summaryrefslogtreecommitdiffstats
path: root/src/video_core
diff options
context:
space:
mode:
Diffstat (limited to 'src/video_core')
-rw-r--r--src/video_core/engines/shader_bytecode.h2
-rw-r--r--src/video_core/morton.cpp4
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.cpp22
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp241
-rw-r--r--src/video_core/renderer_opengl/renderer_opengl.h2
-rw-r--r--src/video_core/surface.cpp7
-rw-r--r--src/video_core/surface.h92
7 files changed, 173 insertions, 197 deletions
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index eb703bb5a..e53c77f2b 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -1049,7 +1049,7 @@ union Instruction {
BitField<49, 1, u64> nodep_flag;
BitField<50, 3, u64> component_mask_selector;
BitField<53, 4, u64> texture_info;
- BitField<60, 1, u64> fp32_flag;
+ BitField<59, 1, u64> fp32_flag;
TextureType GetTextureType() const {
// The TEXS instruction has a weird encoding for the texture type.
diff --git a/src/video_core/morton.cpp b/src/video_core/morton.cpp
index 47e76d8fe..b68f4fb13 100644
--- a/src/video_core/morton.cpp
+++ b/src/video_core/morton.cpp
@@ -66,8 +66,6 @@ static constexpr ConversionArray morton_to_linear_fns = {
MortonCopy<true, PixelFormat::BC6H_UF16>,
MortonCopy<true, PixelFormat::BC6H_SF16>,
MortonCopy<true, PixelFormat::ASTC_2D_4X4>,
- MortonCopy<true, PixelFormat::G8R8U>,
- MortonCopy<true, PixelFormat::G8R8S>,
MortonCopy<true, PixelFormat::BGRA8>,
MortonCopy<true, PixelFormat::RGBA32F>,
MortonCopy<true, PixelFormat::RG32F>,
@@ -138,8 +136,6 @@ static constexpr ConversionArray linear_to_morton_fns = {
MortonCopy<false, PixelFormat::BC6H_SF16>,
// TODO(Subv): Swizzling ASTC formats are not supported
nullptr,
- MortonCopy<false, PixelFormat::G8R8U>,
- MortonCopy<false, PixelFormat::G8R8S>,
MortonCopy<false, PixelFormat::BGRA8>,
MortonCopy<false, PixelFormat::RGBA32F>,
MortonCopy<false, PixelFormat::RG32F>,
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 352c391a9..d3dcb9a46 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -299,8 +299,6 @@ static constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex
{GL_COMPRESSED_RGB_BPTC_SIGNED_FLOAT, GL_RGB, GL_UNSIGNED_INT_8_8_8_8, ComponentType::Float,
true}, // BC6H_SF16
{GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // ASTC_2D_4X4
- {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // G8R8U
- {GL_RG8, GL_RG, GL_BYTE, ComponentType::SNorm, false}, // G8R8S
{GL_RGBA8, GL_BGRA, GL_UNSIGNED_BYTE, ComponentType::UNorm, false}, // BGRA8
{GL_RGBA32F, GL_RGBA, GL_FLOAT, ComponentType::Float, false}, // RGBA32F
{GL_RG32F, GL_RG, GL_FLOAT, ComponentType::Float, false}, // RG32F
@@ -626,18 +624,6 @@ static void ConvertS8Z24ToZ24S8(std::vector<u8>& data, u32 width, u32 height, bo
}
}
-static void ConvertG8R8ToR8G8(std::vector<u8>& data, u32 width, u32 height) {
- constexpr auto bpp{GetBytesPerPixel(PixelFormat::G8R8U)};
- for (std::size_t y = 0; y < height; ++y) {
- for (std::size_t x = 0; x < width; ++x) {
- const std::size_t offset{bpp * (y * width + x)};
- const u8 temp{data[offset]};
- data[offset] = data[offset + 1];
- data[offset + 1] = temp;
- }
- }
-}
-
/**
* Helper function to perform software conversion (as needed) when loading a buffer from Switch
* memory. This is for Maxwell pixel formats that cannot be represented as-is in OpenGL or with
@@ -670,12 +656,6 @@ static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelForma
// Convert the S8Z24 depth format to Z24S8, as OpenGL does not support S8Z24.
ConvertS8Z24ToZ24S8(data, width, height, false);
break;
-
- case PixelFormat::G8R8U:
- case PixelFormat::G8R8S:
- // Convert the G8R8 color format to R8G8, as OpenGL does not support G8R8.
- ConvertG8R8ToR8G8(data, width, height);
- break;
}
}
@@ -687,8 +667,6 @@ static void ConvertFormatAsNeeded_LoadGLBuffer(std::vector<u8>& data, PixelForma
static void ConvertFormatAsNeeded_FlushGLBuffer(std::vector<u8>& data, PixelFormat pixel_format,
u32 width, u32 height) {
switch (pixel_format) {
- case PixelFormat::G8R8U:
- case PixelFormat::G8R8S:
case PixelFormat::ASTC_2D_4X4:
case PixelFormat::ASTC_2D_8X8:
case PixelFormat::ASTC_2D_4X4_SRGB:
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index 4e685fa2c..1bb09e61b 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -347,6 +347,15 @@ public:
BuildInputList();
}
+ void SetConditionalCodesFromExpression(const std::string& expresion) {
+ SetInternalFlag(InternalFlag::ZeroFlag, "(" + expresion + ") == 0");
+ LOG_WARNING(HW_GPU, "Condition codes implementation is incomplete.");
+ }
+
+ void SetConditionalCodesFromRegister(const Register& reg, u64 dest_elem = 0) {
+ SetConditionalCodesFromExpression(GetRegister(reg, static_cast<u32>(dest_elem)));
+ }
+
/**
* Returns code that does an integer size conversion for the specified size.
* @param value Value to perform integer size conversion on.
@@ -401,14 +410,24 @@ public:
* @param dest_num_components Number of components in the destination.
* @param value_num_components Number of components in the value.
* @param is_saturated Optional, when True, saturates the provided value.
+ * @param sets_cc Optional, when True, sets the corresponding values to the implemented
+ * condition flags.
* @param dest_elem Optional, the destination element to use for the operation.
*/
void SetRegisterToFloat(const Register& reg, u64 elem, const std::string& value,
u64 dest_num_components, u64 value_num_components,
- bool is_saturated = false, u64 dest_elem = 0, bool precise = false) {
-
- SetRegister(reg, elem, is_saturated ? "clamp(" + value + ", 0.0, 1.0)" : value,
- dest_num_components, value_num_components, dest_elem, precise);
+ bool is_saturated = false, bool sets_cc = false, u64 dest_elem = 0,
+ bool precise = false) {
+ const std::string clamped_value = is_saturated ? "clamp(" + value + ", 0.0, 1.0)" : value;
+ SetRegister(reg, elem, clamped_value, dest_num_components, value_num_components, dest_elem,
+ precise);
+ if (sets_cc) {
+ if (reg == Register::ZeroIndex) {
+ SetConditionalCodesFromExpression(clamped_value);
+ } else {
+ SetConditionalCodesFromRegister(reg, dest_elem);
+ }
+ }
}
/**
@@ -419,25 +438,29 @@ public:
* @param dest_num_components Number of components in the destination.
* @param value_num_components Number of components in the value.
* @param is_saturated Optional, when True, saturates the provided value.
+ * @param sets_cc Optional, when True, sets the corresponding values to the implemented
+ * condition flags.
* @param dest_elem Optional, the destination element to use for the operation.
* @param size Register size to use for conversion instructions.
*/
void SetRegisterToInteger(const Register& reg, bool is_signed, u64 elem,
const std::string& value, u64 dest_num_components,
u64 value_num_components, bool is_saturated = false,
- u64 dest_elem = 0, Register::Size size = Register::Size::Word,
- bool sets_cc = false) {
+ bool sets_cc = false, u64 dest_elem = 0,
+ Register::Size size = Register::Size::Word) {
UNIMPLEMENTED_IF(is_saturated);
-
+ const std::string final_value = ConvertIntegerSize(value, size);
const std::string func{is_signed ? "intBitsToFloat" : "uintBitsToFloat"};
- SetRegister(reg, elem, func + '(' + ConvertIntegerSize(value, size) + ')',
- dest_num_components, value_num_components, dest_elem, false);
+ SetRegister(reg, elem, func + '(' + final_value + ')', dest_num_components,
+ value_num_components, dest_elem, false);
if (sets_cc) {
- const std::string zero_condition = "( " + ConvertIntegerSize(value, size) + " == 0 )";
- SetInternalFlag(InternalFlag::ZeroFlag, zero_condition);
- LOG_WARNING(HW_GPU, "Condition codes implementation is incomplete.");
+ if (reg == Register::ZeroIndex) {
+ SetConditionalCodesFromExpression(final_value);
+ } else {
+ SetConditionalCodesFromRegister(reg, dest_elem);
+ }
}
}
@@ -470,10 +493,10 @@ public:
// pack. I couldn't test this on hardware but it shouldn't really matter since most
// of the time when a Mrg_* flag is used both components will be mirrored. That
// being said, it deserves a test.
- return "((" + GetRegisterAsInteger(reg, 0, false) +
+ return "uintBitsToFloat((" + GetRegisterAsInteger(reg, 0, false) +
" & 0xffff0000) | (packHalf2x16(" + value + ") & 0x0000ffff))";
case Tegra::Shader::HalfMerge::Mrg_H1:
- return "((" + GetRegisterAsInteger(reg, 0, false) +
+ return "uintBitsToFloat((" + GetRegisterAsInteger(reg, 0, false) +
" & 0x0000ffff) | (packHalf2x16(" + value + ") & 0xffff0000))";
default:
UNREACHABLE();
@@ -1275,7 +1298,7 @@ private:
void WriteLogicOperation(Register dest, LogicOperation logic_op, const std::string& op_a,
const std::string& op_b,
Tegra::Shader::PredicateResultMode predicate_mode,
- Tegra::Shader::Pred predicate) {
+ Tegra::Shader::Pred predicate, const bool set_cc) {
std::string result{};
switch (logic_op) {
case LogicOperation::And: {
@@ -1299,7 +1322,7 @@ private:
}
if (dest != Tegra::Shader::Register::ZeroIndex) {
- regs.SetRegisterToInteger(dest, true, 0, result, 1, 1);
+ regs.SetRegisterToInteger(dest, true, 0, result, 1, 1, false, set_cc);
}
using Tegra::Shader::PredicateResultMode;
@@ -1319,7 +1342,8 @@ private:
}
void WriteLop3Instruction(Register dest, const std::string& op_a, const std::string& op_b,
- const std::string& op_c, const std::string& imm_lut) {
+ const std::string& op_c, const std::string& imm_lut,
+ const bool set_cc) {
if (dest == Tegra::Shader::Register::ZeroIndex) {
return;
}
@@ -1342,7 +1366,7 @@ private:
result += ')';
- regs.SetRegisterToInteger(dest, true, 0, result, 1, 1);
+ regs.SetRegisterToInteger(dest, true, 0, result, 1, 1, false, set_cc);
}
void WriteTexsInstructionFloat(const Instruction& instr, const std::string& texture) {
@@ -1357,12 +1381,12 @@ private:
if (written_components < 2) {
// Write the first two swizzle components to gpr0 and gpr0+1
- regs.SetRegisterToFloat(instr.gpr0, component, texture, 1, 4, false,
+ regs.SetRegisterToFloat(instr.gpr0, component, texture, 1, 4, false, false,
written_components % 2);
} else {
ASSERT(instr.texs.HasTwoDestinations());
// Write the rest of the swizzle components to gpr28 and gpr28+1
- regs.SetRegisterToFloat(instr.gpr28, component, texture, 1, 4, false,
+ regs.SetRegisterToFloat(instr.gpr28, component, texture, 1, 4, false, false,
written_components % 2);
}
@@ -1755,7 +1779,7 @@ private:
instr.tlds.GetTextureProcessMode() == Tegra::Shader::TextureProcessMode::LL;
constexpr std::array<const char*, 4> coord_container{
- {"", "int coord = (", "ivec2 coord = ivec2(", "ivec3 coord = ivec3("}};
+ {"", "int coords = (", "ivec2 coords = ivec2(", "ivec3 coords = ivec3("}};
std::string coord = coord_container[total_coord_count];
@@ -1871,8 +1895,6 @@ private:
instr.fmul.tab5c68_0 != 1, "FMUL tab5cb8_0({}) is not implemented",
instr.fmul.tab5c68_0
.Value()); // SMO typical sends 1 here which seems to be the default
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in FMUL is not implemented");
op_b = GetOperandAbsNeg(op_b, false, instr.fmul.negate_b);
@@ -1896,20 +1918,17 @@ private:
}
regs.SetRegisterToFloat(instr.gpr0, 0, op_a + " * " + op_b + postfactor_op, 1, 1,
- instr.alu.saturate_d, 0, true);
+ instr.alu.saturate_d, instr.generates_cc, 0, true);
break;
}
case OpCode::Id::FADD_C:
case OpCode::Id::FADD_R:
case OpCode::Id::FADD_IMM: {
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in FADD is not implemented");
-
op_a = GetOperandAbsNeg(op_a, instr.alu.abs_a, instr.alu.negate_a);
op_b = GetOperandAbsNeg(op_b, instr.alu.abs_b, instr.alu.negate_b);
regs.SetRegisterToFloat(instr.gpr0, 0, op_a + " + " + op_b, 1, 1,
- instr.alu.saturate_d, 0, true);
+ instr.alu.saturate_d, instr.generates_cc, 0, true);
break;
}
case OpCode::Id::MUFU: {
@@ -1917,31 +1936,31 @@ private:
switch (instr.sub_op) {
case SubOp::Cos:
regs.SetRegisterToFloat(instr.gpr0, 0, "cos(" + op_a + ')', 1, 1,
- instr.alu.saturate_d, 0, true);
+ instr.alu.saturate_d, false, 0, true);
break;
case SubOp::Sin:
regs.SetRegisterToFloat(instr.gpr0, 0, "sin(" + op_a + ')', 1, 1,
- instr.alu.saturate_d, 0, true);
+ instr.alu.saturate_d, false, 0, true);
break;
case SubOp::Ex2:
regs.SetRegisterToFloat(instr.gpr0, 0, "exp2(" + op_a + ')', 1, 1,
- instr.alu.saturate_d, 0, true);
+ instr.alu.saturate_d, false, 0, true);
break;
case SubOp::Lg2:
regs.SetRegisterToFloat(instr.gpr0, 0, "log2(" + op_a + ')', 1, 1,
- instr.alu.saturate_d, 0, true);
+ instr.alu.saturate_d, false, 0, true);
break;
case SubOp::Rcp:
regs.SetRegisterToFloat(instr.gpr0, 0, "1.0 / " + op_a, 1, 1,
- instr.alu.saturate_d, 0, true);
+ instr.alu.saturate_d, false, 0, true);
break;
case SubOp::Rsq:
regs.SetRegisterToFloat(instr.gpr0, 0, "inversesqrt(" + op_a + ')', 1, 1,
- instr.alu.saturate_d, 0, true);
+ instr.alu.saturate_d, false, 0, true);
break;
case SubOp::Sqrt:
regs.SetRegisterToFloat(instr.gpr0, 0, "sqrt(" + op_a + ')', 1, 1,
- instr.alu.saturate_d, 0, true);
+ instr.alu.saturate_d, false, 0, true);
break;
default:
UNIMPLEMENTED_MSG("Unhandled MUFU sub op={0:x}",
@@ -1952,8 +1971,9 @@ private:
case OpCode::Id::FMNMX_C:
case OpCode::Id::FMNMX_R:
case OpCode::Id::FMNMX_IMM: {
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in FMNMX is not implemented");
+ UNIMPLEMENTED_IF_MSG(
+ instr.generates_cc,
+ "Condition codes generation in FMNMX is partially implemented");
op_a = GetOperandAbsNeg(op_a, instr.alu.abs_a, instr.alu.negate_a);
op_b = GetOperandAbsNeg(op_b, instr.alu.abs_b, instr.alu.negate_b);
@@ -1964,7 +1984,7 @@ private:
regs.SetRegisterToFloat(instr.gpr0, 0,
'(' + condition + ") ? min(" + parameters + ") : max(" +
parameters + ')',
- 1, 1, false, 0, true);
+ 1, 1, false, instr.generates_cc, 0, true);
break;
}
case OpCode::Id::RRO_C:
@@ -1989,18 +2009,16 @@ private:
break;
}
case OpCode::Id::FMUL32_IMM: {
- UNIMPLEMENTED_IF_MSG(instr.op_32.generates_cc,
- "Condition codes generation in FMUL32 is not implemented");
-
- regs.SetRegisterToFloat(instr.gpr0, 0,
- regs.GetRegisterAsFloat(instr.gpr8) + " * " +
- GetImmediate32(instr),
- 1, 1, instr.fmul32.saturate, 0, true);
+ regs.SetRegisterToFloat(
+ instr.gpr0, 0,
+ regs.GetRegisterAsFloat(instr.gpr8) + " * " + GetImmediate32(instr), 1, 1,
+ instr.fmul32.saturate, instr.op_32.generates_cc, 0, true);
break;
}
case OpCode::Id::FADD32I: {
- UNIMPLEMENTED_IF_MSG(instr.op_32.generates_cc,
- "Condition codes generation in FADD32I is not implemented");
+ UNIMPLEMENTED_IF_MSG(
+ instr.op_32.generates_cc,
+ "Condition codes generation in FADD32I is partially implemented");
std::string op_a = regs.GetRegisterAsFloat(instr.gpr8);
std::string op_b = GetImmediate32(instr);
@@ -2021,7 +2039,8 @@ private:
op_b = "-(" + op_b + ')';
}
- regs.SetRegisterToFloat(instr.gpr0, 0, op_a + " + " + op_b, 1, 1, false, 0, true);
+ regs.SetRegisterToFloat(instr.gpr0, 0, op_a + " + " + op_b, 1, 1, false,
+ instr.op_32.generates_cc, 0, true);
break;
}
}
@@ -2035,16 +2054,14 @@ private:
switch (opcode->get().GetId()) {
case OpCode::Id::BFE_IMM: {
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in BFE is not implemented");
-
std::string inner_shift =
'(' + op_a + " << " + std::to_string(instr.bfe.GetLeftShiftValue()) + ')';
std::string outer_shift =
'(' + inner_shift + " >> " +
std::to_string(instr.bfe.GetLeftShiftValue() + instr.bfe.shift_position) + ')';
- regs.SetRegisterToInteger(instr.gpr0, true, 0, outer_shift, 1, 1);
+ regs.SetRegisterToInteger(instr.gpr0, true, 0, outer_shift, 1, 1, false,
+ instr.generates_cc);
break;
}
default: {
@@ -2055,8 +2072,6 @@ private:
break;
}
case OpCode::Type::Bfi: {
- UNIMPLEMENTED_IF(instr.generates_cc);
-
const auto [base, packed_shift] = [&]() -> std::tuple<std::string, std::string> {
switch (opcode->get().GetId()) {
case OpCode::Id::BFI_IMM_R:
@@ -2071,9 +2086,10 @@ private:
const std::string offset = '(' + packed_shift + " & 0xff)";
const std::string bits = "((" + packed_shift + " >> 8) & 0xff)";
const std::string insert = regs.GetRegisterAsInteger(instr.gpr8, 0, false);
- regs.SetRegisterToInteger(
- instr.gpr0, false, 0,
- "bitfieldInsert(" + base + ", " + insert + ", " + offset + ", " + bits + ')', 1, 1);
+ regs.SetRegisterToInteger(instr.gpr0, false, 0,
+ "bitfieldInsert(" + base + ", " + insert + ", " + offset +
+ ", " + bits + ')',
+ 1, 1, false, instr.generates_cc);
break;
}
case OpCode::Type::Shift: {
@@ -2095,9 +2111,6 @@ private:
case OpCode::Id::SHR_C:
case OpCode::Id::SHR_R:
case OpCode::Id::SHR_IMM: {
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in SHR is not implemented");
-
if (!instr.shift.is_signed) {
// Logical shift right
op_a = "uint(" + op_a + ')';
@@ -2105,7 +2118,7 @@ private:
// Cast to int is superfluous for arithmetic shift, it's only for a logical shift
regs.SetRegisterToInteger(instr.gpr0, true, 0, "int(" + op_a + " >> " + op_b + ')',
- 1, 1);
+ 1, 1, false, instr.generates_cc);
break;
}
case OpCode::Id::SHL_C:
@@ -2113,7 +2126,8 @@ private:
case OpCode::Id::SHL_IMM:
UNIMPLEMENTED_IF_MSG(instr.generates_cc,
"Condition codes generation in SHL is not implemented");
- regs.SetRegisterToInteger(instr.gpr0, true, 0, op_a + " << " + op_b, 1, 1);
+ regs.SetRegisterToInteger(instr.gpr0, true, 0, op_a + " << " + op_b, 1, 1, false,
+ instr.generates_cc);
break;
default: {
UNIMPLEMENTED_MSG("Unhandled shift instruction: {}", opcode->get().GetName());
@@ -2127,18 +2141,17 @@ private:
switch (opcode->get().GetId()) {
case OpCode::Id::IADD32I:
- UNIMPLEMENTED_IF_MSG(instr.op_32.generates_cc,
- "Condition codes generation in IADD32I is not implemented");
+ UNIMPLEMENTED_IF_MSG(
+ instr.op_32.generates_cc,
+ "Condition codes generation in IADD32I is partially implemented");
if (instr.iadd32i.negate_a)
op_a = "-(" + op_a + ')';
regs.SetRegisterToInteger(instr.gpr0, true, 0, op_a + " + " + op_b, 1, 1,
- instr.iadd32i.saturate != 0);
+ instr.iadd32i.saturate, instr.op_32.generates_cc);
break;
case OpCode::Id::LOP32I: {
- UNIMPLEMENTED_IF_MSG(instr.op_32.generates_cc,
- "Condition codes generation in LOP32I is not implemented");
if (instr.alu.lop32i.invert_a)
op_a = "~(" + op_a + ')';
@@ -2148,7 +2161,7 @@ private:
WriteLogicOperation(instr.gpr0, instr.alu.lop32i.operation, op_a, op_b,
Tegra::Shader::PredicateResultMode::None,
- Tegra::Shader::Pred::UnusedIndex);
+ Tegra::Shader::Pred::UnusedIndex, instr.op_32.generates_cc);
break;
}
default: {
@@ -2177,7 +2190,7 @@ private:
case OpCode::Id::IADD_R:
case OpCode::Id::IADD_IMM: {
UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in IADD is not implemented");
+ "Condition codes generation in IADD is partially implemented");
if (instr.alu_integer.negate_a)
op_a = "-(" + op_a + ')';
@@ -2186,14 +2199,15 @@ private:
op_b = "-(" + op_b + ')';
regs.SetRegisterToInteger(instr.gpr0, true, 0, op_a + " + " + op_b, 1, 1,
- instr.alu.saturate_d);
+ instr.alu.saturate_d, instr.generates_cc);
break;
}
case OpCode::Id::IADD3_C:
case OpCode::Id::IADD3_R:
case OpCode::Id::IADD3_IMM: {
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in IADD3 is not implemented");
+ UNIMPLEMENTED_IF_MSG(
+ instr.generates_cc,
+ "Condition codes generation in IADD3 is partially implemented");
std::string op_c = regs.GetRegisterAsInteger(instr.gpr39);
@@ -2249,14 +2263,16 @@ private:
result = '(' + op_a + " + " + op_b + " + " + op_c + ')';
}
- regs.SetRegisterToInteger(instr.gpr0, true, 0, result, 1, 1);
+ regs.SetRegisterToInteger(instr.gpr0, true, 0, result, 1, 1, false,
+ instr.generates_cc);
break;
}
case OpCode::Id::ISCADD_C:
case OpCode::Id::ISCADD_R:
case OpCode::Id::ISCADD_IMM: {
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in ISCADD is not implemented");
+ UNIMPLEMENTED_IF_MSG(
+ instr.generates_cc,
+ "Condition codes generation in ISCADD is partially implemented");
if (instr.alu_integer.negate_a)
op_a = "-(" + op_a + ')';
@@ -2267,7 +2283,8 @@ private:
const std::string shift = std::to_string(instr.alu_integer.shift_amount.Value());
regs.SetRegisterToInteger(instr.gpr0, true, 0,
- "((" + op_a + " << " + shift + ") + " + op_b + ')', 1, 1);
+ "((" + op_a + " << " + shift + ") + " + op_b + ')', 1, 1,
+ false, instr.generates_cc);
break;
}
case OpCode::Id::POPC_C:
@@ -2291,8 +2308,6 @@ private:
case OpCode::Id::LOP_C:
case OpCode::Id::LOP_R:
case OpCode::Id::LOP_IMM: {
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in LOP is not implemented");
if (instr.alu.lop.invert_a)
op_a = "~(" + op_a + ')';
@@ -2301,15 +2316,13 @@ private:
op_b = "~(" + op_b + ')';
WriteLogicOperation(instr.gpr0, instr.alu.lop.operation, op_a, op_b,
- instr.alu.lop.pred_result_mode, instr.alu.lop.pred48);
+ instr.alu.lop.pred_result_mode, instr.alu.lop.pred48,
+ instr.generates_cc);
break;
}
case OpCode::Id::LOP3_C:
case OpCode::Id::LOP3_R:
case OpCode::Id::LOP3_IMM: {
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in LOP3 is not implemented");
-
const std::string op_c = regs.GetRegisterAsInteger(instr.gpr39);
std::string lut;
@@ -2319,15 +2332,16 @@ private:
lut = '(' + std::to_string(instr.alu.lop3.GetImmLut48()) + ')';
}
- WriteLop3Instruction(instr.gpr0, op_a, op_b, op_c, lut);
+ WriteLop3Instruction(instr.gpr0, op_a, op_b, op_c, lut, instr.generates_cc);
break;
}
case OpCode::Id::IMNMX_C:
case OpCode::Id::IMNMX_R:
case OpCode::Id::IMNMX_IMM: {
UNIMPLEMENTED_IF(instr.imnmx.exchange != Tegra::Shader::IMinMaxExchange::None);
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in IMNMX is not implemented");
+ UNIMPLEMENTED_IF_MSG(
+ instr.generates_cc,
+ "Condition codes generation in IMNMX is partially implemented");
const std::string condition =
GetPredicateCondition(instr.imnmx.pred, instr.imnmx.negate_pred != 0);
@@ -2335,7 +2349,7 @@ private:
regs.SetRegisterToInteger(instr.gpr0, instr.imnmx.is_signed, 0,
'(' + condition + ") ? min(" + parameters + ") : max(" +
parameters + ')',
- 1, 1);
+ 1, 1, false, instr.generates_cc);
break;
}
case OpCode::Id::LEA_R2:
@@ -2396,7 +2410,8 @@ private:
UNIMPLEMENTED_IF_MSG(instr.lea.pred48 != static_cast<u64>(Pred::UnusedIndex),
"Unhandled LEA Predicate");
const std::string value = '(' + op_a + " + (" + op_b + "*(1 << " + op_c + ")))";
- regs.SetRegisterToInteger(instr.gpr0, true, 0, value, 1, 1);
+ regs.SetRegisterToInteger(instr.gpr0, true, 0, value, 1, 1, false,
+ instr.generates_cc);
break;
}
@@ -2501,7 +2516,7 @@ private:
UNIMPLEMENTED_IF_MSG(instr.ffma.tab5980_1 != 0, "FFMA tab5980_1({}) not implemented",
instr.ffma.tab5980_1.Value());
UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in FFMA is not implemented");
+ "Condition codes generation in FFMA is partially implemented");
switch (opcode->get().GetId()) {
case OpCode::Id::FFMA_CR: {
@@ -2532,7 +2547,7 @@ private:
}
regs.SetRegisterToFloat(instr.gpr0, 0, "fma(" + op_a + ", " + op_b + ", " + op_c + ')',
- 1, 1, instr.alu.saturate_d, 0, true);
+ 1, 1, instr.alu.saturate_d, instr.generates_cc, 0, true);
break;
}
case OpCode::Type::Hfma2: {
@@ -2603,16 +2618,14 @@ private:
}
regs.SetRegisterToInteger(instr.gpr0, instr.conversion.is_output_signed, 0, op_a, 1,
- 1, instr.alu.saturate_d, 0, instr.conversion.dest_size,
- instr.generates_cc.Value() != 0);
+ 1, instr.alu.saturate_d, instr.generates_cc, 0,
+ instr.conversion.dest_size);
break;
}
case OpCode::Id::I2F_R:
case OpCode::Id::I2F_C: {
UNIMPLEMENTED_IF(instr.conversion.dest_size != Register::Size::Word);
UNIMPLEMENTED_IF(instr.conversion.selector);
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in I2F is not implemented");
std::string op_a;
if (instr.is_b_gpr) {
@@ -2635,14 +2648,12 @@ private:
op_a = "-(" + op_a + ')';
}
- regs.SetRegisterToFloat(instr.gpr0, 0, op_a, 1, 1);
+ regs.SetRegisterToFloat(instr.gpr0, 0, op_a, 1, 1, false, instr.generates_cc);
break;
}
case OpCode::Id::F2F_R: {
UNIMPLEMENTED_IF(instr.conversion.dest_size != Register::Size::Word);
UNIMPLEMENTED_IF(instr.conversion.src_size != Register::Size::Word);
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in F2F is not implemented");
std::string op_a = regs.GetRegisterAsFloat(instr.gpr20);
if (instr.conversion.abs_a) {
@@ -2674,14 +2685,13 @@ private:
break;
}
- regs.SetRegisterToFloat(instr.gpr0, 0, op_a, 1, 1, instr.alu.saturate_d);
+ regs.SetRegisterToFloat(instr.gpr0, 0, op_a, 1, 1, instr.alu.saturate_d,
+ instr.generates_cc);
break;
}
case OpCode::Id::F2I_R:
case OpCode::Id::F2I_C: {
UNIMPLEMENTED_IF(instr.conversion.src_size != Register::Size::Word);
- UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in F2I is not implemented");
std::string op_a{};
if (instr.is_b_gpr) {
@@ -2724,7 +2734,8 @@ private:
}
regs.SetRegisterToInteger(instr.gpr0, instr.conversion.is_output_signed, 0, op_a, 1,
- 1, false, 0, instr.conversion.dest_size);
+ 1, false, instr.generates_cc, 0,
+ instr.conversion.dest_size);
break;
}
default: {
@@ -2887,7 +2898,7 @@ private:
shader.AddLine(coord);
if (depth_compare) {
- regs.SetRegisterToFloat(instr.gpr0, 0, texture, 1, 1, false);
+ regs.SetRegisterToFloat(instr.gpr0, 0, texture, 1, 1);
} else {
shader.AddLine("vec4 texture_tmp = " + texture + ';');
std::size_t dest_elem{};
@@ -2896,7 +2907,7 @@ private:
// Skip disabled components
continue;
}
- regs.SetRegisterToFloat(instr.gpr0, elem, "texture_tmp", 1, 4, false,
+ regs.SetRegisterToFloat(instr.gpr0, elem, "texture_tmp", 1, 4, false, false,
dest_elem);
++dest_elem;
}
@@ -2982,7 +2993,7 @@ private:
// Skip disabled components
continue;
}
- regs.SetRegisterToFloat(instr.gpr0, elem, "texture_tmp", 1, 4, false,
+ regs.SetRegisterToFloat(instr.gpr0, elem, "texture_tmp", 1, 4, false, false,
dest_elem);
++dest_elem;
}
@@ -3231,7 +3242,7 @@ private:
}
case OpCode::Type::PredicateSetRegister: {
UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in PSET is not implemented");
+ "Condition codes generation in PSET is partially implemented");
const std::string op_a =
GetPredicateCondition(instr.pset.pred12, instr.pset.neg_pred12 != 0);
@@ -3248,10 +3259,11 @@ private:
const std::string result = '(' + predicate + ") " + combiner + " (" + second_pred + ')';
if (instr.pset.bf == 0) {
const std::string value = '(' + result + ") ? 0xFFFFFFFF : 0";
- regs.SetRegisterToInteger(instr.gpr0, false, 0, value, 1, 1);
+ regs.SetRegisterToInteger(instr.gpr0, false, 0, value, 1, 1, false,
+ instr.generates_cc);
} else {
const std::string value = '(' + result + ") ? 1.0 : 0.0";
- regs.SetRegisterToFloat(instr.gpr0, 0, value, 1, 1);
+ regs.SetRegisterToFloat(instr.gpr0, 0, value, 1, 1, false, instr.generates_cc);
}
break;
}
@@ -3368,14 +3380,11 @@ private:
") " + combiner + " (" + second_pred + "))";
if (instr.fset.bf) {
- regs.SetRegisterToFloat(instr.gpr0, 0, predicate + " ? 1.0 : 0.0", 1, 1);
+ regs.SetRegisterToFloat(instr.gpr0, 0, predicate + " ? 1.0 : 0.0", 1, 1, false,
+ instr.generates_cc);
} else {
regs.SetRegisterToInteger(instr.gpr0, false, 0, predicate + " ? 0xFFFFFFFF : 0", 1,
- 1);
- }
- if (instr.generates_cc.Value() != 0) {
- regs.SetInternalFlag(InternalFlag::ZeroFlag, predicate);
- LOG_WARNING(HW_GPU, "FSET Condition Code is incomplete");
+ 1, false, instr.generates_cc);
}
break;
}
@@ -3462,7 +3471,7 @@ private:
UNIMPLEMENTED_IF(instr.xmad.sign_a);
UNIMPLEMENTED_IF(instr.xmad.sign_b);
UNIMPLEMENTED_IF_MSG(instr.generates_cc,
- "Condition codes generation in XMAD is not implemented");
+ "Condition codes generation in XMAD is partially implemented");
std::string op_a{regs.GetRegisterAsInteger(instr.gpr8, 0, instr.xmad.sign_a)};
std::string op_b;
@@ -3548,7 +3557,8 @@ private:
sum = "((" + sum + " & 0xFFFF) | (" + src2 + "<< 16))";
}
- regs.SetRegisterToInteger(instr.gpr0, is_signed, 0, sum, 1, 1);
+ regs.SetRegisterToInteger(instr.gpr0, is_signed, 0, sum, 1, 1, false,
+ instr.generates_cc);
break;
}
default: {
@@ -3752,8 +3762,7 @@ private:
}
regs.SetRegisterToInteger(instr.gpr0, result_signed, 1, result, 1, 1,
- instr.vmad.saturate == 1, 0, Register::Size::Word,
- instr.vmad.cc);
+ instr.vmad.saturate, instr.vmad.cc);
break;
}
case OpCode::Id::VSETP: {
diff --git a/src/video_core/renderer_opengl/renderer_opengl.h b/src/video_core/renderer_opengl/renderer_opengl.h
index 067fad81b..b85cc262f 100644
--- a/src/video_core/renderer_opengl/renderer_opengl.h
+++ b/src/video_core/renderer_opengl/renderer_opengl.h
@@ -17,7 +17,7 @@ class EmuWindow;
}
namespace Layout {
-class FramebufferLayout;
+struct FramebufferLayout;
}
namespace OpenGL {
diff --git a/src/video_core/surface.cpp b/src/video_core/surface.cpp
index a97b1562b..1a344229f 100644
--- a/src/video_core/surface.cpp
+++ b/src/video_core/surface.cpp
@@ -196,11 +196,14 @@ PixelFormat PixelFormatFromTextureFormat(Tegra::Texture::TextureFormat format,
LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
UNREACHABLE();
case Tegra::Texture::TextureFormat::G8R8:
+ // TextureFormat::G8R8 is actually ordered red then green, as such we can use
+ // PixelFormat::RG8U and PixelFormat::RG8S. This was tested with The Legend of Zelda: Breath
+ // of the Wild, which uses this format to render the hearts on the UI.
switch (component_type) {
case Tegra::Texture::ComponentType::UNORM:
- return PixelFormat::G8R8U;
+ return PixelFormat::RG8U;
case Tegra::Texture::ComponentType::SNORM:
- return PixelFormat::G8R8S;
+ return PixelFormat::RG8S;
}
LOG_CRITICAL(HW_GPU, "Unimplemented component_type={}", static_cast<u32>(component_type));
UNREACHABLE();
diff --git a/src/video_core/surface.h b/src/video_core/surface.h
index e23cfecbc..c2259c3c2 100644
--- a/src/video_core/surface.h
+++ b/src/video_core/surface.h
@@ -38,57 +38,55 @@ enum class PixelFormat {
BC6H_UF16 = 20,
BC6H_SF16 = 21,
ASTC_2D_4X4 = 22,
- G8R8U = 23,
- G8R8S = 24,
- BGRA8 = 25,
- RGBA32F = 26,
- RG32F = 27,
- R32F = 28,
- R16F = 29,
- R16U = 30,
- R16S = 31,
- R16UI = 32,
- R16I = 33,
- RG16 = 34,
- RG16F = 35,
- RG16UI = 36,
- RG16I = 37,
- RG16S = 38,
- RGB32F = 39,
- RGBA8_SRGB = 40,
- RG8U = 41,
- RG8S = 42,
- RG32UI = 43,
- R32UI = 44,
- ASTC_2D_8X8 = 45,
- ASTC_2D_8X5 = 46,
- ASTC_2D_5X4 = 47,
- BGRA8_SRGB = 48,
- DXT1_SRGB = 49,
- DXT23_SRGB = 50,
- DXT45_SRGB = 51,
- BC7U_SRGB = 52,
- ASTC_2D_4X4_SRGB = 53,
- ASTC_2D_8X8_SRGB = 54,
- ASTC_2D_8X5_SRGB = 55,
- ASTC_2D_5X4_SRGB = 56,
- ASTC_2D_5X5 = 57,
- ASTC_2D_5X5_SRGB = 58,
- ASTC_2D_10X8 = 59,
- ASTC_2D_10X8_SRGB = 60,
+ BGRA8 = 23,
+ RGBA32F = 24,
+ RG32F = 25,
+ R32F = 26,
+ R16F = 27,
+ R16U = 28,
+ R16S = 29,
+ R16UI = 30,
+ R16I = 31,
+ RG16 = 32,
+ RG16F = 33,
+ RG16UI = 34,
+ RG16I = 35,
+ RG16S = 36,
+ RGB32F = 37,
+ RGBA8_SRGB = 38,
+ RG8U = 39,
+ RG8S = 40,
+ RG32UI = 41,
+ R32UI = 42,
+ ASTC_2D_8X8 = 43,
+ ASTC_2D_8X5 = 44,
+ ASTC_2D_5X4 = 45,
+ BGRA8_SRGB = 46,
+ DXT1_SRGB = 47,
+ DXT23_SRGB = 48,
+ DXT45_SRGB = 49,
+ BC7U_SRGB = 50,
+ ASTC_2D_4X4_SRGB = 51,
+ ASTC_2D_8X8_SRGB = 52,
+ ASTC_2D_8X5_SRGB = 53,
+ ASTC_2D_5X4_SRGB = 54,
+ ASTC_2D_5X5 = 55,
+ ASTC_2D_5X5_SRGB = 56,
+ ASTC_2D_10X8 = 57,
+ ASTC_2D_10X8_SRGB = 58,
MaxColorFormat,
// Depth formats
- Z32F = 61,
- Z16 = 62,
+ Z32F = 59,
+ Z16 = 60,
MaxDepthFormat,
// DepthStencil formats
- Z24S8 = 63,
- S8Z24 = 64,
- Z32FS8 = 65,
+ Z24S8 = 61,
+ S8Z24 = 62,
+ Z32FS8 = 63,
MaxDepthStencilFormat,
@@ -149,8 +147,6 @@ constexpr std::array<u32, MaxPixelFormat> compression_factor_table = {{
4, // BC6H_UF16
4, // BC6H_SF16
4, // ASTC_2D_4X4
- 1, // G8R8U
- 1, // G8R8S
1, // BGRA8
1, // RGBA32F
1, // RG32F
@@ -232,8 +228,6 @@ constexpr std::array<u32, MaxPixelFormat> block_width_table = {{
4, // BC6H_UF16
4, // BC6H_SF16
4, // ASTC_2D_4X4
- 1, // G8R8U
- 1, // G8R8S
1, // BGRA8
1, // RGBA32F
1, // RG32F
@@ -309,8 +303,6 @@ constexpr std::array<u32, MaxPixelFormat> block_height_table = {{
4, // BC6H_UF16
4, // BC6H_SF16
4, // ASTC_2D_4X4
- 1, // G8R8U
- 1, // G8R8S
1, // BGRA8
1, // RGBA32F
1, // RG32F
@@ -386,8 +378,6 @@ constexpr std::array<u32, MaxPixelFormat> bpp_table = {{
128, // BC6H_UF16
128, // BC6H_SF16
128, // ASTC_2D_4X4
- 16, // G8R8U
- 16, // G8R8S
32, // BGRA8
128, // RGBA32F
64, // RG32F
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-24 05:57:32 +0100