shader: Fix VertexA Shaders.

src/shader_recompiler/frontend/maxwell/program.cpp

@@ -171,20 +171,29 @@ IR::Program MergeDualVertexPrograms(IR::Program& vertex_a, IR::Program& vertex_b
     IR::Program result{};
     Optimization::VertexATransformPass(vertex_a);
     Optimization::VertexBTransformPass(vertex_b);
-    std::swap(result.blocks, vertex_a.blocks);
-    result.blocks.insert(result.blocks.end(), vertex_b.blocks.begin(), vertex_b.blocks.end());
+    for (const auto& term : vertex_a.syntax_list) {
+        if (term.type == IR::AbstractSyntaxNode::Type::Return) {
+            continue;
+        }
+        result.syntax_list.push_back(term);
+    }
+    for (const auto& term : vertex_b.syntax_list) {
+        result.syntax_list.push_back(term);
+    }
+    result.blocks = GenerateBlocks(result.syntax_list);
+    result.post_order_blocks = vertex_b.post_order_blocks;
+    for (const auto& block : vertex_a.post_order_blocks) {
+        result.post_order_blocks.push_back(block);
+    }
     result.stage = Stage::VertexB;
     result.info = vertex_a.info;
     result.local_memory_size = std::max(vertex_a.local_memory_size, vertex_b.local_memory_size);
-
     for (size_t index = 0; index < 32; ++index) {
         result.info.input_generics[index].used |= vertex_b.info.input_generics[index].used;
         result.info.stores_generics[index] |= vertex_b.info.stores_generics[index];
     }
     Optimization::JoinTextureInfo(result.info, vertex_b.info);
     Optimization::JoinStorageInfo(result.info, vertex_b.info);
-    Optimization::DualVertexJoinPass(result);
-    result.post_order_blocks = PostOrder(result.syntax_list.front());
     Optimization::DeadCodeEliminationPass(result);
     Optimization::VerificationPass(result);
     Optimization::CollectShaderInfoPass(env_vertex_b, result);

src/shader_recompiler/ir_opt/dual_vertex_pass.cpp

@@ -13,16 +13,24 @@
 
 namespace Shader::Optimization {
 
-void VertexATransformPass(IR::Program&) {
-    throw NotImplementedException("VertexA pass");
+void VertexATransformPass(IR::Program& program) {
+  for (IR::Block* const block : program.blocks) {
+      for (IR::Inst& inst : block->Instructions()) {
+          if (inst.GetOpcode() == IR::Opcode::Epilogue) {
+              return inst.Invalidate();
+          }
+      }
+  }
 }
 
-void VertexBTransformPass(IR::Program&) {
-    throw NotImplementedException("VertexA pass");
+void VertexBTransformPass(IR::Program& program) {
+  for (IR::Block* const block : program.blocks) {
+      for (IR::Inst& inst : block->Instructions()) {
+          if (inst.GetOpcode() == IR::Opcode::Prologue) {
+              return inst.Invalidate();
+          }
+      }
   }
-
-void DualVertexJoinPass(IR::Program&) {
-    throw NotImplementedException("VertexA pass");
 }
 
 } // namespace Shader::Optimization

src/shader_recompiler/ir_opt/passes.h

@@ -25,7 +25,6 @@ void VerificationPass(const IR::Program& program);
 // Dual Vertex
 void VertexATransformPass(IR::Program& program);
 void VertexBTransformPass(IR::Program& program);
-void DualVertexJoinPass(IR::Program& program);
 void JoinTextureInfo(Info& base, Info& source);
 void JoinStorageInfo(Info& base, Info& source);
 

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -40,6 +40,7 @@ namespace OpenGL {
 namespace {
 using Shader::Backend::GLASM::EmitGLASM;
 using Shader::Backend::SPIRV::EmitSPIRV;
+using Shader::Maxwell::MergeDualVertexPrograms;
 using Shader::Maxwell::TranslateProgram;
 using VideoCommon::ComputeEnvironment;
 using VideoCommon::FileEnvironment;
@@ -446,6 +447,8 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline(
     size_t env_index{};
     u32 total_storage_buffers{};
     std::array<Shader::IR::Program, Maxwell::MaxShaderProgram> programs;
+    const bool uses_vertex_a{key.unique_hashes[0] != 0};
+    const bool uses_vertex_b{key.unique_hashes[1] != 0};
     for (size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
         if (key.unique_hashes[index] == 0) {
             continue;
@@ -454,12 +457,23 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline(
         ++env_index;
 
         const u32 cfg_offset{static_cast<u32>(env.StartAddress() + sizeof(Shader::ProgramHeader))};
-        Shader::Maxwell::Flow::CFG cfg(env, pools.flow_block, cfg_offset);
+        Shader::Maxwell::Flow::CFG cfg(env, pools.flow_block, cfg_offset, index == 0);
+        if (!uses_vertex_a || index != 1) {
+            // Normal path
             programs[index] = TranslateProgram(pools.inst, pools.block, env, cfg);
 
             for (const auto& desc : programs[index].info.storage_buffers_descriptors) {
                 total_storage_buffers += desc.count;
             }
+        } else {
+            // VertexB path when VertexA is present.
+            Shader::IR::Program& program_va{programs[0]};
+            Shader::IR::Program program_vb{TranslateProgram(pools.inst, pools.block, env, cfg)};
+            for (const auto& desc : program_vb.info.storage_buffers_descriptors) {
+                total_storage_buffers += desc.count;
+            }
+            programs[index] = MergeDualVertexPrograms(program_va, program_vb, env);
+        }
     }
     const u32 glasm_storage_buffer_limit{device.GetMaxGLASMStorageBufferBlocks()};
     const bool glasm_use_storage_buffers{total_storage_buffers <= glasm_storage_buffer_limit};
@@ -472,7 +486,9 @@ std::unique_ptr<GraphicsPipeline> ShaderCache::CreateGraphicsPipeline(
     if (!device.UseAssemblyShaders()) {
         source_program.handle = glCreateProgram();
     }
-    for (size_t index = 0; index < Maxwell::MaxShaderProgram; ++index) {
+
+    for (size_t index = uses_vertex_a && uses_vertex_b ? 1 : 0; index < Maxwell::MaxShaderProgram;
+         ++index) {
         if (key.unique_hashes[index] == 0) {
             continue;
         }