summaryrefslogtreecommitdiffstats
path: root/src
diff options
context:
space:
mode:
authorReinUsesLisp <reinuseslisp@airmail.cc>2018-12-19 04:31:58 +0100
committerReinUsesLisp <reinuseslisp@airmail.cc>2019-01-15 21:54:52 +0100
commit4fd06efeb94ff5fc5af4c5e4b9e8a4fa95d3b383 (patch)
tree85161efb42877cfa211d6c19b3aad3d0dd7af1cc /src
parentshader_decode: Implement LOP3 (diff)
downloadyuzu-4fd06efeb94ff5fc5af4c5e4b9e8a4fa95d3b383.tar
yuzu-4fd06efeb94ff5fc5af4c5e4b9e8a4fa95d3b383.tar.gz
yuzu-4fd06efeb94ff5fc5af4c5e4b9e8a4fa95d3b383.tar.bz2
yuzu-4fd06efeb94ff5fc5af4c5e4b9e8a4fa95d3b383.tar.lz
yuzu-4fd06efeb94ff5fc5af4c5e4b9e8a4fa95d3b383.tar.xz
yuzu-4fd06efeb94ff5fc5af4c5e4b9e8a4fa95d3b383.tar.zst
yuzu-4fd06efeb94ff5fc5af4c5e4b9e8a4fa95d3b383.zip
Diffstat (limited to 'src')
-rw-r--r--src/video_core/shader/decode/arithmetic_integer.cpp61
1 files changed, 61 insertions, 0 deletions
diff --git a/src/video_core/shader/decode/arithmetic_integer.cpp b/src/video_core/shader/decode/arithmetic_integer.cpp
index 145bbcfc8..3b9b9d6d9 100644
--- a/src/video_core/shader/decode/arithmetic_integer.cpp
+++ b/src/video_core/shader/decode/arithmetic_integer.cpp
@@ -9,6 +9,7 @@
namespace VideoCommon::Shader {
+using Tegra::Shader::IAdd3Height;
using Tegra::Shader::Instruction;
using Tegra::Shader::OpCode;
using Tegra::Shader::Register;
@@ -42,6 +43,66 @@ u32 ShaderIR::DecodeArithmeticInteger(BasicBlock& bb, u32 pc) {
SetRegister(bb, instr.gpr0, Operation(OperationCode::IAdd, PRECISE, op_a, op_b));
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");
+
+ Node op_c = GetRegister(instr.gpr39);
+
+ const auto ApplyHeight = [&](IAdd3Height height, Node value) {
+ switch (height) {
+ case IAdd3Height::None:
+ return value;
+ case IAdd3Height::LowerHalfWord:
+ return Operation(OperationCode::IBitwiseAnd, NO_PRECISE, value, Immediate(0xffff));
+ case IAdd3Height::UpperHalfWord:
+ return Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, value,
+ Immediate(16));
+ default:
+ UNIMPLEMENTED_MSG("Unhandled IADD3 height: {}", static_cast<u32>(height));
+ return Immediate(0);
+ }
+ };
+
+ if (opcode->get().GetId() == OpCode::Id::IADD3_R) {
+ op_a = ApplyHeight(instr.iadd3.height_a, op_a);
+ op_b = ApplyHeight(instr.iadd3.height_b, op_b);
+ op_c = ApplyHeight(instr.iadd3.height_c, op_c);
+ }
+
+ op_a = GetOperandAbsNegInteger(op_a, false, instr.iadd3.neg_a, true);
+ op_b = GetOperandAbsNegInteger(op_b, false, instr.iadd3.neg_b, true);
+ op_c = GetOperandAbsNegInteger(op_c, false, instr.iadd3.neg_c, true);
+
+ const Node value = [&]() {
+ const Node add_ab = Operation(OperationCode::IAdd, NO_PRECISE, op_a, op_b);
+ if (opcode->get().GetId() != OpCode::Id::IADD3_R) {
+ return Operation(OperationCode::IAdd, NO_PRECISE, add_ab, op_c);
+ }
+ const Node shifted = [&]() {
+ switch (instr.iadd3.mode) {
+ case Tegra::Shader::IAdd3Mode::RightShift:
+ // TODO(tech4me): According to
+ // https://envytools.readthedocs.io/en/latest/hw/graph/maxwell/cuda/int.html?highlight=iadd3
+ // The addition between op_a and op_b should be done in uint33, more
+ // investigation required
+ return Operation(OperationCode::ILogicalShiftRight, NO_PRECISE, add_ab,
+ Immediate(16));
+ case Tegra::Shader::IAdd3Mode::LeftShift:
+ return Operation(OperationCode::ILogicalShiftLeft, NO_PRECISE, add_ab,
+ Immediate(16));
+ default:
+ return add_ab;
+ }
+ }();
+ return Operation(OperationCode::IAdd, NO_PRECISE, shifted, op_c);
+ }();
+
+ SetRegister(bb, instr.gpr0, value);
+ break;
+ }
case OpCode::Id::ISCADD_C:
case OpCode::Id::ISCADD_R:
case OpCode::Id::ISCADD_IMM: {