shader_ir/warp: Implement FSWZADD

5 files changed, 44 insertions, 0 deletions

diff --git a/src/video_core/engines/shader_bytecode.h b/src/video_core/engines/shader_bytecode.h
index 8f6bc76eb..312617f71 100644
--- a/src/video_core/engines/shader_bytecode.h
+++ b/src/video_core/engines/shader_bytecode.h
@@ -616,6 +616,14 @@ union Instruction {
     } shfl;
 
     union {
+        BitField<44, 1, u64> ftz;
+        BitField<39, 2, u64> tab5cb8_2;
+        BitField<38, 1, u64> ndv;
+        BitField<47, 1, u64> cc;
+        BitField<28, 8, u64> swizzle;
+    } fswzadd;
+
+    union {
         BitField<8, 8, Register> gpr;
         BitField<20, 24, s64> offset;
     } gmem;
@@ -1590,6 +1598,7 @@ public:
         DEPBAR,
         VOTE,
         SHFL,
+        FSWZADD,
         BFE_C,
         BFE_R,
         BFE_IMM,
@@ -1888,6 +1897,7 @@ private:
             INST("1111000011110---", Id::DEPBAR, Type::Synch, "DEPBAR"),
             INST("0101000011011---", Id::VOTE, Type::Warp, "VOTE"),
             INST("1110111100010---", Id::SHFL, Type::Warp, "SHFL"),
+            INST("0101000011111---", Id::FSWZADD, Type::Warp, "FSWZADD"),
             INST("1110111111011---", Id::LD_A, Type::Memory, "LD_A"),
             INST("1110111101001---", Id::LD_S, Type::Memory, "LD_S"),
             INST("1110111101000---", Id::LD_L, Type::Memory, "LD_L"),
diff --git a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
index ca4e6e468..21c137ec5 100644
--- a/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
+++ b/src/video_core/renderer_opengl/gl_shader_decompiler.cpp
@@ -1379,6 +1379,20 @@ private:
         return GenerateUnary(operation, "float", Type::Float, type);
     }
 
+    Expression FSwizzleAdd(Operation operation) {
+        const std::string op_a = VisitOperand(operation, 0).AsFloat();
+        const std::string op_b = VisitOperand(operation, 1).AsFloat();
+        const std::string instr_mask = VisitOperand(operation, 2).AsUint();
+
+        const std::string mask = code.GenerateTemporary();
+        code.AddLine("uint {} = {} >> ((gl_SubGroupInvocationARB & 3) << 1);", mask, instr_mask);
+
+        const std::string modifier_a = fmt::format("fswzadd_modifiers_a[{} & 3]", mask);
+        const std::string modifier_b = fmt::format("fswzadd_modifiers_b[{} & 3]", mask);
+        return {fmt::format("(({} * {}) + ({} * {}))", op_a, modifier_a, op_b, modifier_b),
+                Type::Float};
+    }
+
     Expression ICastFloat(Operation operation) {
         return GenerateUnary(operation, "int", Type::Int, Type::Float);
     }
@@ -1991,6 +2005,7 @@ private:
         &GLSLDecompiler::FTrunc,
         &GLSLDecompiler::FCastInteger<Type::Int>,
         &GLSLDecompiler::FCastInteger<Type::Uint>,
+        &GLSLDecompiler::FSwizzleAdd,
 
         &GLSLDecompiler::Add<Type::Int>,
         &GLSLDecompiler::Mul<Type::Int>,
@@ -2460,6 +2475,9 @@ bvec2 HalfFloatNanComparison(bvec2 comparison, vec2 pair1, vec2 pair2) {
     bvec2 is_nan2 = isnan(pair2);
     return bvec2(comparison.x || is_nan1.x || is_nan2.x, comparison.y || is_nan1.y || is_nan2.y);
 }
+
+const float fswzadd_modifiers_a[] = float[4](-1.0f,  1.0f, -1.0f,  0.0f );
+const float fswzadd_modifiers_b[] = float[4](-1.0f, -1.0f,  1.0f, -1.0f );
 )";
 }
 
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 383720ea1..2850d5b59 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -783,6 +783,11 @@ private:
         return {};
     }
 
+    Id FSwizzleAdd(Operation operation) {
+        UNIMPLEMENTED();
+        return {};
+    }
+
     Id HNegate(Operation operation) {
         UNIMPLEMENTED();
         return {};
@@ -1363,6 +1368,7 @@ private:
         &SPIRVDecompiler::Unary<&Module::OpTrunc, Type::Float>,
         &SPIRVDecompiler::Unary<&Module::OpConvertSToF, Type::Float, Type::Int>,
         &SPIRVDecompiler::Unary<&Module::OpConvertUToF, Type::Float, Type::Uint>,
+        &SPIRVDecompiler::FSwizzleAdd,
 
         &SPIRVDecompiler::Binary<&Module::OpIAdd, Type::Int>,
         &SPIRVDecompiler::Binary<&Module::OpIMul, Type::Int>,
diff --git a/src/video_core/shader/decode/warp.cpp b/src/video_core/shader/decode/warp.cpp
index c2875eb2b..d98d0e1dd 100644
--- a/src/video_core/shader/decode/warp.cpp
+++ b/src/video_core/shader/decode/warp.cpp
@@ -94,6 +94,15 @@ u32 ShaderIR::DecodeWarp(NodeBlock& bb, u32 pc) {
             Operation(OperationCode::ShuffleIndexed, GetRegister(instr.gpr8), src_thread_id));
         break;
     }
+    case OpCode::Id::FSWZADD: {
+        UNIMPLEMENTED_IF(instr.fswzadd.ndv);
+
+        Node op_a = GetRegister(instr.gpr8);
+        Node op_b = GetRegister(instr.gpr20);
+        Node mask = Immediate(static_cast<u32>(instr.fswzadd.swizzle));
+        SetRegister(bb, instr.gpr0, Operation(OperationCode::FSwizzleAdd, op_a, op_b, mask));
+        break;
+    }
     default:
         UNIMPLEMENTED_MSG("Unhandled warp instruction: {}", opcode->get().GetName());
         break;
diff --git a/src/video_core/shader/node.h b/src/video_core/shader/node.h
index bd3547e0d..54217e6a4 100644
--- a/src/video_core/shader/node.h
+++ b/src/video_core/shader/node.h
@@ -47,6 +47,7 @@ enum class OperationCode {
     FTrunc,        /// (MetaArithmetic, float a) -> float
     FCastInteger,  /// (MetaArithmetic, int a) -> float
     FCastUInteger, /// (MetaArithmetic, uint a) -> float
+    FSwizzleAdd,   /// (float a, float b, uint mask) -> float
 
     IAdd,                  /// (MetaArithmetic, int a, int b) -> int
     IMul,                  /// (MetaArithmetic, int a, int b) -> int