summaryrefslogtreecommitdiff
path: root/src
diff options
context:
space:
mode:
Diffstat (limited to 'src')
-rw-r--r--src/video_core/engines/shader_bytecode.h19
-rw-r--r--src/video_core/renderer_opengl/gl_rasterizer_cache.cpp34
-rw-r--r--src/video_core/renderer_opengl/gl_shader_decompiler.cpp128
3 files changed, 92 insertions, 89 deletions
diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index ceef8087c..cefd57f4c 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -230,10 +230,18 @@ union Instruction {
} fmnmx;
union {
+ BitField<39, 1, u64> invert_a;
+ BitField<40, 1, u64> invert_b;
+ BitField<41, 2, LogicOperation> operation;
+ BitField<44, 2, u64> unk44;
+ BitField<48, 3, Pred> pred48;
+ } lop;
+
+ union {
BitField<53, 2, LogicOperation> operation;
BitField<55, 1, u64> invert_a;
BitField<56, 1, u64> invert_b;
- } lop;
+ } lop32i;
float GetImm20_19() const {
float result{};
@@ -476,6 +484,9 @@ public:
I2I_C,
I2I_R,
I2I_IMM,
+ LOP_C,
+ LOP_R,
+ LOP_IMM,
LOP32I,
MOV_C,
MOV_R,
@@ -518,7 +529,6 @@ public:
ArithmeticInteger,
ArithmeticIntegerImmediate,
Bfe,
- Logic,
Shift,
Ffma,
Flow,
@@ -676,7 +686,10 @@ private:
INST("0100110000000---", Id::BFE_C, Type::Bfe, "BFE_C"),
INST("0101110000000---", Id::BFE_R, Type::Bfe, "BFE_R"),
INST("0011100-00000---", Id::BFE_IMM, Type::Bfe, "BFE_IMM"),
- INST("000001----------", Id::LOP32I, Type::Logic, "LOP32I"),
+ INST("0100110001000---", Id::LOP_C, Type::ArithmeticInteger, "LOP_C"),
+ INST("0101110001000---", Id::LOP_R, Type::ArithmeticInteger, "LOP_R"),
+ INST("0011100001000---", Id::LOP_IMM, Type::ArithmeticInteger, "LOP_IMM"),
+ INST("000001----------", Id::LOP32I, Type::ArithmeticIntegerImmediate, "LOP32I"),
INST("0100110001001---", Id::SHL_C, Type::Shift, "SHL_C"),
INST("0101110001001---", Id::SHL_R, Type::Shift, "SHL_R"),
INST("0011100-01001---", Id::SHL_IMM, Type::Shift, "SHL_IMM"),
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index ff48a2669..e61960cc0 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -888,9 +888,6 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, ScaleMatc
// Use GetSurfaceSubRect instead
ASSERT(params.width == params.stride);
- ASSERT(!params.is_tiled ||
- (params.GetActualWidth() % 8 == 0 && params.GetActualHeight() % 8 == 0));
-
// Check for an exact match in existing surfaces
Surface surface =
FindMatch<MatchFlags::Exact | MatchFlags::Invalid>(surface_cache, params, match_res_scale);
@@ -1048,8 +1045,13 @@ Surface RasterizerCacheOpenGL::GetTextureSurface(const Tegra::Texture::FullTextu
if (config.tic.IsTiled()) {
params.block_height = config.tic.BlockHeight();
- params.width = Common::AlignUp(params.width, params.block_height);
- params.height = Common::AlignUp(params.height, params.block_height);
+
+ // TODO(bunnei): The below align up is a hack. This is here because some compressed textures
+ // are not a multiple of their own compression factor, and so this accounts for that. This
+ // could potentially result in an extra row of 4px being decoded if a texture is not a
+ // multiple of 4.
+ params.width = Common::AlignUp(params.width, 4);
+ params.height = Common::AlignUp(params.height, 4);
} else {
// Use the texture-provided stride value if the texture isn't tiled.
params.stride = static_cast<u32>(params.PixelsInBytes(config.tic.Pitch()));
@@ -1057,26 +1059,6 @@ Surface RasterizerCacheOpenGL::GetTextureSurface(const Tegra::Texture::FullTextu
params.UpdateParams();
- if (params.GetActualWidth() % 8 != 0 || params.GetActualHeight() % 8 != 0 ||
- params.stride != params.width) {
- Surface src_surface;
- MathUtil::Rectangle<u32> rect;
- std::tie(src_surface, rect) = GetSurfaceSubRect(params, ScaleMatch::Ignore, true);
-
- rect = rect.Scale(params.GetCompresssionFactor());
-
- params.res_scale = src_surface->res_scale;
- Surface tmp_surface = CreateSurface(params);
-
- auto dst_rect = tmp_surface->GetScaledRect().Scale(params.GetCompresssionFactor());
- BlitTextures(src_surface->texture.handle, rect, tmp_surface->texture.handle, dst_rect,
- SurfaceParams::GetFormatType(params.pixel_format), read_framebuffer.handle,
- draw_framebuffer.handle);
-
- remove_surfaces.emplace(tmp_surface);
- return tmp_surface;
- }
-
return GetSurface(params, ScaleMatch::Ignore, true);
}
@@ -1251,7 +1233,7 @@ void RasterizerCacheOpenGL::ValidateSurface(const Surface& surface, Tegra::GPUVA
const auto interval = *it & validate_interval;
// Look for a valid surface to copy from
- SurfaceParams params = surface->FromInterval(interval);
+ SurfaceParams params = *surface;
Surface copy_surface =
FindMatch<MatchFlags::Copy>(surface_cache, params, ScaleMatch::Ignore, interval);
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index af68c3bda..6ec0a0742 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -16,6 +16,7 @@ namespace Decompiler {
using Tegra::Shader::Attribute;
using Tegra::Shader::Instruction;
+using Tegra::Shader::LogicOperation;
using Tegra::Shader::OpCode;
using Tegra::Shader::Register;
using Tegra::Shader::Sampler;
@@ -759,6 +760,31 @@ private:
return (absolute_offset % SchedPeriod) == 0;
}
+ void WriteLogicOperation(Register dest, LogicOperation logic_op, const std::string& op_a,
+ const std::string& op_b) {
+ switch (logic_op) {
+ case LogicOperation::And: {
+ regs.SetRegisterToInteger(dest, true, 0, '(' + op_a + " & " + op_b + ')', 1, 1);
+ break;
+ }
+ case LogicOperation::Or: {
+ regs.SetRegisterToInteger(dest, true, 0, '(' + op_a + " | " + op_b + ')', 1, 1);
+ break;
+ }
+ case LogicOperation::Xor: {
+ regs.SetRegisterToInteger(dest, true, 0, '(' + op_a + " ^ " + op_b + ')', 1, 1);
+ break;
+ }
+ case LogicOperation::PassB: {
+ regs.SetRegisterToInteger(dest, true, 0, op_b, 1, 1);
+ break;
+ }
+ default:
+ NGLOG_CRITICAL(HW_GPU, "Unimplemented logic operation: {}", static_cast<u32>(logic_op));
+ UNREACHABLE();
+ }
+ }
+
/**
* Compiles a single instruction from Tegra to GLSL.
* @param offset the offset of the Tegra shader instruction.
@@ -942,55 +968,6 @@ private:
break;
}
- case OpCode::Type::Logic: {
- std::string op_a = regs.GetRegisterAsInteger(instr.gpr8, 0, true);
-
- if (instr.alu.lop.invert_a)
- op_a = "~(" + op_a + ')';
-
- switch (opcode->GetId()) {
- case OpCode::Id::LOP32I: {
- u32 imm = static_cast<u32>(instr.alu.imm20_32.Value());
-
- if (instr.alu.lop.invert_b)
- imm = ~imm;
-
- std::string op_b = std::to_string(imm);
-
- switch (instr.alu.lop.operation) {
- case Tegra::Shader::LogicOperation::And: {
- regs.SetRegisterToInteger(instr.gpr0, true, 0, '(' + op_a + " & " + op_b + ')',
- 1, 1);
- break;
- }
- case Tegra::Shader::LogicOperation::Or: {
- regs.SetRegisterToInteger(instr.gpr0, true, 0, '(' + op_a + " | " + op_b + ')',
- 1, 1);
- break;
- }
- case Tegra::Shader::LogicOperation::Xor: {
- regs.SetRegisterToInteger(instr.gpr0, true, 0, '(' + op_a + " ^ " + op_b + ')',
- 1, 1);
- break;
- }
- case Tegra::Shader::LogicOperation::PassB: {
- regs.SetRegisterToInteger(instr.gpr0, true, 0, op_b, 1, 1);
- break;
- }
- default:
- NGLOG_CRITICAL(HW_GPU, "Unimplemented lop32i operation: {}",
- static_cast<u32>(instr.alu.lop.operation.Value()));
- UNREACHABLE();
- }
- break;
- }
- default: {
- NGLOG_CRITICAL(HW_GPU, "Unhandled logic instruction: {}", opcode->GetName());
- UNREACHABLE();
- }
- }
- break;
- }
case OpCode::Type::Shift: {
std::string op_a = regs.GetRegisterAsInteger(instr.gpr8, 0, true);
@@ -1036,17 +1013,26 @@ private:
case OpCode::Type::ArithmeticIntegerImmediate: {
std::string op_a = regs.GetRegisterAsInteger(instr.gpr8);
-
- if (instr.iadd32i.negate_a)
- op_a = '-' + op_a;
-
- std::string op_b = '(' + std::to_string(instr.alu.imm20_32.Value()) + ')';
+ std::string op_b = std::to_string(instr.alu.imm20_32.Value());
switch (opcode->GetId()) {
case OpCode::Id::IADD32I:
+ 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);
break;
+ case OpCode::Id::LOP32I: {
+ if (instr.alu.lop32i.invert_a)
+ op_a = "~(" + op_a + ')';
+
+ if (instr.alu.lop32i.invert_b)
+ op_b = "~(" + op_b + ')';
+
+ WriteLogicOperation(instr.gpr0, instr.alu.lop32i.operation, op_a, op_b);
+ break;
+ }
default: {
NGLOG_CRITICAL(HW_GPU, "Unhandled ArithmeticIntegerImmediate instruction: {}",
opcode->GetName());
@@ -1057,12 +1043,7 @@ private:
}
case OpCode::Type::ArithmeticInteger: {
std::string op_a = regs.GetRegisterAsInteger(instr.gpr8);
-
- if (instr.alu_integer.negate_a)
- op_a = '-' + op_a;
-
- std::string op_b = instr.alu_integer.negate_b ? "-" : "";
-
+ std::string op_b;
if (instr.is_b_imm) {
op_b += '(' + std::to_string(instr.alu.GetSignedImm20_20()) + ')';
} else {
@@ -1078,6 +1059,12 @@ private:
case OpCode::Id::IADD_C:
case OpCode::Id::IADD_R:
case OpCode::Id::IADD_IMM: {
+ if (instr.alu_integer.negate_a)
+ op_a = "-(" + op_a + ')';
+
+ if (instr.alu_integer.negate_b)
+ op_b = "-(" + op_b + ')';
+
regs.SetRegisterToInteger(instr.gpr0, true, 0, op_a + " + " + op_b, 1, 1,
instr.alu.saturate_d);
break;
@@ -1085,12 +1072,33 @@ private:
case OpCode::Id::ISCADD_C:
case OpCode::Id::ISCADD_R:
case OpCode::Id::ISCADD_IMM: {
+ if (instr.alu_integer.negate_a)
+ op_a = "-(" + op_a + ')';
+
+ if (instr.alu_integer.negate_b)
+ op_b = "-(" + op_b + ')';
+
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);
break;
}
+ case OpCode::Id::LOP_C:
+ case OpCode::Id::LOP_R:
+ case OpCode::Id::LOP_IMM: {
+ ASSERT_MSG(!instr.alu.lop.unk44, "Unimplemented");
+ ASSERT_MSG(instr.alu.lop.pred48 == Pred::UnusedIndex, "Unimplemented");
+
+ if (instr.alu.lop.invert_a)
+ op_a = "~(" + op_a + ')';
+
+ if (instr.alu.lop.invert_b)
+ op_b = "~(" + op_b + ')';
+
+ WriteLogicOperation(instr.gpr0, instr.alu.lop.operation, op_a, op_b);
+ break;
+ }
default: {
NGLOG_CRITICAL(HW_GPU, "Unhandled ArithmeticInteger instruction: {}",
opcode->GetName());
generated by cgit v1.2.3 (git 2.25.1) at 2025-05-22 14:29:13 +0000