diff options
Diffstat (limited to 'src/video_core/shader')
33 files changed, 746 insertions, 669 deletions
diff --git a/src/video_core/shader/decode.cpp b/src/video_core/shader/decode.cpp index 2da595c0d..a0554c97e 100644 --- a/src/video_core/shader/decode.cpp +++ b/src/video_core/shader/decode.cpp @@ -11,6 +11,7 @@ #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" #include "video_core/engines/shader_header.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/arithmetic.cpp b/src/video_core/shader/decode/arithmetic.cpp index b4859bc1e..87d8fecaa 100644 --- a/src/video_core/shader/decode/arithmetic.cpp +++ b/src/video_core/shader/decode/arithmetic.cpp @@ -6,6 +6,7 @@ #include "common/common_types.h" #include "common/logging/log.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/arithmetic_half.cpp b/src/video_core/shader/decode/arithmetic_half.cpp index 3a29c4a46..b06cbe441 100644 --- a/src/video_core/shader/decode/arithmetic_half.cpp +++ b/src/video_core/shader/decode/arithmetic_half.cpp @@ -6,6 +6,7 @@ #include "common/common_types.h" #include "common/logging/log.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/arithmetic_half_immediate.cpp b/src/video_core/shader/decode/arithmetic_half_immediate.cpp index 5341e460f..7bcf38f23 100644 --- a/src/video_core/shader/decode/arithmetic_half_immediate.cpp +++ b/src/video_core/shader/decode/arithmetic_half_immediate.cpp @@ -6,6 +6,7 @@ #include "common/common_types.h" #include "common/logging/log.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/arithmetic_immediate.cpp b/src/video_core/shader/decode/arithmetic_immediate.cpp index 3095f2fd4..f1875967c 100644 --- a/src/video_core/shader/decode/arithmetic_immediate.cpp +++ b/src/video_core/shader/decode/arithmetic_immediate.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp index 9fd4b273e..c8c1a7f40 100644 --- a/src/video_core/shader/decode/arithmetic_integer.cpp +++ b/src/video_core/shader/decode/arithmetic_integer.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/arithmetic_integer_immediate.cpp b/src/video_core/shader/decode/arithmetic_integer_immediate.cpp index 679ac0d4e..73880db0e 100644 --- a/src/video_core/shader/decode/arithmetic_integer_immediate.cpp +++ b/src/video_core/shader/decode/arithmetic_integer_immediate.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/bfe.cpp b/src/video_core/shader/decode/bfe.cpp index 1ae192c6a..e02bcd097 100644 --- a/src/video_core/shader/decode/bfe.cpp +++ b/src/video_core/shader/decode/bfe.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/bfi.cpp b/src/video_core/shader/decode/bfi.cpp index 0b12a0d08..8be1119df 100644 --- a/src/video_core/shader/decode/bfi.cpp +++ b/src/video_core/shader/decode/bfi.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/conversion.cpp b/src/video_core/shader/decode/conversion.cpp index b5ec9a6f5..4221f0c58 100644 --- a/src/video_core/shader/decode/conversion.cpp +++ b/src/video_core/shader/decode/conversion.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/ffma.cpp b/src/video_core/shader/decode/ffma.cpp index a1d04c6e5..29be25ca3 100644 --- a/src/video_core/shader/decode/ffma.cpp +++ b/src/video_core/shader/decode/ffma.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/float_set.cpp b/src/video_core/shader/decode/float_set.cpp index cc522f1de..f5013e44a 100644 --- a/src/video_core/shader/decode/float_set.cpp +++ b/src/video_core/shader/decode/float_set.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/float_set_predicate.cpp b/src/video_core/shader/decode/float_set_predicate.cpp index 9d2322a1d..2323052b0 100644 --- a/src/video_core/shader/decode/float_set_predicate.cpp +++ b/src/video_core/shader/decode/float_set_predicate.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/half_set.cpp b/src/video_core/shader/decode/half_set.cpp index 755f2ec44..48ca7a4af 100644 --- a/src/video_core/shader/decode/half_set.cpp +++ b/src/video_core/shader/decode/half_set.cpp @@ -8,6 +8,7 @@ #include "common/common_types.h" #include "common/logging/log.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/half_set_predicate.cpp b/src/video_core/shader/decode/half_set_predicate.cpp index fba44d714..d59d15bd8 100644 --- a/src/video_core/shader/decode/half_set_predicate.cpp +++ b/src/video_core/shader/decode/half_set_predicate.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/hfma2.cpp b/src/video_core/shader/decode/hfma2.cpp index a425f9eb7..c3bcf1ae9 100644 --- a/src/video_core/shader/decode/hfma2.cpp +++ b/src/video_core/shader/decode/hfma2.cpp @@ -7,6 +7,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/integer_set.cpp b/src/video_core/shader/decode/integer_set.cpp index a4cdaf74d..46e3d5905 100644 --- a/src/video_core/shader/decode/integer_set.cpp +++ b/src/video_core/shader/decode/integer_set.cpp @@ -4,6 +4,7 @@ #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/integer_set_predicate.cpp b/src/video_core/shader/decode/integer_set_predicate.cpp index a6a1fb632..dd20775d7 100644 --- a/src/video_core/shader/decode/integer_set_predicate.cpp +++ b/src/video_core/shader/decode/integer_set_predicate.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp index e6a010a7d..80fc0ccfc 100644 --- a/src/video_core/shader/decode/memory.cpp +++ b/src/video_core/shader/decode/memory.cpp @@ -10,6 +10,7 @@ #include "common/common_types.h" #include "common/logging/log.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { @@ -169,7 +170,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { const Node it_offset = Immediate(i * 4); const Node real_address = Operation(OperationCode::UAdd, NO_PRECISE, real_address_base, it_offset); - const Node gmem = StoreNode(GmemNode(real_address, base_address, descriptor)); + const Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor); SetTemporal(bb, i, gmem); } @@ -262,7 +263,7 @@ u32 ShaderIR::DecodeMemory(NodeBlock& bb, u32 pc) { const Node it_offset = Immediate(i * 4); const Node real_address = Operation(OperationCode::UAdd, NO_PRECISE, real_address_base, it_offset); - const Node gmem = StoreNode(GmemNode(real_address, base_address, descriptor)); + const Node gmem = MakeNode<GmemNode>(real_address, base_address, descriptor); bb.push_back(Operation(OperationCode::Assign, gmem, GetTemporal(i + 1))); } @@ -298,9 +299,9 @@ std::tuple<Node, Node, GlobalMemoryBase> ShaderIR::TrackAndGetGlobalMemory(NodeB const Node base_address{ TrackCbuf(addr_register, global_code, static_cast<s64>(global_code.size()))}; - const auto cbuf = std::get_if<CbufNode>(base_address); + const auto cbuf = std::get_if<CbufNode>(&*base_address); ASSERT(cbuf != nullptr); - const auto cbuf_offset_imm = std::get_if<ImmediateNode>(cbuf->GetOffset()); + const auto cbuf_offset_imm = std::get_if<ImmediateNode>(&*cbuf->GetOffset()); ASSERT(cbuf_offset_imm != nullptr); const auto cbuf_offset = cbuf_offset_imm->GetValue(); diff --git a/src/video_core/shader/decode/other.cpp b/src/video_core/shader/decode/other.cpp index a6c123573..d46a8ab82 100644 --- a/src/video_core/shader/decode/other.cpp +++ b/src/video_core/shader/decode/other.cpp @@ -6,6 +6,7 @@ #include "common/common_types.h" #include "common/logging/log.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { @@ -108,22 +109,20 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) { UNIMPLEMENTED_IF_MSG(instr.bra.constant_buffer != 0, "Constant buffer flow is not supported"); - // The SSY opcode tells the GPU where to re-converge divergent execution paths, it sets the - // target of the jump that the SYNC instruction will make. The SSY opcode has a similar - // structure to the BRA opcode. + // The SSY opcode tells the GPU where to re-converge divergent execution paths with SYNC. const u32 target = pc + instr.bra.GetBranchTarget(); - bb.push_back(Operation(OperationCode::PushFlowStack, Immediate(target))); + bb.push_back( + Operation(OperationCode::PushFlowStack, MetaStackClass::Ssy, Immediate(target))); break; } case OpCode::Id::PBK: { UNIMPLEMENTED_IF_MSG(instr.bra.constant_buffer != 0, "Constant buffer PBK is not supported"); - // PBK pushes to a stack the address where BRK will jump to. This shares stack with SSY but - // using SYNC on a PBK address will kill the shader execution. We don't emulate this because - // it's very unlikely a driver will emit such invalid shader. + // PBK pushes to a stack the address where BRK will jump to. const u32 target = pc + instr.bra.GetBranchTarget(); - bb.push_back(Operation(OperationCode::PushFlowStack, Immediate(target))); + bb.push_back( + Operation(OperationCode::PushFlowStack, MetaStackClass::Pbk, Immediate(target))); break; } case OpCode::Id::SYNC: { @@ -132,7 +131,7 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) { static_cast<u32>(cc)); // The SYNC opcode jumps to the address previously set by the SSY opcode - bb.push_back(Operation(OperationCode::PopFlowStack)); + bb.push_back(Operation(OperationCode::PopFlowStack, MetaStackClass::Ssy)); break; } case OpCode::Id::BRK: { @@ -141,7 +140,7 @@ u32 ShaderIR::DecodeOther(NodeBlock& bb, u32 pc) { static_cast<u32>(cc)); // The BRK opcode jumps to the address previously set by the PBK opcode - bb.push_back(Operation(OperationCode::PopFlowStack)); + bb.push_back(Operation(OperationCode::PopFlowStack, MetaStackClass::Pbk)); break; } case OpCode::Id::IPA: { diff --git a/src/video_core/shader/decode/predicate_set_predicate.cpp b/src/video_core/shader/decode/predicate_set_predicate.cpp index 71844c42b..9290d22eb 100644 --- a/src/video_core/shader/decode/predicate_set_predicate.cpp +++ b/src/video_core/shader/decode/predicate_set_predicate.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/predicate_set_register.cpp b/src/video_core/shader/decode/predicate_set_register.cpp index 387491bd3..febbfeb50 100644 --- a/src/video_core/shader/decode/predicate_set_register.cpp +++ b/src/video_core/shader/decode/predicate_set_register.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/register_set_predicate.cpp b/src/video_core/shader/decode/register_set_predicate.cpp index f8659e48e..e6c9d287e 100644 --- a/src/video_core/shader/decode/register_set_predicate.cpp +++ b/src/video_core/shader/decode/register_set_predicate.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/shift.cpp b/src/video_core/shader/decode/shift.cpp index 44ae87ece..2ac16eeb0 100644 --- a/src/video_core/shader/decode/shift.cpp +++ b/src/video_core/shader/decode/shift.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp index 5b033126d..4a356dbd4 100644 --- a/src/video_core/shader/decode/texture.cpp +++ b/src/video_core/shader/decode/texture.cpp @@ -11,6 +11,7 @@ #include "common/common_types.h" #include "common/logging/log.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { @@ -291,8 +292,8 @@ const Sampler& ShaderIR::GetBindlessSampler(const Tegra::Shader::Register& reg, const Node sampler_register = GetRegister(reg); const Node base_sampler = TrackCbuf(sampler_register, global_code, static_cast<s64>(global_code.size())); - const auto cbuf = std::get_if<CbufNode>(base_sampler); - const auto cbuf_offset_imm = std::get_if<ImmediateNode>(cbuf->GetOffset()); + const auto cbuf = std::get_if<CbufNode>(&*base_sampler); + const auto cbuf_offset_imm = std::get_if<ImmediateNode>(&*cbuf->GetOffset()); ASSERT(cbuf_offset_imm != nullptr); const auto cbuf_offset = cbuf_offset_imm->GetValue(); const auto cbuf_index = cbuf->GetIndex(); @@ -388,8 +389,8 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type, Node array, Node depth_compare, u32 bias_offset, std::vector<Node> aoffi, std::optional<Tegra::Shader::Register> bindless_reg) { - const bool is_array = array; - const bool is_shadow = depth_compare; + const auto is_array = static_cast<bool>(array); + const auto is_shadow = static_cast<bool>(depth_compare); const bool is_bindless = bindless_reg.has_value(); UNIMPLEMENTED_IF_MSG((texture_type == TextureType::Texture3D && (is_array || is_shadow)) || diff --git a/src/video_core/shader/decode/video.cpp b/src/video_core/shader/decode/video.cpp index cb9ab72b1..97fc6f9b1 100644 --- a/src/video_core/shader/decode/video.cpp +++ b/src/video_core/shader/decode/video.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/decode/xmad.cpp b/src/video_core/shader/decode/xmad.cpp index 04a776398..93dee77d1 100644 --- a/src/video_core/shader/decode/xmad.cpp +++ b/src/video_core/shader/decode/xmad.cpp @@ -5,6 +5,7 @@ #include "common/assert.h" #include "common/common_types.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h new file mode 100644 index 000000000..3cfb911bb --- /dev/null +++ b/src/video_core/shader/node.h @@ -0,0 +1,519 @@ +// Copyright 2019 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include <array> +#include <cstddef> +#include <memory> +#include <string> +#include <tuple> +#include <utility> +#include <variant> +#include <vector> + +#include "common/common_types.h" +#include "video_core/engines/shader_bytecode.h" + +namespace VideoCommon::Shader { + +enum class OperationCode { + Assign, /// (float& dest, float src) -> void + + Select, /// (MetaArithmetic, bool pred, float a, float b) -> float + + FAdd, /// (MetaArithmetic, float a, float b) -> float + FMul, /// (MetaArithmetic, float a, float b) -> float + FDiv, /// (MetaArithmetic, float a, float b) -> float + FFma, /// (MetaArithmetic, float a, float b, float c) -> float + FNegate, /// (MetaArithmetic, float a) -> float + FAbsolute, /// (MetaArithmetic, float a) -> float + FClamp, /// (MetaArithmetic, float value, float min, float max) -> float + FMin, /// (MetaArithmetic, float a, float b) -> float + FMax, /// (MetaArithmetic, float a, float b) -> float + FCos, /// (MetaArithmetic, float a) -> float + FSin, /// (MetaArithmetic, float a) -> float + FExp2, /// (MetaArithmetic, float a) -> float + FLog2, /// (MetaArithmetic, float a) -> float + FInverseSqrt, /// (MetaArithmetic, float a) -> float + FSqrt, /// (MetaArithmetic, float a) -> float + FRoundEven, /// (MetaArithmetic, float a) -> float + FFloor, /// (MetaArithmetic, float a) -> float + FCeil, /// (MetaArithmetic, float a) -> float + FTrunc, /// (MetaArithmetic, float a) -> float + FCastInteger, /// (MetaArithmetic, int a) -> float + FCastUInteger, /// (MetaArithmetic, uint a) -> float + + IAdd, /// (MetaArithmetic, int a, int b) -> int + IMul, /// (MetaArithmetic, int a, int b) -> int + IDiv, /// (MetaArithmetic, int a, int b) -> int + INegate, /// (MetaArithmetic, int a) -> int + IAbsolute, /// (MetaArithmetic, int a) -> int + IMin, /// (MetaArithmetic, int a, int b) -> int + IMax, /// (MetaArithmetic, int a, int b) -> int + ICastFloat, /// (MetaArithmetic, float a) -> int + ICastUnsigned, /// (MetaArithmetic, uint a) -> int + ILogicalShiftLeft, /// (MetaArithmetic, int a, uint b) -> int + ILogicalShiftRight, /// (MetaArithmetic, int a, uint b) -> int + IArithmeticShiftRight, /// (MetaArithmetic, int a, uint b) -> int + IBitwiseAnd, /// (MetaArithmetic, int a, int b) -> int + IBitwiseOr, /// (MetaArithmetic, int a, int b) -> int + IBitwiseXor, /// (MetaArithmetic, int a, int b) -> int + IBitwiseNot, /// (MetaArithmetic, int a) -> int + IBitfieldInsert, /// (MetaArithmetic, int base, int insert, int offset, int bits) -> int + IBitfieldExtract, /// (MetaArithmetic, int value, int offset, int offset) -> int + IBitCount, /// (MetaArithmetic, int) -> int + + UAdd, /// (MetaArithmetic, uint a, uint b) -> uint + UMul, /// (MetaArithmetic, uint a, uint b) -> uint + UDiv, /// (MetaArithmetic, uint a, uint b) -> uint + UMin, /// (MetaArithmetic, uint a, uint b) -> uint + UMax, /// (MetaArithmetic, uint a, uint b) -> uint + UCastFloat, /// (MetaArithmetic, float a) -> uint + UCastSigned, /// (MetaArithmetic, int a) -> uint + ULogicalShiftLeft, /// (MetaArithmetic, uint a, uint b) -> uint + ULogicalShiftRight, /// (MetaArithmetic, uint a, uint b) -> uint + UArithmeticShiftRight, /// (MetaArithmetic, uint a, uint b) -> uint + UBitwiseAnd, /// (MetaArithmetic, uint a, uint b) -> uint + UBitwiseOr, /// (MetaArithmetic, uint a, uint b) -> uint + UBitwiseXor, /// (MetaArithmetic, uint a, uint b) -> uint + UBitwiseNot, /// (MetaArithmetic, uint a) -> uint + UBitfieldInsert, /// (MetaArithmetic, uint base, uint insert, int offset, int bits) -> uint + UBitfieldExtract, /// (MetaArithmetic, uint value, int offset, int offset) -> uint + UBitCount, /// (MetaArithmetic, uint) -> uint + + HAdd, /// (MetaArithmetic, f16vec2 a, f16vec2 b) -> f16vec2 + HMul, /// (MetaArithmetic, f16vec2 a, f16vec2 b) -> f16vec2 + HFma, /// (MetaArithmetic, f16vec2 a, f16vec2 b, f16vec2 c) -> f16vec2 + HAbsolute, /// (f16vec2 a) -> f16vec2 + HNegate, /// (f16vec2 a, bool first, bool second) -> f16vec2 + HClamp, /// (f16vec2 src, float min, float max) -> f16vec2 + HUnpack, /// (Tegra::Shader::HalfType, T value) -> f16vec2 + HMergeF32, /// (f16vec2 src) -> float + HMergeH0, /// (f16vec2 dest, f16vec2 src) -> f16vec2 + HMergeH1, /// (f16vec2 dest, f16vec2 src) -> f16vec2 + HPack2, /// (float a, float b) -> f16vec2 + + LogicalAssign, /// (bool& dst, bool src) -> void + LogicalAnd, /// (bool a, bool b) -> bool + LogicalOr, /// (bool a, bool b) -> bool + LogicalXor, /// (bool a, bool b) -> bool + LogicalNegate, /// (bool a) -> bool + LogicalPick2, /// (bool2 pair, uint index) -> bool + LogicalAll2, /// (bool2 a) -> bool + LogicalAny2, /// (bool2 a) -> bool + + LogicalFLessThan, /// (float a, float b) -> bool + LogicalFEqual, /// (float a, float b) -> bool + LogicalFLessEqual, /// (float a, float b) -> bool + LogicalFGreaterThan, /// (float a, float b) -> bool + LogicalFNotEqual, /// (float a, float b) -> bool + LogicalFGreaterEqual, /// (float a, float b) -> bool + LogicalFIsNan, /// (float a) -> bool + + LogicalILessThan, /// (int a, int b) -> bool + LogicalIEqual, /// (int a, int b) -> bool + LogicalILessEqual, /// (int a, int b) -> bool + LogicalIGreaterThan, /// (int a, int b) -> bool + LogicalINotEqual, /// (int a, int b) -> bool + LogicalIGreaterEqual, /// (int a, int b) -> bool + + LogicalULessThan, /// (uint a, uint b) -> bool + LogicalUEqual, /// (uint a, uint b) -> bool + LogicalULessEqual, /// (uint a, uint b) -> bool + LogicalUGreaterThan, /// (uint a, uint b) -> bool + LogicalUNotEqual, /// (uint a, uint b) -> bool + LogicalUGreaterEqual, /// (uint a, uint b) -> bool + + Logical2HLessThan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HEqual, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HLessEqual, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HGreaterThan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HNotEqual, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HGreaterEqual, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HLessThanWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HEqualWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HLessEqualWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HGreaterThanWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HNotEqualWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + Logical2HGreaterEqualWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 + + Texture, /// (MetaTexture, float[N] coords) -> float4 + TextureLod, /// (MetaTexture, float[N] coords) -> float4 + TextureGather, /// (MetaTexture, float[N] coords) -> float4 + TextureQueryDimensions, /// (MetaTexture, float a) -> float4 + TextureQueryLod, /// (MetaTexture, float[N] coords) -> float4 + TexelFetch, /// (MetaTexture, int[N], int) -> float4 + + Branch, /// (uint branch_target) -> void + PushFlowStack, /// (uint branch_target) -> void + PopFlowStack, /// () -> void + Exit, /// () -> void + Discard, /// () -> void + + EmitVertex, /// () -> void + EndPrimitive, /// () -> void + + YNegate, /// () -> float + LocalInvocationIdX, /// () -> uint + LocalInvocationIdY, /// () -> uint + LocalInvocationIdZ, /// () -> uint + WorkGroupIdX, /// () -> uint + WorkGroupIdY, /// () -> uint + WorkGroupIdZ, /// () -> uint + + Amount, +}; + +enum class InternalFlag { + Zero = 0, + Sign = 1, + Carry = 2, + Overflow = 3, + Amount = 4, +}; + +enum class MetaStackClass { + Ssy, + Pbk, +}; + +class OperationNode; +class ConditionalNode; +class GprNode; +class ImmediateNode; +class InternalFlagNode; +class PredicateNode; +class AbufNode; +class CbufNode; +class LmemNode; +class GmemNode; +class CommentNode; + +using NodeData = + std::variant<OperationNode, ConditionalNode, GprNode, ImmediateNode, InternalFlagNode, + PredicateNode, AbufNode, CbufNode, LmemNode, GmemNode, CommentNode>; +using Node = std::shared_ptr<NodeData>; +using Node4 = std::array<Node, 4>; +using NodeBlock = std::vector<Node>; + +class Sampler { +public: + /// This constructor is for bound samplers + explicit Sampler(std::size_t offset, std::size_t index, Tegra::Shader::TextureType type, + bool is_array, bool is_shadow) + : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow}, + is_bindless{false} {} + + /// This constructor is for bindless samplers + explicit Sampler(u32 cbuf_index, u32 cbuf_offset, std::size_t index, + Tegra::Shader::TextureType type, bool is_array, bool is_shadow) + : offset{(static_cast<u64>(cbuf_index) << 32) | cbuf_offset}, index{index}, type{type}, + is_array{is_array}, is_shadow{is_shadow}, is_bindless{true} {} + + /// This constructor is for serialization/deserialization + explicit Sampler(std::size_t offset, std::size_t index, Tegra::Shader::TextureType type, + bool is_array, bool is_shadow, bool is_bindless) + : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow}, + is_bindless{is_bindless} {} + + std::size_t GetOffset() const { + return offset; + } + + std::size_t GetIndex() const { + return index; + } + + Tegra::Shader::TextureType GetType() const { + return type; + } + + bool IsArray() const { + return is_array; + } + + bool IsShadow() const { + return is_shadow; + } + + bool IsBindless() const { + return is_bindless; + } + + std::pair<u32, u32> GetBindlessCBuf() const { + return {static_cast<u32>(offset >> 32), static_cast<u32>(offset)}; + } + + bool operator<(const Sampler& rhs) const { + return std::tie(index, offset, type, is_array, is_shadow, is_bindless) < + std::tie(rhs.index, rhs.offset, rhs.type, rhs.is_array, rhs.is_shadow, + rhs.is_bindless); + } + +private: + /// Offset in TSC memory from which to read the sampler object, as specified by the sampling + /// instruction. + std::size_t offset{}; + std::size_t index{}; ///< Value used to index into the generated GLSL sampler array. + Tegra::Shader::TextureType type{}; ///< The type used to sample this texture (Texture2D, etc) + bool is_array{}; ///< Whether the texture is being sampled as an array texture or not. + bool is_shadow{}; ///< Whether the texture is being sampled as a depth texture or not. + bool is_bindless{}; ///< Whether this sampler belongs to a bindless texture or not. +}; + +struct GlobalMemoryBase { + u32 cbuf_index{}; + u32 cbuf_offset{}; + + bool operator<(const GlobalMemoryBase& rhs) const { + return std::tie(cbuf_index, cbuf_offset) < std::tie(rhs.cbuf_index, rhs.cbuf_offset); + } +}; + +/// Parameters describing an arithmetic operation +struct MetaArithmetic { + bool precise{}; ///< Whether the operation can be constraint or not +}; + +/// Parameters describing a texture sampler +struct MetaTexture { + const Sampler& sampler; + Node array; + Node depth_compare; + std::vector<Node> aoffi; + Node bias; + Node lod; + Node component{}; + u32 element{}; +}; + +/// Parameters that modify an operation but are not part of any particular operand +using Meta = std::variant<MetaArithmetic, MetaTexture, MetaStackClass, Tegra::Shader::HalfType>; + +/// Holds any kind of operation that can be done in the IR +class OperationNode final { +public: + explicit OperationNode(OperationCode code) : OperationNode(code, Meta{}) {} + + explicit OperationNode(OperationCode code, Meta meta) + : OperationNode(code, meta, std::vector<Node>{}) {} + + explicit OperationNode(OperationCode code, std::vector<Node> operands) + : OperationNode(code, Meta{}, std::move(operands)) {} + + explicit OperationNode(OperationCode code, Meta meta, std::vector<Node> operands) + : code{code}, meta{std::move(meta)}, operands{std::move(operands)} {} + + template <typename... Args> + explicit OperationNode(OperationCode code, Meta meta, Args&&... operands) + : code{code}, meta{std::move(meta)}, operands{operands...} {} + + OperationCode GetCode() const { + return code; + } + + const Meta& GetMeta() const { + return meta; + } + + std::size_t GetOperandsCount() const { + return operands.size(); + } + + const Node& operator[](std::size_t operand_index) const { + return operands.at(operand_index); + } + +private: + OperationCode code{}; + Meta meta{}; + std::vector<Node> operands; +}; + +/// Encloses inside any kind of node that returns a boolean conditionally-executed code +class ConditionalNode final { +public: + explicit ConditionalNode(Node condition, std::vector<Node>&& code) + : condition{std::move(condition)}, code{std::move(code)} {} + + const Node& GetCondition() const { + return condition; + } + + const std::vector<Node>& GetCode() const { + return code; + } + +private: + Node condition; ///< Condition to be satisfied + std::vector<Node> code; ///< Code to execute +}; + +/// A general purpose register +class GprNode final { +public: + explicit constexpr GprNode(Tegra::Shader::Register index) : index{index} {} + + u32 GetIndex() const { + return static_cast<u32>(index); + } + +private: + Tegra::Shader::Register index{}; +}; + +/// A 32-bits value that represents an immediate value +class ImmediateNode final { +public: + explicit constexpr ImmediateNode(u32 value) : value{value} {} + + u32 GetValue() const { + return value; + } + +private: + u32 value{}; +}; + +/// One of Maxwell's internal flags +class InternalFlagNode final { +public: + explicit constexpr InternalFlagNode(InternalFlag flag) : flag{flag} {} + + InternalFlag GetFlag() const { + return flag; + } + +private: + InternalFlag flag{}; +}; + +/// A predicate register, it can be negated without additional nodes +class PredicateNode final { +public: + explicit constexpr PredicateNode(Tegra::Shader::Pred index, bool negated) + : index{index}, negated{negated} {} + + Tegra::Shader::Pred GetIndex() const { + return index; + } + + bool IsNegated() const { + return negated; + } + +private: + Tegra::Shader::Pred index{}; + bool negated{}; +}; + +/// Attribute buffer memory (known as attributes or varyings in GLSL terms) +class AbufNode final { +public: + // Initialize for standard attributes (index is explicit). + explicit AbufNode(Tegra::Shader::Attribute::Index index, u32 element, Node buffer = {}) + : buffer{std::move(buffer)}, index{index}, element{element} {} + + // Initialize for physical attributes (index is a variable value). + explicit AbufNode(Node physical_address, Node buffer = {}) + : physical_address{std::move(physical_address)}, buffer{std::move(buffer)} {} + + Tegra::Shader::Attribute::Index GetIndex() const { + return index; + } + + u32 GetElement() const { + return element; + } + + const Node& GetBuffer() const { + return buffer; + } + + bool IsPhysicalBuffer() const { + return static_cast<bool>(physical_address); + } + + const Node& GetPhysicalAddress() const { + return physical_address; + } + +private: + Node physical_address; + Node buffer; + Tegra::Shader::Attribute::Index index{}; + u32 element{}; +}; + +/// Constant buffer node, usually mapped to uniform buffers in GLSL +class CbufNode final { +public: + explicit CbufNode(u32 index, Node offset) : index{index}, offset{std::move(offset)} {} + + u32 GetIndex() const { + return index; + } + + const Node& GetOffset() const { + return offset; + } + +private: + u32 index{}; + Node offset; +}; + +/// Local memory node +class LmemNode final { +public: + explicit LmemNode(Node address) : address{std::move(address)} {} + + const Node& GetAddress() const { + return address; + } + +private: + Node address; +}; + +/// Global memory node +class GmemNode final { +public: + explicit GmemNode(Node real_address, Node base_address, const GlobalMemoryBase& descriptor) + : real_address{std::move(real_address)}, base_address{std::move(base_address)}, + descriptor{descriptor} {} + + const Node& GetRealAddress() const { + return real_address; + } + + const Node& GetBaseAddress() const { + return base_address; + } + + const GlobalMemoryBase& GetDescriptor() const { + return descriptor; + } + +private: + Node real_address; + Node base_address; + GlobalMemoryBase descriptor; +}; + +/// Commentary, can be dropped +class CommentNode final { +public: + explicit CommentNode(std::string text) : text{std::move(text)} {} + + const std::string& GetText() const { + return text; + } + +private: + std::string text; +}; + +} // namespace VideoCommon::Shader diff --git a/src/video_core/shader/node_helper.cpp b/src/video_core/shader/node_helper.cpp new file mode 100644 index 000000000..6fccbbba3 --- /dev/null +++ b/src/video_core/shader/node_helper.cpp @@ -0,0 +1,99 @@ +// Copyright 2019 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include <cstring> +#include <vector> + +#include "common/common_types.h" +#include "video_core/shader/node_helper.h" +#include "video_core/shader/shader_ir.h" + +namespace VideoCommon::Shader { + +Node Conditional(Node condition, std::vector<Node> code) { + return MakeNode<ConditionalNode>(condition, std::move(code)); +} + +Node Comment(std::string text) { + return MakeNode<CommentNode>(std::move(text)); +} + +Node Immediate(u32 value) { + return MakeNode<ImmediateNode>(value); +} + +Node Immediate(s32 value) { + return Immediate(static_cast<u32>(value)); +} + +Node Immediate(f32 value) { + u32 integral; + std::memcpy(&integral, &value, sizeof(u32)); + return Immediate(integral); +} + +OperationCode SignedToUnsignedCode(OperationCode operation_code, bool is_signed) { + if (is_signed) { + return operation_code; + } + switch (operation_code) { + case OperationCode::FCastInteger: + return OperationCode::FCastUInteger; + case OperationCode::IAdd: + return OperationCode::UAdd; + case OperationCode::IMul: + return OperationCode::UMul; + case OperationCode::IDiv: + return OperationCode::UDiv; + case OperationCode::IMin: + return OperationCode::UMin; + case OperationCode::IMax: + return OperationCode::UMax; + case OperationCode::ICastFloat: + return OperationCode::UCastFloat; + case OperationCode::ICastUnsigned: + return OperationCode::UCastSigned; + case OperationCode::ILogicalShiftLeft: + return OperationCode::ULogicalShiftLeft; + case OperationCode::ILogicalShiftRight: + return OperationCode::ULogicalShiftRight; + case OperationCode::IArithmeticShiftRight: + return OperationCode::UArithmeticShiftRight; + case OperationCode::IBitwiseAnd: + return OperationCode::UBitwiseAnd; + case OperationCode::IBitwiseOr: + return OperationCode::UBitwiseOr; + case OperationCode::IBitwiseXor: + return OperationCode::UBitwiseXor; + case OperationCode::IBitwiseNot: + return OperationCode::UBitwiseNot; + case OperationCode::IBitfieldInsert: + return OperationCode::UBitfieldInsert; + case OperationCode::IBitCount: + return OperationCode::UBitCount; + case OperationCode::LogicalILessThan: + return OperationCode::LogicalULessThan; + case OperationCode::LogicalIEqual: + return OperationCode::LogicalUEqual; + case OperationCode::LogicalILessEqual: + return OperationCode::LogicalULessEqual; + case OperationCode::LogicalIGreaterThan: + return OperationCode::LogicalUGreaterThan; + case OperationCode::LogicalINotEqual: + return OperationCode::LogicalUNotEqual; + case OperationCode::LogicalIGreaterEqual: + return OperationCode::LogicalUGreaterEqual; + case OperationCode::INegate: + UNREACHABLE_MSG("Can't negate an unsigned integer"); + return {}; + case OperationCode::IAbsolute: + UNREACHABLE_MSG("Can't apply absolute to an unsigned integer"); + return {}; + default: + UNREACHABLE_MSG("Unknown signed operation with code={}", static_cast<u32>(operation_code)); + return {}; + } +} + +} // namespace VideoCommon::Shader diff --git a/src/video_core/shader/node_helper.h b/src/video_core/shader/node_helper.h new file mode 100644 index 000000000..0c2aa749b --- /dev/null +++ b/src/video_core/shader/node_helper.h @@ -0,0 +1,65 @@ +// Copyright 2019 yuzu Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include <memory> +#include <string> +#include <tuple> +#include <type_traits> +#include <utility> +#include <vector> + +#include "common/common_types.h" +#include "video_core/shader/node.h" + +namespace VideoCommon::Shader { + +/// This arithmetic operation cannot be constraint +inline constexpr MetaArithmetic PRECISE = {true}; +/// This arithmetic operation can be optimized away +inline constexpr MetaArithmetic NO_PRECISE = {false}; + +/// Creates a conditional node +Node Conditional(Node condition, std::vector<Node> code); + +/// Creates a commentary node +Node Comment(std::string text); + +/// Creates an u32 immediate +Node Immediate(u32 value); + +/// Creates a s32 immediate +Node Immediate(s32 value); + +/// Creates a f32 immediate +Node Immediate(f32 value); + +/// Converts an signed operation code to an unsigned operation code +OperationCode SignedToUnsignedCode(OperationCode operation_code, bool is_signed); + +template <typename T, typename... Args> +Node MakeNode(Args&&... args) { + static_assert(std::is_convertible_v<T, NodeData>); + return std::make_shared<NodeData>(T(std::forward<Args>(args)...)); +} + +template <typename... Args> +Node Operation(OperationCode code, Args&&... args) { + if constexpr (sizeof...(args) == 0) { + return MakeNode<OperationNode>(code); + } else if constexpr (std::is_convertible_v<std::tuple_element_t<0, std::tuple<Args...>>, + Meta>) { + return MakeNode<OperationNode>(code, std::forward<Args>(args)...); + } else { + return MakeNode<OperationNode>(code, Meta{}, std::forward<Args>(args)...); + } +} + +template <typename... Args> +Node SignedOperation(OperationCode code, bool is_signed, Args&&... args) { + return Operation(SignedToUnsignedCode(code, is_signed), std::forward<Args>(args)...); +} + +} // namespace VideoCommon::Shader diff --git a/src/video_core/shader/shader_ir.cpp b/src/video_core/shader/shader_ir.cpp index 8a6ee5cf5..11b545cca 100644 --- a/src/video_core/shader/shader_ir.cpp +++ b/src/video_core/shader/shader_ir.cpp @@ -9,6 +9,7 @@ #include "common/common_types.h" #include "common/logging/log.h" #include "video_core/engines/shader_bytecode.h" +#include "video_core/shader/node_helper.h" #include "video_core/shader/shader_ir.h" namespace VideoCommon::Shader { @@ -28,30 +29,11 @@ ShaderIR::ShaderIR(const ProgramCode& program_code, u32 main_offset) ShaderIR::~ShaderIR() = default; -Node ShaderIR::StoreNode(NodeData&& node_data) { - auto store = std::make_unique<NodeData>(node_data); - const Node node = store.get(); - stored_nodes.push_back(std::move(store)); - return node; -} - -Node ShaderIR::Conditional(Node condition, std::vector<Node>&& code) { - return StoreNode(ConditionalNode(condition, std::move(code))); -} - -Node ShaderIR::Comment(std::string text) { - return StoreNode(CommentNode(std::move(text))); -} - -Node ShaderIR::Immediate(u32 value) { - return StoreNode(ImmediateNode(value)); -} - Node ShaderIR::GetRegister(Register reg) { if (reg != Register::ZeroIndex) { used_registers.insert(static_cast<u32>(reg)); } - return StoreNode(GprNode(reg)); + return MakeNode<GprNode>(reg); } Node ShaderIR::GetImmediate19(Instruction instr) { @@ -69,7 +51,7 @@ Node ShaderIR::GetConstBuffer(u64 index_, u64 offset_) { const auto [entry, is_new] = used_cbufs.try_emplace(index); entry->second.MarkAsUsed(offset); - return StoreNode(CbufNode(index, Immediate(offset))); + return MakeNode<CbufNode>(index, Immediate(offset)); } Node ShaderIR::GetConstBufferIndirect(u64 index_, u64 offset_, Node node) { @@ -80,7 +62,7 @@ Node ShaderIR::GetConstBufferIndirect(u64 index_, u64 offset_, Node node) { entry->second.MarkAsUsedIndirect(); const Node final_offset = Operation(OperationCode::UAdd, NO_PRECISE, node, Immediate(offset)); - return StoreNode(CbufNode(index, final_offset)); + return MakeNode<CbufNode>(index, final_offset); } Node ShaderIR::GetPredicate(u64 pred_, bool negated) { @@ -89,7 +71,7 @@ Node ShaderIR::GetPredicate(u64 pred_, bool negated) { used_predicates.insert(pred); } - return StoreNode(PredicateNode(pred, negated)); + return MakeNode<PredicateNode>(pred, negated); } Node ShaderIR::GetPredicate(bool immediate) { @@ -98,12 +80,12 @@ Node ShaderIR::GetPredicate(bool immediate) { Node ShaderIR::GetInputAttribute(Attribute::Index index, u64 element, Node buffer) { used_input_attributes.emplace(index); - return StoreNode(AbufNode(index, static_cast<u32>(element), buffer)); + return MakeNode<AbufNode>(index, static_cast<u32>(element), buffer); } Node ShaderIR::GetPhysicalInputAttribute(Tegra::Shader::Register physical_address, Node buffer) { uses_physical_attributes = true; - return StoreNode(AbufNode(GetRegister(physical_address), buffer)); + return MakeNode<AbufNode>(GetRegister(physical_address), buffer); } Node ShaderIR::GetOutputAttribute(Attribute::Index index, u64 element, Node buffer) { @@ -115,11 +97,11 @@ Node ShaderIR::GetOutputAttribute(Attribute::Index index, u64 element, Node buff } used_output_attributes.insert(index); - return StoreNode(AbufNode(index, static_cast<u32>(element), buffer)); + return MakeNode<AbufNode>(index, static_cast<u32>(element), buffer); } Node ShaderIR::GetInternalFlag(InternalFlag flag, bool negated) { - const Node node = StoreNode(InternalFlagNode(flag)); + const Node node = MakeNode<InternalFlagNode>(flag); if (negated) { return Operation(OperationCode::LogicalNegate, node); } @@ -127,7 +109,7 @@ Node ShaderIR::GetInternalFlag(InternalFlag flag, bool negated) { } Node ShaderIR::GetLocalMemory(Node address) { - return StoreNode(LmemNode(address)); + return MakeNode<LmemNode>(address); } Node ShaderIR::GetTemporal(u32 id) { @@ -393,68 +375,4 @@ Node ShaderIR::BitfieldExtract(Node value, u32 offset, u32 bits) { Immediate(bits)); } -/*static*/ OperationCode ShaderIR::SignedToUnsignedCode(OperationCode operation_code, - bool is_signed) { - if (is_signed) { - return operation_code; - } - switch (operation_code) { - case OperationCode::FCastInteger: - return OperationCode::FCastUInteger; - case OperationCode::IAdd: - return OperationCode::UAdd; - case OperationCode::IMul: - return OperationCode::UMul; - case OperationCode::IDiv: - return OperationCode::UDiv; - case OperationCode::IMin: - return OperationCode::UMin; - case OperationCode::IMax: - return OperationCode::UMax; - case OperationCode::ICastFloat: - return OperationCode::UCastFloat; - case OperationCode::ICastUnsigned: - return OperationCode::UCastSigned; - case OperationCode::ILogicalShiftLeft: - return OperationCode::ULogicalShiftLeft; - case OperationCode::ILogicalShiftRight: - return OperationCode::ULogicalShiftRight; - case OperationCode::IArithmeticShiftRight: - return OperationCode::UArithmeticShiftRight; - case OperationCode::IBitwiseAnd: - return OperationCode::UBitwiseAnd; - case OperationCode::IBitwiseOr: - return OperationCode::UBitwiseOr; - case OperationCode::IBitwiseXor: - return OperationCode::UBitwiseXor; - case OperationCode::IBitwiseNot: - return OperationCode::UBitwiseNot; - case OperationCode::IBitfieldInsert: - return OperationCode::UBitfieldInsert; - case OperationCode::IBitCount: - return OperationCode::UBitCount; - case OperationCode::LogicalILessThan: - return OperationCode::LogicalULessThan; - case OperationCode::LogicalIEqual: - return OperationCode::LogicalUEqual; - case OperationCode::LogicalILessEqual: - return OperationCode::LogicalULessEqual; - case OperationCode::LogicalIGreaterThan: - return OperationCode::LogicalUGreaterThan; - case OperationCode::LogicalINotEqual: - return OperationCode::LogicalUNotEqual; - case OperationCode::LogicalIGreaterEqual: - return OperationCode::LogicalUGreaterEqual; - case OperationCode::INegate: - UNREACHABLE_MSG("Can't negate an unsigned integer"); - return {}; - case OperationCode::IAbsolute: - UNREACHABLE_MSG("Can't apply absolute to an unsigned integer"); - return {}; - default: - UNREACHABLE_MSG("Unknown signed operation with code={}", static_cast<u32>(operation_code)); - return {}; - } -} - } // namespace VideoCommon::Shader diff --git a/src/video_core/shader/shader_ir.h b/src/video_core/shader/shader_ir.h index ff7472e30..edcf2288e 100644 --- a/src/video_core/shader/shader_ir.h +++ b/src/video_core/shader/shader_ir.h @@ -18,188 +18,14 @@ #include "video_core/engines/maxwell_3d.h" #include "video_core/engines/shader_bytecode.h" #include "video_core/engines/shader_header.h" +#include "video_core/shader/node.h" namespace VideoCommon::Shader { -class OperationNode; -class ConditionalNode; -class GprNode; -class ImmediateNode; -class InternalFlagNode; -class PredicateNode; -class AbufNode; ///< Attribute buffer -class CbufNode; ///< Constant buffer -class LmemNode; ///< Local memory -class GmemNode; ///< Global memory -class CommentNode; - using ProgramCode = std::vector<u64>; -using NodeData = - std::variant<OperationNode, ConditionalNode, GprNode, ImmediateNode, InternalFlagNode, - PredicateNode, AbufNode, CbufNode, LmemNode, GmemNode, CommentNode>; -using Node = const NodeData*; -using Node4 = std::array<Node, 4>; -using NodeBlock = std::vector<Node>; - constexpr u32 MAX_PROGRAM_LENGTH = 0x1000; -enum class OperationCode { - Assign, /// (float& dest, float src) -> void - - Select, /// (MetaArithmetic, bool pred, float a, float b) -> float - - FAdd, /// (MetaArithmetic, float a, float b) -> float - FMul, /// (MetaArithmetic, float a, float b) -> float - FDiv, /// (MetaArithmetic, float a, float b) -> float - FFma, /// (MetaArithmetic, float a, float b, float c) -> float - FNegate, /// (MetaArithmetic, float a) -> float - FAbsolute, /// (MetaArithmetic, float a) -> float - FClamp, /// (MetaArithmetic, float value, float min, float max) -> float - FMin, /// (MetaArithmetic, float a, float b) -> float - FMax, /// (MetaArithmetic, float a, float b) -> float - FCos, /// (MetaArithmetic, float a) -> float - FSin, /// (MetaArithmetic, float a) -> float - FExp2, /// (MetaArithmetic, float a) -> float - FLog2, /// (MetaArithmetic, float a) -> float - FInverseSqrt, /// (MetaArithmetic, float a) -> float - FSqrt, /// (MetaArithmetic, float a) -> float - FRoundEven, /// (MetaArithmetic, float a) -> float - FFloor, /// (MetaArithmetic, float a) -> float - FCeil, /// (MetaArithmetic, float a) -> float - FTrunc, /// (MetaArithmetic, float a) -> float - FCastInteger, /// (MetaArithmetic, int a) -> float - FCastUInteger, /// (MetaArithmetic, uint a) -> float - - IAdd, /// (MetaArithmetic, int a, int b) -> int - IMul, /// (MetaArithmetic, int a, int b) -> int - IDiv, /// (MetaArithmetic, int a, int b) -> int - INegate, /// (MetaArithmetic, int a) -> int - IAbsolute, /// (MetaArithmetic, int a) -> int - IMin, /// (MetaArithmetic, int a, int b) -> int - IMax, /// (MetaArithmetic, int a, int b) -> int - ICastFloat, /// (MetaArithmetic, float a) -> int - ICastUnsigned, /// (MetaArithmetic, uint a) -> int - ILogicalShiftLeft, /// (MetaArithmetic, int a, uint b) -> int - ILogicalShiftRight, /// (MetaArithmetic, int a, uint b) -> int - IArithmeticShiftRight, /// (MetaArithmetic, int a, uint b) -> int - IBitwiseAnd, /// (MetaArithmetic, int a, int b) -> int - IBitwiseOr, /// (MetaArithmetic, int a, int b) -> int - IBitwiseXor, /// (MetaArithmetic, int a, int b) -> int - IBitwiseNot, /// (MetaArithmetic, int a) -> int - IBitfieldInsert, /// (MetaArithmetic, int base, int insert, int offset, int bits) -> int - IBitfieldExtract, /// (MetaArithmetic, int value, int offset, int offset) -> int - IBitCount, /// (MetaArithmetic, int) -> int - - UAdd, /// (MetaArithmetic, uint a, uint b) -> uint - UMul, /// (MetaArithmetic, uint a, uint b) -> uint - UDiv, /// (MetaArithmetic, uint a, uint b) -> uint - UMin, /// (MetaArithmetic, uint a, uint b) -> uint - UMax, /// (MetaArithmetic, uint a, uint b) -> uint - UCastFloat, /// (MetaArithmetic, float a) -> uint - UCastSigned, /// (MetaArithmetic, int a) -> uint - ULogicalShiftLeft, /// (MetaArithmetic, uint a, uint b) -> uint - ULogicalShiftRight, /// (MetaArithmetic, uint a, uint b) -> uint - UArithmeticShiftRight, /// (MetaArithmetic, uint a, uint b) -> uint - UBitwiseAnd, /// (MetaArithmetic, uint a, uint b) -> uint - UBitwiseOr, /// (MetaArithmetic, uint a, uint b) -> uint - UBitwiseXor, /// (MetaArithmetic, uint a, uint b) -> uint - UBitwiseNot, /// (MetaArithmetic, uint a) -> uint - UBitfieldInsert, /// (MetaArithmetic, uint base, uint insert, int offset, int bits) -> uint - UBitfieldExtract, /// (MetaArithmetic, uint value, int offset, int offset) -> uint - UBitCount, /// (MetaArithmetic, uint) -> uint - - HAdd, /// (MetaArithmetic, f16vec2 a, f16vec2 b) -> f16vec2 - HMul, /// (MetaArithmetic, f16vec2 a, f16vec2 b) -> f16vec2 - HFma, /// (MetaArithmetic, f16vec2 a, f16vec2 b, f16vec2 c) -> f16vec2 - HAbsolute, /// (f16vec2 a) -> f16vec2 - HNegate, /// (f16vec2 a, bool first, bool second) -> f16vec2 - HClamp, /// (f16vec2 src, float min, float max) -> f16vec2 - HUnpack, /// (Tegra::Shader::HalfType, T value) -> f16vec2 - HMergeF32, /// (f16vec2 src) -> float - HMergeH0, /// (f16vec2 dest, f16vec2 src) -> f16vec2 - HMergeH1, /// (f16vec2 dest, f16vec2 src) -> f16vec2 - HPack2, /// (float a, float b) -> f16vec2 - - LogicalAssign, /// (bool& dst, bool src) -> void - LogicalAnd, /// (bool a, bool b) -> bool - LogicalOr, /// (bool a, bool b) -> bool - LogicalXor, /// (bool a, bool b) -> bool - LogicalNegate, /// (bool a) -> bool - LogicalPick2, /// (bool2 pair, uint index) -> bool - LogicalAll2, /// (bool2 a) -> bool - LogicalAny2, /// (bool2 a) -> bool - - LogicalFLessThan, /// (float a, float b) -> bool - LogicalFEqual, /// (float a, float b) -> bool - LogicalFLessEqual, /// (float a, float b) -> bool - LogicalFGreaterThan, /// (float a, float b) -> bool - LogicalFNotEqual, /// (float a, float b) -> bool - LogicalFGreaterEqual, /// (float a, float b) -> bool - LogicalFIsNan, /// (float a) -> bool - - LogicalILessThan, /// (int a, int b) -> bool - LogicalIEqual, /// (int a, int b) -> bool - LogicalILessEqual, /// (int a, int b) -> bool - LogicalIGreaterThan, /// (int a, int b) -> bool - LogicalINotEqual, /// (int a, int b) -> bool - LogicalIGreaterEqual, /// (int a, int b) -> bool - - LogicalULessThan, /// (uint a, uint b) -> bool - LogicalUEqual, /// (uint a, uint b) -> bool - LogicalULessEqual, /// (uint a, uint b) -> bool - LogicalUGreaterThan, /// (uint a, uint b) -> bool - LogicalUNotEqual, /// (uint a, uint b) -> bool - LogicalUGreaterEqual, /// (uint a, uint b) -> bool - - Logical2HLessThan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HEqual, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HLessEqual, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HGreaterThan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HNotEqual, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HGreaterEqual, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HLessThanWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HEqualWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HLessEqualWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HGreaterThanWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HNotEqualWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - Logical2HGreaterEqualWithNan, /// (MetaHalfArithmetic, f16vec2 a, f16vec2) -> bool2 - - Texture, /// (MetaTexture, float[N] coords) -> float4 - TextureLod, /// (MetaTexture, float[N] coords) -> float4 - TextureGather, /// (MetaTexture, float[N] coords) -> float4 - TextureQueryDimensions, /// (MetaTexture, float a) -> float4 - TextureQueryLod, /// (MetaTexture, float[N] coords) -> float4 - TexelFetch, /// (MetaTexture, int[N], int) -> float4 - - Branch, /// (uint branch_target) -> void - PushFlowStack, /// (uint branch_target) -> void - PopFlowStack, /// () -> void - Exit, /// () -> void - Discard, /// () -> void - - EmitVertex, /// () -> void - EndPrimitive, /// () -> void - - YNegate, /// () -> float - LocalInvocationIdX, /// () -> uint - LocalInvocationIdY, /// () -> uint - LocalInvocationIdZ, /// () -> uint - WorkGroupIdX, /// () -> uint - WorkGroupIdY, /// () -> uint - WorkGroupIdZ, /// () -> uint - - Amount, -}; - -enum class InternalFlag { - Zero = 0, - Sign = 1, - Carry = 2, - Overflow = 3, - Amount = 4, -}; - /// Describes the behaviour of code path of a given entry point and a return point. enum class ExitMethod { Undetermined, ///< Internal value. Only occur when analyzing JMP loop. @@ -208,71 +34,6 @@ enum class ExitMethod { AlwaysEnd, ///< All code paths reach a END instruction. }; -class Sampler { -public: - // Use this constructor for bounded Samplers - explicit Sampler(std::size_t offset, std::size_t index, Tegra::Shader::TextureType type, - bool is_array, bool is_shadow) - : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow}, - is_bindless{false} {} - - // Use this constructor for bindless Samplers - explicit Sampler(u32 cbuf_index, u32 cbuf_offset, std::size_t index, - Tegra::Shader::TextureType type, bool is_array, bool is_shadow) - : offset{(static_cast<u64>(cbuf_index) << 32) | cbuf_offset}, index{index}, type{type}, - is_array{is_array}, is_shadow{is_shadow}, is_bindless{true} {} - - // Use this only for serialization/deserialization - explicit Sampler(std::size_t offset, std::size_t index, Tegra::Shader::TextureType type, - bool is_array, bool is_shadow, bool is_bindless) - : offset{offset}, index{index}, type{type}, is_array{is_array}, is_shadow{is_shadow}, - is_bindless{is_bindless} {} - - std::size_t GetOffset() const { - return offset; - } - - std::size_t GetIndex() const { - return index; - } - - Tegra::Shader::TextureType GetType() const { - return type; - } - - bool IsArray() const { - return is_array; - } - - bool IsShadow() const { - return is_shadow; - } - - bool IsBindless() const { - return is_bindless; - } - - std::pair<u32, u32> GetBindlessCBuf() const { - return {static_cast<u32>(offset >> 32), static_cast<u32>(offset)}; - } - - bool operator<(const Sampler& rhs) const { - return std::tie(index, offset, type, is_array, is_shadow, is_bindless) < - std::tie(rhs.index, rhs.offset, rhs.type, rhs.is_array, rhs.is_shadow, - rhs.is_bindless); - } - -private: - /// Offset in TSC memory from which to read the sampler object, as specified by the sampling - /// instruction. - std::size_t offset{}; - std::size_t index{}; ///< Value used to index into the generated GLSL sampler array. - Tegra::Shader::TextureType type{}; ///< The type used to sample this texture (Texture2D, etc) - bool is_array{}; ///< Whether the texture is being sampled as an array texture or not. - bool is_shadow{}; ///< Whether the texture is being sampled as a depth texture or not. - bool is_bindless{}; ///< Whether this sampler belongs to a bindless texture or not. -}; - class ConstBuffer { public: explicit ConstBuffer(u32 max_offset, bool is_indirect) @@ -305,268 +66,11 @@ private: bool is_indirect{}; }; -struct GlobalMemoryBase { - u32 cbuf_index{}; - u32 cbuf_offset{}; - - bool operator<(const GlobalMemoryBase& rhs) const { - return std::tie(cbuf_index, cbuf_offset) < std::tie(rhs.cbuf_index, rhs.cbuf_offset); - } -}; - struct GlobalMemoryUsage { bool is_read{}; bool is_written{}; }; -struct MetaArithmetic { - bool precise{}; -}; - -struct MetaTexture { - const Sampler& sampler; - Node array{}; - Node depth_compare{}; - std::vector<Node> aoffi; - Node bias{}; - Node lod{}; - Node component{}; - u32 element{}; -}; - -constexpr MetaArithmetic PRECISE = {true}; -constexpr MetaArithmetic NO_PRECISE = {false}; - -using Meta = std::variant<MetaArithmetic, MetaTexture, Tegra::Shader::HalfType>; - -/// Holds any kind of operation that can be done in the IR -class OperationNode final { -public: - explicit OperationNode(OperationCode code) : code{code} {} - - explicit OperationNode(OperationCode code, Meta&& meta) : code{code}, meta{std::move(meta)} {} - - template <typename... T> - explicit OperationNode(OperationCode code, const T*... operands) - : OperationNode(code, {}, operands...) {} - - template <typename... T> - explicit OperationNode(OperationCode code, Meta&& meta, const T*... operands_) - : code{code}, meta{std::move(meta)}, operands{operands_...} {} - - explicit OperationNode(OperationCode code, Meta&& meta, std::vector<Node>&& operands) - : code{code}, meta{meta}, operands{std::move(operands)} {} - - explicit OperationNode(OperationCode code, std::vector<Node>&& operands) - : code{code}, operands{std::move(operands)} {} - - OperationCode GetCode() const { - return code; - } - - const Meta& GetMeta() const { - return meta; - } - - std::size_t GetOperandsCount() const { - return operands.size(); - } - - Node operator[](std::size_t operand_index) const { - return operands.at(operand_index); - } - -private: - const OperationCode code; - const Meta meta; - std::vector<Node> operands; -}; - -/// Encloses inside any kind of node that returns a boolean conditionally-executed code -class ConditionalNode final { -public: - explicit ConditionalNode(Node condition, std::vector<Node>&& code) - : condition{condition}, code{std::move(code)} {} - - Node GetCondition() const { - return condition; - } - - const std::vector<Node>& GetCode() const { - return code; - } - -private: - const Node condition; ///< Condition to be satisfied - std::vector<Node> code; ///< Code to execute -}; - -/// A general purpose register -class GprNode final { -public: - explicit constexpr GprNode(Tegra::Shader::Register index) : index{index} {} - - u32 GetIndex() const { - return static_cast<u32>(index); - } - -private: - const Tegra::Shader::Register index; -}; - -/// A 32-bits value that represents an immediate value -class ImmediateNode final { -public: - explicit constexpr ImmediateNode(u32 value) : value{value} {} - - u32 GetValue() const { - return value; - } - -private: - const u32 value; -}; - -/// One of Maxwell's internal flags -class InternalFlagNode final { -public: - explicit constexpr InternalFlagNode(InternalFlag flag) : flag{flag} {} - - InternalFlag GetFlag() const { - return flag; - } - -private: - const InternalFlag flag; -}; - -/// A predicate register, it can be negated without additional nodes -class PredicateNode final { -public: - explicit constexpr PredicateNode(Tegra::Shader::Pred index, bool negated) - : index{index}, negated{negated} {} - - Tegra::Shader::Pred GetIndex() const { - return index; - } - - bool IsNegated() const { - return negated; - } - -private: - const Tegra::Shader::Pred index; - const bool negated; -}; - -/// Attribute buffer memory (known as attributes or varyings in GLSL terms) -class AbufNode final { -public: - // Initialize for standard attributes (index is explicit). - explicit constexpr AbufNode(Tegra::Shader::Attribute::Index index, u32 element, - Node buffer = {}) - : buffer{buffer}, index{index}, element{element} {} - - // Initialize for physical attributes (index is a variable value). - explicit constexpr AbufNode(Node physical_address, Node buffer = {}) - : physical_address{physical_address}, buffer{buffer} {} - - Tegra::Shader::Attribute::Index GetIndex() const { - return index; - } - - u32 GetElement() const { - return element; - } - - Node GetBuffer() const { - return buffer; - } - - bool IsPhysicalBuffer() const { - return physical_address != nullptr; - } - - Node GetPhysicalAddress() const { - return physical_address; - } - -private: - Node physical_address{}; - Node buffer{}; - Tegra::Shader::Attribute::Index index{}; - u32 element{}; -}; - -/// Constant buffer node, usually mapped to uniform buffers in GLSL -class CbufNode final { -public: - explicit constexpr CbufNode(u32 index, Node offset) : index{index}, offset{offset} {} - - u32 GetIndex() const { - return index; - } - - Node GetOffset() const { - return offset; - } - -private: - const u32 index; - const Node offset; -}; - -/// Local memory node -class LmemNode final { -public: - explicit constexpr LmemNode(Node address) : address{address} {} - - Node GetAddress() const { - return address; - } - -private: - const Node address; -}; - -/// Global memory node -class GmemNode final { -public: - explicit constexpr GmemNode(Node real_address, Node base_address, - const GlobalMemoryBase& descriptor) - : real_address{real_address}, base_address{base_address}, descriptor{descriptor} {} - - Node GetRealAddress() const { - return real_address; - } - - Node GetBaseAddress() const { - return base_address; - } - - const GlobalMemoryBase& GetDescriptor() const { - return descriptor; - } - -private: - const Node real_address; - const Node base_address; - const GlobalMemoryBase descriptor; -}; - -/// Commentary, can be dropped -class CommentNode final { -public: - explicit CommentNode(std::string text) : text{std::move(text)} {} - - const std::string& GetText() const { - return text; - } - -private: - std::string text; -}; - class ShaderIR final { public: explicit ShaderIR(const ProgramCode& program_code, u32 main_offset); @@ -663,26 +167,6 @@ private: u32 DecodeXmad(NodeBlock& bb, u32 pc); u32 DecodeOther(NodeBlock& bb, u32 pc); - /// Internalizes node's data and returns a managed pointer to a clone of that node - Node StoreNode(NodeData&& node_data); - - /// Creates a conditional node - Node Conditional(Node condition, std::vector<Node>&& code); - /// Creates a commentary - Node Comment(std::string text); - /// Creates an u32 immediate - Node Immediate(u32 value); - /// Creates a s32 immediate - Node Immediate(s32 value) { - return Immediate(static_cast<u32>(value)); - } - /// Creates a f32 immediate - Node Immediate(f32 value) { - u32 integral; - std::memcpy(&integral, &value, sizeof(u32)); - return Immediate(integral); - } - /// Generates a node for a passed register. Node GetRegister(Tegra::Shader::Register reg); /// Generates a node representing a 19-bit immediate value @@ -827,37 +311,6 @@ private: std::tuple<Node, Node, GlobalMemoryBase> TrackAndGetGlobalMemory( NodeBlock& bb, Tegra::Shader::Instruction instr, bool is_write); - template <typename... T> - Node Operation(OperationCode code, const T*... operands) { - return StoreNode(OperationNode(code, operands...)); - } - - template <typename... T> - Node Operation(OperationCode code, Meta&& meta, const T*... operands) { - return StoreNode(OperationNode(code, std::move(meta), operands...)); - } - - Node Operation(OperationCode code, std::vector<Node>&& operands) { - return StoreNode(OperationNode(code, std::move(operands))); - } - - Node Operation(OperationCode code, Meta&& meta, std::vector<Node>&& operands) { - return StoreNode(OperationNode(code, std::move(meta), std::move(operands))); - } - - template <typename... T> - Node SignedOperation(OperationCode code, bool is_signed, const T*... operands) { - return StoreNode(OperationNode(SignedToUnsignedCode(code, is_signed), operands...)); - } - - template <typename... T> - Node SignedOperation(OperationCode code, bool is_signed, Meta&& meta, const T*... operands) { - return StoreNode( - OperationNode(SignedToUnsignedCode(code, is_signed), std::move(meta), operands...)); - } - - static OperationCode SignedToUnsignedCode(OperationCode operation_code, bool is_signed); - const ProgramCode& program_code; const u32 main_offset; @@ -868,8 +321,6 @@ private: std::map<u32, NodeBlock> basic_blocks; NodeBlock global_code; - std::vector<std::unique_ptr<NodeData>> stored_nodes; - std::set<u32> used_registers; std::set<Tegra::Shader::Pred> used_predicates; std::set<Tegra::Shader::Attribute::Index> used_input_attributes; diff --git a/src/video_core/shader/track.cpp b/src/video_core/shader/track.cpp index 19ede1eb9..fc957d980 100644 --- a/src/video_core/shader/track.cpp +++ b/src/video_core/shader/track.cpp @@ -16,12 +16,12 @@ std::pair<Node, s64> FindOperation(const NodeBlock& code, s64 cursor, OperationCode operation_code) { for (; cursor >= 0; --cursor) { const Node node = code.at(cursor); - if (const auto operation = std::get_if<OperationNode>(node)) { + if (const auto operation = std::get_if<OperationNode>(&*node)) { if (operation->GetCode() == operation_code) { return {node, cursor}; } } - if (const auto conditional = std::get_if<ConditionalNode>(node)) { + if (const auto conditional = std::get_if<ConditionalNode>(&*node)) { const auto& conditional_code = conditional->GetCode(); const auto [found, internal_cursor] = FindOperation( conditional_code, static_cast<s64>(conditional_code.size() - 1), operation_code); @@ -35,11 +35,11 @@ std::pair<Node, s64> FindOperation(const NodeBlock& code, s64 cursor, } // namespace Node ShaderIR::TrackCbuf(Node tracked, const NodeBlock& code, s64 cursor) const { - if (const auto cbuf = std::get_if<CbufNode>(tracked)) { + if (const auto cbuf = std::get_if<CbufNode>(&*tracked)) { // Cbuf found, but it has to be immediate return std::holds_alternative<ImmediateNode>(*cbuf->GetOffset()) ? tracked : nullptr; } - if (const auto gpr = std::get_if<GprNode>(tracked)) { + if (const auto gpr = std::get_if<GprNode>(&*tracked)) { if (gpr->GetIndex() == Tegra::Shader::Register::ZeroIndex) { return nullptr; } @@ -51,7 +51,7 @@ Node ShaderIR::TrackCbuf(Node tracked, const NodeBlock& code, s64 cursor) const } return TrackCbuf(source, code, new_cursor); } - if (const auto operation = std::get_if<OperationNode>(tracked)) { + if (const auto operation = std::get_if<OperationNode>(&*tracked)) { for (std::size_t i = 0; i < operation->GetOperandsCount(); ++i) { if (const auto found = TrackCbuf((*operation)[i], code, cursor)) { // Cbuf found in operand @@ -60,7 +60,7 @@ Node ShaderIR::TrackCbuf(Node tracked, const NodeBlock& code, s64 cursor) const } return nullptr; } - if (const auto conditional = std::get_if<ConditionalNode>(tracked)) { + if (const auto conditional = std::get_if<ConditionalNode>(&*tracked)) { const auto& conditional_code = conditional->GetCode(); return TrackCbuf(tracked, conditional_code, static_cast<s64>(conditional_code.size())); } @@ -75,7 +75,7 @@ std::optional<u32> ShaderIR::TrackImmediate(Node tracked, const NodeBlock& code, if (!found) { return {}; } - if (const auto immediate = std::get_if<ImmediateNode>(found)) { + if (const auto immediate = std::get_if<ImmediateNode>(&*found)) { return immediate->GetValue(); } return {}; @@ -88,11 +88,11 @@ std::pair<Node, s64> ShaderIR::TrackRegister(const GprNode* tracked, const NodeB if (!found_node) { return {}; } - const auto operation = std::get_if<OperationNode>(found_node); + const auto operation = std::get_if<OperationNode>(&*found_node); ASSERT(operation); const auto& target = (*operation)[0]; - if (const auto gpr_target = std::get_if<GprNode>(target)) { + if (const auto gpr_target = std::get_if<GprNode>(&*target)) { if (gpr_target->GetIndex() == tracked->GetIndex()) { return {(*operation)[1], new_cursor}; } |