Opaque type ; ATM do not copy anything
+		return std::vector<MipmapLevelInfo>();
 	}
 	throw EXCEPTION("Wrong format %d", format);
 }

rpcs3/Emu/RSX/D3D12/D3D12FragmentProgramDecompiler.cpp

@@ -169,6 +169,8 @@ void D3D12FragmentDecompiler::insertMainStart(std::stringstream & OS)
 					continue;
 				}
 			}
+			if (PI.name == "ssa")
+				continue;
 			OS << "	" << PT.type << " " << PI.name << " = In." << PI.name << ";" << std::endl;
 		}
 	}
@@ -176,6 +178,7 @@ void D3D12FragmentDecompiler::insertMainStart(std::stringstream & OS)
 	OS << "	" << "float4 gl_FragCoord = In.Position;" << std::endl;
 	if (m_prog.origin_mode == rsx::window_origin::bottom)
 		OS << "	gl_FragCoord.y = (" << std::to_string(m_prog.height) << " - gl_FragCoord.y);\n";
+	OS << "	float4 ssa = is_front_face ? float4(1., 1., 1., 1.) : float4(-1., -1., -1., -1.);\n";
 	// Declare output
 	for (const ParamType &PT : m_parr.params[PF_PARAM_NONE])
 	{

rpcs3/Emu/RSX/D3D12/D3D12VertexProgramDecompiler.cpp

@@ -44,7 +44,7 @@ namespace
 		{
 			if (static_cast<size_t>(real_input.location) != std::get<0>(attribute))
 				continue;
-			OS << "Buffer<float4> " << std::get<1>(attribute) << "_buffer : register(t" << reg++ << ");\n";
+			OS << "Buffer<" << (real_input.int_type ? "int4" : "float4") << "> " << std::get<1>(attribute) << "_buffer : register(t" << reg++ << ");\n";
 			return true;
 		}
 		return false;

rpcs3/Emu/RSX/GCM.cpp

@@ -1124,13 +1124,13 @@ namespace
 	{
 		switch (rsx::to_vertex_base_type(type))
 		{
-		case rsx::vertex_base_type::s1: return "Short";
+		case rsx::vertex_base_type::s1: return "Signed short normalized";
 		case rsx::vertex_base_type::f: return "Float";
 		case rsx::vertex_base_type::sf: return "Half float";
-		case rsx::vertex_base_type::ub: return "Unsigned byte";
+		case rsx::vertex_base_type::ub: return "Unsigned byte normalized";
-		case rsx::vertex_base_type::s32k: return "Signed int";
+		case rsx::vertex_base_type::s32k: return "Signed short unormalized";
 		case rsx::vertex_base_type::cmp: return "CMP";
-		case rsx::vertex_base_type::ub256: return "UB256";
+		case rsx::vertex_base_type::ub256: return "Unsigned byte unormalized";
 		}
 	}
 

rpcs3/Emu/RSX/GCM.h

@@ -30,10 +30,10 @@ namespace rsx
 		s1, ///< signed byte
 		f, ///< float
 		sf, ///< half float
-		ub, ///< unsigned byte
+		ub, ///< unsigned byte interpreted as 0.f and 1.f
 		s32k, ///< signed 32bits int
 		cmp, ///< compressed aka X11G11Z10 and always 1. W.
-		ub256,
+		ub256, ///< unsigned byte interpreted as between 0 and 255.
 	};
 
 	vertex_base_type to_vertex_base_type(u8 in);

rpcs3/Emu/RSX/GL/GLFragmentProgram.cpp

@@ -116,6 +116,8 @@ void GLFragmentDecompilerThread::insertMainStart(std::stringstream & OS)
 			OS << ";" << std::endl;
 		}
 	}
+
+	OS << "	vec4 ssa = gl_FrontFacing ? vec4(1.) : vec4(-1.);\n";
 }
 
 void GLFragmentDecompilerThread::insertMainEnd(std::stringstream & OS)

rpcs3/Emu/RSX/GL/GLGSRender.cpp

@@ -37,10 +37,10 @@ namespace
 		 * The buffer texture spec only allows fetches aligned to 8, 16, 32, etc...
 		 * This rules out most 3-component formats, except for the 32-wide RGB32F, RGB32I, RGB32UI
 		 */
-		const u32 vec1_types[] = { GL_R16, GL_R32F, GL_R16F, GL_R8, GL_R32I, GL_R16, GL_R8UI };
+		const u32 vec1_types[] = { GL_R16, GL_R32F, GL_R16F, GL_R8, GL_R16I, GL_R16, GL_R8UI };
-		const u32 vec2_types[] = { GL_RG16, GL_RG32F, GL_RG16F, GL_RG8, GL_RG32I, GL_RG16, GL_RG8UI };
+		const u32 vec2_types[] = { GL_RG16, GL_RG32F, GL_RG16F, GL_RG8, GL_RG16I, GL_RG16, GL_RG8UI };
-		const u32 vec3_types[] = { GL_RGBA16, GL_RGB32F, GL_RGBA16F, GL_RGBA8, GL_RGB32I, GL_RGBA16, GL_RGBA8UI };	//VEC3 COMPONENTS NOT SUPPORTED!
+		const u32 vec3_types[] = { GL_RGBA16, GL_RGB32F, GL_RGBA16F, GL_RGBA8, GL_RGBA16I, GL_RGBA16, GL_RGBA8UI };	//VEC3 COMPONENTS NOT SUPPORTED!
-		const u32 vec4_types[] = { GL_RGBA16, GL_RGBA32F, GL_RGBA16F, GL_RGBA8, GL_RGBA32I, GL_RGBA16, GL_RGBA8UI };
+		const u32 vec4_types[] = { GL_RGBA16, GL_RGBA32F, GL_RGBA16F, GL_RGBA8, GL_RGBA16I, GL_RGBA16, GL_RGBA8UI };
 
 		const u32* vec_selectors[] = { 0, vec1_types, vec2_types, vec3_types, vec4_types };
 

rpcs3/Emu/RSX/RSXThread.cpp

@@ -127,6 +127,7 @@ namespace rsx
 		switch (type)
 		{
 		case vertex_base_type::s1:
+		case vertex_base_type::s32k:
 			switch (size)
 			{
 			case 1:
@@ -136,8 +137,8 @@ namespace rsx
 			case 3:
 				return sizeof(u16) * 4;
 			}
-			throw new EXCEPTION("Wrong vector size");
+			throw EXCEPTION("Wrong vector size");
-		case vertex_base_type::f:     return sizeof(f32) * size;
+		case vertex_base_type::f: return sizeof(f32) * size;
 		case vertex_base_type::sf:
 			switch (size)
 			{
@@ -148,7 +149,7 @@ namespace rsx
 			case 3:
 				return sizeof(f16) * 4;
 			}
-			throw new EXCEPTION("Wrong vector size");
+			throw EXCEPTION("Wrong vector size");
 		case vertex_base_type::ub:
 			switch (size)
 			{
@@ -159,15 +160,11 @@ namespace rsx
 			case 3:
 				return sizeof(u8) * 4;
 			}
-			throw new EXCEPTION("Wrong vector size");
+			throw EXCEPTION("Wrong vector size");
-		case vertex_base_type::s32k:  return sizeof(u32) * size;
+		case vertex_base_type::cmp: return sizeof(u16) * 4;
-		case vertex_base_type::cmp:   return sizeof(u16) * 4;
+		case vertex_base_type::ub256: Expects(size == 4); return sizeof(u8) * 4;
-		case vertex_base_type::ub256: return sizeof(u8) * 4;
-
-		default:
-			throw new EXCEPTION("RSXVertexData::GetTypeSize: Bad vertex data type (%d)!", type);
-			return 0;
 		}
+		throw EXCEPTION("RSXVertexData::GetTypeSize: Bad vertex data type (%d)!", type);
 	}
 
 	void tiled_region::write(const void *src, u32 width, u32 height, u32 pitch)
@@ -571,6 +568,25 @@ namespace rsx
 		}
 	}
 
+	namespace
+	{
+		bool is_int_type(rsx::vertex_base_type type)
+		{
+			switch (type)
+			{
+			case rsx::vertex_base_type::s32k:
+			case rsx::vertex_base_type::ub256:
+				return true;
+			case rsx::vertex_base_type::f:
+			case rsx::vertex_base_type::cmp:
+			case rsx::vertex_base_type::sf:
+			case rsx::vertex_base_type::s1:
+			case rsx::vertex_base_type::ub:
+				return false;
+			}
+		}
+	}
+
 	std::array<u32, 4> thread::get_color_surface_addresses() const
 	{
 		u32 offset_color[] =
@@ -639,7 +655,8 @@ namespace rsx
 					vertex_arrays_info[index].size,
 					vertex_arrays_info[index].frequency,
 					!!((modulo_mask >> index) & 0x1),
-					true
+					true,
+					is_int_type(vertex_arrays_info[index].type)
 				}
 				);
 			}
@@ -651,7 +668,8 @@ namespace rsx
 					register_vertex_info[index].size,
 					register_vertex_info[index].frequency,
 					!!((modulo_mask >> index) & 0x1),
-					false
+					false,
+					is_int_type(vertex_arrays_info[index].type)
 				}
 				);
 			}

rpcs3/Emu/RSX/RSXVertexProgram.h

@@ -197,10 +197,11 @@ struct rsx_vertex_input
 	u16 frequency;
 	bool is_modulo; // either modulo frequency or divide frequency
 	bool is_array; // false if "reg value"
+	bool int_type;
 
 	bool operator==(const rsx_vertex_input other) const
 	{
-		return location == other.location && size == other.size && frequency == other.frequency && is_modulo == other.is_modulo && is_array == other.is_array;
+		return location == other.location && size == other.size && frequency == other.frequency && is_modulo == other.is_modulo && is_array == other.is_array && int_type == other.int_type;
 	}
 };