shader_decode: Implement LD_A

1 files changed, 39 insertions, 1 deletions

diff --git a/src/video_core/shader/decode/memory.cpp b/src/video_core/shader/decode/memory.cpp
index d6086004b..30e2b33a3 100644
--- a/src/video_core/shader/decode/memory.cpp
+++ b/src/video_core/shader/decode/memory.cpp
@@ -9,14 +9,52 @@
 
 namespace VideoCommon::Shader {
 
+using Tegra::Shader::Attribute;
 using Tegra::Shader::Instruction;
 using Tegra::Shader::OpCode;
+using Tegra::Shader::Register;
 
 u32 ShaderIR::DecodeMemory(BasicBlock& bb, u32 pc) {
     const Instruction instr = {program_code[pc]};
     const auto opcode = OpCode::Decode(instr);
 
-    UNIMPLEMENTED();
+    switch (opcode->get().GetId()) {
+    case OpCode::Id::LD_A: {
+        // Note: Shouldn't this be interp mode flat? As in no interpolation made.
+        UNIMPLEMENTED_IF_MSG(instr.gpr8.Value() != Register::ZeroIndex,
+                             "Indirect attribute loads are not supported");
+        UNIMPLEMENTED_IF_MSG((instr.attribute.fmt20.immediate.Value() % sizeof(u32)) != 0,
+                             "Unaligned attribute loads are not supported");
+
+        Tegra::Shader::IpaMode input_mode{Tegra::Shader::IpaInterpMode::Perspective,
+                                          Tegra::Shader::IpaSampleMode::Default};
+
+        u64 next_element = instr.attribute.fmt20.element;
+        auto next_index = static_cast<u64>(instr.attribute.fmt20.index.Value());
+
+        const auto LoadNextElement = [&](u32 reg_offset) {
+            const Node buffer = GetRegister(instr.gpr39);
+            const Node attribute = GetInputAttribute(static_cast<Attribute::Index>(next_index),
+                                                     next_element, input_mode, buffer);
+
+            SetRegister(bb, instr.gpr0.Value() + reg_offset, attribute);
+
+            // Load the next attribute element into the following register. If the element
+            // to load goes beyond the vec4 size, load the first element of the next
+            // attribute.
+            next_element = (next_element + 1) % 4;
+            next_index = next_index + (next_element == 0 ? 1 : 0);
+        };
+
+        const u32 num_words = static_cast<u32>(instr.attribute.fmt20.size.Value()) + 1;
+        for (u32 reg_offset = 0; reg_offset < num_words; ++reg_offset) {
+            LoadNextElement(reg_offset);
+        }
+        break;
+    }
+    default:
+        UNIMPLEMENTED_MSG("Unhandled memory instruction: {}", opcode->get().GetName());
+    }
 
     return pc;
 }