12 files changed, 111 insertions, 45 deletions

diff --git a/src/shader_recompiler/frontend/ir/ir_emitter.cpp b/src/shader_recompiler/frontend/ir/ir_emitter.cpp
index 8f120a2f6..34c2f67fb 100644
--- a/src/shader_recompiler/frontend/ir/ir_emitter.cpp
+++ b/src/shader_recompiler/frontend/ir/ir_emitter.cpp
@@ -468,11 +468,11 @@ F16F32F64 IREmitter::FPFma(const F16F32F64& a, const F16F32F64& b, const F16F32F
 
 F16F32F64 IREmitter::FPAbs(const F16F32F64& value) {
     switch (value.Type()) {
-    case Type::U16:
+    case Type::F16:
         return Inst<F16>(Opcode::FPAbs16, value);
-    case Type::U32:
+    case Type::F32:
         return Inst<F32>(Opcode::FPAbs32, value);
-    case Type::U64:
+    case Type::F64:
         return Inst<F64>(Opcode::FPAbs64, value);
     default:
         ThrowInvalidType(value.Type());
@@ -481,11 +481,11 @@ F16F32F64 IREmitter::FPAbs(const F16F32F64& value) {
 
 F16F32F64 IREmitter::FPNeg(const F16F32F64& value) {
     switch (value.Type()) {
-    case Type::U16:
+    case Type::F16:
         return Inst<F16>(Opcode::FPNeg16, value);
-    case Type::U32:
+    case Type::F32:
         return Inst<F32>(Opcode::FPNeg32, value);
-    case Type::U64:
+    case Type::F64:
         return Inst<F64>(Opcode::FPNeg64, value);
     default:
         ThrowInvalidType(value.Type());
@@ -495,10 +495,10 @@ F16F32F64 IREmitter::FPNeg(const F16F32F64& value) {
 F16F32F64 IREmitter::FPAbsNeg(const F16F32F64& value, bool abs, bool neg) {
     F16F32F64 result{value};
     if (abs) {
-        result = FPAbs(value);
+        result = FPAbs(result);
     }
     if (neg) {
-        result = FPNeg(value);
+        result = FPNeg(result);
     }
     return result;
 }
diff --git a/src/shader_recompiler/frontend/ir/pred.h b/src/shader_recompiler/frontend/ir/pred.h
index c6f2f82bf..4e7f32423 100644
--- a/src/shader_recompiler/frontend/ir/pred.h
+++ b/src/shader_recompiler/frontend/ir/pred.h
@@ -19,8 +19,8 @@ enum class Pred : u64 {
     PT,
 };
 
-constexpr size_t NUM_USER_PREDS = 6;
-constexpr size_t NUM_PREDS = 7;
+constexpr size_t NUM_USER_PREDS = 7;
+constexpr size_t NUM_PREDS = 8;
 
 [[nodiscard]] constexpr size_t PredIndex(Pred pred) noexcept {
     return static_cast<size_t>(pred);
diff --git a/src/shader_recompiler/frontend/maxwell/program.cpp b/src/shader_recompiler/frontend/maxwell/program.cpp
index 16cdc12e2..ed5dbf41f 100644
--- a/src/shader_recompiler/frontend/maxwell/program.cpp
+++ b/src/shader_recompiler/frontend/maxwell/program.cpp
@@ -56,12 +56,12 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
             .post_order_blocks{},
         });
     }
+    fmt::print(stdout, "{}\n", IR::DumpProgram(program));
     Optimization::LowerFp16ToFp32(program);
     for (IR::Function& function : functions) {
         function.post_order_blocks = PostOrder(function.blocks);
         Optimization::SsaRewritePass(function.post_order_blocks);
     }
-    fmt::print(stdout, "{}\n", IR::DumpProgram(program));
     Optimization::GlobalMemoryToStorageBufferPass(program);
     for (IR::Function& function : functions) {
         Optimization::PostOrderInvoke(Optimization::ConstantPropagationPass, function);
diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_add.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_add.cpp
index cb3a326cf..219ffcc6a 100644
--- a/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_add.cpp
+++ b/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_add.cpp
@@ -9,7 +9,6 @@
 
 namespace Shader::Maxwell {
 namespace {
-
 void FADD(TranslatorVisitor& v, u64 insn, bool sat, bool cc, bool ftz, FpRounding fp_rounding,
           const IR::F32& src_b, bool abs_a, bool neg_a, bool abs_b, bool neg_b) {
     union {
@@ -18,9 +17,6 @@ void FADD(TranslatorVisitor& v, u64 insn, bool sat, bool cc, bool ftz, FpRoundin
         BitField<8, 8, IR::Reg> src_a;
     } const fadd{insn};
 
-    if (sat) {
-        throw NotImplementedException("FADD SAT");
-    }
     if (cc) {
         throw NotImplementedException("FADD CC");
     }
@@ -31,7 +27,11 @@ void FADD(TranslatorVisitor& v, u64 insn, bool sat, bool cc, bool ftz, FpRoundin
         .rounding{CastFpRounding(fp_rounding)},
         .fmz_mode{ftz ? IR::FmzMode::FTZ : IR::FmzMode::None},
     };
-    v.F(fadd.dest_reg, v.ir.FPAdd(op_a, op_b, control));
+    IR::F32 value{v.ir.FPAdd(op_a, op_b, control)};
+    if (sat) {
+        value = v.ir.FPSaturate(value);
+    }
+    v.F(fadd.dest_reg, value);
 }
 
 void FADD(TranslatorVisitor& v, u64 insn, const IR::F32& src_b) {
@@ -53,15 +53,15 @@ void FADD(TranslatorVisitor& v, u64 insn, const IR::F32& src_b) {
 } // Anonymous namespace
 
 void TranslatorVisitor::FADD_reg(u64 insn) {
-    FADD(*this, insn, GetReg20F(insn));
+    FADD(*this, insn, GetRegFloat20(insn));
 }
 
-void TranslatorVisitor::FADD_cbuf(u64) {
-    throw NotImplementedException("FADD (cbuf)");
+void TranslatorVisitor::FADD_cbuf(u64 insn) {
+    FADD(*this, insn, GetFloatCbuf(insn));
 }
 
-void TranslatorVisitor::FADD_imm(u64) {
-    throw NotImplementedException("FADD (imm)");
+void TranslatorVisitor::FADD_imm(u64 insn) {
+    FADD(*this, insn, GetFloatImm20(insn));
 }
 
 void TranslatorVisitor::FADD32I(u64) {
diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_conversion_integer.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_conversion_integer.cpp
index 4d82a0009..81175627f 100644
--- a/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_conversion_integer.cpp
+++ b/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_conversion_integer.cpp
@@ -158,7 +158,7 @@ void TranslatorVisitor::F2I_cbuf(u64 insn) {
         case SrcFormat::F16:
             return IR::F16{ir.CompositeExtract(ir.UnpackFloat2x16(GetCbuf(insn)), f2i.half)};
         case SrcFormat::F32:
-            return GetCbufF(insn);
+            return GetFloatCbuf(insn);
         case SrcFormat::F64: {
             return UnpackCbuf(*this, insn);
         }
diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_fused_multiply_add.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_fused_multiply_add.cpp
index 1464f2807..758700d3c 100644
--- a/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_fused_multiply_add.cpp
+++ b/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_fused_multiply_add.cpp
@@ -51,7 +51,7 @@ void FFMA(TranslatorVisitor& v, u64 insn, const IR::F32& src_b, const IR::F32& s
 } // Anonymous namespace
 
 void TranslatorVisitor::FFMA_reg(u64 insn) {
-    FFMA(*this, insn, GetReg20F(insn), GetReg39F(insn));
+    FFMA(*this, insn, GetRegFloat20(insn), GetRegFloat39(insn));
 }
 
 void TranslatorVisitor::FFMA_rc(u64) {
@@ -59,7 +59,7 @@ void TranslatorVisitor::FFMA_rc(u64) {
 }
 
 void TranslatorVisitor::FFMA_cr(u64 insn) {
-    FFMA(*this, insn, GetCbufF(insn), GetReg39F(insn));
+    FFMA(*this, insn, GetFloatCbuf(insn), GetRegFloat39(insn));
 }
 
 void TranslatorVisitor::FFMA_imm(u64) {
diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_multiply.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_multiply.cpp
index 1b1d38be7..5c38d3fc1 100644
--- a/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_multiply.cpp
+++ b/src/shader_recompiler/frontend/maxwell/translate/impl/floating_point_multiply.cpp
@@ -91,7 +91,7 @@ void FMUL(TranslatorVisitor& v, u64 insn, const IR::F32& src_b) {
 } // Anonymous namespace
 
 void TranslatorVisitor::FMUL_reg(u64 insn) {
-    return FMUL(*this, insn, GetReg20F(insn));
+    return FMUL(*this, insn, GetRegFloat20(insn));
 }
 
 void TranslatorVisitor::FMUL_cbuf(u64) {
diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/impl.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/impl.cpp
index 079e3497f..be17bb0d9 100644
--- a/src/shader_recompiler/frontend/maxwell/translate/impl/impl.cpp
+++ b/src/shader_recompiler/frontend/maxwell/translate/impl/impl.cpp
@@ -48,11 +48,11 @@ IR::U32 TranslatorVisitor::GetReg39(u64 insn) {
     return X(reg.index);
 }
 
-IR::F32 TranslatorVisitor::GetReg20F(u64 insn) {
+IR::F32 TranslatorVisitor::GetRegFloat20(u64 insn) {
     return ir.BitCast<IR::F32>(GetReg20(insn));
 }
 
-IR::F32 TranslatorVisitor::GetReg39F(u64 insn) {
+IR::F32 TranslatorVisitor::GetRegFloat39(u64 insn) {
     return ir.BitCast<IR::F32>(GetReg39(insn));
 }
 
@@ -73,7 +73,7 @@ IR::U32 TranslatorVisitor::GetCbuf(u64 insn) {
     return ir.GetCbuf(binding, byte_offset);
 }
 
-IR::F32 TranslatorVisitor::GetCbufF(u64 insn) {
+IR::F32 TranslatorVisitor::GetFloatCbuf(u64 insn) {
     return ir.BitCast<IR::F32>(GetCbuf(insn));
 }
 
@@ -88,6 +88,17 @@ IR::U32 TranslatorVisitor::GetImm20(u64 insn) {
     return ir.Imm32(value);
 }
 
+IR::F32 TranslatorVisitor::GetFloatImm20(u64 insn) {
+    union {
+        u64 raw;
+        BitField<20, 19, u64> value;
+        BitField<56, 1, u64> is_negative;
+    } const imm{insn};
+    const f32 positive_value{Common::BitCast<f32>(static_cast<u32>(imm.value) << 12)};
+    const f32 value{imm.is_negative != 0 ? -positive_value : positive_value};
+    return ir.Imm32(value);
+}
+
 IR::U32 TranslatorVisitor::GetImm32(u64 insn) {
     union {
         u64 raw;
diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/impl.h b/src/shader_recompiler/frontend/maxwell/translate/impl/impl.h
index 27aba2cf8..4d4cf2ebf 100644
--- a/src/shader_recompiler/frontend/maxwell/translate/impl/impl.h
+++ b/src/shader_recompiler/frontend/maxwell/translate/impl/impl.h
@@ -304,13 +304,14 @@ public:
     [[nodiscard]] IR::U32 GetReg8(u64 insn);
     [[nodiscard]] IR::U32 GetReg20(u64 insn);
     [[nodiscard]] IR::U32 GetReg39(u64 insn);
-    [[nodiscard]] IR::F32 GetReg20F(u64 insn);
-    [[nodiscard]] IR::F32 GetReg39F(u64 insn);
+    [[nodiscard]] IR::F32 GetRegFloat20(u64 insn);
+    [[nodiscard]] IR::F32 GetRegFloat39(u64 insn);
 
     [[nodiscard]] IR::U32 GetCbuf(u64 insn);
-    [[nodiscard]] IR::F32 GetCbufF(u64 insn);
+    [[nodiscard]] IR::F32 GetFloatCbuf(u64 insn);
 
     [[nodiscard]] IR::U32 GetImm20(u64 insn);
+    [[nodiscard]] IR::F32 GetFloatImm20(u64 insn);
 
     [[nodiscard]] IR::U32 GetImm32(u64 insn);
 
diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/integer_add.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/integer_add.cpp
index 623e78ff8..1493e1815 100644
--- a/src/shader_recompiler/frontend/maxwell/translate/impl/integer_add.cpp
+++ b/src/shader_recompiler/frontend/maxwell/translate/impl/integer_add.cpp
@@ -84,8 +84,8 @@ void TranslatorVisitor::IADD_cbuf(u64 insn) {
     IADD(*this, insn, GetCbuf(insn));
 }
 
-void TranslatorVisitor::IADD_imm(u64) {
-    throw NotImplementedException("IADD (imm)");
+void TranslatorVisitor::IADD_imm(u64 insn) {
+    IADD(*this, insn, GetImm20(insn));
 }
 
 void TranslatorVisitor::IADD32I(u64 insn) {
diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/move_predicate_to_register.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/move_predicate_to_register.cpp
new file mode 100644
index 000000000..4324fd443
--- /dev/null
+++ b/src/shader_recompiler/frontend/maxwell/translate/impl/move_predicate_to_register.cpp
@@ -0,0 +1,66 @@
+// Copyright 2021 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/bit_field.h"
+#include "shader_recompiler/exception.h"
+#include "shader_recompiler/frontend/maxwell/translate/impl/impl.h"
+
+namespace Shader::Maxwell {
+namespace {
+enum class Mode : u64 {
+    PR,
+    CC,
+};
+} // Anonymous namespace
+
+void TranslatorVisitor::P2R_reg(u64) {
+    throw NotImplementedException("P2R (reg)");
+}
+
+void TranslatorVisitor::P2R_cbuf(u64) {
+    throw NotImplementedException("P2R (cbuf)");
+}
+
+void TranslatorVisitor::P2R_imm(u64 insn) {
+    union {
+        u64 raw;
+        BitField<0, 8, IR::Reg> dest_reg;
+        BitField<8, 8, IR::Reg> src;
+        BitField<40, 1, Mode> mode;
+        BitField<41, 2, u64> byte_selector;
+    } const p2r{insn};
+
+    const u32 mask{GetImm20(insn).U32()};
+    const bool pr_mode{p2r.mode == Mode::PR};
+    const u32 num_items{pr_mode ? 7U : 4U};
+    const u32 offset{static_cast<u32>(p2r.byte_selector) * 8};
+    IR::U32 insert{ir.Imm32(0)};
+    for (u32 index = 0; index < num_items; ++index) {
+        if (((mask >> index) & 1) == 0) {
+            continue;
+        }
+        const IR::U1 cond{[this, index, pr_mode] {
+            if (pr_mode) {
+                return ir.GetPred(IR::Pred{index});
+            }
+            switch (index) {
+            case 0:
+                return ir.GetZFlag();
+            case 1:
+                return ir.GetSFlag();
+            case 2:
+                return ir.GetCFlag();
+            case 3:
+                return ir.GetOFlag();
+            }
+            throw LogicError("Unreachable P2R index");
+        }()};
+        const IR::U32 bit{ir.Select(cond, ir.Imm32(1U << (index + offset)), ir.Imm32(0))};
+        insert = ir.BitwiseOr(insert, bit);
+    }
+    const IR::U32 masked_out{ir.BitwiseAnd(X(p2r.src), ir.Imm32(~(mask << offset)))};
+    X(p2r.dest_reg, ir.BitwiseOr(masked_out, insert));
+}
+
+} // namespace Shader::Maxwell
diff --git a/src/shader_recompiler/frontend/maxwell/translate/impl/not_implemented.cpp b/src/shader_recompiler/frontend/maxwell/translate/impl/not_implemented.cpp
index 6b2a1356b..628cf1c14 100644
--- a/src/shader_recompiler/frontend/maxwell/translate/impl/not_implemented.cpp
+++ b/src/shader_recompiler/frontend/maxwell/translate/impl/not_implemented.cpp
@@ -633,18 +633,6 @@ void TranslatorVisitor::OUT_imm(u64) {
     ThrowNotImplemented(Opcode::OUT_imm);
 }
 
-void TranslatorVisitor::P2R_reg(u64) {
-    ThrowNotImplemented(Opcode::P2R_reg);
-}
-
-void TranslatorVisitor::P2R_cbuf(u64) {
-    ThrowNotImplemented(Opcode::P2R_cbuf);
-}
-
-void TranslatorVisitor::P2R_imm(u64) {
-    ThrowNotImplemented(Opcode::P2R_imm);
-}
-
 void TranslatorVisitor::PBK() {
     // PBK is a no-op
 }