N-archive/suyu
1
0
Fork 0
mirror of https://git.suyu.dev/suyu/suyu.git synced 2024-11-27 09:12:46 +01:00
Code Issues Projects Releases Packages Wiki Activity Actions

shader_decode: Implement LOP3

Browse source
This commit is contained in:
ReinUsesLisp 2018-12-18 22:18:44 -03:00
parent b184ca9089
commit a40fd07516
2 changed files with 62 additions and 0 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

60
src/video_core/shader/decode/arithmetic_integer.cpp
View file

@ -11,6 +11,7 @@ namespace VideoCommon::Shader {
using Tegra::Shader::Instruction;
using Tegra::Shader::OpCode;
using Tegra::Shader::Register;
u32 ShaderIR::DecodeArithmeticInteger(BasicBlock& bb, u32 pc) {
const Instruction instr = {program_code[pc]};
@ -79,6 +80,24 @@ u32 ShaderIR::DecodeArithmeticInteger(BasicBlock& bb, u32 pc) {
instr.alu.lop.pred_result_mode, instr.alu.lop.pred48);
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 Node op_c = GetRegister(instr.gpr39);
const Node lut = [&]() {
if (opcode->get().GetId() == OpCode::Id::LOP3_R) {
return Immediate(instr.alu.lop3.GetImmLut28());
} else {
return Immediate(instr.alu.lop3.GetImmLut48());
}
}();
WriteLop3Instruction(bb, instr.gpr0, op_a, op_b, op_c, lut);
break;
}
case OpCode::Id::IMNMX_C:
case OpCode::Id::IMNMX_R:
case OpCode::Id::IMNMX_IMM: {
@ -102,4 +121,45 @@ u32 ShaderIR::DecodeArithmeticInteger(BasicBlock& bb, u32 pc) {
return pc;
}
void ShaderIR::WriteLop3Instruction(BasicBlock& bb, Register dest, Node op_a, Node op_b, Node op_c,
Node imm_lut) {
constexpr u32 lop_iterations = 32;
const Node one = Immediate(1);
const Node two = Immediate(2);
Node value{};
for (u32 i = 0; i < lop_iterations; ++i) {
const Node shift_amount = Immediate(i);
const Node a = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_c, shift_amount);
const Node pack_0 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, a, one);
const Node b = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_b, shift_amount);
const Node c = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, b, one);
const Node pack_1 = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, c, one);
const Node d = Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, op_a, shift_amount);
const Node e = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, d, one);
const Node pack_2 = Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, e, two);
const Node pack_01 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, pack_0, pack_1);
const Node pack_012 = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, pack_01, pack_2);
const Node shifted_bit =
Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, imm_lut, pack_012);
const Node bit = Operation(OperationCode::IBitwiseAnd, NO_PRECISE, shifted_bit, one);
const Node right =
Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, bit, shift_amount);
if (i > 0) {
value = Operation(OperationCode::IBitwiseOr, NO_PRECISE, value, right);
} else {
value = right;
}
}
SetRegister(bb, dest, value);
}
} // namespace VideoCommon::Shader

2
src/video_core/shader/shader_ir.h
View file

@ -701,6 +701,8 @@ private:
Tegra::Shader::LogicOperation logic_op, Node op_a, Node op_b,
Tegra::Shader::PredicateResultMode predicate_mode,
Tegra::Shader::Pred predicate);
void WriteLop3Instruction(BasicBlock& bb, Tegra::Shader::Register dest, Node op_a, Node op_b,
Node op_c, Node imm_lut);
template <typename... T>
inline Node Operation(OperationCode code, const T*... operands) {