Merge remote-tracking branch 'origin/feat/nx2-mesh-shaders' into test/linux-all-prs

1 files changed, 23 insertions, 2 deletions

diff --git a/src/video_core/shader_environment.cpp b/src/video_core/shader_environment.cpp
index 9d189ad4f0..78438769f5 100644
--- a/src/video_core/shader_environment.cpp
+++ b/src/video_core/shader_environment.cpp
@@ -470,8 +470,16 @@ GraphicsEnvironment::GraphicsEnvironment(Tegra::Engines::Maxwell3D& maxwell3d_,
     gpu_memory->ReadBlock(program_base + start_address, &sph, sizeof(sph));
     initial_offset = sizeof(sph);
     texture_bound = maxwell3d->regs.bindless_texture_const_buffer_slot;
-    if (maxwell3d->GetNativeMacroEngine().native_turing) {
-        if (program != Maxwell::ShaderType::VertexB && program != Maxwell::ShaderType::Pixel) {
+    const auto& regs{maxwell3d->regs};
+    const bool native{maxwell3d->GetNativeMacroEngine().native_turing};
+    // Task and mesh programs: the task in the vertex slot, its mesh program in the tessellation
+    // slot (without a task, the mesh program is in the vertex slot).
+    const bool mesh_slot{native && regs.IsMeshEnabled() &&
+                         (program == Maxwell::ShaderType::VertexB ||
+                          program == Maxwell::ShaderType::Tessellation)};
+    if (native) {
+        if (program != Maxwell::ShaderType::VertexB && program != Maxwell::ShaderType::Pixel &&
+            !mesh_slot) {
             throw Shader::RuntimeError("Unsupported native graphics shader stage {}", static_cast<u32>(program));
         }
         if (std::getenv("EDEN_NATIVE_DRAW_DIAGNOSTICS")) {
@@ -516,6 +524,19 @@ GraphicsEnvironment::GraphicsEnvironment(Tegra::Engines::Maxwell3D& maxwell3d_,
         ASSERT_MSG(false, "Invalid program={}", program);
         break;
     }
+    if (mesh_slot) {
+        const bool has_task{regs.IsShaderConfigEnabled(Maxwell::ShaderType::Tessellation)};
+        stage = program == Maxwell::ShaderType::VertexB && has_task ? Shader::Stage::Task
+                                                                    : Shader::Stage::Mesh;
+        mesh_registers = {regs.reg_array[Maxwell::MeshShaderA], regs.reg_array[Maxwell::MeshShaderB],
+                          regs.reg_array[Maxwell::MeshInitShader]};
+        if (stage == Shader::Stage::Mesh && regs.IsShaderConfigEnabled(Maxwell::ShaderType::Geometry)) {
+            const auto& geometry{regs.pipelines[static_cast<size_t>(Maxwell::ShaderType::Geometry)]};
+            const GPUVAddr address{(static_cast<u64>(geometry.reserved[0] & 0xff) << 32) |
+                                   geometry.reserved[1]};
+            gpu_memory->ReadBlock(address, &mesh_primitive_header, sizeof(mesh_primitive_header));
+        }
+    }
     const u64 local_size{sph.LocalMemorySize()};
     ASSERT(local_size <= (std::numeric_limits<u32>::max)());
     local_memory_size = static_cast<u32>(local_size) + sph.common3.shader_local_memory_crs_size;