nx2: title 4's intro cutscene renders correctly

Native shader capture counted CALL.REL.NOINC as a function call, so a program
using it waited for a RET after its EXIT and read on into the following
shaders until a garbage branch target made it fail. NOINC calls push no
return address (the compiler uses them as jumps; the lowering already treats
them so): count them as forward branches.

That shader builds title 4's 3D colour-grading LUT once per scene; with it
rejected the LUT stayed empty and the tonemap turned every frame black. The
intro cutscene now renders lit and graded, with subtitles, through the
bilinear stand-in for the tensor-core upscaler.

Also: rejected captures dump to EDEN_NX2_SHADER_DUMP as rejected-<addr>.bin
and report the offending instruction.

2 files changed, 17 insertions, 3 deletions

diff --git a/src/shader_recompiler/frontend/ampere/program_capture.h b/src/shader_recompiler/frontend/ampere/program_capture.h
index ea7dfb7ef6..f09bf9735d 100644
--- a/src/shader_recompiler/frontend/ampere/program_capture.h
+++ b/src/shader_recompiler/frontend/ampere/program_capture.h
@@ -19,7 +19,10 @@ public:
             const s64 target = s64{index} * 16 + 16 + relative;
             if (target < 0 || target % 16 || target >= MaximumInstructions * 16)
                 return Result::InvalidTarget;
-            if (opcode == 0x944) {
+            // CALL.REL.NOINC (bit 86) pushes no return address: the compiler uses it as a jump
+            // (the lowering treats it as one), so it never pairs with a RET.
+            const bool noinc_call = opcode == 0x944 && (instruction.words[2] & 0x400000) != 0;
+            if (opcode == 0x944 && !noinc_call) {
                 last_call_target = std::max(last_call_target, static_cast<u32>(target / 16));
             } else {
                 last_forward_target = std::max(last_forward_target, static_cast<u32>(target / 16));
diff --git a/src/video_core/shader_environment.cpp b/src/video_core/shader_environment.cpp
index fc2a26a568..5a14211073 100644
--- a/src/video_core/shader_environment.cpp
+++ b/src/video_core/shader_environment.cpp
@@ -338,8 +338,19 @@ void GenericEnvironment::InitializeNativeCode(u32 source_header_size) {
         auto& instruction = native.emplace_back();
         gpu_memory->ReadBlock(address, &instruction, sizeof(instruction));
         const auto status = capture.Observe(instruction, index);
-        if (status == Shader::Ampere::ProgramCaptureState::Result::InvalidTarget)
-            throw Shader::RuntimeError("Native shader branch outside capture bounds");
+        if (status == Shader::Ampere::ProgramCaptureState::Result::InvalidTarget) {
+            if (const char* dir = std::getenv("EDEN_NX2_SHADER_DUMP")) {
+                const auto path = fmt::format("{}/rejected-{:x}.bin", dir, program_base + native_header_size);
+                if (std::FILE* file = std::fopen(path.c_str(), "wb")) {
+                    std::fwrite(native.data(), sizeof(Instruction), std::min<size_t>(native.size(), 4096), file);
+                    std::fclose(file);
+                }
+            }
+            throw Shader::RuntimeError("Native shader branch outside capture bounds at {:#x} +{:#x} [{:08x} {:08x} {:08x} {:08x}]",
+                                       program_base + native_header_size, index * 16,
+                                       instruction.words[0], instruction.words[1],
+                                       instruction.words[2], instruction.words[3]);
+        }
         if (status == Shader::Ampere::ProgramCaptureState::Result::Complete) break;
     }
     auto lowered = Shader::Ampere::LowerToMaxwell(native);