summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--src/video_core/engines/shader_bytecode.h2
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp1
-rw-r--r--src/video_core/renderer_vulkan/shaders/blit.frag24
-rw-r--r--src/video_core/renderer_vulkan/shaders/blit.vert28
-rw-r--r--src/video_core/renderer_vulkan/shaders/quad_array.comp37
-rw-r--r--src/video_core/renderer_vulkan/shaders/uint8.comp33
-rw-r--r--src/video_core/renderer_vulkan/vk_shader_decompiler.cpp24
-rw-r--r--src/video_core/shader/decode/conversion.cpp15
-rw-r--r--src/video_core/shader/decode/memory.cpp38
-rw-r--r--src/video_core/shader/decode/texture.cpp13
10 files changed, 179 insertions, 36 deletions
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index d6a2cc8b8..dfb12cd2d 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -1973,7 +1973,7 @@ private:
INST("1101-01---------", Id::TLDS, Type::Texture, "TLDS"),
INST("110010----111---", Id::TLD4, Type::Texture, "TLD4"),
INST("1101111011111---", Id::TLD4_B, Type::Texture, "TLD4_B"),
- INST("11011111--00----", Id::TLD4S, Type::Texture, "TLD4S"),
+ INST("11011111-0------", Id::TLD4S, Type::Texture, "TLD4S"),
INST("110111110110----", Id::TMML_B, Type::Texture, "TMML_B"),
INST("1101111101011---", Id::TMML, Type::Texture, "TMML"),
INST("11011110011110--", Id::TXD_B, Type::Texture, "TXD_B"),
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index d1ae4be6d..0389c2143 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -399,6 +399,7 @@ public:
DeclareConstantBuffers();
DeclareGlobalMemory();
DeclareSamplers();
+ DeclareImages();
DeclarePhysicalAttributeReader();
code.AddLine("void execute_{}() {{", suffix);
diff --git a/src/video_core/renderer_vulkan/shaders/blit.frag b/src/video_core/renderer_vulkan/shaders/blit.frag
new file mode 100644
index 000000000..a06ecd24a
--- /dev/null
+++ b/src/video_core/renderer_vulkan/shaders/blit.frag
@@ -0,0 +1,24 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+/*
+ * Build instructions:
+ * $ glslangValidator -V $THIS_FILE -o output.spv
+ * $ spirv-opt -O --strip-debug output.spv -o optimized.spv
+ * $ xxd -i optimized.spv
+ *
+ * Then copy that bytecode to the C++ file
+ */
+
+#version 460 core
+
+layout (location = 0) in vec2 frag_tex_coord;
+
+layout (location = 0) out vec4 color;
+
+layout (binding = 1) uniform sampler2D color_texture;
+
+void main() {
+ color = texture(color_texture, frag_tex_coord);
+}
diff --git a/src/video_core/renderer_vulkan/shaders/blit.vert b/src/video_core/renderer_vulkan/shaders/blit.vert
new file mode 100644
index 000000000..c64d9235a
--- /dev/null
+++ b/src/video_core/renderer_vulkan/shaders/blit.vert
@@ -0,0 +1,28 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+/*
+ * Build instructions:
+ * $ glslangValidator -V $THIS_FILE -o output.spv
+ * $ spirv-opt -O --strip-debug output.spv -o optimized.spv
+ * $ xxd -i optimized.spv
+ *
+ * Then copy that bytecode to the C++ file
+ */
+
+#version 460 core
+
+layout (location = 0) in vec2 vert_position;
+layout (location = 1) in vec2 vert_tex_coord;
+
+layout (location = 0) out vec2 frag_tex_coord;
+
+layout (set = 0, binding = 0) uniform MatrixBlock {
+ mat4 modelview_matrix;
+};
+
+void main() {
+ gl_Position = modelview_matrix * vec4(vert_position, 0.0, 1.0);
+ frag_tex_coord = vert_tex_coord;
+}
diff --git a/src/video_core/renderer_vulkan/shaders/quad_array.comp b/src/video_core/renderer_vulkan/shaders/quad_array.comp
new file mode 100644
index 000000000..5a5703308
--- /dev/null
+++ b/src/video_core/renderer_vulkan/shaders/quad_array.comp
@@ -0,0 +1,37 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+/*
+ * Build instructions:
+ * $ glslangValidator -V $THIS_FILE -o output.spv
+ * $ spirv-opt -O --strip-debug output.spv -o optimized.spv
+ * $ xxd -i optimized.spv
+ *
+ * Then copy that bytecode to the C++ file
+ */
+
+#version 460 core
+
+layout (local_size_x = 1024) in;
+
+layout (std430, set = 0, binding = 0) buffer OutputBuffer {
+ uint output_indexes[];
+};
+
+layout (push_constant) uniform PushConstants {
+ uint first;
+};
+
+void main() {
+ uint primitive = gl_GlobalInvocationID.x;
+ if (primitive * 6 >= output_indexes.length()) {
+ return;
+ }
+
+ const uint quad_map[6] = uint[](0, 1, 2, 0, 2, 3);
+ for (uint vertex = 0; vertex < 6; ++vertex) {
+ uint index = first + primitive * 4 + quad_map[vertex];
+ output_indexes[primitive * 6 + vertex] = index;
+ }
+}
diff --git a/src/video_core/renderer_vulkan/shaders/uint8.comp b/src/video_core/renderer_vulkan/shaders/uint8.comp
new file mode 100644
index 000000000..a320f3ae0
--- /dev/null
+++ b/src/video_core/renderer_vulkan/shaders/uint8.comp
@@ -0,0 +1,33 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+/*
+ * Build instructions:
+ * $ glslangValidator -V $THIS_FILE -o output.spv
+ * $ spirv-opt -O --strip-debug output.spv -o optimized.spv
+ * $ xxd -i optimized.spv
+ *
+ * Then copy that bytecode to the C++ file
+ */
+
+#version 460 core
+#extension GL_EXT_shader_16bit_storage : require
+#extension GL_EXT_shader_8bit_storage : require
+
+layout (local_size_x = 1024) in;
+
+layout (std430, set = 0, binding = 0) readonly buffer InputBuffer {
+ uint8_t input_indexes[];
+};
+
+layout (std430, set = 0, binding = 1) writeonly buffer OutputBuffer {
+ uint16_t output_indexes[];
+};
+
+void main() {
+ uint id = gl_GlobalInvocationID.x;
+ if (id < input_indexes.length()) {
+ output_indexes[id] = uint16_t(input_indexes[id]);
+ }
+}
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 833145971..a8baf91de 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -2557,29 +2557,7 @@ public:
}
Id operator()(const ExprCondCode& expr) {
- const Node cc = decomp.ir.GetConditionCode(expr.cc);
- Id target;
-
- if (const auto pred = std::get_if<PredicateNode>(&*cc)) {
- const auto index = pred->GetIndex();
- switch (index) {
- case Tegra::Shader::Pred::NeverExecute:
- target = decomp.v_false;
- break;
- case Tegra::Shader::Pred::UnusedIndex:
- target = decomp.v_true;
- break;
- default:
- target = decomp.predicates.at(index);
- break;
- }
- } else if (const auto flag = std::get_if<InternalFlagNode>(&*cc)) {
- target = decomp.internal_flags.at(static_cast<u32>(flag->GetFlag()));
- } else {
- UNREACHABLE();
- }
-
- return decomp.OpLoad(decomp.t_bool, target);
+ return decomp.AsBool(decomp.Visit(decomp.ir.GetConditionCode(expr.cc)));
}
Id operator()(const ExprVar& expr) {
diff --git a/src/video_core/shader/decode/conversion.cpp b/src/video_core/shader/decode/conversion.cpp
index 32facd6ba..0eeb75559 100644
--- a/src/video_core/shader/decode/conversion.cpp
+++ b/src/video_core/shader/decode/conversion.cpp
@@ -63,12 +63,11 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
case OpCode::Id::I2F_R:
case OpCode::Id::I2F_C:
case OpCode::Id::I2F_IMM: {
- UNIMPLEMENTED_IF(instr.conversion.int_src.selector != 0);
UNIMPLEMENTED_IF(instr.conversion.dst_size == Register::Size::Long);
UNIMPLEMENTED_IF_MSG(instr.generates_cc,
"Condition codes generation in I2F is not implemented");
- Node value = [&]() {
+ Node value = [&] {
switch (opcode->get().GetId()) {
case OpCode::Id::I2F_R:
return GetRegister(instr.gpr20);
@@ -81,7 +80,19 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
return Immediate(0);
}
}();
+
const bool input_signed = instr.conversion.is_input_signed;
+
+ if (instr.conversion.src_size == Register::Size::Byte) {
+ const u32 offset = static_cast<u32>(instr.conversion.int_src.selector) * 8;
+ if (offset > 0) {
+ value = SignedOperation(OperationCode::ILogicalShiftRight, input_signed,
+ std::move(value), Immediate(offset));
+ }
+ } else {
+ UNIMPLEMENTED_IF(instr.conversion.int_src.selector != 0);
+ }
+
value = ConvertIntegerSize(value, instr.conversion.src_size, input_signed);
value = GetOperandAbsNegInteger(value, instr.conversion.abs_a, false, input_signed);
value = SignedOperation(OperationCode::FCastInteger, input_signed, PRECISE, value);
diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp
index 78e92f52e..c934d0719 100644
--- a/src/video_core/shader/decode/memory.cpp
+++ b/src/video_core/shader/decode/memory.cpp
@@ -22,7 +22,23 @@ using Tegra::Shader::Register;
namespace {
-u32 GetUniformTypeElementsCount(Tegra::Shader::UniformType uniform_type) {
+u32 GetLdgMemorySize(Tegra::Shader::UniformType uniform_type) {
+ switch (uniform_type) {
+ case Tegra::Shader::UniformType::UnsignedByte:
+ case Tegra::Shader::UniformType::Single:
+ return 1;
+ case Tegra::Shader::UniformType::Double:
+ return 2;
+ case Tegra::Shader::UniformType::Quad:
+ case Tegra::Shader::UniformType::UnsignedQuad:
+ return 4;
+ default:
+ UNIMPLEMENTED_MSG("Unimplemented size={}!", static_cast<u32>(uniform_type));
+ return 1;
+ }
+}
+
+u32 GetStgMemorySize(Tegra::Shader::UniformType uniform_type) {
switch (uniform_type) {
case Tegra::Shader::UniformType::Single:
return 1;
@@ -170,7 +186,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
const auto [real_address_base, base_address, descriptor] =
TrackGlobalMemory(bb, instr, false);
- const u32 count = GetUniformTypeElementsCount(type);
+ const u32 count = GetLdgMemorySize(type);
if (!real_address_base || !base_address) {
// Tracking failed, load zeroes.
for (u32 i = 0; i < count; ++i) {
@@ -181,12 +197,22 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
for (u32 i = 0; i < count; ++i) {
const Node it_offset = Immediate(i * 4);
- const Node real_address =
- Operation(OperationCode::UAdd, NO_PRECISE, real_address_base, it_offset);
- const Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor);
+ const Node real_address = Operation(OperationCode::UAdd, real_address_base, it_offset);
+ Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor);
+
+ if (type == Tegra::Shader::UniformType::UnsignedByte) {
+ // To handle unaligned loads get the byte used to dereferenced global memory
+ // and extract that byte from the loaded uint32.
+ Node byte = Operation(OperationCode::UBitwiseAnd, real_address, Immediate(3));
+ byte = Operation(OperationCode::ULogicalShiftLeft, std::move(byte), Immediate(3));
+
+ gmem = Operation(OperationCode::UBitfieldExtract, std::move(gmem), std::move(byte),
+ Immediate(8));
+ }
SetTemporary(bb, i, gmem);
}
+
for (u32 i = 0; i < count; ++i) {
SetRegister(bb, instr.gpr0.Value() + i, GetTemporary(i));
}
@@ -276,7 +302,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) {
break;
}
- const u32 count = GetUniformTypeElementsCount(type);
+ const u32 count = GetStgMemorySize(type);
for (u32 i = 0; i < count; ++i) {
const Node it_offset = Immediate(i * 4);
const Node real_address = Operation(OperationCode::UAdd, real_address_base, it_offset);
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 994c05611..dff01a541 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -743,13 +743,18 @@ Node4 ShaderIR::GetTldsCode(Instruction instr, TextureType texture_type, bool is
// When lod is used always is in gpr20
const Node lod = lod_enabled ? GetRegister(instr.gpr20) : Immediate(0);
- // Fill empty entries from the guest sampler.
+ // Fill empty entries from the guest sampler
const std::size_t entry_coord_count = GetCoordCount(sampler.GetType());
if (type_coord_count != entry_coord_count) {
LOG_WARNING(HW_GPU, "Bound and built texture types mismatch");
- }
- for (std::size_t i = type_coord_count; i < entry_coord_count; ++i) {
- coords.push_back(GetRegister(Register::ZeroIndex));
+
+ // When the size is higher we insert zeroes
+ for (std::size_t i = type_coord_count; i < entry_coord_count; ++i) {
+ coords.push_back(GetRegister(Register::ZeroIndex));
+ }
+
+ // Then we ensure the size matches the number of entries (dropping unused values)
+ coords.resize(entry_coord_count);
}
Node4 values;
generated by cgit v1.2.3 (git 2.25.1) at 2024-08-03 18:12:51 +0200